.R-12/058F I September 2013 I www.er
United States
Environmental Protection
Agency
           atershed modeling to assess the
         sensitivity of streamflow, nutrient and
         sediment loads to potential climate
         change and urban development in
         20 U.S. watersheds

-------
                                                         EPA/600/R-12/058F
                                                             September 2013
Watershed Modeling to Assess the Sensitivity of Streamflow, Nutrient, and
 Sediment Loads to Potential Climate Change and Urban Development in
                          20 U.S. Watersheds
                 National Center for Environmental Assessment
                     Office of Research and Development
                     U.S. Environmental Protection Agency
                           Washington, DC 20460

-------
                                     DISCLAIMER
This document has been reviewed in accordance with U.S. Environmental Protection Agency
policy and approved for publication. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
                                      ABSTRACT
Watershed modeling was conducted in 20 large, U.S. watersheds to characterize the sensitivity of
streamflow, nutrient (nitrogen and phosphorus), and sediment loading to a range of plausible
mid-21st century climate change and urban development scenarios.  The study also provides an
improved understanding of methodological challenges associated with integrating existing tools
(e.g., climate models, downscaling approaches, and watershed models) and data sets to address
these scientific questions.  The study uses a scenario-analysis approach with a consistent set of
watershed models and scenarios applied to multiple locations throughout the nation.  Study areas
were selected to represent a range of geographic, hydrologic, and climatic characteristics.
Watershed simulations were conducted using the Soil Water Assessment Tool (SWAT) and
Hydrologic Simulation Program—FORTRAN (HSPF) models.  Scenarios of future climate
change were developed based on statistically and dynamically downscaled climate model
simulations representative of the period 2041-2070.  Scenarios of urban and residential
development for this same period were developed from the EPA's Integrated Climate and Land
Use Scenarios (ICLUS) project. Future changes in agriculture and human use and management
of water were not evaluated.

Results  provide an improved understanding of the complex and context-dependent relationships
between climate change, land-use change, and water resources in different regions of the nation.
As a first-order conclusion, results indicate that in many locations future conditions are likely to
be different from past experience. Results also provide a plausible envelope on the range of
streamflow and water quality responses to mid-21st century climate change and urban
development in different regions of the nation.  In addition, in many study areas the simulations
suggest  a likely direction of change of streamflow and water quality endpoints. Sensitivity
studies evaluating the implications of different methodological choices help to improve the
scientific foundation for conducting  climate change impacts assessments, thus building the
capacity of the water management community to understand and respond to climate change.
This information is useful to inform  and guide the development of response strategies for
managing risk.
Preferred Citation:
U.S. EPA (Environmental Protection Agency). (2013) Watershed modeling to assess the sensitivity of streamflow,
nutrient, and sediment loads to potential climate change and urban development in 20 U.S. watersheds. National
Center for Environmental Assessment, Washington, DC; EPA/600/R-12/058F. Available from the National
Technical Information Service, Alexandria, VA, and online at http://www.epa.gov/ncea.
                                            11

-------
                            TABLE OF CONTENTS
LIST OF APPENDICES	iv
LIST OF TABLES	vi
LIST OF FIGURES	ix
LIST OF ABBREVIATIONS	xii
PREFACE	xiv
AUTHORS, CONTRIBUTORS AND REVIEWERS	xv
ACKNOWLEDGEMENTS	xvi

1.     EXECUTIVE SUMMARY	1-1

2.     INTRODUCTION	2-1
      2.1.   About This Report	2-2

3.     STUDY AREAS	3-1

4.     MODELING APPROACH	4-1
      4.1.   MODEL BACKGROUND	4-2
            4.1.1.  HSPF	4-2
            4.1.2.  SWAT	4-4
      4.2.   MODEL SETUP	4-6
            4.2.1.  SWAT Setup Process	4-7
            4.2.2.  HSPF Setup Process	4-8
            4.2.3.  Watershed Data Sources	4-9
            4.2.4.  Baseline Meteorology Representation	4-14
      4.3.   SFMULATION OUTPUT AND ENDPOINTS	4-16
      4.4.   MODEL CALIBRATION AND VALIDATION	4-18
            4.4.1.  Hydrology	4-19
            4.4.2.  Water Quality	4-22
            4.4.3.  Accuracy of the Watershed Models	4-23

5.     CLIMATE CHANGE AND URBAN DEVELOPMENT SCENARIOS	5-1
      5.1.   SCENARIO-BASED APPROACH	5-1
      5.2.   CLIMATE CHANGE SCENARIOS	5-2
            5.2.1.  Future Climate Models, Sources, and Downscaling	5-2
            5.2.2.  Translation of Climate Model Projections to Watershed Model Weather
                  Inputs	5-4
      5.3.   URBAN AND RESIDENTIAL DEVELOPMENT SCENARIOS	5-12
            5.3.1.  ICLUS Urban and Residential Development Scenarios	5-14
            5.3.2.  Mapping ICLUS Housing Density Projections to NLCD Land Use
                  Categories	5-14

6.     STREAMFLOW AND WATER QUALITY SENSITIVITY TO DIFFERENT
      METHODOLOGICAL CHOICES: ANALYSIS IN THE FIVE PILOT STUDY
      AREAS	6-1
      6.1.   COMPARISON OF WATERSHED MODELS	6-1
                                      in

-------
                       TABLE OF CONTENTS (continued)
            6.1.1.  Comparison of Model Calibration and Validation Performance	6-2
            6.1.2.  Comparison of Simulated Changes Using SWAT and HSPF	6-7
            6.1.3.  Sensitivity to Increased Atmospheric CO2	6-11
      6.2.   SENSITIVITY TO DIFFERENT METHODS OF DOWNSCALING
            GCM OUTPUT	6-15
            6.2.1.  Climate Model Energy Inputs and PET Estimates	6-15
            6.2.2.  "Degraded" NARCCAP Climate Scenarios	6-16
            6.2.3.  Sensitivity of Flow and Water Quality to Approaches for Downscaling
                  GCM Projections	6-17

7.     REGIONAL SENSITIVITY OF STREAMFLOW AND WATER QUALITY TO
      CLIMATE CHANGE AND LAND DEVELOPMENT: RESULTS IN ALL
      20 WATERSHEDS	7-1
      7.1.   SELECTION OF WATERSHED MODEL FOR USE IN ALL STUDY
            AREAS	7-2
      7.2.   SENSITIVITY TO CLIMATE CHANGE SCENARIOS	7-3
      7.3.   SENSITIVITY TO URBAN AND RESIDENTIAL DEVELOPMENT
            SCENARIOS	7-18
      7.4.   RELATIVE EFFECTS OF CLIMATE CHANGE AND URBAN
            DEVELOPMENT SCENARIOS	7-20
      7.5.   SENSITIVITY TO COMBINED CLIMATE CHANGE AND URBAN
            DEVELOPMENT SCENARIOS	7-24
      7.6.   WATER BALANCE INDICATORS	7-42
      7.7.   MODELING AS SUMPTIONS AND LIMITATIONS	7-60
            7.7.1.  Model Calibration	7-62
            7.7.2.  Watershed Model Selection	7-63

8. SUMMARY AND CONCLUSIONS	8-1

REFERENCES	R-l
                            LIST OF APPENDICES
Appendix A.  Model Setup Process                                             A-l
Appendix B.  Quality Assurance Project Plan (QAPP) Section 8: Model Calibration       B-l
Appendix C.  Climate Change and the Frequency and Intensity of Precipitation Events     C-l
Appendix D.  Model Configuration, Calibration and Validation for the ACF River Basin   D-l
Appendix E.  Model Configuration, Calibration and Validation for the Arizona
            (Salt,Verde, and San Pedro) Basins                                  E-l
Appendix F.  Model Configuration, Calibration and Validation for the Susquehanna
            River Basin                                                     F-l
                                      IV

-------
                          LIST OF APPENDICES (continued)
Appendix G.  Model Configuration, Calibration and Validation for the Minnesota River
             Basin                                                                  G-l
Appendix H.  Model Configuration, Calibration and Validation for the Willamette River
             Basin                                                                  H-l
Appendix I.   Model Configuration, Calibration and Validation for the Lake
             Pontchartrain Drainages                                                  1-1
Appendix J.   Model Configuration, Calibration and Validation for the Tar and Neuse
             River Basins                                                            J-l
Appendix K.  Model Configuration, Calibration and Validation for the Nebraska (Loup
             and Elkhorn River) Basins                                               K-l
Appendix L.  Model Configuration, Calibration and Validation for the Cook Inlet Basin     L-l
Appendix M. Model Configuration, Calibration and Validation for the Georgia-Florida
             Coastal Basins                                                         M-l
Appendix N.  Model Configuration, Calibration and Validation for the Illinois River
             Basin                                                                  N-l
Appendix O.  Model Configuration, Calibration and Validation for the Lake Erie
             Drainages                                                              O-l
Appendix P.  Model Configuration, Calibration and Validation for the New England
             Coastal Basins                                                          P-l
Appendix Q.  Model Configuration, Calibration and Validation for the Rio Grande
             Valley                                                                 Q-l
Appendix R.  Model Configuration, Calibration and Validation for the Sacramento River
             Basin                                                                  R-l
Appendix S.  Model Configuration, Calibration and Validation for the Southern
             California Coastal Basins                                                S-l
Appendix T.  Model Configuration, Calibration and Validation for the South Platte
             River Basin                                                            T-l
Appendix U.  Model Configuration, Calibration and Validation for the Trinity River
             Basin                                                                  U-l
Appendix V.  Model Configuration, Calibration and Validation for the Upper Colorado
             River Basin                                                            V-l
Appendix W. Model Configuration, Calibration and Validation for the Powder and
             Tongue River Basins                                                   W-l
Appendix X.  Scenario Results for the Five Pilot Study Areas                            X-l
Appendix Y.  Scenario Results for the 15 Nonpilot Study Areas                           Y-l
Appendix Z.  Overview of Climate Scenario Monthly Temperature, Precipitation, and
             Potential Evapotranspiration                                             Z-l

-------
                                  LIST OF TABLES
3-1.    Summary of the 20 study areas	3-4
3-2.    Current (2001) land use and land cover in the 20 study areas	3-7
4-1.    Regrouping of the NLCD 2001 land-use classes for the HSPF and SWAT models	4-11
4-2.    Calculated fraction impervious cover within each developed land class for each
       study area based on NLCD 2001	4-12
4-3.    Characteristics of NRCS soil hydrologic groups	4-12
4-4.    Weather station statistics for the 20 study areas (1971-2000)	4-16
4-5.    Summary of streamflow and water quality endpoints	4-18
4-6.    Performance targets for hydrologic simulation (magnitude of annual and seasonal
       relative mean error) from Donigian (2000)	4-20
4-7.    Key hydrology calibration parameters for HSPF	4-22
4-8.    Key hydrology calibration parameters for SWAT	4-22
4-9.    Summary of SWAT model fit for initial calibration site (20 study areas)	4-27
4-10.   Summary of HSPF model  fit for initial calibration sites (five pilot study areas)	4-28
5-1.    Climate models and source of model data used to develop climate change
       scenarios	5-3
5-2.    Climate change data available from each source used to develop climate scenarios	5-6
5-3.    SWAT weather generator parameters and adjustments  applied for scenarios	5-12
5-4.    Comparison of PET estimation between different downscaling approaches	5-13
5-5.    ICLUS projected changes  in developed land area within different imperviousness
       classes by 2050	5-16
6-1.    Percent error in simulated  total streamflow volume for 10-year calibration and
       validation periods at initial and downstream calibration gages	6-3
6-2.    Nash-Sutcliffe coefficient  of model fit efficiency (E) for daily streamflow
       predictions, 10-year calibration and validation periods  at initial and downstream
       calibration gages	6-3
6-3.    Statistical comparison of HSPF and SWAT outputs at downstream station for the
       five pilot sites across all climate scenarios	6-8
6-4.    Effects of omitting simulated auxiliary meteorological  time series on Penman-
       Monteith reference crop PET estimates for "degraded" climate scenarios	6-17
6-5.    Summary of SWAT-simulated total streamflow in the five pilot study areas for
       scenarios representing different methods of downscaling	6-18
6-6.    Summary of SWAT-simulated streamflow and water quality in the Minnesota
       River study area for scenarios representing different methods of downscaling	6-19
6-7.    Range of SWAT-projected changes in annual streamflow and pollutant loads for
       combined mid-21st century NARCCAP climate change and ICLUS urban and
       residential development scenarios	6-22
7-1.    Downstream stations within each study area where simulation results are
       presented	7-1
7-2.    Average annual precipitation (in/yr and percent of baseline) for current conditions
       and mid-21st century climate scenarios	7-4
7-3.    Average annual temperature (°F and change from baseline) for current conditions
       and mid-21st century climate scenarios	7-5
                                          VI

-------
                             LIST OF TABLES (continued)
7-4.    Average annual PET (in/yr and percent of baseline) for current conditions and
       mid-21st century climate scenarios	7-6
7-5.    Average annual SWAT-simulated actual ET (in/yr and percent of baseline) for
       current conditions and mid-21st century climate scenarios	7-7
7-6.    Changes in precipitation intensity for NARCCAP mid-21st century climate
       scenarios	7-9
7-7.    Simulated total streamflow volume (climate scenarios only; percent relative to
       current conditions) for selected downstream stations	7-10
7-8.    Simulated 7-day low flow (climate scenarios only; percent relative to current
       conditions) for selected downstream  stations	7-11
7-9.    Simulated 100-year peak flow (log-Pearson III; climate scenarios only; percent
       relative to current conditions) for selected downstream stations	7-12
7-10.   Simulated changes in the number of days to streamflow centroid (climate
       scenarios only; relative to current conditions) for selected downstream stations	7-13
7-11.   Simulated Richards-Baker flashiness index (climate scenarios only; percent
       relative to current conditions) for selected downstream stations	7-14
7-12.   Simulated total suspended solids load (climate scenarios  only; percent relative to
       current conditions) for selected downstream stations	7-15
7-13.   Simulated total phosphorus load (climate scenarios only; percent relative to
       current conditions) for selected downstream stations	7-16
7-14.   Simulated total nitrogen load (climate scenarios only; percent relative to current
       conditions) for selected downstream  stations	7-17
7-15.   Projected mid-21st century impervious cover changes in study areas from ICLUS
       for A2 emissions storyline	7-21
7-16.   Simulated response to projected 2050 changes in urban and residential
       development (percent or days relative to current conditions) for selected
       downstream stations	7-22
7-17.   Simulated range of responses of mean annual streamflow to mid-21st century
       climate and land-use change at the HUC-8 and larger spatial scale	7-25
7-18.   Simulated total streamflow volume (climate and land-use change scenarios;
       percent relative to current conditions) for selected downstream stations	7-27
7-19.   Simulated 7-day low flow (climate and land-use change scenarios; percent
       relative to current conditions) for selected downstream stations	7-30
7-20.   Simulated 100-year peak flow (log-Pearson III; climate and land-use change
       scenarios; percent relative to current  conditions) for selected downstream stations	7-33
7-21.   Simulated change in the number of days to streamflow centroid (climate and land-
       use change scenarios; relative to current conditions) for selected downstream
       stations	7-36
7-22.   Simulated Richards-Baker flashiness index (climate and land-use change
       scenarios; percent relative to current  conditions) for selected downstream stations	7-39
7-23.   Coefficient of variation of SWAT-simulated changes in streamflow by study area
       in response to the six NARCCAP climate change scenarios for selected
       downstream stations	7-43
                                           vn

-------
                             LIST OF TABLES (continued)
7-24.   Coefficient of variation of SWAT-simulated changes in streamflow by
       NARCCAP climate scenario for selected downstream stations	7-44
7-25.   Simulated total suspended solids load (climate and land-use change scenarios;
       percent relative to current conditions) for selected downstream stations	7-45
7-26.   Simulated total phosphorus load (climate and land-use change scenarios; percent
       relative to current conditions) for selected downstream stations	7-48
7-27.   Simulated total nitrogen load (climate and land-use change scenarios; percent
       relative to current conditions) for selected downstream stations	7-51
7-28.   Simulated percent changes in water balance statistics for study areas (NARCCAP
       climate with land-use change scenarios; median percent change relative to current
       conditions)	7-54
                                          Vlll

-------
                                  LIST OF FIGURES
3-1.    Locations of the 20 study areas with HUC 8-digit watershed boundaries	3-2
3-2.    Distribution of precipitation and temperature among the study areas	3-3
3-3.    Apalachicola-Chattahoochee-Flint basins study area	3-9
3-4.    Arizona: Salt and Verde River section of study area	3-10
3-5.    Arizona: San Pedro River section of study area	3-11
3-6.    Cook Inlet basin study area	3-12
3-7.    Georgia-Florida Coastal Plain study area	3-13
3-8.    Illinois River basin study area	3-14
3-9.    Lake Erie drainages study area	3-15
3-10.   Lake Pontchartrain drainage study area	3-16
3-11.   Minnesota River basin study area	3-17
3-12.   Nebraska: Loup and Elkhorn River basins study area	3-18
3-13.   New England Coastal basins study area	3-19
3-14.   Powder and Tongue River basins  study area	3-20
3-15.   Rio Grande Valley study area	3-21
3-16.   Sacramento River basin study area	3-22
3-17.   Southern California Coastal basins study area	3-23
3-18.   South Platte River basin study area	3-24
3-19.   Susquehanna River basin study area	3-25
3-20.   Tar andNeuse River basins study area	3-26
3-21.   Trinity River basin study area	3-27
3-22.   Upper Colorado River basin study area	3-28
3-23.   Willamette River basin study area	3-29
4-1.    Example of weak correlation of rainfall and flow in the Dismal River at Thedford,
       NE (USGS 06775900) in the Loup River basin	4-24
6-1.    Comparison of model calibration fit to streamflow for the calibration initial site	6-4
6-2.    Sensitivity of model fit for total streamflow volume to temporal change	6-5
6-3.    Sensitivity of model fit for streamflow to spatial change	6-5
6-4.    Comparison of baseline adjusted model fit efficiency for total suspended solids
       monthly loads for calibration site  (left) and downstream site (right)	6-6
6-5.    Comparison of baseline adjusted model fit efficiency for total phosphorus
       monthly loads for calibration site  (left) and downstream site (right)	6-6
6-6.    Comparison of baseline adjusted model fit efficiency for total nitrogen monthly
       loads for calibration site (left) and downstream site (right)	6-6
6-7.    SWAT and HSPF simulated changes in total  streamflow in pilot watersheds
       (expressed relative to current conditions)	6-7
6-8.    SWAT and HSPF simulated changes in TSS  at downstream station in pilot
       watersheds (expressed relative to current conditions)	6-9
6-9.    SWAT and HSPF simulated changes in total  phosphorus load in pilot watersheds
       (expressed relative to current conditions)	6-11
6-10.   SWAT and HSPF simulated changes in total  nitrogen load in pilot watersheds
       (expressed relative to current conditions)	6-12
                                           IX

-------
                           LIST OF FIGURES (continued)


6-11.  Differences between SWAT projections of mid-21st century streamflow and water
      quality (median across six NARCCAP scenarios) with and without representation
      of increased atmospheric CC>2	6-14
6-12.  Consistency in SWAT model projections of mean annual streamflow at
      downstream stations with downscaled (NARCCAP, BCSD) and GCM projections
      oftheGFDLGCM	6-20
6-13.  Consistency in SWAT model projections of mean annual streamflow at
      downstream stations with downscaled (NARCCAP, BCSD) and GCM projections
      oftheCGCMS GCM	6-21
7-1.   Ratio of winter (January-March) to summer (July-September) runoff volume
      under current and mid-21st century NARCCAP climate scenarios	7-19
7-2.   Box plots of the distribution of the ratio of winter (January-March) to summer
      (July-September) runoff volume normalized to the ratio under current conditions	7-20
7-3.   Comparison of simulated responses of mean annual streamflow to urban
      development and climate change scenarios—HSPF model	7-24
7-4.   Simulated total future streamflow volume relative to current conditions
      (NARCCAP climate scenarios with urban development) for selected stations	7-28
7-5.   Median simulated percent changes in total future streamflow volume for six
      NARCCAP  scenarios relative to current conditions by JTUC-8 (median of
      NARCCAP  climate scenarios with urban development)	7-29
7-6.   Simulated 7-day low flow relative to current conditions (NARCCAP climate
      scenarios with urban development) for selected downstream stations	7-31
7-7.   Median simulated percent changes in 7-day average low flow volume for six
      NARCCAP  scenarios relative to current conditions by JTUC-8 (median of
      NARCCAP  climate scenarios with urban development)	7-32
7-8.   Simulated 100-year peak flow relative to current conditions (NARCCAP climate
      scenarios with urban development) for selected downstream stations	7-34
7-9.   Median simulated percent changes in 100-year peak flow for six NARCCAP
      scenarios relative to current conditions by JTUC-8 (median of NARCCAP climate
      scenarios with urban development)	7-35
7-10.  Simulated change in days to streamflow centroid relative to current conditions
      (NARCCAP climate scenarios with urban development) for selected downstream
      stations	7-37
7-11.  Median simulated change in the number of days to streamflow centroid for six
      NARCCAP  scenarios relative to current conditions by JTUC-8 (median of
      NARCCAP  climate scenarios with urban development)	7-38
7-12.  Simulated Richards-Baker flashiness index relative to current conditions
      (NARCCAP climate scenarios with urban development) for selected downstream
      stations	7-40
7-13.  Simulated absolute changes in the Richards-Baker flashiness index for six
      NARCCAP  scenarios relative to current conditions by JTUC-8 (median of
      NARCCAP  climate scenarios with urban development)	7-41

-------
                            LIST OF FIGURES (continued)
7-14.   Simulated total suspended solids load relative to current conditions (NARCCAP
       climate scenarios with urban development) for selected downstream stations	7-46
7-15.   Median simulated percent changes in total suspended solids loads for six
       NARCCAP scenarios relative to current conditions by HUC-8 (median of
       NARCCAP climate scenarios with urban development) for selected downstream
       stations	7-47
7-16.   Simulated total phosphorus load relative to current conditions (NARCCAP
       climate scenarios with urban development) for selected downstream stations	7-49
7-17.   Median simulated percent changes in total phosphorus loads for six NARCCAP
       scenarios relative to current conditions by HUC-8 (median of NARCCAP climate
       scenarios with urban development)	7-50
7-18.   Simulated total nitrogen load relative to current conditions (NARCCAP climate
       scenarios with urban development) for selected downstream stations	7-52
7-19.   Median simulated percent changes in total nitrogen loads for six NARCCAP
       scenarios relative to current conditions by HUC-8 (median of NARCCAP climate
       scenarios with urban development)	7-53
7-20.   Median simulated percent changes in watershed Dryness Ratio for six NARCCAP
       scenarios relative to current conditions (median of NARCCAP  climate scenarios
       with urban development)	7-55
7-21.   Median simulated percent changes in watershed Low Flow Sensitivity for six
       NARCCAP scenarios relative to current conditions (median of NARCCAP
       climate scenarios with urban development)	7-56
7-22.   Median simulated percent changes in watershed Surface Runoff Fraction for six
       NARCCAP scenarios relative to current conditions (median of NARCCAP
       climate scenarios with urban development)	7-57
7-23.   Median simulated percent changes in watershed Snowmelt Fraction for six
       NARCCAP scenarios relative to current conditions (median of NARCCAP
       climate scenarios with urban development)	7-58
7-24.   Median simulated percent changes in watershed Deep Recharge for six
       NARCCAP scenarios relative to current conditions (median of NARCCAP
       climate scenarios with urban development)	7-59
                                          XI

-------
                            LIST OF ABBREVIATIONS
ACE         air, climate, and energy
AET         actual evapotranspiration
ANOVA     analysis of variance
BASINS     Better Assessment Science Integrating Point and Nonpoint Sources
BCSD       bias-corrected and statistically downscaled
CAT         Climate Assessment Tool
CCSM       Community Climate System Model
cfs           cubic feet per second
CGCM3      Third Generation Coupled Global Climate Model
CMIP3       Coupled Model Intercomparison Project Phase 3
CN          curve number
CRCM       Canadian Regional Climate Model
CV          coefficient of variation
DEM        digital elevation model
E            Nash-Sutcliffe coefficient of model fit efficiency
EI'           Garrick's baseline adjusted coefficient of model fit efficiency
ET           evapotranspiration
FTable       hydraulic functional table (in HSPF)
GCM        global climate model
GFDL       Geophysical Fluid Dynamics Laboratory global climate model
GFDL hi res  Geophysical Fluid Dynamics Lab. 50-km global atmospheric time slice model
GIS          geographic information system
HadCM3     Hadley Centre Coupled Model, version 3
HRM3       Hadley Region Model 3
HRU        hydrologic response unit
HRU        Hydrologic response unit
HSG         hydrologic soil group
HSPF        Hydrologic Simulation Program—FORTRAN
HUC         hydrologic unit code
HUC-2       HUC 2-digit watershed
HUC-4       HUC 4-digit watershed
HUC-8       HUC 8-digit watershed
HUC-10      HUC 10-digit watershed
ICLUS       Integrated Climate and Land Use Scenarios
EVIPLND     impervious land segment (in HSPF)
INFILT      nominal infiltration rate parameter (in HSPF)
IPCC        Intergovernmental Panel on Climate Change
LZETP       lower zone evapotranspiration parameter
LZSN       lower soil zone nominal soil moisture storage
MSL         mean sea level
MUSLE      Modified Universal Soil Loss Equation
NARCCAP   North American Regional Climate Change Assessment Program
NARR       North American Regional Reanalysis
NCAR       National Center for Atmospheric Research
                                         xn

-------
                      LIST OF ABBREVIATIONS (continued)
ND          no data
NHD         National Hydrography Dataset
NLCD       National Land Cover Dataset
NRCS       Natural Resource Conservation Service
PCS         Permit Compliance System
PERLND     pervious land segment (in HSPF)
PET         potential evapotranspiration
PRMS       Precipitation Runoff Modeling System
QAPP       Quality Assurance Project Plan
RCHRES     stream reach segment (in HSPF)
RCM         regional climate model
RCM3       Regional Climate Model, version 3
SERGoM     Spatially Explicit Regional Growth Model
SPARROW   Spatially-Referenced Regression On Watershed attributes
STATSGO    State Soil Geographic Database
SWAT       Soil Water Assessment Tool
TMDL       total maximum daily load
TN          total nitrogen
TP           total phosphorus
TSS         total suspended solids
UCI         user control input file (in HSPF)
U.S. EPA     U.S. Environmental Protection Agency
USGS       U.S. Geological Survey
USLE       Universal Soil Loss Equation
WDM       watershed data management binary file (for HSPF)
WinHSPF     Windows interface to Hydrologic Simulation Program—FORTRAN
WRFP       Weather Research and Forecasting Model
WXGEN     weather generator (in SWAT)
                                        Xlll

-------
                                      PREFACE
This report was prepared by U.S. Environmental Protection Agency (EPA)'s Air, Climate, and
Energy (ACE) research program, located within the Office of Research and Development. The
ACE research program is designed to address the increasingly complex environmental issues we
face in the 21st century.  The overarching vision of ACE is to provide the cutting-edge scientific
information and tools to support EPA's strategic goals of protecting and improving air quality
and taking action on climate change in a sustainable manner.

Climate change presents  a risk to the availability and quality of water resources necessary to
support people and the environment. EPA, with Contractor support from Tetra Tech, Inc.,
recently completed a large-scale modeling effort to assess the sensitivity of streamflow and water
quality in different regions of the nation to a range of mid-21st century climate change and urban
development scenarios.  This report describes the methods, models, scenarios, and results of this
project.

Responding to climate change is a complex issue. The information in this report is intended to
inform and help build the capacity of EPA and EPA clients to understand and respond to the
challenge of climate change. This final report reflects consideration of peer review and public
comments received on an External Review Draft report released in March, 2013 (EPA/600/R-
12/058A).
                                          xiv

-------
                  AUTHORS, CONTRIBUTORS AND REVIEWERS
The National Center for Environmental Assessment, Office of Research and Development, was
responsible for preparing this final report. An earlier draft report was prepared by Tetra Tech,
Inc., under EPA Contracts EP-C-05-061 and EP-C-08-004.
AUTHORS

Tetra Tech, Inc.
Jonathan Butcher
Andrew Parker
Saumya Sarkar
Scott Job
Mustafa Faizullabhoy
Peter Cada
Jeremy Wyss

Texas A&M University
Raghavan Srinivasan
Pushpa Tuppad
Deb Debjani

AQUA TERRA Consultants
Anthony Donigian
John Imhoff
Jack Kittle
Brian Bicknell
Paul Hummel
Paul Duda

U.S. Environmental Protection Agency, Office of Research and Development
Thomas Johnson
Chris Weaver
Meredith Warren (ORISE Fellow)
Daniel Nover (AAAS Fellow)
REVIEWERS
This report was much improved by many excellent and thoughtful comments provided by
reviewers Dao Nguyen Khoi, Timothy Randhir, Susanna Tak Yung Tong, and Chong-Yu Xu.
We are also grateful for comments on an earlier draft of this report provided by EPA staff David
Bylsma, Chris Clark, and Steve Klein.
                                        xv

-------
                              ACKNOWLEDGEMENTS
We acknowledge and thank the entire project team at Tetra Tech, Inc., Texas A&M University,
AQUA TERRA, Stratus Consulting, and FTN Associates for their support contributing to the
development of this report.  We also thank Seth McGinnis of the National Center for
Atmospheric Research (NCAR) for processing the North American Regional Climate Change
Assessment Program (NARCCAP) output into change statistics for use in the watershed
modeling. NCAR is supported by the National Science Foundation. We acknowledge the
modeling groups, the Program for Climate Model Diagnosis and Intercomparison and the
WCRP's Working Group on Coupled Modeling for their roles in making available the WCRP
Coupled Model Intercomparison Project Phase 3 (CMIP3) multimodel data set. Support of this
data set is provided by the Office of Science, U.S. Department of Energy.
                                         xvi

-------
                             1. EXECUTIVE SUMMARY
There is growing concern about the potential effects of climate change on water resources.  The
2007 Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC)
states that warming of the climate system is now unequivocal (IPCC, 2007). Regionally variable
changes in the amount and intensity of precipitation have also been observed in much of the
United States (Groisman et al., 2012). Climate modeling experiments suggest these trends will
continue throughout the 21st century,  with continued warming accompanied by a general
intensification of the global hydrologic cycle (IPCC, 2007; Karl et al., 2009; Kharin et al., 2013).
Over the same time horizon, human population is expected to continue to increase, with
accompanying changes in land use and increased demand on water resources. In many areas,
climate change is expected to exacerbate current stresses on water resources from population
growth and economic and land-use change, including urbanization (IPCC, 2007).  Responding to
this challenge requires an improved understanding of how we are vulnerable and development of
strategies for managing future risk.

This report describes watershed modeling in 20 large, U.S. drainage basins (6,000-27,000 mi2)
to characterize the sensitivity of streamflow, nutrient (nitrogen  and phosphorus), and sediment
loading to a range of potential mid-21st century climate futures; to assess the potential interaction
of climate change and urbanization in these basins; and to improve our understanding of
methodological  challenges associated with integrating existing tools (e.g., climate models,
downscaling approaches, and watershed models) and data sets to address these scientific
questions.

Study areas were selected to represent a range of geographic, hydroclimatic, physiographic, and
land-use conditions, while also meeting practical criteria such as the availability of data to
calibrate and validate watershed models. Climate change scenarios are based on mid-21st
century climate  model projections downscaled with regional climate models (RCMs) from the
North American Regional Climate Change Assessment Program (NARCCAP; Mearns, 2009)
and the bias-corrected and statistically downscaled (BCSD) data set described by Maurer et al.
(2007).  Urban and residential development scenarios are based on the U.S. Environmental
Protection Agency (EPA)'s national-scale Integrated Climate and Land Use Scenarios (ICLUS)
project (U.S. EPA, 2009c).  Watershed modeling was conducted using the Hydrologic
Simulation Program—FORTRAN (HSPF) and Soil and  Water  Assessment Tool (SWAT)
watershed models.

Climate change  scenarios based on global climate model (GCM) simulations in the NARCCAP
and BCSD data  sets project a continued general warming trend throughout the nation over the
next century, although the magnitude of the warming varies from place to place. Wetter winters
and earlier snowmelt are likely in many of the northern and higher elevation watersheds.
Changes in other aspects of local climate, such as the timing and intensity of precipitation, show
greater variability and uncertainty. ICLUS urban and residential development scenarios project
continued growth in urban and developed land over the next century throughout the nation with
most growth occurring in and around existing urban areas. Model simulations of watershed
response to these changes provide a national-scale perspective on the range of potential changes
in streamflow and water quality in different regions of the nation.  Simulations evaluating the
                                          1-1

-------
variability in watershed response using different approaches for downscaling climate data and
different watershed models provide guidance on the use of existing models and data sets for
assessing climate change impacts.  Key findings are summarized below.

There is a high degree of regional variability in the model simulated responses of different
streamflow and water quality endpoints to a range of potential mid-21st century climatic
conditions throughout the nation. Comparison of watershed simulations in all 20 study areas
for the 2041-2070 time horizon suggests the following hydrologic changes may occur:
   •   Potential streamflow volume decreases in the Rockies and interior southwest, and
       increases in the east and southeast coasts.

   •   Higher peak streamflow will increase erosion and sediment transport; loads of nitrogen
       and phosphorus are also likely to increase in many watersheds.

   •   Many watersheds are likely to experience significant changes in the timing of streamflow
       and pollutant delivery.  In particular, there will be a tendency to shift from
       snowmelt-dominated spring runoff systems to rain-dominated systems with greater
       winter runoff.

   •   Changes in nutrient and sediment loads are generally correlated with changes in
       hydrology.
Changes in watershed water balance and hydrologic processes are likely in many regions of
the nation. Changes in streamflow are determined by the interaction of changes in precipitation
and evapotranspiration (ET).  Model simulations in this study suggest that in many regions of the
nation, the fraction of streamflow derived from surface stormflow will increase, while
groundwater-supported baseflow and recharge to deep groundwater aquifers may decrease.

The simulated responses of streamflow and water quality endpoints to climate change
scenarios based on different climate models and downscaling methodologies span a wide
range in many cases and sometimes do not agree in the direction of change.  The ultimate
significance of any given simulation of future change will depend on local context, including the
historical range of variability, thresholds and management targets, management options, and
interaction with other stressors. The simulation results in this study do, however, clearly
illustrate that the potential streamflow and water quality response in many areas could be large.
Given these uncertainties, successful climate change adaptation strategies will need to
encompass practices and decisions to reduce vulnerabilities and risk across a range of potential
future climatic conditions.

Simulated responses to increased urban development scenarios are small relative to those
resulting from climate change at the scale of modeling in this study. This is likely due to the
relatively small changes in developed lands as a percent of total watershed area at the large
spatial scale of watersheds in this study.  The finest spatial scale reported in this study is that of
an 8-digit hydrologic unit  code (HUC), and most urbanized areas are located on larger rivers
                                          1-2

-------
downstream of multiple 8-digit HUCs. Over the whole of individual study areas, urban and
residential growth scenarios represented changes in the amount of developed land on the order of
<1 to about 12% of total watershed area, and increases in impervious surfaces on the order of
0 to 5% of total watershed area. As would be expected, such small changes in development did
not have a large effect on streamflow or water quality at larger spatial scales.  It is well
documented, however, that urban and residential development at higher levels can have
significant impacts on streamflow and water quality. At smaller spatial scales where changes in
developed lands represent a larger percentage of watershed area, the effects of urbanization are
likely to be greater.  The scale at which urbanization effects may become comparable to the
effects of a changing climate is uncertain.

Simulation results are sensitive to methodological choices such as different approaches for
downscaling global climate change simulations and use of different watershed models.
Watershed simulations in this study suggest that the variability in watershed response resulting
from a single GCM downscaled using different RCMs can be of the same order of magnitude as
the ensemble variability between the different GCMs evaluated. Watershed simulations using
different models with different structures and methods for representing watershed processes
(HSPF and SWAT in this study) also resulted in increased variability of outcomes. SWAT
simulations accounting for the influence of increased atmospheric carbon dioxide (€62) on
evapotranspiration significantly affected results.  One notable insight from these results is that, in
many watersheds, increases in precipitation amount and/or intensity, urban development, and
atmospheric CO2 can have similar or additive effects on streamflow and pollutant loading (e.g., a
flashier runoff response with higher high and lower low flows).

Significance and next steps. The model simulations in this study contribute to a growing
understanding of the complex and context-dependent relationships between climate change,
land-use change, and water resources in different regions of the nation. As a first order
conclusion, results indicate that in many locations future conditions are likely to be different
from past experience.  In the context of decision making, being aware and planning for this
uncertainty is preferable to accepting a position that later turns out to be incorrect. Results also
provide a plausible envelope on the range  of streamflow and water quality responses to  mid-21st
century climate change and urban development in different regions of the nation. In addition, in
many study areas the simulations suggest a likely direction of change of streamflow and water
quality endpoints. This information can be useful in planning for anticipated but uncertain future
conditions. Sensitivity studies evaluating the implications of different methodological choices
help to improve the scientific foundation for conducting climate change impacts assessments,
thus building the capacity of the water management community to understand and respond to
climate change.

Understanding and responding to climate change is complex, and this study is only an
incremental step towards fully addressing these questions. It must be stressed that results are
conditional upon the methods, models, and scenarios used in this  study. Scenarios represent a
plausible range but are not comprehensive of all possible futures.  Several of the study areas are
also complex, highly managed systems; all infrastructure and operational aspects of water
management are not represented in full detail. Successful climate change adaptation strategies
will need to encompass practices and decisions to reduce vulnerabilities across a wide range of
plausible future climatic conditions.  It is the ultimate goal of this study to build awareness of the
                                           1-3

-------
potential range of future watershed response so that where simulations suggest large and
potentially disruptive changes, the management community will respond to build climate
resiliency.
                                           1-4

-------
                                 2. INTRODUCTION
It is now generally accepted that human activities including the combustion of fossil fuels and
land-use change have resulted, and will continue to result, in long-term changes in climate
(IPCC, 2007; Karl et al., 2009). The 2007 Fourth Assessment Report of the IPCC states that
"warming of the climate system is unequivocal, as is now evident from observations of increases
in global average  air and ocean temperatures, widespread melting of snow and ice and rising
global average sea level" (IPCC, 2007). Regionally variable changes in the amount and intensity
of precipitation have also been observed in much of the United States (Allan and Soden, 2008;
Groisman et al., 2012). Climate modeling experiments suggest these trends will continue
throughout the 21st century, with continued warming accompanied by a general intensification of
the global hydrologic cycle (IPCC, 2007; Karl et al., 2009; Kharin et al., 2013; Emori and
Brown,  2005).  While uncertainty remains, particularly for precipitation changes at regional
spatial scales, the  presence of long-term trends in the record suggests many parts of the United
States could experience future climatic conditions unprecedented in recent history.

Water managers are faced with important questions concerning the implications of climate
change for water resources. Changes in climate will vary over space and time.  The hydrologic
response to climate change will be further influenced by the attributes of specific watersheds,
including physiographic setting, land use, pollutant sources, and human use and management of
water. Runoff is generally expected to increase at higher latitudes and in some wet tropical
areas, and decrease over dry and semiarid regions at mid-latitudes due to decreases in rainfall
and higher rates of evapotranspiration (IPCC, 2007; Karl et al., 2009).  Northern and
mountainous areas that receive snow in the winter are likely to see increased precipitation
occurring as rain versus snow. In addition, most regions of the United States are anticipated to
experience increasing intensity of precipitation events; that is, warming-induced intensification
of the global hydrologic cycle will increase the fraction of total precipitation occurring in large
magnitude events. Precipitation changes can result in hydrologic effects that include changes in
the amount and seasonal timing of streamflow, changes in soil moisture and groundwater
recharge, changes in land cover and watershed biogeochemical cycling, changes in nonpoint
pollutant loading to water bodies, and increased demands on water infrastructure,  including
urban stormwater  and other engineered systems.  Regions exposed to increased storm intensity
could experience increased coastal and inland flooding.  Such changes challenge the assumption
of stationarity that has been the foundation for water management for decades (e.g., Milly et al.,
2008).

Changes in climate and hydrology will also affect water quality.  Although less studied, potential
effects include changes in stream temperature and hydrologic controls on nutrient, sediment, and
dissolved constituent loads to water bodies.  Hydrologic changes associated with climate change
could also influence pollutant loading from urban and agricultural lands.  Previous studies
illustrate the sensitivity of stream nutrient loads,  sediment loads, and ecologically relevant
streamflow characteristics to changes in climate (e.g., see Poff et al., 1996; Williams  et al., 1996;
Murdoch et al., 2000; Monteith et al., 2000; Chang et al., 2001; Bouraoui et al., 2002; SWCS,
2003; Marshall and Randhir, 2008; Wilson and Weng, 2011; Tong et al., 2011). A review
(Whitehead et al., 2009) details progress on these questions but emphasizes that still relatively
little is known about the link between climate change and water quality.
                                           2-1

-------
Many watersheds are currently impacted by existing stressors, including land-use change, water
withdrawals, pollutant discharges, and other factors.  It is important to recognize that climate
change will not act independently, but will interact in complex and poorly understood ways with
existing and future changes in nonclimatic stressors.  One area of concern is the interaction of
climate change and urban development in different watershed settings.  Throughout this century,
urban and residential development is expected to increase throughout much of the nation
(U.S. EPA, 2009c). Stormwater runoff from roads, rooftops, and other impervious surfaces in
urban and residential environments is a well-known cause of stream impairment (Walsh et al.,
2005; Paul and Meyer, 2001).  Changes in rainfall associated with climate change will have a
direct effect on stormwater runoff (Pyke et al., 2011). More generally, changes in climate could
exacerbate or ameliorate the impacts of other nonclimatic stressors. This understanding is
particularly important because  in many situations, the only viable management strategies for
adapting to future climatic conditions involve improved methods for managing and addressing
nonclimatic stressors.

Understanding and adapting to climate change is complicated by the scale, complexity, and
inherent uncertainty of the problem. We currently have a limited ability to predict long-term
(multidecadal) future climate at the local and regional scales needed by  decision makers
(Sarewitz  et al., 2000).  It is therefore not possible to know with certainty the future climatic
conditions to which a particular watershed will be exposed. Scenario analysis using simulation
models is  a useful and common approach for assessing vulnerability to plausible but uncertain
future conditions (Lempert et al., 2006; Sarewitz et al.,  2000; Volkery and Ribeiro, 2009).
Evaluation of multiple scenarios can provide understanding of the complex interactions
associated with watershed response to climate  change and other watershed stressors, and identify
uncertainties associated with changes in different drivers  (such as climate and land-use change)
and uncertainties associated with different analytical approaches and methods. This information
is useful for developing an improved understanding of system behavior  and sensitivity to a wide
range of plausible future climatic conditions and events, identifying how we are vulnerable to
these changes, and ultimately to guide the development of robust strategies for reducing risk in
the face of changing climatic conditions (Sarewitz et al., 2000; Lempert et al., 2006; Johnson and
Weaver, 2009).

2.1.  ABOUT  THIS REPORT
This report describes the structure—including  methods, models, scenarios, and results—of a
large-scale watershed modeling study designed to address gaps in our knowledge of the
sensitivity of U.S. streamflow,  nutrient (nitrogen and phosphorus), and sediment loading to
potential mid-21st century climate change.  Modeling also considers the potential interaction of
climate change with future urban and residential development in these watersheds and provides
insights concerning the effects  of different methodological choices (e.g., method of downscaling
climate change data, choice of  watershed model, etc.) on  simulation results.

Watershed modeling was conducted in 20 large U.S. watersheds using a scenario analysis
approach.  Study sites were selected to represent a range of geographic,  hydrologic,  and  climatic
characteristics throughout the nation.

Model projections consider the effects of climate change  alone, urban and residential
development alone, and the combined effects of climate change and urban development  on

                                           _

-------
streamflow, total nitrogen (TN), total phosphorus (TP), and total suspended solids (TSS) loads.
Climate change scenarios were based on downscaled climate model projections from two
sources; the NARCCAP and the BCSD archive from the Bureau of Reclamation/Santa Clara
University/Lawrence Livermore.  Scenarios of urban and residential development were based on
projections from EPA's ICLUS project.

All 20 watersheds were modeled with the SWAT model using a consistent set of climate and
land-use change scenarios. In a subset of five study watersheds, referred to as pilot sites,
additional simulations were conducted to address methodological questions related to the
conduct of climate change impacts assessments. In these watersheds, a second watershed model,
the HSPF, was run using the same climate and land-use scenarios used with SWAT to assess the
influence of different watershed models on watershed simulations. Pilot watersheds were also
evaluated for additional climate change scenarios to assess hydroclimatic sensitivity to different
methods of downscaling climate data. All watershed models are constructed at a scale
approximating HUC-lOs, but the finest spatial resolution of model calibration and output was on
the order of HUC-8 watersheds.

As with any study of this type, simulation results are conditional on the specific methods,
models, and scenarios used. Given the difficulty and level of effort involved with modeling at
this scale,  it was necessary to standardize model development for efficiency. Several of the
study areas are complex, highly managed systems. We do not attempt to represent all these
operational aspects in full detail. Future changes in agriculture and human use and management
of water were also not evaluated.

This report consists of a main volume and 26 appendices. The main volume describes the study
methods, models,  scenarios, and  results.  The appendices contain additional information on
model setup, calibration, and additional modeling results (at HUC 8-digit spatial  scale) not
included in the main report. Supplementary data sets summarizing SWAT simulation results at
all 20 study areas are also available at EPA's ICLUS web page
http://map3.epa.gov/ICLUSonline/.
                                          2-3

-------
                                  3.  STUDY AREAS
This project evaluates watershed response to climate change and urban development scenarios in
20 large drainage basins, ranging in size from approximately 6,000 to 27,000 mi2, located
throughout the contiguous United States and Alaska (see Figure 3-1 below).  Study areas were
selected based on both geographic and practical considerations. Sites were selected to represent
a broad range of geographic, physiographic, land use, and hydroclimatic settings (see Table 3-1).
Site selection also considered the availability of necessary data for calibration and validation of
watershed models, including a selection of U.S. Geological Survey (USGS) streamflow
monitoring gages (at varying spatial scales) and an adequate set of water quality monitoring data
(e.g., USGS National Water Quality Assessment study areas).  Finally, study areas were selected
to leverage, where possible, preexisting calibrated watershed models.

The 20 study areas selected cover a wide range of geology and climate (see Table 3-1), with
elevations ranging from sea level to over 14,000 feet, average annual temperatures from 34 to
68°F, and average annual precipitation ranging from 15 to 66 inches. Figure 3-2 shows the
distribution of average annual precipitation and temperature among the study sites, indicating a
wide range of climatic conditions, from dry to wet and cold to warm. The ratio of winter
(January-March) to summer (July-September) precipitation varies from about 0.1 to 11 while
the fraction of runoff derived from snowmelt ranges from 0 to 54%. The study areas also sample
all of the Level I ecoregions in the contiguous United States (CECWG, 1997), with the exception
of the Tropical Wet Forests ecoregion (present within the contiguous United States only in
southern Florida). Many of the study areas are in the Eastern Temperate Forests ecoregion, but
this region occupies most of the eastern half of the contiguous United States.

The selected study areas also cover a range of land-use conditions, with agricultural land
occupying from 0 to 78% of the land area and urbanized areas (impervious plus developed
pervious land) occupying up to 38%. Overall imperviousness  of the study areas (at
approximately the FIUC-4 scale) ranges from near zero to about 14%; however, individual
subwatersheds within a study area have substantially greater imperviousness.  For instance,
within the Apalachicola-Chattahoochee-Flint River watersheds (ACF) study area the individual
modeling subbasins  (at approximately the FIUC-10 scale) range from 0.15 to 27.44%
impervious.

A detailed summary of current land use and land cover in the 20 study areas is shown in Table
3-2, based on 2001 data from the National Land Cover Dataset (NLCD).
                                           5-1

-------
to
                                        Powder and Tongue
                                                           Nebraska: Loup and
                                                           Elkhorn River Basin
                                                 Lake Erie
                                                 Drainages
                                              Minnesota
                                              River Basin
               Willamette
               River
               Basin
                                                         New England
                                                         Coastal Basins
                                   Upper Colorado
                                   River Basin
                                                                      Susquehanna
                                                                      River Basin
                                                                                                                      Tar and Neuse
                                                                                                                      River Basins
                Southern
                California
                Coastal
                Basins
                                                                         Atlantic
                                                                          Ocean
      Rio Grande
      Valley
                                                                                                                 Apalachicola-
                                                                                                                 Chattahoochee-
                                                                                                                 Flint Basins
                                                Arizona:         Trinity
                                                Salt, Verde,      Rjver Basjn
                                                and San Pedro
                                                                                          Georgia-Florida
                                                                                          Coastal Plain
                                                                          Lake
                                                                          Pontchartrain
                                                                          Drainage
                                                                                       Gulf of
                                                                                       Mexico
                                                                                                        1,200
                                                                                                            Kilometers
   Study Areas

NAD 1983 Albers meters
Gulf of
Alaska
            Figure 3-1. Locations of the 20 study areas with HUC 8-digit watershed boundaries.

-------
70
60
"c"
750 -
0
s
t40 -
CU
£
"ro
| 30 -
cu
I20
10
0 -
3

» LPont
» Willa
* ACF* GaFIa
.. r * TarNeu
* NewEng

* Susq • Trin
* lllin * Sac
LErie
* Cook » Minn
* Neb
SOP';V°wTon *Ariz*S°Cal
RioGra







0 40 50 60 70 80
Average Annual Temperature (F)
       Figure 3-2. Distribution of precipitation and temperature among the study
       areas.

Note: Precipitation and temperature are averages over the weather stations used in simulation for the modeling period
(approximately 1970-2000, depending on model area).

The USGS (Seaber et al., 1987) has classified watershed drainage areas in a hierarchical system
in which each hydrologic unit is assigned a Hydrologic Unit Code (HUC).  The first four levels
of the hierarchy (occupying eight digits) identify the region (HUC-2), subregion (HUC-4), basin
(HUC-6), and subbasin (HUC-8). The United States contains 222 HUC-4s with an average size
of 16,800 mi2. The 20 study areas selected for this study are of a similar scale to HUC-4 basins,
ranging in size from approximately 6,000 to 27,000 mi , but do not correspond exactly with the
boundaries of established HUC-4 basins. Each study area comprises from 7 to 19 HUC  8-digit
watersheds.  The individual HUC 8-digit watersheds in the study areas have a median size of
1,164 mi2, and an interquartile range from 805 to 1,808 mi2. In some cases study areas are
composed of a single, contiguous watershed. In other cases, study areas include several  adjacent
but noncontiguous watersheds (e.g., separate rivers draining to the coast).  Where possible,
watersheds strongly influenced by upstream dams, diversions, or other human interventions were
avoided to simplify modeling.

Maps of the individual study areas are provided in Figures  3-3 through 3-23.  Detailed
descriptions of each study area are presented in Appendices D through W, which describe model
development and calibration for the individual study areas.
                                            5-3

-------
Table 3-1. Summary of the 20 study areas

Study area
Apalachicola-
Chattahoochee-Flint Basins
(Pilot Site)
Arizona: Salt, Verde, and
San Pedro (Pilot Site)

Cook Inlet Basin

Georgia-Florida Coastal
Plain
Illinois River Basin

Lake Erie Drainages

Lake Pontchartrain
Drainage
Minnesota River Basin
(Pilot Site)

Site ID
ACF

Ariz

Cook

GaFla
Illin

LErie

LPont
Minn


Location
(states)
GA, AL,
FL

AZ

AK

GA,FL
IL, IN, WI

OH, IN,
MI

LA, MS
MN,IA,
SD

Total
area
(mi2)
19,283

14,910

22,243

17,541
17,004

11,682

5,852
16,989


Elevation
range
(ft MSL)
0-4,347

1,918-11,407

0-18,882

0-485
365-1,183

339-1,383

0-502
683-2,134


Average
precip
(in/yr)
54.26

19.67

28.50

53.21
38.25

38.15

66.33
28.26


Averag
etemp
(°F)
63.43

56.81

34.16

68.24
49.00

49.10

66.64
43.90


Ratio winter
to summer
runoff
2.01

2.06

0.11

1.29
1.24

2.60

1.70
0.50

Fraction of
runoff as
snowmelt
(%)
0.7

9.3

53.8

0.1
13.3

13.4

0.5
14.8


Level I
ecoregions
Eastern Temperate
Forests

Temperate Sierras,
Southern Semi-arid
Highlands, North
America Deserts
Marine West Coast
Forests, Northwest
Forested Mountains
Eastern Temperate
Forests
Eastern Temperate
Forests

Eastern Temperate
Forests

Eastern Temperate
Forests
Great Plains,
Eastern Temperate
Forests

Major cities
Atlanta, GA

Flagstaff, AZ;
Sierra Vista, AZ

Anchorage, AK

Tallahassee, FL;
Tampa, FL
Chicago, IL;
Milwaukee, WI;
Peoria, IL
Fort Wayne, IN;
Cleveland, OH;
Akron, OH
New Orleans,
LA;
Baton Rouge, LA
Mankato, MN,
Minneapolis, MN

-------
Table 3-1. Summary of the 20 study areas (continued)
Study area
Nebraska: Loup and
Elkhom River Basins
New England Coastal
Basins
Powder and Tongue River
Basins


Rio Grande Valley

Sacramento River Basin


Southern California Coastal
Basins
South Platte River Basin

Susquehanna River Basin
(Pilot Site)
Site ID
Neb
NewEng
PowTon


RioGra

Sac


SoCal
SoPlat

Susq
Location
(states)
NE
MA,NH,
ME
MT,WY


NM,CO

CA


CA
CO,WY

PA, NY,
MD
Total
area
(mi2)
22,095
10,359
18,800


18,959

8,316


8,322
14,668

27,504
Elevation
range
(ft MSL)
1,069-4,292
0-5,422
2,201-13,138


4,726-14,173

17-10,424


0-11,488
4,291-14,261

0-3,141
Average
precip
(in/yr)
26.10
48.45
17.70


15.18

37.47


20.21
16.82

41.30
Average
temp
48.35
46.23
44.15


44.71

57.45


61.20
43.46

48.26
Ratio winter
to summer
runoff
0.91
1.41
1.18


0.52

1.61


5.94
0.49

2.06
Fraction of
runoff as
snowmelt (%)
12.6
21.1
30.2


23.8

17.6


4.9
28.3

16.6
Level I ecoregions
Great Plains
Northern Forests,
Eastern Temperate
Forests
Great Plains, North
American Deserts,
Northwestern
Forested Mountains
Northwest Forested
Mountains, North
American Deserts,
Temperate Sierras
Mediterranean
California,
Northwest Forested
Mountains
Mediterranean
California
Great Plains,
Northwest Forested
Mountains
Eastern Temperate
Forests, Northern
Forests
Major cities
No major cities
Portland, ME,
Greater Boston,
MA
No major cities


Santa Fe, NM;
Albuquerque, NM

Chico, CA;
Reading, CA


Greater Los
Angeles, CA
Fort Collins, CO;
Denver, CO
Scranton, PA;
Harrisburg, PA

-------
        Table 3-1.  Summary of the 20 study areas (continued)
Study area
Tar and Neuse River
Basins
Trinity River Basin
Upper Colorado River
Basin
Willamette River Basin
(Pilot Site)
Site ID
TarNeu
Trin
UppCol
Willa
Location
(states)
NC
TX
CO,UT
OR
Total
area
(mi2)
9,972
17,949
17,865
11,209
Elevation
range
(ft MSL)
0-854
0-2,150
4,323-14,303
8-10,451
Average
precip
(in/yr)
49.91
40.65
16.36
58.38
Average
temp
(°F)
59.91
64.78
41.73
51.19
Ratio winter
to summer
runoff
1.59
1.45
0.31
10.99
Fraction of
runoff as
snowmelt (%)
3.3
1.6
42.4
4.5
Level I ecoregions
Eastern Temperate
Forests
Great Plains,
Eastern Temperate
Forests
Great Plains,
Eastern Temperate
Forests
Marine West Coast
Forests, Northwest
Forested Mountains
Major cities
Raleigh, NC;
Durham, NC;
Greenville, NC
Dallas, TX
Grand Junction,
CO;
Edwards, CO
Portland, OR;
Salem, OR;
Eugene, OR
MSL = mean sea level

Notes: Precipitation and temperature are averages over the weather stations used in simulation for the modeling period (approximately 1970-2000, depending on model area).
The ratio of winter (January-March) to summer (July-September) runoff and the fraction of runoff as snowmelt are derived from the calibrated SWAT model applications
described in this report.

-------
Table 3-2. Current (2001) land use and land cover in the 20 study areas
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Total area
(mi2)
19,283
14,910
22,243
17,541
17,004
11,682
5,852
16,989
22,095
10,359
18,800
18,959
Water
(%)
1.8
0.2
2.55
0.9
1.9
1.1
3.3
3.0
0.8
4.2
0.1
0.3
Barren
(%)
0.4
0.3
18.97
0.4
0.1
0.1
0.4
0.1
0.1
0.5
0.7
1.0
Wetland
(%)
9.3
0.3
7.59
25.7
1.4
2.7
32.3
4.9
3.2
7.6
1.7
2.1
Forest
(%)
47.9
41.9
24.10
33.5
10.3
13.0
23.1
2.9
1.1
63.6
10.0
35.3
Shrub
(%)
9.6
56.0
38.11
10.1
2.1
1.5
14.3
4.6
64.5
2.2
85.5
54.2
Pasture/hay
(%)
9.1
0.1
0.05
7.2
3.6
5.8
10.3
5.9
1.1
4.5
0.6
4.1
Cultivated
(%)
12.4
0.1
0.11
10.9
62.6
61.2
4.5
72.1
26.5
1.1
1.0
0.7
Developed
pervious*
(%)
7.3
1.0
0.58
8.8
11.9
11.2
8.5
5.5
2.4
10.8
0.4
1.7
Impervious
(%)
2.0
0.2
0.24
2.5
6.2
3.5
3.2
1.1
0.4
5.6
0.1
0.5
Snow/ice
(%)
0.0
0.0
7.70
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

-------
             Table 3-2. Current (2001) land use and land cover in the 20 study areas (continued)
Study area
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
Total area
(mi2)
8,316
8,322
14,668
27,504
9,972
17,949
17,865
11,209
Water
(%)
0.5
0.6
0.9
1.1
4.5
3.7
0.5
0.9
Barren
(%)
0.5
0.6
1.0
0.4
0.2
0.3
3.8
0.9
Wetland
(%)
2.0
0.4
2.3
1.2
14.1
7.8
1.6
1.8
Forest
(%)
22.4
10.6
23.7
61.1
33.5
16.4
53.9
56.2
Shrub
(%)
48.3
50.9
46.4
1.8
10.0
30.6
33.9
12.3
Pasture/hay
(%)
2.3
1.0
1.5
17.1
7.3
20.6
3.2
12.5
Cultivated
(%)
19.7
2.8
16.5
9.8
21.1
7.0
1.1
8.2
Developed
pervious*
(%)
3.6
19.4
5.0
5.9
7.7
9.4
1.0
4.7
Impervious
(%)
0.7
13.8
2.1
1.5
1.7
4.2
0.4
2.5
Snow/ice
(%)
0.0
0.0
0.7
0.0
0.0
0.0
0.7
0.0
oo
     *Developed pervious land includes the pervious portion of open space and low, medium, and high density land uses.

-------
                                                                          Legend
                                                                              Hydrography
                                                                              Water (Nal. Atlas Dataset)
                                                                              US Census Populated Places
                                                                          ^•1 Municipalities (pop 2 50.000)
                                                                              County Boundaries
                                                                              Watershed with HUCSs
                                                          AthensCarke
                                                          County
                          ChaUahoochee-
                       __   Lake Harding
                                             03130005 t    ^
-------
    Hydrography

    Water (Nat. Alias Dataset)

    US Census Populated Places

^•1 Municipalities (pop > 50 COO)

   | County Boundaries

    Watershed withHUCSs
                                 Salt and Verde
                                 River Basins
                                                                      San Pedro
                                                                      River Basin
              Big
       C h i no -W i 11 i a m s o n
            Valley
          (15060201
                                  Lower Verd
                                  (15060203)
                                           Tonto
                                        (15060105)
                                                                        Carrizo
                                                                       (15060104) ^

                                                                                  White
                                                                                (15060102)
Upper Salt
(15060103)
seven
ervoir
                                                                                    Black
                                                                                  (15060101)
GCRP Model Areas - Salt and Verde River Basins
                Base Map
  Figure 3-4.  Arizona: Salt and Verde River section of study area.
                                                3-10

-------
                                                                                      Salt and Verde
                                                                                      River Basins
                                                                      San Pedro
                                                                      River Basin
                                Lower San Pedro
                                  (15050203)
                                                                  San Pedro
                                                                  River
                                       Upper San Pedro
                                          (15050202)
                                                                                   Hydrography
                                                                                   Interstate
                                                                                   Water (Nat. Atlas Dataset)
                                                                                   US Census Populated Places
                                                                                   Municipalities (pop a 50,000)
                                                                                   County Boundaries
                                                                                   Watershed with HUC8s
GCRP Model Areas - San Pedro River Basin
              Base Map
 Figure 3-5. Arizona: San Pedro River section of study area.
                                                3-11

-------
      Hydrography

 ^H Water (Nat. Atlas Dataset)

      US Census Populated Places

 ^H Municipalities (pop 2 50,000)

      County Boundaries
 Chulitna
  River
(19020502)
    ] Watershed with HUCSs
                                                                   Talkeetna
                                                                     River
                                                                  (19020503)
                             Upper Susitna
                                River
                              (19020501)
                                                  Lower Susitna
                                                      River
                                                    (19020505)
                               Upper Kenai
                                Peninsula
                               (19020302)
 Anchor Point
                                      Gw/fof
                                      Alaska
                                          GCRP Model Areas - Cook Inlet Basin
                                                      Base Map
Figure 3-6.  Cook Inlet basin study area.

-------
                                    Alapaha
                                   0311*02021
                                   Hydrography
                                   Water (Nat Atlas Dataset)
                                   US Census Populated Places
                              ^H Municipalities (pop > 50,000)
                              I    I County Boundaries
                               ~~] Watershed with HUCSs
                                                         Withlacoochee
                Ochlockonee
                 (03120002)
                                                       Upper Suwannee
                                                        i (03110201)
                                                                                    Jacksonville
                                   Aucilla
                                  (03110103)
                                                  Lower
                                                 Suwannee
                                                 (03-110205)
                                                                   S (03110206)
                                                                   r  ii—    •
                                                            PithlacKaotee
                                                              (03100207)
                                                     Spring H
                                                    _,^HH
                                                    Tampa
                                    illsboTouqn
                                    0*3100205)
  GULF OF
  MEXICO
St. Petersburg
GCRP Model Areas - Georgia-Florida Coastal Plain
                Base Map
 Figure 3-7. Georgia-Florida Coastal Plain study area.
                                               3-13

-------
    Hydrography

    Water (Nat. Atlas Dalasel)

    US Census Populated Places

•I Municipalities (pop > 50,000)

I   I County Boundaries

    Watershed with HUCSs
         Grand Rapids


           Michigan
                                                                      Lake
                                                                   Michigan
                                     Lower Fox
                                     [071200.07,)
Kankakee
(07J-20'0'01)
                       Lower Mlmois-
                      Senachwine
                                          Vermilion
                                          (07130002)
                               Mackinaw
                               (07130004)
                                 Bloomington
              Lowerlllinpis-
             Lake'Chautauaua
                (07130003)
      GCRP Model Areas - Illnois River Basins
                  Base Map
  Figure 3-8. Illinois River basin study area.
                                                 3-14

-------
 Legend
     Hydrography
     Water (Nat. Atlas Dataset)
     US Census Populated Places
 H Municipalities (pop £ 50.000)
 I   I County Boundaries
     Watershed withHUCSs
                                  Ti,ffin
                                (04100006)
                                                                                                Black-Roky
                                                                                                 ,041100,
Lower Maumee
                                          Huron-
                                         Vermilion
                                         (04100012)
               St. Joseph
               (04100003)
                                                                      Sandusky
                                                                      (04100011)
                                                   Blanchard
                                                   oin oooos)
                                Auglai
                               (04100007)
                                                                            Columbus
                                         GCRP Model Areas - Lake Erie Drainages
                                                      Base Map
Figure 3-9.  Lake Erie drainages study area.

-------
                                                                                Hydrography

                                                                                Water (Nal Atlas Dataset)

                                                                                US Census Populated Places

                                                                            •I Municipalities (pop 2 50.000)

                                                                            I   I County Boundaries

                                                                                Watershed with HUCSs
                                                                 Mississippi
                                                                       »——
                                                                       Louisiana
         Bayou Sara-
          Thompson
          (08070201)
Tangipahoa
 08070205)
                            Amite
                          (08070202)
                                          Tickpaw
                                        -(08070203)
                                                                       Liberty Bayou-
                                                                        Teh efun eta
                                                                        (08090201)
         Baton Rouge.
                                                                    Lake
                                                               Pontchartrain
                                     Lake Maurepas
                                       08070204)
            Eastern Louisiana
                Coastal
               (08090203)
                                                                                    New Orleans
GCRP Model Areas - Lake Pontchartrain Drainage
                Base Map
Figure 3-10. Lake Pontchartrain drainage study area.
                                              3-16

-------
                                                                                Hydrography

                                                                                Water (Nal. Atlas Dataset)

                                                                                US Census Populated Places

                                                                                Municipalities (pop 2 50.000)

                                                                                County Boundaries

                                                                                Watershed with HUCSs
„ North Dakota
        pper Minnesota
         (07020001)
                                                                                       Minneapolis
                         Hawk^Yellow Medicine
                                                                   :ower Minnesota
                                                                     (070200012)  {
                                  edwood
                                (07020006)
                                                  Middle Minnesota
                                                     (07020007)
                                          Cottonwood
                                          (07020008)
                                                                                    -- Mankato
South Dakota
                                                                                  Le Sueur
                                                                                  (07020011)

                                                                                  _L
                                                        Watonwan
                                                        (07020010)
                                                                         Blue Earth
                                                                         (07020009)
   GCRP Model Areas - Minnesota River Basin
                Base Map
  Figure 3-11.  Minnesota River basin study area.
                                               3-17

-------
oo
                   Hydrography

                   Water (Nat Alias Dataset)

                   US Census Populated Places

                   Municipalities (pop 2 50,000)

                   County Boundaries
                 "2 Watershed with HUCSs
                                    Upper North Loup
                                       (10210006)
                                                                    North ^i
                                                                    Fork   ]l    Logan
                                                                   Elkhorn \\ (1022'0004
                                                                   0220002
                                                                                     Upper Elkhorn
                                                                                       (10220001)
                      Upper Middle Loup	
                       '  (10210001)
                           Calamus
                          (10210008)
                                     Dismal
                                   (10210002)
                                                                            Lower Elkhorn
                                                                             (10220003)    /
                                                                                                                           Loup
                                                                                                                         (10210009)
                                                                                                                North Loup
                                                                                                             (102100071)
North    South L'pup
Platte     (10210604)
                                                                                                    Lower Middle Loup
                                                                                                        (10210003)
                                                                                         —X _M(Jd_
                                                                                              (10210005)
                                                  GCRP Model Areas - Nebraska: Loup and Elkhorn River Basins
                                                                       Base Map
              Figure 3-12. Nebraska: Loup and Elkhorn River basins study area.

-------
   - Hydrography

•I Water (Nat. Atlas Dataset)

    US Census Populated Places

^H Municipalities (pop > 50,000)

|   | County Boundaries

    Watershed with HUCSs
 New Hampshire


W-
                                                                    Presumpscot
                                                                     (01060001)
                                                    Stfco
                                                  (01060002)
                        Pemigewasset
                          (01070001)
                                        Winnipesaukee  i     i
                                          (01070002)-
                                                         Pisc'ataqua-Salmon Falls
                                           Merrimack
                                           (01070006)

                                        1
                           Contoocook
                            (01070003)
                                                                    Charles
                                                                   (01090001)
                                                                                   Lowell

                                                                                  Lynn

                                                                                 Boston
             Nashua
            (01070004)
                  Lr
                                                      oncord
                                                    (01070005)
                                                                       Taunton
                                                       Rhode
                                                       Island
 GCRP Model Areas - New England Coastal Basins
                 Base Map
  Figure 3-13. New England Coastal basins study area.
                                                3-19

-------
     Hydrography

•I Water (Nat Atlas Dataset)

     US Census Populated Places

•I Municipalities (pop ? 50.000)

I    I County Boundaries

     Watershed withHUCSs
                                                                        Powder
                                                                       (10090207)
                                            Upper Tongue
                                             (10090101)
                                                                             Little
                                                                            Powder
                                                                          (10090208)
  Clear
(10090206)
                                                Crazy
                                               Woman
                                              (10090205)1 Upper Powder
                                                          (10090202)
                                          iddle
                                         Fork
                                        Powder
                                       (10090201)
                                    South Fork
                                      Powder
                                    (10090203
          Salt
       (10090204)
                                                       Antelope Hills
GCRP Model Areas - Powder and Tongue River Basins
                  Base Map
  Figure 3-14.  Powder and Tongue River basins study area.
                                                 3-20

-------
    Hydrography

•I Water (Nat. Atlas Dataset)

    US Census Populated Places

•I Municipalities (pop > SO COO)

I   I County Boundaries

    Watershed with HUCSs
Saguache
(13010004)
                                             Rio Grande
                                             Headwaters
                                             13010001)
              San Luis
              13010003)
                                                              lamosa-Trinchera
                                                                (13010002)-/
              New Mexico
                                                                    Upper Rio Grande
                                                                       (13020101)
                                                    Rio Chama
                                                    (13020102)
                                             Jemez
                                           (13020202)
                       Albuquerque
                                                       Rio Grande
                                                         Santa Fe
                                                        (13020201)
                                                                Cedar Grove
                                            Rio Grande-
                                           Albuquerque
                                            (13020203)
      GCRP Model Areas - Rio Grande Valley
                  Base Map
  Figure 3-15. Rio Grande Valley study area.
                                                3-21

-------
    Hydrography

    Water (Nat Alias Dataset)

    US Census Populated Places

•I Municipalities (pop > 50.000)

   I County Boundaries

    Model Subbasins
Sacramento-LowerCow-
     Lower Clear
      (18020101)
   otto n wood
  Headwaters
  (18020113)
                                                                         Upper Cow-Battle
                                                                           (18020118)
                                   Upper Elder-
                                  Upper Thomes
                                    (18020114)
                                        Mill-Big Chico
                                          (18020119)
                                                   Lower Thomes
                                                    (18020103)
                                    Upper Stony
                                     (18020115)
                                                                             ^'er Butte
                                                                              (18020120)s
                                                                 Lower Butte
                                                                 ^('18020105)
                                            Sacramento-
                                            Stone Corral
                                                                                       Citrus Heights

                                                                                       Folsom
                  Sacramento
    GCRP Model Areas - Sacramento River Basin
                  Base Map
  Figure 3-16.  Sacramento River basin study area.
                                                 3-22

-------
to
                                                                                                                           Legend

                                                                                                                              - Hydrography

                                                                                                                               Water (Nat. Atlas Dasascti

                                                                                                                               US Census Populated Places

                                                                                                                           ^H Municipalities (pop 2 90,000)

                                                                                                                           I   I County Boundaries

                                                                                                                           f~~| Watershed with HUC8s
                                                      Santa Clarita
                                                                 Lancaster

                                                                          Palmdale
Santa Clara
(18070102)
 Ventura
18070101)
                                                                       Los Angeles
                                                                          ,8070105)
                                                                           I
Thousand Oaks

        Los Angeles
                                                                                    San Gabriel
                                                                                    (18070106)
                                                                                          San Bernardino
                                                Santa" rVlonica.Bav
                                                   (18070104)
                                            Santa Monica
                                                                                              Santa Ana
                                                                                              (18070203)
                                              Glendale
                                               Pasadena
                                                                                                                               Redlands
                                                                                                                                  -7---.
                                                                                                                                  Riverside
                                             Seal Beat
                                             (180702011)
                                                                                                            San J a cm to
                                                                                                             (18070202)
                                                                       Newport'Bay
                                                                        (18070204)
                                                                                       iso-San Onofre
                                                                  Mission Viejo
                                                                                                             Santa Margarita
                                                                                                               (18070302)
                                                   GCRP Model Areas - Southern California Coastal Basins
                                                                     Base Map
             Figure 3-17. Southern California Coastal basins study area.

-------
                                                                                     Lone Tree, _
                                                                                      (10190008)
                                                                               Crow
                                                                            (10190009)
                                                 Cache
                                               La Poudre
                                               (10190007)
     Hydrography
     Water (Nat. Atlas Dataset)
     US Census Populated Places
     Municipalities (pop > 50,000)

     County Boundaries
     Watershed with HLJCSs
Big Thompson
  (10190006)
                                                                    Middle South Platte-
                                                                       Cherry Creek
                                                                        (10190003)
  St. Vrain
 (10190005)-»p
                            Boulder
                          Denver
                                    Clear
                                  (10190004)
                                                                                   BIJOU
                                                                                 (10190011)
                                            Upper South
                                            '   Platte/
                          South Platte
                          Headwater
                          (10190001)
                                                                                     Kiowa
                                                                                   (10190010)
                                                               Colorado
                                                               Springs
                                                              0    15   30
GCRP Model Areas - South Platte River Basin
              Base Map
  Figure 3-18. South Platte River basin study area.
                                                3-24

-------
                           Lake Ontario
                                                                    Upper Susquehanna
                                                                         (2050101 j  '
                                                    Chenango
                                                    (2050102)
                                             ego-Wappasenin
                                               (2050103)
                           Chemung
                           (2050105)
                          Tioga
                      ••I (2050104)
                                         Upper Susquehanna-Tunkhanno
                    PJne Pennsylvania
                       (2050205)
                                      Middle West Branch
                                         Susquehanna
                                          (2050203)
                                                              UppeJ'
                                                     Susquehanna-lScka wanna
                                                            (ZOSftttfr)
Sinnemahpning
   (205p*20'2)
Lower West Brancru
    usquehanna
     2050206)
            0}
            r         f Bald Eagle
        K-           
-------
to
                                                                                                                             Legend

                                                                                                                                - Hydrography

                                                                                                                                 Water (Nat. Atlas Oataset)

                                                                                                                                 US Census Populated Places

                                                                                                                             ^H Municipalities (pop £ 50,000)

                                                                                                                             [	| County Boundaries

                                                                                                                               H Watershed with HUCSs
                                                              Rocky Mount
North Carolina  '
 Fishing
(03020102)
                                                        Upper Tar
                                                        (03020101)
                                                                                           Lower Tar
                                                                                           (03020103)
                                                                                                                    Pamlico
                                                                                                                   (03020104)
                                                                                    ontentnea
                                                                                   (03020203)
                         Upper Neuse
                          (03020201)
                                                                                                   Middle Neuse
                                                                                                     (03020202)
                                                                                                                             r Neuse
                                                                                                                           3020204)
                                                                                                   Jacksonville
                                                    GCRP Model Areas - Tar and Neuse River Basins
                                                                    Base Map
              Figure 3-20.  Tar and Neuse River basins study area.

-------
                Upper West
                Fork Trinity
                {120301'01)
                                                                                               Longview
                                                                Lower-Trinity
                                                                 Tehuacana
                                                                 (12030201:).
     Hydrography

^H Water (Nat. Atlas Dataset)

     US Census Populated Places

H Municipalities (pop 5 50.000)
I    | County Boundaries

 ~1 Watershed with HUCSs
      GCRP Model Areas - Trinity River Basin
                   Base Map
  Figure 3-21.  Trinity River basin study area.
                                                    3-27

-------
to
oo
                 -  Hydrography

                   Water (Nat. Atlas Datasot)

                   US Census Populated Places

               HI Municipalities (pop £ 50,000)

               I   I County Boundaries

               r"~| Watershed with HUCBs
                                                                            Colorado Headwaters
                                                                                 (14010001)
                                                     Parachute-
                                                       Roan
                                                     (14010006)
                                                                                                        Eagle
                                                                                                      (14010003)
                                                                                     Blue
                                                                                  (14010002
                                                           Colorado Headwaters
1                                                                 Plateau
                                                                 14010005)
                                                                                                        Highlands
                                                                                                        Ranch
                                                                                            Roaring Fork
                                                                                             (14010004)
Grand
Junction
                                                   Lower Gunnison
                                                      (14020005)
                                                                                                 East-
                                                                                                Taylor
                                                                                              (14020001)
                                                                           Upper Gunnison
                                                                             (14020002)
                                                           Uncompahgre
                                                           1114020006)
                                                                                                Tomichl—
                                                                                               (14020003)
                                                   GCRP Model Areas - Upper Colorado River Basin
                                                                   Base Map
              Figure 3-22.  Upper Colorado River basin study area.

-------
                                                                                      Hydrography

                                                                                      Water (Nal. Atlas Dataset)

                                                                                      US Census Populated Places

                                                                                 ^H  Municipalities (pop > 50.000)

                                                                                      County Boundaries

                                                                                      Watershed with HUCSs
                     Tualatin
                   (17090010)
                       S
                          Yamhill
                         (17090008)
                                               Molalla-Puddmg
                                                 (17090009)
                                                Clackatnas
                                                (17090011)
              -Upper Willamette
                  (17090003)
                                          South Santiam
                                           (17090006)
                                                               North Santiam
                                                                 17090005)
                                                      Mckenzie
                                                     (17090004)
                                                                                             Washington
                                       Middle Fork
                                        Willamette
                                        (17090001)
Coastal Fo
 Willamette
 (17090002)
GCRP Model Areas - Willamette River Basin
              Base Map
  Figure 3-23.  Willamette River basin study area.
                                                  3-29

-------
                             4.  MODELING APPROACH
This study uses dynamic watershed models to simulate the watershed response to potential
mid-21st century climate change scenarios, urban and residential development scenarios, and
combined climate change and urban development scenarios.  Watershed models were developed
for 20 large-scale study areas (approximately HUC-4 scale) located throughout the contiguous
United States and Alaska.  The study also evaluates the sensitivity of modeling results to
different methodological choices for assessing climate change impacts, such as the use of climate
change scenarios based on different methods of downscaling GCM projections and the use of
different watershed models.

A watershed model is a useful tool for providing a quantitative linkage between external forcing
and in-stream response.  It is essentially a series of algorithms applied to watershed
characteristics and meteorological data to simulate naturally occurring, land-based processes
over an extended period, including hydrology and pollutant transport. Many watershed models
are also capable of simulating in-stream processes. After a model has been set up and calibrated
for a watershed, it can be used to quantify the existing loading of pollutants from subbasins or
from different land-use categories and can also be used to assess the effects of a variety of
management scenarios.

Five of the 20 sites were selected as "pilot" sites: the Minnesota River watershed (Minn), the
Apalachicola-Chattahoochee-Flint River watersheds (ACF), the Willamette River watershed
(Willa), the Salt/Verde/San Pedro River watershed (Ariz), and the Susquehanna River watershed
(Susq).  Pilot sites were selected in part due to previous experiences of the study team in
applying watershed models in these areas, and in part because they provide a representative cross
section of the full set of 20 study areas from a regional, meteorological, geographic, and land-use
perspective.  Pilot sites were used for testing and comparing model development and application
methods, as well as for evaluating the sensitivity of modeling results to different types of climate
change scenarios and use of different watershed models. Analysis of the pilot site results led to
the selection of a reduced, more streamlined approach for the remaining 15 sites using one
watershed model and a reduced set of climate change scenarios.

Two watershed models were selected  for initial application to the five pilot study sites: HSPF
(Bicknell et al., 2001, 2005) and SWAT (Neitsch et al., 2005).  Each of these models has been
widely used for hydrologic and water  quality applications for regulatory purposes, such as the
development of pollutant load allocations under the Total Maximum Daily Load (TMDL)
provisions of the Clean Water Act.  Both models are also in the public domain with open-source
code, enabling ready replication of results. They both provide dynamic simulation with a
subdaily or daily time step and can be built from readily available spatial coverages, but are
sufficiently efficient to allow implementation of multiple runs for model calibration or scenario
application purposes. Both models have also been used in previous studies of watershed
responses to climate change (e.g., Taner et al., 2011; and Tong et al., 2011 for HSPF; Luo et al.,
2013; Wilson and Weng, 2011; Marshall and Randhir, 2008; and Ficklin et al., 2009 for SWAT).

Application of both HSPF and SWAT to the five pilot watersheds allowed assessment of the
variability associated with use of different watershed models in simulating watershed response to
                                           4-1

-------
climate change. The two model frameworks exhibited similar skill in reproducing observations
at the large spatial scales addressed in this project (see Section 4.4.3); however, SWAT is based
on a plant growth model that can explicitly represent the impacts of altered temperature,
moisture, and CC>2 regimes on plants and the resulting impacts on the water balance and pollutant
transport.  The analysis of the pilot site results (see Section 6) emphasized the potential
importance of these processes.  Therefore, the SWAT model was applied in all 20 study areas.
HSPF and SWAT are each described in more detail below

4.1. MODEL BACKGROUND
4.1.1.  HSPF
The HSPF (Bicknell et al., 2001, 2005) is a comprehensive, dynamic watershed and receiving
water quality modeling framework that was originally developed in the mid-1970s.  During the
past several decades, it has been used to develop hundreds of EPA-approved  TMDLs, and it is
generally considered among the most advanced hydrologic and watershed loading models
available.  The hydrologic portion  of HSPF is based on the Stanford Watershed Model
(Crawford and Linsley, 1966), which was one of the pioneering watershed models developed in
the 1960s.  The HSPF  framework is developed modularly with many different components that
can be assembled in different ways, depending on the objectives of a project.  The model
includes three major modules:
   •   PERLND for simulating watershed processes on pervious land areas

   •   IMPLND for simulating processes on impervious land areas

   •   RCHRES for simulating processes in streams and vertically mixed lakes


All three of these modules include many subroutines that calculate the various hydrologic and
water quality processes in the watershed.  Many options are available for both simplified and
complex process formulations.

HSPF models hydrology as a water balance in multiple surface and subsurface layers and is
typically implemented in large watersheds at an hourly time step. The water balance is simulated
based on Philip's infiltration (Bicknell et  al., 2001, 2005) coupled with multiple surface and
subsurface stores (interception storage, surface storage, upper zone soil storage, lower zone soil
storage, active groundwater, and inactive  [deep] groundwater). Potential evapotranspiration
(PET) is externally specified to the model.

As implemented in HSPF, the infiltration algorithms represent both the continuous variation of
infiltration rate with time as a function of soil moisture and the areal variation of infiltration over
the land segment.  The infiltration capacity, the maximum rate at which soil will accept
infiltration, is a function of both the fixed and variable characteristics of the watershed.  Fixed
characteristics include soil permeability and land slopes, while variables are soil surface
conditions and soil moisture content. A linear probability function is used to account for spatial
variation (Bicknell et al., 2005). The primary parameters controlling infiltration are INFILT, an
index to mean soil infiltration rate (in/hr)  and LZSN, the lower soil zone nominal soil moisture
                                          4-2

-------
storage. Specifically, the mean infiltration capacity over a land segment at any point in time,
IBAR, is calculated as
                          IBAR =
                                      INFILT
                                  (LZS/
                                            VNFEXP
                                     /LZSN'
• INFFAC
                                                                                   4-1
where LZS is the current lower soil zone storage, INFEXP is an exponent typically set to a value
of 2, and INFFAC is an adjustment factor to account for frozen ground effects.

Neither INFILT nor LZSN is directly observable or provided in soils databases and both must be
refined in calibration. As INFILT is not a maximum rate nor an infiltration capacity term, its
values are normally much less than published infiltration rates, soil percolation test results, or
permeability rates from the literature (U.S. EPA, 2000).

Sediment erosion in HSPF uses a method that is formally similar to, but distinct from, the
universal soil loss equation (USLE) sediment-detachment approach coupled with transport
capacity based on overland flow. Nutrients may be simulated at varying levels of complexity,
but are most typically represented by either buildup/washoff or sediment potency approaches on
the land surface coupled with user-specified monthly concentrations in interflow and
groundwater.

Spatially, the watershed is divided into a series of subbasins representing the drainage areas that
contribute to each of the stream reaches. The stream network (RCHRES) links the surface runoff
and groundwater flow contributions from each of the land segments and subbasins and routes
them through water bodies. The stream model includes precipitation and evaporation from the
water surfaces  as well as streamflow contributions from the watershed, tributaries, and upstream
stream reaches. It also simulates a full range of stream sediment and nutrient processes,
including detailed representations of scour, deposition, and algal  growth.

The version of HSPF used in this study is the Windows interface to Hydrologic Simulation
Program—FORTRAN (WinHSPF) as distributed with BASINS version 4.0. WinHSPF is a
Windows interface to HSPF and is a component of the EPA's Better Assessment Science
Integrating point and Nonpoint Sources (BASINS) Version 4.0 (U.S. EPA, 2001, 2009a, 2009c).
WinHSPF itself is a user interface to HSPF that assists the user in building User Control Input
(UCI) files (containing model input parameters) from geographic information system (GIS) data
(Duda et al., 2001). After the UCI file is built, WinHSPF is used to view, understand, and
modify the model  representation of a watershed. HSPF can be run from within WinHSPF. The
actual model executable engine distributed with BASINS is called WinHSPFLt, which can be
run in batch mode independent of the BASINS/WinHSPF interface. The model code for HSPF
is stable and well documented. Detailed descriptions of the model theory and user control input
are provided in Bicknell et al. (2001, 2005).

WinHSPF also provides access to the Climate Assessment Tool (CAT), which is a component of
BASINS 4.0. BASINS CAT facilitates watershed-based  assessments of the potential effects of

                                          _

-------
climate variability and change on water and watershed systems (namely streamflow and pollutant
loads) using the HSPF model (U.S. EPA, 2009a, 2009b).  BASINS CAT is capable of creating
climate change scenarios that allow users to assess a wide range of what //questions related to
climate change.

4.1.2.  SWAT
The SWAT model was developed by the U.S. Department of Agriculture to simulate the effect of
land management practices on water, sediment, and agricultural chemical yields in large,
complex watersheds with varying soils, land use, and management conditions over long periods
of time (Neitsch et al., 2005). SWAT requires data inputs for weather, soils, topography,
vegetation, and land use to model water and sediment movement, nutrient cycling, and numerous
other watershed processes. SWAT is a continuous model appropriate for long-term simulations.

SWAT, as implemented in this study, employs a curve number approach (SCS, 1972) to estimate
surface runoff and then completes the water balance through simulation of subsurface flows,
evapotranspiration, soil storages, and deep seepage losses. The curve number approach requires
a daily time step. PET is typically calculated internally by SWAT based on other weather inputs.

SWAT provides an option for subdaily Green-Ampt infiltration, but this is infrequently used.
The curve number approach is popular because parameters are simple and readily available. The
curve number approach estimates the depth of daily runoff (Q)  from rainfall depth (P),  initial
abstractions (la, depth), a storage parameter (S, depth), and a curve number (CN), as (SCS,
1972):
                                         (P-fo)2
                                        P-Ia + S                                  4_2
la is typically assumed to be 20% of S (indeed, this is a hard-coded default in SWAT). In units
of millimeters for S, this yields:
                                          .
                                (P + 0.8S)        CN
                                                                                    4_3
The curve number is estimated as a function of land use, cover, condition, hydrologic soil group
(HSG), and antecedent soil moisture. SWAT provides capabilities to automatically adjust the
CN based on soil moisture, plant evapotranspiration, slope, and the presence of frozen ground.
The conceptual simplicity of the curve number approach also introduces some potential
problems. Specifically, the curve number was developed as a design methodology to estimate
average runoff volume of a specific return period, given average total event rainfall of the same
return period. It was not designed to predict runoff from specific individual events or runoff
from more frequent smaller events, and applicability to continuous simulation is inexact,
                                          4-4

-------
especially at small spatial scales. For a summary of these issues and their potential implications
in continuous simulation modeling, see Garen and Moore (2005).

Sediment yield and erosion are calculated by SWAT using the Modified Universal Soil Loss
Equation (MUSLE; Williams, 1975). The MUSLE is based on several factors, including surface
runoff volume, peak runoff rate, area of hydrologic response unit (HRU), soil erodibility, land
cover and management, support practice, topography, and a coarse fragment factor.  MUSLE
implicitly combines the processes of sediment detachment and delivery. Nutrient load
generation and movement are simulated using overland runoff and subsurface flow.

A key feature of SWAT is the incorporation of an explicit plant growth model, including plant
interactions with water and nutrient stores. The transformation of various nitrogen and
phosphorus species is simulated in detail in the soil; however, concentrations of nutrients in
groundwater discharges are user specified, as in HSPF.

In-stream  simulation of sediment in SWAT 2005 includes a highly simplified representation of
scour and deposition processes. Nutrient kinetics in receiving waters are based on the numeric
representation used in the QUAL2E model but implemented only at a daily time step.

SWAT is generally considered to be an effective tool for watershed simulation that is especially
appropriate for estimating streamflow and cumulative pollutant loads in agricultural and rural
watersheds (see review by Gassman et al., 2007). Bosch et al. (2011) found that SWAT was an
effective tool for estimating hydrology, sediment, and nutrient loads in Lake Erie watersheds, but
performed less well in urbanized settings. SWAT has some potential weaknesses relative to
HSPF for the simulation of urban lands because it is typically run using a curve number approach
at a daily time step while HSPF is typically run at an hourly time step using Philip infiltration.
The daily time step is insufficient to resolve details of urban runoff hydrographs that have
important implications for stability of small  stream channels, while the curve number approach
can result in poor resolution of surface versus subsurface flow pathways (Garen and Moore,
2005). The impacts of these differences are, however, believed to be minor at the larger spatial
scales addressed in this study.

An important component of the SWAT model is the weather generator (WXGEN).  SWAT
requires daily values of precipitation, maximum and minimum temperature, solar radiation,
relative humidity, and wind speed. The user may read these inputs from a file or generate the
values using SWAT's weather generator model based on monthly average data summarized over
a number of years (Neitsch et al., 2005). The weather generator model (Sharpley and Williams,
1990) can be used to generate climatic data or to fill in gaps in weather data.  The weather
generator first independently  generates  precipitation for the day. Maximum temperature,
minimum temperature, solar radiation, and relative humidity are then generated based on the
presence or absence of rain for the day. Finally, wind speed is generated independently.

The version of SWAT used in this study is SWAT 2005 as distributed with ArcSWAT 2.1,
which was the most recent stable version of SWAT available at the start of this study.
ArcSWAT 2.1 is an ArcGIS-ArcView extension and a graphical user input interface for the
SWAT watershed model (TAMU, 2010).  As with HSPF, the underlying executable code can be
run in batch mode independent of the user interface.  Unlike HSPF, the SWAT code is
                                          4-5

-------
continuously evolving, with frequent enhancements and bug fixes.  For a detailed description of
the version of SWAT used here, see Neitsch et al. (2005).

4.2.  MODEL SETUP
Watershed models were configured to simulate each study area as a series of hydrologically
connected subbasins. Each study area was subdivided into subbasin-scale modeling units.
Continuous simulations of streamflow, total nitrogen, total phosphorus, and total suspended
solids were then made for each unit using meteorological, land use, soil, and stream data.

Many study areas are highly managed systems influenced by humans, including dams, water
transfers and withdrawals,  point source discharges and other factors.  Given the difficulty of
modeling at the large spatial scale in this study, detailed representation of all management was
not possible. The following assumptions were made to simplify modeling among all 20 study
areas:
       External boundary conditions (where needed), such as upstream inflows and pollutant
       loads, are assumed constant.

       Interactions with deep groundwater systems are assumed constant.

       Large-scale shifts in natural cover type in response to climate change are not simulated.

       Point source discharges and water withdrawals are assumed constant at current levels.

       Only large dams that have a significant impact on hydrology at the HUC-8 (subbasin)
       scale are included in the models. Where these dams are simulated, an approximation of
       current operating rules (using a target storage approach) is assumed to apply in all future
       scenarios.

       Human adaptation response to climate change, such as  shifts in water use or cropping
       practices, are not simulated.
The project team consisted of multiple modelers working in different locations.  To ensure
consistency of results, a common set of procedures and assumptions was established (e.g., see
Appendix A). Both HSPF and SWAT were implemented using a HRU approach to upland
simulation. An HRU consists of a unique combination of land use/land cover, soil, and land
management practice characteristics, and thus represents areas of similar hydrologic response.
Individual land parcels included within an HRU are expected to possess similar hydrologic and
load generating characteristics and can thus be simulated as a unit.  The HRU approach is the
default for SWAT but is also good practice with HSPF. Consistent with the broad spatial scale
of the models, the land cover component is aggregated into a relatively small number of
categories (e.g., forest, wetland,  range, grass/pastureland, crop, developed pervious, low-density
impervious, and high-density impervious).
                                          4-6

-------
Initial preparation of spatial data was done primarily in ArcGIS for the entire study area.
Processed GIS inputs were then used in ArcSWAT (which runs as an extension in ArcGIS), and
imported into B ASINS4 (which uses Map Window GIS) to complete the setup of SWAT and
HSPF, respectively. Spatial data sources are discussed in more detail in Section 4.2.3.
Additional initial setup tasks included identification of weather stations, streamflow gaging and
water quality monitoring locations, and major watershed features that significantly affect the
water balance, such as presence of major lakes, reservoirs, and diversions.

4.2.1.  SWAT Setup Process
SWAT model setup used the ArcSWAT extension in ArcGIS.  The general procedure for SWAT
setup is described below; a more detailed modeling protocol used for this project is included in
Appendix A.

Subbasin boundaries and reach hydrography for each study area were generally defined from
NHDPlus catchments (U.S. EPA, 2010) aggregated to approximately the HUC-10 spatial scale.
The subbasin and reach shapefiles were imported into the SWAT interface and subbasin
parameters were calculated automatically.

Study area boundaries were configured to minimize the presence of large reservoirs due to the
difficulty of representing operational rules.  Models included only major reservoirs that have a
significant effect on streamflow at the scale of HUC-8s or greater.  Inclusion of reservoirs was
left to the discretion of individual modelers; however, the reservoirs included are generally those
that drain an area greater than  a single HUC-8 and provide a retention time of half a year or
greater. If a reservoir was located at the terminus of the model area, it was generally ignored so
that the model represented  input to, rather than output from, the terminal reservoir. Models
include point source discharges from major permitted facilities (greater than 1 million gallons
per day [MGD] discharge).  It was also necessary to  define an upstream boundary condition
"point source" for study areas  where the model did not extend to the headwaters (e.g.,
Sacramento River basin).

HRUs were developed from an intersection of land use, slope, and major soils, using the
geospatial data sources described in Section 4.2.3. In the HRU analysis, SWAT was used to
classify the slopes into two categories: above and below  10%.  A single breakpoint was chosen
to represent major differences  in runoff and erosive energy without creating an unmanageable
number of individual HRUs. The State Soil Geographic Database (STATSGO) soils coverage
was assigned using the dominant component method in which each soil polygon is represented
by the properties of the dominant constituent soil.  The NLCD 2001 land use coverage was
loaded directly into ArcSWAT without modification. The default NLCD class to SWAT class
mapping was appropriate for most areas.  Impervious percentage was assigned to developed
land-use classes in the SWAT urban database using values calculated from the NLCD
impervious coverage.  The same assumptions were applied for the future developed land-use
classes (i.e., the future classes  have the same total  and connected impervious fractions as the
corresponding existing urban land uses).  HRUs were created by overlaying land use, soil, and
slope at appropriate cutoff tolerance levels to prevent the creation of large numbers of
insignificant HRUs. Land-use classes were retained  if they occupied at least 5% of the area of a
subbasin (with the exception of developed land uses, which were retained regardless of area).
Soils were retained if they occupied at least 10% of the area within a given land use in a

                                          _

-------
subbasin.  Slope classes were retained if they occupied at least 5% of the area within a given soil
polygon. Land uses, soils, and slope classes that fall below the cutoff value are reapportioned to
the dominant classes so that 100% of the watershed area is modeled (Winchell et al., 2008).

The SWAT models were linked to meteorological stations contained in EPA's BASINS 4
meteorological data set (U.S. EPA, 2008).  The models used observed time series for
precipitation and temperature; other weather data were simulated with the  SWAT weather
generator, as discussed in Section 4.2.4.  Elevation bands were turned on if necessary to account
for orographic effects in areas with a sparse precipitation network and significant elevation
changes. This was generally appropriate where elevations within subbasins spanned a range of
250 m or more.  Daily curve number hydrology with observed precipitation and air temperature
was used.

Land management operations were assigned, primarily to account for agricultural practices.  For
urban lands, the USGS regression method for pollutant load estimation was specified. In-stream
water quality options started with program defaults.

The target time period for simulation was 31 water years, with the first year dropped from
analysis to account for model spinup (initialization). Some weather stations may have been
absent for the spinup year, but SWAT fills in the missing records using the weather generator.
The remaining 30  years span a period for which the supplied weather data were complete and
included the year 2000 (with the exception of the Loup/Elkhorn basins in Nebraska, for which
the simulation period ended in 1999 due to the termination of a number of precipitation gauges
before the end of 2000).

4.2.2. HSPF Setup Process
HSPF models were developed on a common basis with the SWAT models using the same
geospatial data,  but only for the five pilot watersheds.  Subbasin boundaries and reach
hydrography were defined using the same NHDPlus catchments as the SWAT models. The
HRUs for HSPF were calculated from the SWAT HRUs, but differ in that soils were aggregated
into hydrologic  soil group. Pervious (PERLND) and impervious (IMPLND) land areas are
specified and simulated separately in HSPF, whereas SWAT specifies an impervious fraction for
different land-use  categories.

The WinHSPF interface distributed with BASINS (U.S.  EPA, 2001) was used to create the user
control input (UCI) and watershed data management (WDM)  files. A starter UCI file was
prepared that assigned default values for HRUs. Initial parameter values were based on previous
modeling where available. For areas without previous modeling, hydrologic parameters were
based on recommended ranges in BASINS Technical Note 6 (U.S. EPA, 2000) and related to soil
and meteorological characteristics where appropriate.  Snowmelt simulation used the simplified
degree-day method.

The stage-storage-discharge hydraulic functional tables (FTables) for stream reaches were
generated automatically during model creation. The WinHSPF FTable tool calculates the tables
using relationships to drainage area. FTables were adjusted in WinHSPF if specific information
was available to the modeler. Hydraulic characteristics for major reservoirs and flow/load
characteristics for major point sources were defined manually based on available information.
                                          4-8

-------
Nutrients on the land surface were modeled as inorganic nitrogen, inorganic phosphorus, and
total organic matter. The latter was transformed to appropriate fractions of organic nitrogen and
organic phosphorus in the linkage to the stream. The in-stream simulation represented total
nitrogen and total phosphorus as general quality constituents subject to removal approximated as
an exponential decay process.  Initial values for decay rates were taken from studies supporting
the USGS SPARROW model (e.g., Alexander et al., 2008).

4.2.3. Watershed Data Sources
The HSPF and SWAT models each use identical geospatial and other input data sources as
described below.

4.2.3.1. Watershed Boundaries and Reach Hydrography
Subbasin boundaries and reach hydrography (with connectivity) for both SWAT and HSPF were
defined using NFIDPlus data (U.S. EPA, 2010), which is a comprehensive set of digital spatial
data representing the surface water of the United States including lakes, ponds, streams, rivers,
canals, and oceans. NHDPlus provided catchment/reach flow connectivity, allowing for creation
of large model subbasins with automation. NHDPlus incorporates the National Hydrography
Dataset (NHD), the National Elevation Dataset, the NLCD, and the Watershed Boundary
Dataset. A MapWindow script was developed to automate (with supervision) the aggregation of
NHDPlus catchments/reaches into model subbasins and reaches.  The general approach was to
first run the aggregation  script with a smaller target subbasin size (i.e., create several hundred to
a thousand subbasins), then run the script again to create watersheds of the target model size
(comparable to the  HUC-10 spatial scale).  The two-tiered approach  has several benefits; it was
found to be more time efficient, it allowed for greater control over the final basin size, and it
provided a midpoint that could be used to redefine subbasin boundaries to match specified
locations, such as gaging stations and dams/diversions.

Each delineated subbasin was conceptually  represented with a single stream assumed to be a
completely mixed,  one-dimensional segment with a constant cross section.  For the HSPF model,
reach slopes were calculated based on Digital Elevation Model (DEM) data, and stream lengths
were measured from the original NHD stream coverage. Assuming representative trapezoidal
geometry for all streams, mean stream depth and channel width were estimated using regression
curves that relate upstream drainage area to stream dimensions developed for three regions in the
Eastern United States.  Existing and more detailed models provided additional site-specific
information on channel characteristics for some watersheds (e.g., Minnesota River; Tetra Tech,
2008b).

The SWAT model also automatically calculates the initial  stream geometric values based on
subbasin drainage areas, standard channel forms, and elevation, using relationships developed for
numerous areas of the United States.  Channel slope is automatically calculated from the DEM.

4.2.3.2. Elevation
Topography was represented by digital elevation models (OEMs) with a resolution of 30 meters
obtained from USGS' National Elevation Dataset (Gesch et al., 2002). Multiple DEM coverages
were grouped and clipped to the extent of the model watershed area (with a 10-mile buffer to
allow for unforeseen changes to watershed boundaries).
                                          4-9

-------
4.2.3.3. Land Use and Land Cover
The SWAT and HSPF models use a common land use platform representing current (calibration)
conditions and derived from the 2001 NLCD (Homer et al., 2004, 2007).  The 2001 NLCD land
cover was used to ensure consistency between all models for the project.  The 2001 land use was
chosen rather than the 2006 coverage because it is closer in time to the calibration period of the
models, which typically runs through 2002/3.  The 2001 land use is assumed to apply throughout
the baseline model application period.

Some additional processing of the NLCD data was necessary.  Several of the land use classes
were aggregated into more general categories to provide a more manageable set of HRUs. The
developed land classes were kept separate for SWAT but aggregated for HSPF. This is because
SWAT assigns percent imperviousness to total developed area, whereas HSPF explicitly
separates developed pervious and impervious areas.  The regrouping of the NLCD classes for
SWAT and HSPF is shown in Table 4-1.

The percent impervious area was specified for each developed land class from the NLCD Urban
Impervious data coverage. The NLCD 2001 Urban Imperviousness coverage was clipped to the
extent of the model watershed area (with 10-mile buffer) to calculate the impervious area. The
percent impervious area was then specified by combining data from the 2001 NLCD Land Cover
and Urban Impervious data products. Specifically, average percent impervious area was
calculated over the whole basin for each of the four developed land use classes. These
percentages were then used to separate out impervious land. The analysis was performed
separately for each of the 20 study areas, since regional differences occur. Table  4-2 presents the
calculated 2001  impervious areas for each study area.

4.2.3.4. Soils
Soils data were implemented using SWAT's built-in STATSGO (USDA, 1991) national soils
database. The SWAT model uses the full set of characteristics of dominant soil groups directly,
including information on infiltration, water holding capacity, erodibility, and soil  chemistry.  A
key input is infiltration capacity, which is used, among other things, to estimate the runoff curve
number. Curve  numbers are a function of hydrologic soil group, vegetation, land use, cultivation
practice, and antecedent moisture conditions. The Natural Resource Conservation Service
(NRCS; SCS, 1972) has classified more than 4,000 soils into four hydrologic soil groups (HSGs)
according to their minimum infiltration rate for bare soil after prolonged wetting.  The
characteristics associated with each HSG are provided in Table 4-3.

In the HSPF setup the HRUs are not based directly on dominant soils; instead, these were
aggregated to represent HSGs. The HSGs include special agricultural classes  (A/D, B/D, and
C/D) in which the first letter represents conditions with artificial drainage and the second letter
represents conditions without drainage. The first designator was assumed to apply to all crop
land, while the second designator was assumed for all other land uses.
                                         4-10

-------
       Table 4-1. Regrouping of the NLCD 2001 land-use classes for the HSPF and
       SWAT models
NLCD class
1 1 Water"
12 Perennial ice/snow
21 Developed open space
22 Dev. Low Intensity
23 Dev. Med. Intensity
24 Dev. High Intensity
3 1 Barren Land
41 Forest — Deciduous
42 Forest — Evergreen
43 Forest — Mixed
51-52 Shrubland
71-74 Herbaceous Upland
81 Pasture/Hay
82 Cultivated
91-97 Wetland (emergent)
98-99 Wetland (nonemergent)
SWAT class
WATR (water)
WATR (water)
URLD (Urban Residential— Low Density)
URMD (Urban Residential — Medium Density)
URHD (Urban Residential— High Density)
UIDU (Urban Industrial and High Intensity)
SWRN (Range-Southwestern U.S.)
FRSD (Forest— Deciduous)
FRSE (Forest — Evergreen)
FRST (Forest— Mixed)
RNGB (Range— Brush)
RNGE (Range — grasses)
HAY
AGRR (Agricultural Land-Row Crops)
WETF (Wetlands— Forested), WETL (Wetlands),
WETN (Wetlands— Nonforested)
WATR (water)
HSPF class
WATER
BARREN
DEVPERV (Developed Pervious)
IMPERV (Impervious)
BARREN
FOREST
SHRUB
GRASS
BARREN
GRASS
AGRI (Agriculture)
WETL (Wetlands)
WATER
"Water surface area is usually accounted for as reach area.
                                         4-11

-------
       Table 4-2. Calculated fraction impervious cover within each developed land
       class for each study area based on NLCD 2001
Site ID
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
Open space (%)
8.04
7.37
10.11
7.20
8.83
7.30
7.53
6.59
8.34
8.22
7.42
8.76
5.95
7.75
6.41
6.90
7.17
7.74
9.78
9.56
Low intensity (%)
30.16
29.66
29.79
31.87
32.36
32.53
32.91
29.20
29.68
32.81
31.64
32.36
30.02
35.39
33.46
31.26
30.90
31.65
31.89
32.31
Medium intensity (%)
60.71
53.71
61.48
60.14
61.24
60.72
60.11
55.01
60.14
60.90
59.16
60.49
55.41
61.31
60.79
60.90
61.05
60.78
60.48
61.49
High intensity (%)
89.90
73.85
87.17
87.47
88.70
86.75
88.08
83.31
86.59
87.25
85.99
84.32
81.20
88.83
86.76
85.41
87.31
89.15
87.41
88.94
       Table 4-3. Characteristics of NRCS soil hydrologic groups
Soil group
A
B
C
D
Characteristics
Sandy, deep, well drained soils; deep loess; aggregated silty soils
Sandy loams, shallow loess, moderately deep and moderately well drained soils
Clay loam soils, shallow sandy loams with a low permeability horizon impeding
drainage (soils with a high clay content), soils low in organic content
Heavy clay soils with swelling potential (heavy plastic clays), water-logged soils,
certain saline soils, or shallow soils over an impermeable layer
Minimum infiltration
capacity (in/hr)
0.30-0.45
0.15-0.30
0.05-0.15
0.00-0.05
4.2.3.5.  Point Source Discharges

The primary objective of this study is to examine relative changes that are potentially associated
with changes in climate and land use. From that perspective, point source discharges can be
                                         4-12

-------
characterized as a nuisance parameter.  However, point sources that are large enough relative to
receiving waters to affect the observed streamflow and nutrient loads in river systems need to be
included to calibrate the models.  This is done in a simplified way, and the point sources were
then held constant for future conditions, allowing analysis of relative change.  Only the major
dischargers, typically those with a discharge rate greater than 1 MGD were included in the
models. The major dischargers account for the majority of the total flow from all permitted
discharges in most watersheds, so the effect on the calibration of omitting smaller sources is
relatively small, except perhaps during extreme low streamflow conditions. Data were sought
from the EPA's Permit Compliance System database for the major dischargers in the watersheds.
Facilities that were missing TN, TP, or total suspended solids (TSS) concentrations were filled
with a typical pollutant concentration value from the literature based on Standard Industrial
Classification (SIC) code. The major dischargers were represented at long-term average flows,
without accounting for changes over time or seasonal variations.

4.2.3.6. Atmospheric Deposition
Atmospheric deposition can be a significant source of inorganic nitrogen to watersheds and
water bodies. SWAT2005 allows the user to specify wet atmospheric deposition of nitrate
nitrogen. This is specified as a constant concentration across the entire watershed.  Wet
deposition of ammonia and dry deposition of nitrogen is not addressed in the SWAT2005 model.

HSPF allows the specification of both wet and dry deposition of both nitrate nitrogen (NOs) and
ammonium nitrogen (NH4), and both were included in the model. Dry deposition is specified as
a loading series, rather than concentration series. Because wet deposition is specified as a
concentration, it will vary in accordance with precipitation changes in future climate scenarios,
whereas the dry deposition series (HSPF only) is assumed constant for future scenarios.

Total oxidized nitrogen (NOX) emissions in the United  States remained relatively constant to a
first approximation across the model period considered in this study from the early  1970s up
through 2002 (U.S. EPA, 2002).  There is strong geographic variability in atmospheric
deposition, but much smaller year-to-year variability at the national scale over this period
(Suddick and Davidson, 2012). The National Acid Deposition Program (NADP;
http://nadp.sws.uiuc.edu/) monitors wet deposition across the country and produces yearly
gridded maps of NOs and NH4 wet deposition concentrations.  Dry deposition rates are
monitored (and interpreted with models) by the EPA Clean Air Status and Trends Network
(http://epa.gov/castnet/javaweb/index.html). Results for year 2000 were selected as generally
representative and each study watershed was characterized by a spatial average wet deposition
concentration (and dry deposition loading rate for HSPF). Atmospheric deposition of
phosphorus and sediment was not considered a significant potential source and is not addressed
in the models.

4.2.3.7. Impoundments, Diversions, and Withdrawals
The hydrology of many large watersheds in the United States is strongly impacted by
anthropogenic modifications, including large impoundments and withdrawals for consumptive
use. It is necessary to take these factors into account to develop a calibrated model. At the same
time, these anthropogenic factors constitute a problem for evaluating responses to future
changes, as there is no clear basis for evaluating future changes in reservoir operations or water
                                          4-13

-------
withdrawals.  In addition, information on impoundments, withdrawals, and trans-basin water
imports is often difficult to obtain. The approach taken in this project is to minimize the
importance of impoundments and withdrawals by focusing on relative changes between present
and future conditions with these factors held constant. In this way, the results that are presented
are estimates of the change that may be anticipated based on changes to meteorological and land
use forcing within the subject study area, with other factors held constant.  Simulation results do
not account for potential future changes in water management.

The general approach adopted for this project was to select study areas by  avoiding major human
interventions (e.g., reservoirs) in the flow system where possible, to ignore relatively minor
interventions, and where necessary to represent significant interventions in a simplified manner.
In the first instance,  study watersheds were delineated to avoid major reservoirs where possible.
For example, the model of the Verde River watershed (Arizona) is terminated at the inflow to
Horseshoe Reservoir. In some cases, as in the Sacramento River watershed,  an upstream
reservoir is treated as a constant boundary condition because information on  future reservoir
management responses to climate change was not available.

Impoundments, withdrawals, and water imports that do not have a major impact on downstream
streamflow were generally omitted from the large scale models.  Inclusion or omission of such
features was a subjective choice of individual modelers; however, it was generally necessary to
include such features if they resulted in a modification of flow at downstream gages on the order
of 10% or more. Where these features were included they were represented in a simplified
manner: (1) impoundments were represented by simplified (two-season) stage-discharge
operating rules, developed either from documented operational procedures or from analysis of
monitored discharge; (2) large withdrawals were represented  as either annual or monthly
constant average rates; and (3) major trans-basin water imports were also represented as either
annual or monthly constant average rates depending on availability of data. Use of surface water
for irrigation was simulated only in those basins where it was determined during calibration that
it was a significant factor in the overall water balance. These simplifying assumptions decrease
the quality of model fit during calibration and validation, but  provide a stable basis for the
analysis of relative response to climate and land-use change within the basin.

More detailed information about the representation of impoundments and other anthropogenic
influences on hydrology in each study area are presented in Appendices D through W.

4.2.4.  Baseline Meteorology Representation
Time series of observed meteorological data (for both SWAT and HSPF) were obtained from the
2006 BASINS 4 Meteorological Database (U.S. EPA, 2008).  The database contains records for
16,000 stations from 1970-2006, set up on an hourly  basis, and has the advantage of providing a
consistent set of parameters with missing records filled and daily records disaggregated to an
hourly time step. The disaggregation was performed using automated scripts that distribute the
daily data using one of several nearby hourly stations, using the one whose daily total is closest
to the daily value being disaggregated.  If the daily total for the hourly stations being used were
not within a specified tolerance of the daily value, the daily value was distributed using a
triangular distribution centered at the middle of the day.  The process involved extensive quality
control review; however, the true temporal distribution of precipitation at daily stations is
unknown and the automated approach can occasionally result in anomalously high hourly

                                          __

-------
estimates. Both factors introduce some irreducible uncertainty into hydrologic simulations using
disaggregated daily precipitation stations.

A typical site-specific watershed project would assemble additional weather data sources to
address under-represented areas, but this requires significant amounts of additional quality
control and data processing to fill gaps and address accumulated records.  It was assumed that
the use of the BASINS 2006 data was sufficient to produce reasonable results at the broad spatial
scale that is the focus of this project, particularly for evaluating the relative magnitude of change.
Significant orographic variability was accounted for through the use of lapse rates because the
available stations typically under-represent high mountain areas.

The required meteorological time series for both SWAT and HSPF (as implemented for this
project) included precipitation, air temperature, and either calculated PET or time series required
to generate PET. SWAT uses daily meteorological data, while HSPF requires hourly  data.
Stations were selected to provide a common 30-year or more period of record (or one  that could
be filled from an approximately co-located station).

Table 4-4 presents a summary of annual precipitation and temperature observations for each of
the study areas from 1971-2000.  For more specific  details on the meteorological data used in
each of the study areas, refer to the model  calibration reports provided in Appendices  D through
W.

PET is the third major weather time series input to the watershed models. As evapotranspiration
is typically the largest outgoing term in the water balance, watershed models are highly  sensitive
to the specification of PET, particularly for simulating low streamflow conditions and events.
Many watershed modeling efforts perform well with simplified approaches to estimating PET,
such as the Hamon method (included as an option in the BASINS data set), which depend
primarily on air temperature. However, the robustness of watershed model calibrations
conducted with simplified PET is suspect under conditions of climate change, since a  variety of
other factors that influence PET, such as wind speed and cloud cover, are also likely to change.
Therefore, we implemented Penman-Monteith PET,  which employs a full energy balance
(Monteith, 1965; Jensen et al., 1990).  The implementation varies slightly between SWAT and
HSPF. In SWAT, the full Penman-Monteith method (Allen et al., 2005) is implemented as an
internal option in the model and includes feedback from crop height simulated by the  plant
growth model. For HSPF, Penman-Monteith reference evapotranspiration at each weather
station was calculated externally using the SWAT2005 model subroutines with observed
precipitation and temperature. In both cases, the additional inputs to the energy balance (solar
radiation, wind movement, cloud cover, and relative humidity) were provided by the SWAT
weather generator, which relies on monthly conditional probability statistics for each of these
inputs. An evaluation of the BASINS meteorological data set indicated substantial amounts of
missing data for these inputs (especially for solar radiation and cloud cover); hence, the  SWAT
weather generator was preferred to enable consistent 30-year simulations.  HSPF does not
simulate crop growth, so monthly coefficients are incorporated in the model to convert reference
crop PET to values appropriate to different crop stages using the Food and Agriculture
Organization (FAO) method (Allen et al.,  1998).
                                          4-15

-------
       Table 4-4. Weather station statistics for the 20 study areas (1971-2000)
Model area
Apalachicola-Chattahoochee-
Flint Basins
Arizona: Salt, Verde, and San
Pedro
Cook Inlet Basin
Georgia-Florida Coastal
Plain
Illinois River Basin
Lake Erie Drainages
Lake Pontchartrain Drainage
Minnesota River Basin
Nebraska: Loup and Elkhom
River Basins
New England Coastal Basins
Powder and Tongue River
Basins
Rio Grande Valley
Sacramento River Basin
Southern California Coastal
Basins
South Platte River Basin
Susquehanna River Basin
Tar and Neuse River Basins
Trinity River Basin
Upper Colorado River Basin
Willamette River Basin
Number of
precipitation
stations
37
29
14
51
72
57
26
39
81
52
37
53
28
85
50
60
40
64
47
37
Average annual
precipitation total
(inches)
54.26
19.67
28.50
53.21
38.25
38.15
66.33
28.26
26.10
48.45
17.70
15.18
37.47
20.21
16.82
41.30
49.91
40.65
16.36
58.38
Number of
temperature stations
22
25
14
37
47
41
15
32
31
36
30
41
18
33
23
27
28
32
39
29
Average annual
temperature (°F)
63.43
56.81
34.16
68.24
49.00
49.10
66.64
43.90
48.35
46.23
44.15
44.71
57.45
61.20
43.46
48.26
59.91
64.78
41.73
51.19
4.3.  SIMULATION OUTPUT AND ENDPOINTS
Simulations focused on streamflow, total nitrogen, total phosphorus, and total suspended solids
loads. Output from both models was analyzed as daily time series over the 30-year analysis
period.  Several summary metrics or endpoints were also calculated based on the daily time
series. Because of calibration uncertainty inherent in modeling at this scale, estimates of relative
change between historical and future simulations are most relevant.  In addition to basic
streamflow statistics, comparisons are made for  100-year flood peak (fit with Log Pearson type
III distribution; USGS,  1982), average annual 7-day low flow, Richards-Baker flashiness index
(a measure of the frequency and rapidity of short-term changes in streamflow; Baker et al.,
2004), and days to the centroid  of mass for the annual streamflow on a water-year basis (i.e.,
                                          4-16

-------
days from previous October 1 at which half of the streamflow for the water year is achieved, an
important indicator of changes in the snow accumulation and melt cycle).  For the Log Pearson
III estimator, use of a regionalized skew coefficient is not appropriate to climate change scenario
applications as the regional map represents existing climate.  Therefore, the K factor is estimated
using the skew coefficient from the model output only, without any weighting with the regional
estimate.

Each of the streamflow endpoints discussed in the preceding section has been calculated for each
scenario at the output of each HUC-8 contained within a study area.  Several other summary
measures of the water balance, largely drawn from the work of Kurd et al. (1999), are
summarized as averages at the whole-watershed scale. These are the Dryness Ratio (fraction of
precipitation that is lost to ET as reported by the SWAT model), Low Flow Sensitivity
(expressed as the rate of baseflow generation by shallow groundwater, tile drainage, and lateral
subsurface flow pathways in units of cfs/mi2), Surface Runoff Fraction (the fraction of total
streamflow from the uplands that occurs through overland flow pathways), Snowmelt Fraction
(the fraction of total streamflow from the uplands that is generated by melting snow), and Deep
Recharge Rate (the annual average depth of water simulated as recharging deep aquifers that do
not interact with local streams).  Table 4-5 provides a summary of streamflow and water quality
endpoints evaluated in this study.

The mobilization and transport of pollutants will also be affected by climate and land-use
change, both as a direct result of hydrologic changes and through changes in land cover and plant
growth. Monthly and annual loads of sediment, phosphorus, and nitrogen are likely the most
useful and reliable measures of water quality  produced by the analysis. Accordingly, the focus
of comparison among scenarios is on monthly and average annual loads for total suspended
solids, total nitrogen, and total phosphorus. As with the streamflow simulation, it is more robust
to examine relative rather than absolute changes in simulated pollutant loads when comparing
scenarios to current conditions. Thus, we also calculate and express results as percent changes.
All models are calibrated and validated, but in many cases current loads are imprecisely known
due to limited monitoring data.

Because the sediment load in rivers/streams is often dominated by channel adjustment processes,
which are highly site specific and occur at a fine spatial scale, it is anticipated that precision in
the simulation of sediment and sediment-associated pollutant loads will be relatively low.
Nutrient balances can also be strongly affected by biological processes in the channels, which
can only be roughly approximated at the scale of modeling undertaken. It should also be noted
that the modeling makes the following assumptions that limit uses for absolute (as opposed to
relative) simulations of future pollutant loads: (1) external boundary conditions (if needed), such
as upstream inflows and pollutant loads, are constant; (2) point source discharges and water
withdrawals are assumed constant at current rates; (3) no provision is made for human adaptation
in rural land management, such as shifts in crop type  in response to climate change; and (4) plant
growth responses to climate change are simulated to the extent they are represented in the SWAT
plant growth model; however, large-scale shifts in natural cover type in response to climate
change are not simulated.
                                          4-17

-------
       Table 4-5. Summary of streamflow and water quality endpoints
Endpoint
Future Flow Volume
Average Seven Day
Low Flow
100 Year Peak Flow
Days to Flow Centroid
Richards-Baker
Flashiness Index
Dryness Ratio
Low Flow Sensitivity
Surface Runoff
Fraction
Snowmelt Fraction
Deep Recharge
AET
PET
Total Suspended
Solids (TSS)
Total Phosphorus (TP)
Total Nitrogen (TN)
Dimension
\}lt
L3/t
L3/t
f(days)
dimensionless
dimensionless
L/t
dimensionless
dimensionless
L/t
L/t
Lit
mass/t
mass/f
mass/t
Description
Average of simulated streamflow
volume per unit time
Average annual 7-day low
streamflow event volume
Estimated peak streamflow rate
based on annual flow maxima
series, Log Pearson III method
Number of days from the previous
October 1 (start of water year) at
which half of the streamflow
volume for that water year is
achieved
Indicator of the frequency and
rapidity of short term changes in
daily streamflow rates
Fraction of input precipitation lost
toET
Rate of baseflow contributions
from shallow groundwater, tile
drainage, and lateral subsurface
flow pathways, depth per unit time
Fraction of streamflow contributed
by overland flow pathways
Fraction of streamflow contributed
by snowmelt
Depth of water recharging deep
aquifers per unit time
Actual depth of evapotranspiration
lost to the atmosphere per unit time
Theoretical potential
evapotranspiration as depth per unit
time, assuming moisture not
limiting
Mass load of suspended sediment
exiting stream reach per unit time
Mass load of total phosphorus
exiting reach per unit time
Mass load of total nitrogen exiting
stream reach per unit time
Calculation
Sum of annual streamflow volume
simulated by the watershed model
Lowest 7-day-average streamflow
simulated for each year
Log Pearson III extreme value estimate
following USGS (1982), based on
simulated annual maxima series
Count of days to 50% of simulated total
annual streamflow volume for each
water year.
Analyzed by method given in Baker et
al. (2004), applied to daily streamflow
series for each year
Calculated as (precipitation -
outflow)/precipitation for consistency
with Kurd etal. (1999)
Sum of simulated streamflow from
shallow groundwater, tile drainage, and
lateral subsurface flow pathways
divided by area.
Surface runoff divided by total outflow.
Estimated as water equivalent of
simulated snowfall divided by total
precipitation
Total water volume simulated as lost to
deep recharge divided by area
Evapotranspiration simulated by the
watershed model
Potential evapotranspiration simulated
by the Penman-Monteith method
(Jensen et al., 1990)
Sum of simulated mass exiting a stream
reach
Sum of simulated mass exiting a stream
reach
Sum of simulated mass exiting a stream
reach
4.4. MODEL CALIBRATION AND VALIDATION
The watershed models were calibrated and validated in each of the study areas in accordance
with the project Quality Assurance Project Plan (QAPP; see Appendix B).  The following section
                                        4-18

-------
provides a brief summary of calibration and validation methods and results. Detailed description
of calibration and validation methods and results for the individual study areas are presented in
Appendices D through W.

Calibration refers to the adjustment of model parameters to reproduce or fit simulation results to
observed data. Calibration is required for parameters that cannot be deterministically and
uniquely evaluated from topographic, climatic, physical, and chemical characteristics of the
watershed and compounds of interest. Validation is performed by application of the calibrated
model to a different period of observed data to test the robustness of the calibrated parameter set.
If the model exhibited a significant degradation in performance in the validation period, the
calibration process is repeated until results are considered acceptable.

The calibration and validation approach for the study areas was to first focus on a single HUC-8
within the larger study area  (preferably one for which some modeling was already available
along with a good record of flow gaging and water quality monitoring data), and then extend the
calibration to adjacent areas with modifications as needed to achieve a reasonable fit at multiple
spatial scales. Each HUC-8 watershed was generally subdivided into approximately 8 subbasins,
approximating the HUC-10  watershed scale.

The base period of observed data used for calibration and validation was approximately 1970 to
2000, with some variation depending on availability of meteorological data, while the base land
use was from 2001 NLCD.  In watersheds with significant land-use change, moving back too far
from 2001 may not provide  a firm  basis for calibration. Therefore, calibration generally focused
on approximately the 1991-2001 time period, although the full 1971-2000 period was used for
comparison to future changes. Validation was typically performed on the period before 1991
and/or data from post-1991  at different locations.

4.4.1. Hydrology
The goal of hydrologic calibration for both HSPF and SWAT was to achieve error statistics for
total streamflow volume, seasonal  streamflow volume, and high and low streamflow within the
range recommended by Lumb et al. (1994) and Donigian (2000) while also maximizing the
Nash-Sutcliffe coefficient of model fit efficiency (£).  Standardized spreadsheet tools were
developed to help ensure consistency in the calibration and validation process across watersheds,
and to provide a standardized set of statistics and graphical comparisons to data. These statistics
were used to adjust appropriate model parameters until a good statistical match was shown
between the model output and observed data.

Lumb et al. (1994) and Donigian (2000) recommend performance targets for HSPF based on
relative  mean errors calculated from simulated and observed daily average streamflow.
Donigian classified these into qualitative ranges, which were modified slightly in this project for
application to both HSPF and SWAT (see Table 4-6).  In general, hydrologic calibration
endeavored to achieve a "good" level of model fit where possible.  It is important to note that the
tolerance ranges are intended to be applied to mean values and that individual events or
observations may show larger differences and still be acceptable (Donigian, 2000).
                                          4-19

-------
       Table 4-6.  Performance targets for hydrologic simulation (magnitude of
       annual and seasonal relative mean error) from Donigian (2000)
Model component
1 . Error in total volume
2. Error in 50% lowest streamflow volumes
3. Error in 10% highest streamflow volumes
4. Error in storm volume
5. Winter volume error
6. Spring volume error
7. Summer volume error
8. Fall volume error
9. Error in summer storm volumes
Very good (%)
<5
<5
<10
<10
<15
<15
<15
<15
<25
Good (%)
5-10
5-10
10-15
10-20
15-30
15-30
15-30
15-30
25-50
Fair (%)
10-15
10-25
15-25
20-30
30-50
30-50
30-50
30-50
50-75
Poor (%)
>15
>25
>25
>30
>50
>50
>50
>50
>75
The Nash-Sutcliffe coefficient of model fit efficiency (E) is also widely used to evaluate the
performance of models that predict time series.  Nash and Sutcliffe (1970) define E as:
                                          Zfa-^)
                                          t(o,-o)2
                                                                                     4-4
where Ot andPj represent members of a set of n paired time series observations and predictions,
respectively, and O is the mean of the observed values.  E ranges from minus infinity to 1.0, with
higher values indicating better agreement. The coefficient represents the ratio of the mean
square error to the variance in the observed data, subtracted from unity (Wilcox et al., 1990). A
value of zero for E indicates that the observed mean is as good a predictor of time series values
as the model,  while negative values indicate that the observed mean is a better predictor than the
model.  A value of E greater than 0.7 is often  taken as an indicator of a good model fit
(Donigian, 2000). Note, however, that the value depends on the time basis on which the
coefficient is evaluated.  That is, values of E for monthly average streamflow are typically
noticeably greater than values of E for daily streamflow, as watershed models, in the face of
uncertainty in the representativeness of precipitation records, are often better predictors of
interseasonal trends than of intraseasonal variability.  Moriasi et al., (2007) recommend a Nash-
Sutcliffe E of 0.50 or better (applied to monthly sums) as an indicator of adequate hydrologic
calibration when accompanied by a relative error of 25% or less.

A potential problem with the use of E is that it depends  on squared differences, making it overly
sensitive to extreme values (Legates and McCabe, 1999).  This is particularly problematic for
sparse time series, such as water quality observations, in which poor estimation of one or a few
high outliers may strongly influence the resulting statistic.  It is an even  greater problem for the
                                          4-20

-------
comparison of model output to load estimates based on sparse concentration data, as these
estimates are themselves highly uncertain (using point-in-time grab samples to represent daily
averages and interpolating to unobserved days), further increasing the leverage associated with
high outliers.

To address these issues and lessen the effect of outliers, Garrick et al. (1978) proposed use of a
baseline adjusted coefficient of model fit efficiency, E\, which depends on absolute differences
rather than squared differences:
                                                                                      4-5
Garrick's proposed statistic is actually more general, allowing 0 ' to be a baseline value that may
be a function of time or of other variables, rather than simply the mean. E\ 'may be similar to or
greater or less than E for a given set of predictions and measurements depending on the type of
outliers that are present.

For most watershed models, E is an appropriate measure for the fit of streamflow time series in
which complete series of observations are known with reasonable precision. E\ 'is a more
appropriate and stable measure for the comparison of simulated pollutant loads to estimates
based on sparse observed data.

4.4.1.1. Flow Calibration Adjustments
HSPF and SWAT hydrology calibration adjustments were made for a range of sensitive model
parameters selected to represent key watershed processes affecting runoff (U.S. EPA, 2000;
Neitsch et al., 2005; see Tables 4-7 and 4-8, respectively, for selected key parameters most
frequently adjusted). The adjustment of other parameters and the degree of adjustment to each
parameter vary by watershed. Details are provided in the individual calibration reports for each
of the watersheds in Appendices D through W.
                                          4-21

-------
       Table 4-7.  Key hydrology calibration parameters for HSPF
Parameter name
INFILT
AGWRC
LZSN
BASETP
KMELT
PET factor
DEEPFR
LZETP
Definition
Nominal infiltration rate parameter
Ground-water recession rate
Lower zone nominal soil moisture storage
ET by riparian vegetation
Degree-day melt factor
Potential evapotranspiration
Fraction of groundwater inflow that will enter deep groundwater
Lower zone E-T parameter
       Table 4-8.  Key hydrology calibration parameters for SWAT
Parameter name
CN
ESCO
SLTRLAG
ALPHA_BF
GW_DELAY
CANMAX
OV_N, CH_N2, CH_N1
Sol_AWC
Bank storage and recession rates
Snow parameters SFTMP, SMTMP,
SMFMX and SMFMN
TIMP
CHJC1
Definition
Curve numbers — varied systematically by land use
Soil evaporation compensation factor
Surface runoff lag coefficient
Baseflow alpha factor
Groundwater delay time
Maximum canopy storage
Manning's "w" values for overland flow, main channels, and tributary channels
Available water capacity of the soil layer, mm water/mm of soil
Bank storage and recession rates
Snowfall temperature, snowmelt base temperature, maximum melt rate for snow
during year, and minimum melt rate for snow during year
Snow pack temperature lag factor
Effective hydraulic conductivity in tributary channel alluvium
4.4.2. Water Quality
The models in this study are designed to simulate total nitrogen, total phosphorus, and total
suspended solids. The first objective of calibration was to reduce the relative absolute deviation
between simulated and estimated loads to below 25% if possible.  The water quality calibration
focuses on the replication of monthly loads, as specified in the project QAPP (see Appendix B).
While a close match to individual, instantaneous concentration observations cannot be expected
given the approach taken in the model simulations of water quality, the calibration also examined
the general relationship of observed and predicted concentrations with the intent of minimizing
bias relative to streamflow regime or time of year.  Comparison to monthly loads presents
challenges, as monthly loads are not observed. Instead, monthly loads must be estimated from
                                          4-22

-------
scattered concentration grab samples and continuous streamflow records.  As a result, the
monthly load calibration is inevitably based on the comparison of two uncertain numbers.
Nonetheless, calibration is able to achieve a reasonable agreement.  The direct comparison of
estimated and simulated monthly loads was supplemented by detailed examinations of the
relationships of streamflow to loads and concentrations, and the distribution of concentration
prediction errors versus streamflow, time, and season to help minimize bias in the calibration.

For application on a nationwide basis, it was assumed that total suspended solids and total
phosphorus loads will likely exhibit a strong positive correlation to streamflow (and associated
erosive processes), while total nitrogen loads, which often have a dominant subsurface loading
component, will  not (Allan, 1986; Burwell et al., 1975; Follett, 1995). Accordingly, total
suspended solids and total phosphorus loads were estimated from observations using a
flow-stratified log-log regression approach, while total nitrogen loads were estimated using a
flow-stratified averaging estimator, consistent with the findings of Preston et al. (1989).

4.4.2.1. Water Quality Calibration Adjustments
Water quality calibration began with sediment transport processes.  Observed suspended solids
concentrations are the result of multiple processes, including sediment detachment, sediment
transport in overland flow, and channel scour and deposition processes. The sediment
detachment routines for both SWAT and HSPF were related to USLE parameters available in the
soils database. For most basins, calibration focuses on sediment transport in overland flow,
using the peak rate or transport rate factors available in both models.  Channel scour and
deposition processes were modified where needed to achieve a fit to observations or where
detailed work with prior models provided a basis for modifying the default parameters.

In HSPF, nitrogen loading from the land surface was simulated as a buildup/washoff process,
while phosphorus was simulated as sediment-associated. Both nitrogen and phosphorus also
were simulated with dissolved-phase loads from interflow and groundwater discharge.
Calibration for nutrients in HSPF primarily addressed adjustments of the buildup/washoff
coefficients or sediment potency (concentration relative to sediment load) factors and  monthly
subsurface discharge concentrations. In SWAT, the nutrient simulation is intimately linked to
the plant growth model, but is sensitive to initial nutrient concentrations and the ability of plants
to withdraw nutrients from various  soil layers. In watersheds where significant channel  scour
was simulated, the nutrient content of scoured sediment was also an important calibration
parameter.

4.4.3. Accuracy of the Watershed Models
The quality of model fit varies with the study area and parameter considered. In general, the full
suite of SWAT models for the 20 watersheds—after calibration—provide a good to excellent
representation of the water balance at the monthly scale and a fair to good representation of
hydrology at the daily scale (see Table 4-9 for the initial calibration site results). The  quality of
model fit to hydrology as measured at multiple stations (HUC-8 spatial scale and larger)
throughout the watershed was, not surprisingly, better when a spatial  calibration approach was
used.  At all calibration and validation sites, the median  monthly Nash-Sutcliffe E coefficient
from the SWAT  models was 0.74 for both the pilot and nonpilot study areas. More detailed
calibration and validation results for each study area are provided in Appendices D through W.
                                          4-23

-------
Less precise model fit to observations resulted in several study areas for various reasons.  In
addition to differences in individual modeler preferences and skill, Low E coefficients in the Rio
Grande Valley likely reflect insufficient knowledge of operations of the many reservoirs in the
basin.  Calibrating watershed hydrology was problematic in systems dominated by large-scale
interactions with regional groundwater systems—notably, Verde River in Arizona and the
Loup/Elkhorn River system in the Nebraska sandhills. Both HSPF and SWAT use simplistic
storage reservoir representations of groundwater in which water can percolate from the soil
profile into local shallow groundwater storage, from which it is gradually released following an
exponential decay pattern characterized by a recession coefficient. Perennial streamflow in the
Verde River is sustained by groundwater discharges of nonlocal origin that derive from the
upstream Chino basin.  The Loup and Elkhorn Rivers drain highly porous sands where surface
runoff is minimal and streamflow in some tributary rivers is nearly constant and only weakly
correlated to rainfall patterns (e.g., see Figure 4-1), a situation that is difficult to address in a
rainfall-runoff model without linking to a true groundwater simulation model.
                                                Avg Monthly Rainfall (in)

                                                -Avg Observed Flow (10/1/1989 to 9/30/1999)
    300
    200
 o
    100
      O-89       A-91
O-92
A-94
O-95
A-97
O-98
                                       Month
       Figure 4-1.  Example of weak correlation of rainfall and flow in the Dismal
       River at Thedford, NE (USGS 06775900) in the Loup River basin.
Different modelers handled the situation in these two regions in different ways. For the Verde
River (where both HSPF and SWAT were applied) the regional groundwater inflow was
specified as an external forcing time series. This has the advantage of allowing the model
calibration to focus on rainfall-runoff events that are responsible for most year-to-year variability
in streamflow and most pollutant transport. The major disadvantage is that there is not a clear
means to specify how this groundwater forcing might respond to changes in climate. Instead,
results for the Verde River show relative changes that would be expected under the assumption
that the regional groundwater discharge does not change.

For the Loup and Elkhorn River basins, a reasonable fit to both calibration and validation periods
was obtained by specifying extremely slow groundwater recession rates in conjunction with use
                                          4-24

-------
of the soil crack flow option (which allows a fraction of rainfall to flow directly to groundwater)
in the sandhill region.  This approach can replicate the major observed features of the water
balance, although it does not achieve a high degree of precision in explaining day-to-day
variability in observed streamflow. Further, the simulated groundwater discharges are
responsive to changes in climate forcing. However, use of this approach comes at a cost due to
the way that groundwater is simulated in the SWAT model. Specifically, SWAT simulates
baseflow discharge on a given day as a function of discharge on the previous day, modified by
the recession coefficient, plus the effects of new recharge to groundwater.  Groundwater
discharge at the start of the simulation is constrained to be zero. Use of a very slow recession
rate gives a reasonable fit to the calibration and validation periods in this study area;  however, it
also results in very slow convergence of estimated groundwater discharge from the initial zero.
This resulted in a situation in which it took approximately 10 years for streamflow to reach levels
in line with observations. Thus, simulated streamflow for the early years are often zero. Adding
a longer spinup period does not resolve the problem as the low recession rate results  in a
nonstationary solution in which baseflow continues to gradually increase over time and the
simulated streamflow eventually overshoots observations during the calibration period  if the
spinup period is extended. Due to this issue, change scenario results are presented only for the
20-year calibration and validation periods in the Loup and Elkhorn River study area.

Calibration and validation for water quality is subject to higher uncertainty than streamflow
calibration due to limited amounts of monitoring data and a simplified representation of the
multiple complex processes that determine in-stream pollutant concentrations.  The primary
objective of water quality simulation in this project is to assess relative changes in pollutant
loads, but loads are not directly observed. Inferring loads from point-in-time concentration data
and streamflow introduces another layer of uncertainty into the calibration process. Calibration
also examined observed versus predicted concentrations;  however, SWAT, as a daily curve
number model, does not have a high level of skill in simulating instantaneous concentrations,
particularly during high flow events, and is better suited to the simulation of loads at the weekly
to monthly scale.

As with the hydrology calibration, the reliability of the models for simulating changes in water
quality appears to increase with calibration at multiple locations. In general, it is more difficult
to obtain a high level of precision for simulated water quality than for hydrology in a watershed
model, as the processes are complex, the data typically sparse, and any errors in hydrology tend
to be amplified in the water quality simulation.  The water quality calibration is based on loads,
but loads are not directly observed. Instead, loads are inferred from sparse concentration
monitoring data and streamflow gaging.  Thus, both the simulated and "observed" loads are
subject to considerable uncertainty. Comparison based on concentrations can also be
problematic, as most water quality samples are grab samples that represent points in  time and
space, whereas model output is integrated over a stream segment and may produce large
apparent errors due to small shifts in timing. Finally, most stations at the HUC-8 scale include
upstream point sources, which  often have a strong influence on low-flow concentrations and load
estimates. Limited knowledge about point source loads thus also creates a challenge for the
water quality calibration. In most cases, the pollutant load simulations from the SWAT model
appear to be in the fair to good range (see Table 4-9)—except in a few cases where parameters
were extended from one station to another watershed without adjustment, giving poor results.
This suggests limits to the reliability of simulation results in the portions of watersheds for which
                                          4-25

-------
calibration was not pursued. Nonetheless, simulations of the relative response to climate change
and land development scenarios are more reliable than for the actual observed future values—as
long as the significant processes that determine pollutant load and transport within a watershed
are represented.

HSPF model calibration for the five pilot sites provided a somewhat stronger fit to daily
streamflow in four of the five watersheds (see Table 4-10), presumably at least in part due to
HSPF's use of subdaily precipitation.  In two models, the fit to total suspended solids load was
notably worse for HSPF, apparently due to the difficulties in adjusting the more complex channel
scour and  deposition routines of this model with limited data and on a compressed schedule.
                                          4-26

-------
           Table 4-9.  Summary of SWAT model fit for initial calibration site (20 study areas)
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Initial calibration/
validation watershed
Upper Flint River
Verde River
Kenai River
Ochlockonee River
Iroquois River
Cuyahoga River
Amite River
Cottonwood River
Elkhorn River
Saco River
Tongue River
Saguache Creek
Sacramento River
Santa Ana River
South Platte River
Initial
calibration/
validation
USGS gage
02349605
09504000
15266300
02329000
05526000
04208000
07378500
05317000
06800500
01066000
06306300
08227000
11377100
11066460
06714000
Hydrology
caL/val. yr
1993-2002/
1983-1992
1992-2002/
1982-1992
1992-2001/
1982-1991
1992-2002/
1982-1992
1992-2001/
1982-1992
1990-2000/
1980-1990
1995-2004/
1985-1994
1992-2002/
1982-1992
1989-1999/
1978-1988
1993-2003/
1983-1993
1993-2003/
1983-1993
1993-2003/
1983-1993
1992-2001/
1983-1992
1991-2001/
1981-1991
1991-2000/
1981-1990
Total volume
cal./val.
(daily and
monthly E)
0.62/0.56
0.88/0.83
0.03/-1.0
0.88/0.32
0.68/0.55
0.80/0.75
0.71/0.80
0.79/0.90
0.70/0.67
0.77/0.71
0.61/0.62
0.70/0.73
0.79/0.69
0.95/0.90
0.79/0.74
0.91/0.83
0.42/0.52
0.70/0.66
0.61/0.76
0.71/0.84
0.72/0.7
0.83/0.82
0.47/0.07
0.53/0.31
0.75/0.57
0.94/0.92
0.63/0.59
0.75/0.68
0.74/0.52
0.86/0.63
Total
volume
cal./val.
(% error)
7.28/3.33
-2.46/5.68
-18.96/19.49
4.25/-5.S4
-16.99/-2.98
-3.32/-13.38
-1.61/-0.93
-5.41/-0.84
-2.59/-8.81
1.08/0.67
9.26/-9.9S
-4.92/32.99
10.23/10.06
3.71/1.61
9.82/-16.28
Water quality
cal./val yr
1999-2002/
1991-1998
1993-2002/
1986-1992
1985-2001/
1972-1984
1992-2002/
1982-1992
1985-2001/
1978-1984
1990-2000/
1980-1990
1984-1994/ND
1993-2000/
1986-1992
1990-1995/
1979-1989
1993-2003/
1983-1993
1993-2003/
1982-2002
1985-2003/
1973-1984
1997-200 I/
1973-1996
1998-2000/ND
1993-2000/ND
TSS monthly
load cal./val.
(% error)
-9/17
16.9/-42.6
66.4/64.1
9.S/-6.6
38/39
67.9/69.8
9.2/NA
9.2/9
59.6/66.8
-9/3.2
-21.8/-3.4
57.3/41
-2/-5S
19/NA
86.6/NA
TP monthly
load cal./val.
(% error)
-50/-30
83.5/31.4
83.2/82.18
-7.4/-S.8
5/-1
23.9/-12.5
2.4/NA
9.3/-21.6
24.2/34.9
9.6/-11.5
8.8/35.1
-46.9/-6S3.98
-8/-33
-14.7/NA
-14/NA
TN monthly
load cal./val.
(% error)
-18/9
-14.4/-15.9
57.3/50.4
-8/-5
56/60
35.8/13.7
-8.9/NA
-8.9/-1.3
28.1/18.1
27.5/26.3
3.9/31.5
-28.3/-909.1
-135/-156
-5.5/NA
6.1/NA
to

-------
           Table 4-9. Summary of SWAT model fit for initial calibration site (20 study areas) (continued)
Study area
Susq
TarNeu
Trin
UppCol
Willa
Initial calibration/
validation watershed
Raystown Branch of the
Juniata River
Contentnea Creek
Trinity River
Colorado River
Tualatin River
Initial
calibration/
validation
USGS gage
02050303
02091500
08066500
09070500
14207500
Hydrology
caL/val. yr
1995-2005/
1985-1995
1993-2003/
1983-1993
1992-2001/
1982-1991
1992-2002/
1982-1992
1995-2005/
1985-1995
Total volume
cal./val.
(daily and
monthly E)
0.29/0.42
0.67/0.66
0.68/0.64
0.86/0.74
0.62/0.47
0.74/0.76
0.83/0.78
0.86/0.82
0.49/0.39
0.88/0.81
Total
volume
cal./val.
(% error)
-5.41/16.3
-3.98/-1.18
-6.88/0.70
8.18/0.93
-4.76/-12.1
Water quality
cal./val yr
1991-2000/ 1990
1993-2003/
1983-1993
1985-2001/
1972-1984
1992-2002/ —
ND
1991-1995/
1986-1990
TSS monthly
load cal./val.
(% error)
-10.1/-33.6
-19.9/9.9
9.2/-17.4
0.4/NA
-12/-7
TP monthly
load cal./val.
(% error)
-0.5/-9.2
15.9/5.3
3/-21.58
47.4/NA
-114/-105
TN monthly
load cal./val.
(% error)
28.6/43.9
-5.6/5.3
-3.8/-31.9
15.1/NA
-727-66
to
oo
Table 4-10.  Summary of HSPF model fit for initial calibration sites (five pilot study areas)
Study area
ACF
Ariz
Minn
Susq
Willa
Initial calibration/
validation watershed
Upper Flint River
Verde River
Cottonwood River
Raystown Branch of the
Juniata River
Tualatin River
Initial
calibration/
validation
USGS gage
02349605
09504000
05317000
02050303
14207500
Hydrology
caL/val. yr
1993-2002/
1983-1992
1992-2002/
1982-1992
1992-2002/
1982-1992
1995-2005/
1985-1995
1995-2005/
1985-1995
Total volume
cal./val. (daily
and monthly E)
0.71/0.65
0.93/0.90
0.48/0.45
0.85/0.66
0.75/0.78
0.69/0.86
0.70/0.55
0.90/0.87
0.73/0.81
0.96/0.92
Total
volume
cal./val.
(% error)
5.50/5.79
2.43/6.31
1.61/14.78
-0.16/-8.0
-3.92/-9.80
Water quality
cal./val yr
1999-2002/
1991-1998
1993-2002/
1986-1992
1993-2002/
1986-1992
1991-2000/ 1990
1991-1995/
1986-1990
TSS monthly
load cal./val.
(% error)
-117/-78
31/-41
7.5/13.1
-78.2/-89J
3.0/4.8
TP monthly
load cal./val.
(% error)
-S9/-23
87/66
23/15.8
26.0/21.5
-1.2/-9.3
TN monthly
load cal./val.
(% error)
-30/-22
1.6/-2.7
15.4/16.2
7.0/17.2
2.21-63

-------
        5. CLIMATE CHANGE AND URBAN DEVELOPMENT SCENARIOS
Watershed simulations were conducted using SWAT and HSPF in each study area to assess the
sensitivity of streamflow, total nitrogen, total phosphorus, and total suspended solids loads to a
range of plausible mid-21st century climate change and urban development scenarios.  Climate
change scenarios are based on downscaled climate model projections for mid-21st century from
the NARCCAP and BCSD (Maurer et al., 2007) data sets. Fourteen climate scenarios were
applied to the five pilot sites, and a subset of 6 climate scenarios from the NARCCAP archive
were applied to the nonpilot sites. Scenarios of urban and residential development were based on
projections from EPA's ICLUS project (U.S. EPA, 2009c).

Simulations  were conducted to assess the response to climate change scenarios alone, urban and
residential development scenarios alone, and combined climate change and urban development
scenarios. The following sections discuss the use and implementation of climate change and
urban development scenarios in this study.

5.1.  SCENARIO-BASED APPROACH
The scientific uncertainties related to our understanding of the physical climate system are large,
and they will continue to be large for the foreseeable future.  It is beyond our current capabilities
to predict with accuracy decadal (and longer) climate changes at the regional spatial scales of
relevance for watershed processes (e.g., see Cox and Stephenson,  2007; Stainforth et al., 2007;
Raisanen, 2007; Hawkins and Sutton, 2009; among many others). The uncertainties associated
with socioeconomic trajectories, technological advances, and regulatory changes that will drive
greenhouse gas emissions changes (and land-use changes) are even larger and less potentially
tractable.

Faced with this uncertainty, an appropriate strategy is to take a scenario-based approach to the
problem of understanding climate change impacts on water quality.  A scenario is a plausible
description of how the future may develop, based on a coherent and internally consistent set of
assumptions about driving forces and key relationships (IPCC, 2007). Scenarios are used in
assessments  to provide alternative views of future conditions considered likely to influence a
given system or activity. By systematically exploring the implications of a wide range of
plausible alternative futures, or scenarios, we can reveal where the greatest vulnerabilities lie.
This information can be used by decision makers to help understand and guide the development
of response strategies for managing climate risk.  A critical step in this approach is to create a
number of plausible future states that span the key uncertainties in the problem.  The goal is not
to estimate a single, "most likely" future trajectory for each study watershed, but instead to
understand, to the extent feasible, how big an envelope of potential future impacts we are unable
to discount and must therefore incorporate into future planning.

Note that for climate change studies, the word "scenario" is often  used in the context of the IPCC
greenhouse gas storylines.  The IPCC emissions scenarios describe alternative development
pathways, covering a range of demographic, economic, and technological driving forces that
affect greenhouse gas emissions. This can produce some confusion when phrases like "climate
change scenarios" are used to refer to the future climates simulated using these greenhouse gas
                                          5-1

-------
storylines. For the purposes of this study, "scenario" is a generic term that can be applied to any
defined future, including a climate future or a land-use future, among others.

5.2.  CLIMATE CHANGE SCENARIOS
It is standard practice when assessing climate change impacts to consider an ensemble of climate
change scenarios based on different climate models and emissions pathways.  Use of a single
model run is not considered scientifically rigorous because different GCMs often produce very
different results, and there is no consensus in the climate modeling community that any model is
comprehensively better or more accurate than the others (e.g., see Gleckler et al., 2008).
Different methods of "downscaling" GCM model output to finer spatial scales can also influence
the variability among models.

5.2.1. Future Climate Models, Sources, and Downscaling
To sample across this model-based uncertainty, this project focused on six climate change
scenarios derived from four GCMs covered by the regional downscaling efforts of the
NARCCAP (http://www.narccap.ucar.edu). NARCCAP uses higher-resolution RCMs to
dynamically downscale output from four of the GCMs used in the  IPCC 4th Assessment Report
(IPCC, 2007) to a 50 x 50 km  grid over North America. This downscaled output is archived for
the two 30-year periods (1971-2000 and 2041-2070) at a temporal resolution of 3 hours.
NARCCAP uses the IPCC's A2 greenhouse gas storyline (which at the time of development was
a relatively "pessimistic" future greenhouse gas trajectory, but is now more middle-of-the-road
compared to current trends and the most recently developed scenarios).  We note that, by mid-
21st Century, the different IPCC greenhouse gas storylines  have not yet diverged much, so
impact of the choice of any one particular storyline is diminished compared to later in the
century.

At the time we initiated the watershed modeling, six downscaled scenarios were available from
NARCCAP, and we are using these six as our common set of climate scenarios across all the
20 watersheds, as listed in Table 5-1.

One of the objectives of this work was to investigate the influence of downscaling approaches on
watershed model simulations.  To evaluate the sensitivity of our results to downscaling
methodology, we ran the watershed models in the five pilot sites with eight additional scenarios
(also listed in Table 5-1)  derived from the same four GCMs used in NARCCAP: four scenarios
interpolated to station locations directly from the GCM output (without downscaling), and four
scenarios based on the BCSD statistically downscaled climate projections described by Maurer et
al. (2007), and served at: http://gdo-dcp.ucllnl.org/downscaled_cmip3_projections/.  The BCSD
data provides monthly mean surface air temperature and precipitation rates for the contiguous
United States (along with portions of Canada and Northern Mexico) at a horizontal grid spacing
of 1/8 degree (roughly 12 x 12 km2) for the period 1950-2099.

The BCSD climate projections use statistical downscaling to interpret GCMs to a finer resolution
based on  current observations. The principal potential weakness of this approach is an
assumption of stationarity. That is, the  assumption is made that the relationship between large-
scale precipitation and temperature and local precipitation and temperature in the future will be
the same  as in the past. Thus, the method can successfully account for orographic effects that are
                                          5-2

-------
observed in current data, but not for impacts that might result from the interaction of changed
wind direction and orographic effects.  A second assumption included in the bias-correction step
of the BCSD method is that any biases exhibited by a GCM for the historical period will also be
exhibited in simulations of future periods.
        Table 5-1. Climate models and source of model data used to develop climate
        change scenarios
Scenario #
Climate model(s) (GCM/RCM)
NARCCAP (dynamically downscaled)
1
2
3
4
5
6
CGCM3/CRCM
HadCM3/HRM3
GFDL/RCM3
GFDL/GFDL hi res
CGCM3/RCM3
CCSM/WRFP
GCM (without downscaling)
7
8
9
10
CGCM3
HadCM3
GFDL
CCSM
BCSD (statistically downscaled)
11
12
13
14
CGCM3
HadCM3
GFDL
CCSM
Model Abbreviations:
CGCM3:         Third Generation Coupled Global Climate Model
                http://www.ee.gc.ca/ccmac-cccma/default.asp?lang=En&n=4A642EDE-l
HadCM3:        Hadley Centre Coupled Model, version 3
                http://www-pcmdi.llnl.gov/ipcc/model_documentation/HadCM3.htm
GFDL:           Geophysical Fluid Dynamics Laboratory GCM
                http://www-pcmdi.llnl.gov/ipcc/model_documentation/GFDL-cm2.htm
CCSM:           Community Climate System Model
                http://www-pcmdi.llnl.gov/ipcc/model_documentation/CCSM3.htm
CRCM:          Canadian Regional Climate Model
                http://www.ee.gc.ca/ccmac-cccma/default.asp?lang=En&n=4A642EDE-l
RCM3:           Regional Climate Model, version 3
                http://users.ictp.it/~pubregcm/RegCM3/
HRM3:           Hadley Region Model 3
                http://precis.metoffice.com/
WRFP:           Weather Research and Forecasting Model
                http://www.wrf-model.org/index.php
GFDL hi res:      Geophysical Fluid Dynamics Laboratory 50-km global atmospheric time slice
                http://www-pcmdi.llnl.gov/ipcc/model_documentation/GFDL-cm2.htm
                                                 5-3

-------
The BCSD scenarios, while all derived from the A2 climate storyline, do not in all cases use the
output of the exact same GCM run that was used to construct the NARCCAP archive.
Specifically, the BCSD results for the Geophysical Fluid Dynamics Laboratory global climate
model (GFDL) and Third Generation Coupled Global Climate Model (CGCM3) GCMs use
exactly the same GCM output as NARCCAP, but BCSD results for Hadley Centre Coupled
Model, version 3 (HadCMS) and Community Climate System Model (CCSM) use different runs
of the A2 scenario than used by NARCCAP. The HadCMS run used in NARCCAP was a
custom run generated specifically for NARCCAP and has not been downscaled for the BCSD
archive. The CCSM run used in NARCCAP is run number 5, which is not available in the
CMIP3 archive used by BCSD. Instead, BCSD uses the HadCMS run 1  and  CCSM run 4 from
the CMIP3 archive for the A2 scenario.  As a result, the most direct comparisons between the
NARCCAP and BCSD data sets are for the GFDL and CGCM3 GCM models.  However, we
still expect comparisons between NARCCAP and BCSD for the HadCMS and CCSM to provide
useful insights when considered along with the GFDL and CGCM3 comparisons.  These
scenarios were evaluated only at the five pilot study areas.

Scenarios for the five pilot sites also examined use of the direct output from the GCM runs used
to drive the NARCCAP downscaling (i.e., no downscaling). Comparison of results from these
scenarios to full dynamical downscaling is expected to inform the accuracy with which simpler
methods can be used to address watershed response.  These scenarios were evaluated only at the
five pilot study areas.

Table 5-1 summarizes the  climate change scenarios used in this study and also contains a
numbering key for shorthand reference to climate scenarios. For example, climate scenario 2
refers to the HadCMS GCM, downscaled with the Hadley Region Model 3 (HRM3) RCM.  All
14 scenarios are applied in the five pilot sites. Only scenarios 1 through 6 are applied for the
nonpilot sites.

5.2.2.  Translation of Climate Model Projections to Watershed Model Weather Inputs
Even the 50-km NARCCAP scale is relatively coarse for watershed modeling. In this study,
meteorological time series for input to the watershed models were created using a "change
factor" or "delta change" method (Anandhi et al., 2011). Using this approach, a period of
baseline observed weather data was selected for each study area (to which the watershed models
have been  calibrated), and the data series adjusted or perturbed to represent a specific type of
climate change projected by a climate model (i.e., a climate change scenario). The benefits  of
the change factor approach include its simplicity, elimination of the need for bias correction, and
ability to create spatially variable climate change scenarios that maintain the  observed historical
spatial correlation structure among different watershed locations. Specifically, there is a
tendency for GCMs to generate too many low-intensity events and to under-simulate the
intensity of heavy events (Sun et al., 2006; Dai, 2006). The frequency and duration of large
events can have significant effects on hydrology, pollutant loading, and other watershed
processes.  Applying the model-derived change factors to the observed precipitation time series
mitigates this problem. Limitations of this approach include the inability to adjust the number
and timing of precipitation events (e.g., to add precipitation events on dry days), and potential
bias introduced through the selection of an arbitrary historical base period as the template for
                                         5-4

-------
future climate time series.  In addition, climate models do not necessarily archive all the
meteorological forcing variables required to run watershed models.

Monthly change factors derived from climate models for each climate change scenario were
calculated by comparing simulated monthly average values for baseline (1971-2000) and future
(2041-2070) climate conditions.  It should be noted that the intention is not to simulate the
impacts of change in land use and climate that occurred over the decades from 1971 to 2000.
Rather, the 1971-2000 meteorological data is assumed to provide a static estimate of natural
climate variability under "current" land-use conditions, which are defined by the selection of the
2001 NLCD baseline land cover.

Change statistics from the climate models were interpolated to locations corresponding to each
of the BASINS meteorological stations and SWAT weather generator stations used in the
watershed models.  Change factors were used to perturb existing records of hourly observed
precipitation and temperature using the CAT  (U.S. EPA, 2009b).  CAT permits the sequential
modification of weather records to introduce a number of alterations, each reflecting various
assumptions concerning the regional manifestations of climate change. Precipitation records can
be modified by (1) multiplying all records by an empirical constant reflecting projected climate
change to  simulate a shift in total precipitation, applied uniformly to all periods and intensity
classes,  (2) selective application of such a multiplier to specific seasons or months, (3) selective
application of the multiplier to a range of months or years within the record, and (4) selective
application of the multiplier to storm events of a specific size or intensity class.  Modification
can be iteratively applied to more than one event size class, allowing changes in frequency and
intensity as well as changes in overall volume of precipitation to be represented. Temperature
records can be modified by adding or subtracting a constant to all values in the record, or
selective application to certain months or years within the record.

The third meteorological time series required by the watershed models is PET, which is
calculated based on  other meteorological time series as described in Section 5.2.2.3.

The full suite of statistics available to calculate PET using the Penman-Monteith energy balance
method  is  not available for the statistically downscaled model runs or the nondownscaled GCM
archives. Data availability is summarized in Table 5-2 and assumptions for creating PET time
series in the absence of specific data sets is discussed in Section 5.2.2.3.

It is important to note that using this approach, multiyear climate change scenarios created by
perturbing multiple years of historical weather data are representative of a single, future time
period and do not represent continuous climatic change during this period (i.e., they are not
transient simulations). Instead, the variability in multiyear scenarios created in this way provides
a snapshot of the natural variability in climate based on historical conditions.

5.2.2.1.  Temperature Changes
Monthly variations (deltas) to the temperature time series throughout the entire time period were
applied using the BASINS CAT.  Monthly adjustments based on each scenario were used and a
modified HSPF binary data (WDM) file was created. The temperature time series were adjusted
based on an additive change using the monthly deltas (temperature difference in Kelvin [K])
calculated from the 2041-2070 to 1971-2000 climate simulation comparison. Beginning with
                                           5-5

-------
the HSPF WDM, an automated script then creates the SWAT observed temperature files (daily
maximum and daily minimum).
       Table 5-2. Climate change data available from each source used to develop
       climate scenarios
Scenario
#
RCM
GCM
Temp.
Prec.
Dew point
temp
Solar
radiation
Wind
speed
Min
temp.
Max
temp.
Prec. bin
data
NARCCAP RCM-downscaled scenarios
1
2
3
4
5
6
CRCM
HRM3
RCM3
GFDL hi res
RCM3
WRFP
CGCM3
HadCM3
GFDL
GFDL
CGCM3
CCSM
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
n/a
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Driving GCMs of the NARCCAP andBCSD scenarios (Le., no downscaling)
1
8
9
10




CGCM3
HadCM3
GFDL
CCSM
X
X
X
X
X
X
X
X
X
n/a
n/a
X
X
n/a
X
X
X
n/a
X
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
BCSD statistically downscaled scenarios
11
12
13
14




CGCM3
HadCM3
GFDL
CCSM
X
X
X
X
X
X
X
X
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
Note: X indicates data are available; n/a indicates not available.

5.2.2.2. Precipitation Changes
Relative changes in the frequency and intensity of precipitation events associated with climate
change may prove to be more influential in determining future patterns of discharge than changes
in overall (annual, seasonal) precipitation. Appendix C provides a summary review of recent
literature on potential changes in the precipitation regime, including volume and intensity, and
the ability of climate models to simulate these changes.

As a general pattern, warming of the lower atmosphere is projected to lead to a more vigorous
hydrologic cycle, characterized by increases in global precipitation,  and proportionally larger
increases in high-intensity precipitation events (Trenberth et al., 2007). Much of the United
States is anticipated to experience an increasing proportion of annual precipitation as larger,
more intense events (Kundzewicz et al., 2007; Groisman et al., 2012). Increasing intensity of
precipitation could increase direct runoff during events and increase nonpoint source loading of
sediment, nutrients, and other pollutants to streams (Gutowski et al., 2008). To ensure that
model  simulations embody the most important dimensions of climate change affecting watershed
                                           5-6

-------
response, it is important that climate change scenarios represent potential changes in
precipitation intensity-frequency-duration relationships.

The most rigorous approach to applying the downscaled climate scenario results to modification
of the existing precipitation series would be to undertake a detailed analysis (by month) of the
distribution of precipitation event volumes and intensities.  Working on an event basis is
important because many of the existing precipitation time series in the BASINS meteorological
data set are disaggregated from daily totals.  However, analyzing volume-event data for each of
the climate scenarios for all the precipitation stations was not feasible and the ability of the
climate models to correctly simulate event durations is suspect.

Using the change factor method,  future climate time series are constructed by applying changes
to observed precipitation time series that represent the ratio between historical simulations and
future climate simulations in a given climate model. No modifications were made to the number
of rainfall events in the observed record.  The following approach was developed to apply
changes in intensity in the baseline precipitation time series.

Total accumulated precipitation data for different percentile bins (for each station location by
month) were provided by NARCCAP for the dynamically downscaled climate change scenarios.
The data consisted of total  simulated precipitation volume (over 30 years) and the 0-25, 25-50,
50-70, and 70-90, and >90 percentile  bins of the 3-hour intensity distribution (relative to the
existing intensity distribution). These intensity percentiles yield information on where
precipitation intensification occurs, but represent fixed 3-hour windows, not discrete event
volumes, as required  for the CAT program. Most of the climate scenarios showed increases in
precipitation volume  in the larger events, while volume in the smaller events remained constant
or decreased. The net effect of this was an increase in the proportion of annual precipitation
occurring in larger events.  Analysis of the comprehensive (percentile, total volume) climate
scenario data showed that, for most weather stations, the change in the lower percentiles of the
intensity distribution  appeared to be relatively  small compared to the changes above the 70th
percentile.  However, in some cases (e.g., in Arizona), there is greater change in the 25-50*
percentile bin.

Analyses of observed changes in precipitation  during the 20* and early 21st century indicate that
more than half of the precipitation increase has occurred in the top 10 or 5% of events (Karl and
Knight, 1998; Alexander et al., 2006).  However,  GCMs have been shown to systematically
underestimate the frequency of heavy events in the top few percentages (Trenberth et al., 2003;
Sun et al., 2006; Dai, 2006). Therefore, the top 30% range is selected as a compromise that
accounts for intensification but remains within the general  skill of the climate models.

To account for changes in intensity, climate change scenarios were thus created using the delta
method by applying climate change adjustments separately to precipitation events >70*
percentile and events <70*  percentile, while maintaining the appropriate mass balance as
described below.

Percentile bin-intensity data were available only for climate scenarios 1 through 6
(RCM-downscaled scenarios).  Bin data were not available for climate scenarios 7 through 14
(GCM and statistically downscaled scenarios). Two approaches were developed to  account for
                                           5-7

-------
intensification of precipitation, depending on whether precipitation bin data were available.
Each approach is discussed in detail below.

Approach 1: Precipitation Bin Data Are Available
For scenarios where bin data was available (the six NARCCAP scenarios) the following
approach was used.  For these data, the change in the volume above the 70* percentile intensity
can be taken as an index of the change in the top 30% of events. At the same time, it is necessary
to maintain mass balance by honoring the predicted relative change in total volume.  This can be
accomplished mathematically as follows:

Let the ratio of total volume in a climate scenario (¥2) relative to the baseline scenario volume
(Vi) be given by  r = (V^Vi).  Further assume that the total event volume (V) can be decomposed
into the top 30% (Vn) and bottom 70% (Vi).  These may be related by a ratio s = VH/VL.  To
conserve the total volume we must have:
                                        V,=rV,                                      5-1
Equation 5-1 can be rewritten to account for intensification of the top 30% of events (Vn) by
introducing an intensification parameter, q:
                                                                                     5-2
Substituting for the first instance of VH,I=S VL,\ in eq 5-2 yields:
                                                                                     5-3
In eq 5-3 the first term represents the change in the volume of the lower 70% of events and the
second term the change in the top 30%.  This provides multiplicative factors that can be applied
to event ranges using the BASINS CAT program on a month-by-month basis.

The intensification parameter, q, can be calculated by defining it relative to the lower 70% of
values (i.e., from 0 to 70*  percentile). Specifically (r - rqs), which represents the events below
the 70th percentile, can be written as the ratio of the sum of the volumes below the 70th percentile
in a climate scenario relative to the sum of the volumes below the 70th percentile for the current
condition:
                                                  (C?7o)
                                                    7ol                              5.4

-------
where (Qio)i and (1370)2 are the sum of the volumes reported up to the 70* percentile for a month
for the current condition and future condition respectively.


Solving eq 5-4 for q yields:


                                       q = (\-Alr)ls                                     5_5


where A is defined as A =
                        (670)1

In sum, for each month at each station the following were calculated:

           V,
       r =
                            from the summary of the climate scenario output,

       s -  H/ftT _T/r ^       from the existing observed precipitation data for the station, sorted
            /\y ~*H)
                            into events and postprocessed to evaluate the top 30% (F#) and
                            bottom 70% (VL) event volumes. The numerator is calculated as
                            the difference between total volume and the top 30% volume,
                            rather than directly from VL to correct for analyses in which some
                            scattered precipitation is not included within defined "events."
                            The s value was calculated by month and percentile (for every
                            station, every month) using the observed precipitation time-series
                            data that forms the template for the delta method representation of
                            future climate time series.

       q = (1 - AIT) IS         where A is obtained from the percentile bin climate scenario output
                            summary
The multiplicative adjustment factors for use in the CAT tool can then be assembled as:
       r (1 - qs), for the events below the 70* percentile, and
       r (1 + q\ for the events above the 70th percentile.
In addition to the typical pattern of increasing rainfall occurring in large events, this approach is
applicable for the cases in which there is a relative increase in the low-percentile intensities.  In
those cases, the change in the 70th percentile intensity is relatively small and tends to be less than
current conditions under the future scenario, resulting in q being a small negative number.  In
such cases, application of the method results in a decrease in the fraction of the total volume
                                           5-9

-------
belonging to the larger events, with a shift to the smaller events—thus approximating observed
increases in intensity for smaller events.

In general, it is necessary to have -\ 1  (increasing precipitation), the future volume representing the
climate scenario (Vi  can  be defined as:
                                    V2=VlL+r • VIH                                  5_6

where r*is the change applied to the upper range (>30%), VH is the volume in the top 30%, and
VL is the volume in the bottom 70% of events.
Rearranging eq 5-6 and expressing r = r + -(r -1) •  IL/VIH , the overall change is satisfied, as:



          V =V   + r*  -V   = V  + r-V  -V  + r-V  =  r (V   +  V  ) =  r-  V
          V1   'IL  ^ '    V\H    'IL ^ ' V\H   'IL ^ '  V\L    ' \V\H ^  '\L/    '   V\^       5-7


Further, as r > 7, r is always positive.

For the case of r < 1 (decreasing precipitation), an across-the-board decrease in precipitation was
applied as follows:


                                    V2 = r • V\L + r • V\H                                 S-8
                                          5-10

-------
The adjustment factors can then be assembled as follows:

       For the events above the 70th percentile, if
              r > 1, then user*
              r < 1, then use r.

       For the events below the 70th percentile, if
              r > 1, then use 1 (no change)
              r < 1, then use r.
5.2.2.3. Potential Evapotranspiration Changes
Potential evapotranspiration is an important parameter that is sensitive to climate change and
urban development. In this study, PET is simulated with the Penman-Monteith energy balance
method.  In addition to temperature and precipitation, the Penman-Monteith method requires dew
point (or relative humidity), solar radiation, and wind as inputs. Because only a few stations
have time series for all four additional variables that are complete over the entire 1971-2000
period, these variables are derived from the SWAT 2005 statistical weather generator (Neitsch et
al., 2005). This is done internally by SWAT. For HSPF implementation a stand-alone version of
the weather generator code was created and used to create time series for each of the needed
variables at each BASINS meteorological station based on the nearest SWAT weather generator
station after applying an elevation correction.

The SWAT weather generator database (.wgn) contains the statistical data needed to generate
representative daily climate data for the different stations. Adjustments to the wgn file
parameters were made using monthly change statistics for the NARCCAP dynamically
downscaled scenarios. Specifically solar radiation, dew point temperature, and wind speed were
adjusted for each scenario (see Table 5-3).

The probability of a wet day following a dry day in the month and the probability of a wet day
following a wet day in the month were kept the same as in the original SWAT climate generator
file for the station.  Climate models showed a systematic bias, likely introduced by the scale
mismatch (between a 50-km grid and a station observation) for weather generator parameters
like wet day/dry  day timing, resulting in too many trace precipitation events relative to observed.
Thus it was not possible to use climate models to determine changes in these parameters. Also,
an analysis of the dynamically downscaled 3-hourly time series for the Canadian Regional
Climate Model (CRCM) downscaling of the CGCM3 GCM at five randomly selected locations
in the southeast,  southwest, mid-Atlantic, upper Midwest, and Pacific Northwest demonstrated
that the probability that a rainy day is followed by a rainy day (transition probability) in  the
model output did not change significantly at any of the sample locations.

For the BCSD climate scenarios, information on these additional meteorological variables is not
available. Many of these outputs are also unavailable from the archived nondownscaled GCM
output. For these scenarios it was assumed that the statistical parameters remained unchanged at
current conditions. While the lack of change is not physically realistic (e.g., changes in  rainfall
will be associated with changes in cloud cover and thus with changes in direct solar radiation
                                          5-11

-------
reaching the land surface), this reflects the way in which output from these models is typically
used.
       Table 5-3. SWAT weather generator parameters and adjustments applied
       for scenarios
SWAT wgn file
parameter
SOLARAV1
DEWPT1
WNDAV1
Description
Average daily solar radiation for month (MJ/m2/day)
Average daily dew point temperature in month (°C)
Average daily wind speed in month (m/s)
Adjustment applied
Adjusted based on Surface Downwelling Shortwave
Radiation change (%)
Additive Delta value provided for climate scenario for
each month
Adjusted based on 10-meter Wind Speed change (%)
Inconsistencies in the available data among different scenarios required special treatment. One
of theNARCCAP scenario archives (Scenario 5: CGCM3 downscaled with regional climate
model, version 3 [RCM3]) does not include solar radiation, which may be affected by changes in
cloud cover. Current condition statistics for solar radiation contained in the weather generator
were used for this scenario.  This does not appear to introduce a significant bias as the resulting
changes in PET fall within the range of those derived from the other NARCCAP scenarios.

Table 5-4  compares the reference crop estimates of Penman-Monteith PET for the five pilot
watersheds.  This is the PET used directly by the HSPF model, while the  SWAT model performs
an identical calculation internally, and then adjusts actual evapotranspiration (AET) for crop
height and leaf area development. Because PET is most strongly a function of temperature, a
fairly consistent increase in PET is simulated for most basins. It can be seen from the figures in
Appendix Z, however, that the statistically downscaled and nondownscaled GCM scenarios
(scenarios 7-14) that do not include solar radiation, dew point, and wind  time series consistent
with the simulated precipitation and temperature, generally provide higher estimates of PET than
do the dynamically downscaled models.  This issue is explored in more detail in Section 6.2.
5.3. URBAN AND RESIDENTIAL DEVELOPMENT SCENARIOS
Watershed simulations were also conducted to assess the sensitivity of study areas to potential
mid-21st century changes in urban and residential development.
                                         5-12

-------
Table 5-4.  Comparison of PET estimation between different downscaling approaches
Scenario type
Climate scenario
ACF
(GA, AL, FL)
Minnesota River
(MN, SD)
Salt/Verde/San
Pedro (AZ)
Susquehanna
(PA, NY, MD)
Willamette (OR)
annual average PET
(in)
difference from
NARCCAP mean
annual average PET
(in)
difference from
NARCCAP mean
annual average PET
(in)
difference from
NARCCAP mean
annual average PET
(in)
difference from
NARCCAP mean
annual average PET
(in)
difference from
NARCCAP mean
NARCCAP dynamically
downscaled
1. CRCM-CGCM3
60.32
1.46%
58.57
2.92%
83.67
0.47%
43.78
1.79%
44.18
-0.37%
5.RCM3-
CGCM3
58.59
-1.46%
55.24
-2.92%
82.89
-0.47%
42.24
-1.79%
44.51
0.37%
Nondownscaled
GCM
7. CGCM3
59.85
0.67%
56.22
-1.21%
84.19
1.09%
42.91
-0.23%
45.24
2.01%
BCSD
statistically
downscaled
11. CGCM3
64.75
8.90%
63.90
12.29%
85.01
2.07%
51.15
18.94%
50.73
14.41%
NARCCAP dynamically
downscaled
3. RCM3-GFDL
60.46
2.81%
54.92
-4.44%
81.32
-0.98%
43.06
0.43%
45.44
1.70%
4. GFDL
(high res)
57.16
-2.81%
60.02
4.44%
82.93
0.98%
42.69
-0.43%
43.91
-1.70%
Nondownscaled
GCM
9. GFDL
67.88
15.42%
64.99
13.08%
86.73
5.60%
50.18
17.05%
49.16
10.04%
BCSD
statistically
downscaled
13. GFDL
65.97
12.17%
63.65
10.75%
84.74
3.18%
50.17
17.02%
49.17
10.06%

-------
5.3.1. ICLUS Urban and Residential Development Scenarios
Projected changes in urban and residential development were acquired from EPA's ICLUS
project (U.S. EPA, 2009c). ICLUS has produced seamless, national-scale change scenarios for
developed land that are compatible with the assumptions about population growth, migration,
and economic development that underlie the IPCC greenhouse gas emissions storylines. ICLUS
projections were developed using a demographic model coupled with a spatial allocation model
that distributes the population as housing units across the landscape. Specifically, population is
allocated to 1-hectare (ha) pixels, by county, using the Spatially Explicit Regional Growth Model
(SERGoM).  The model is run for the conterminous United States and output is available for
each emissions storyline by decade to 2100. The final spatial data sets provide decadal
projections of housing density and impervious surface cover as a function of population for the
period 2000 through 2100 (U.S. EPA, 2009c).

Data from the ICLUS project are composed of grid-based housing density estimates with 100-m
cells, whose values are set equal to units/ha x 1,000.  Existing housing densities were estimated
using a variety of sources and models, and future housing densities developed under various
scenarios for each decade through 2100.  For the existing housing density grid, two types of
"undevelopable" area where residential development was precluded were masked out during the
production—a comprehensive spatial data set of protected lands (including land placed in
conservation easements), and land assumed to be commercial/industrial under current conditions.
Undevelopable commercial/industrial land use was masked out according to the SERGoM
method (U.S. EPA, 2009c) that eliminated commercial, industrial, and transportation areas that
preclude residential development, identified as "locations (1-ha cells) that had >25%
urban/built-up land cover with lower than suburban levels of housing density."

The ICLUS projections used in this study thus do not account for potential growth in
commercial/industrial land use. It is also important to note that the ICLUS projections do not
explicitly account for changes in rural or agricultural land uses. These categories change in the
analysis based on ICLUS only when they convert to developed land.

5.3.2. Mapping ICLUS Housing Density Projections to NLCD Land Use Categories
The ICLUS projections used in this study are for changes in housing density and impervious
cover.  This data cannot be used directly with the SWAT and HSPF watershed models, which
require land use data consistent with the NLCD.  It was therefore necessary to translate between
ICLUS projections and NLCD land-use classes.

In addition, ICLUS housing density class estimates and the NLCD developed classes do not have
a one-to-one spatial relationship because they are constructed on different underlying scales.
ICLUS represents housing density based largely on the scale of census block groups.  As a result,
it represents the overall density within a relatively large geographic area when compared to the
30x30 meter resolution of NLCD 2001 land cover and can represent a mix of different NLCD
classes. Therefore, land-use changes must be evaluated on a spatially aggregated basis at the
scale of model subbasins.

Baseline land use, derived from the 2001 NLCD, contains four developed land classifications
(NLCD classes 21 through 24), nominally representing "developed, open space" (less than 20%
                                         5-14

-------
impervious), developed, low intensity (20-49% impervious), developed, medium intensity
(50-79% impervious), and developed, high intensity (greater than 80% impervious).  Impervious
fractions within each developed NLCD land-use class were estimated separately for each study
area, using the 2001 NLCD Land Cover and Urban Impervious data products.  ICLUS land-use
change scenarios were implemented by modifying the existing land-use distribution in the
watershed models.

ICLUS estimates housing density on a continuous scale.  To process the data more efficiently,
the data were reclassified into 10 housing density ranges.  In each study area, the ICLUS housing
density ranges were cross-tabulated with NLCD 2001 classes based on percent imperviousness.
It was assumed that the number of housing units changes, but that the characteristic percent
impervious values for each NLCD developed class remains constant. The change in land area
needed to account for the change in impervious area was then back calculated.

To represent the net change in future land cover, the change in developed land use was added (or
subtracted) from the existing totals in each subbasin. Land area was then removed from each
undeveloped NLCD class (excluding water and wetlands) according to their relative ratios in
each subbasin to account for increases in developed area.  If the undeveloped land area was not
sufficient to accommodate the projected growth, development on wetlands was allowed. The
reductions in undeveloped land were distributed proportionately among modeled soils (in
SWAT) or hydrologic soil groups (in HSPF). The new developed lands were then assumed to
have the parameters of the most dominant soil and lowest HRU slope in the  subbasin.  For
HSPF, the changed area was implemented directly in the area table of the user control input
(.uci) file.  For SWAT, the land-use change was implemented by custom code that directly
modified the SWAT geodatabase that creates the model input files.

The gains (and losses) in NLCD class interpreted from ICLUS were tabulated  separately for each
subbasin.  In almost every case, the gains far exceeded the losses and a net increase was
projected in all four NLCD developed classes. However, in a few cases there was an overall loss
of the lowest density NLCD class. This tended to occur when a subbasin was already built out,
and ICLUS projected redevelopment at a higher density.

The projected overall changes in developed land for 2050 as interpreted to the NLCD land-cover
classes and used for modeling are presented in Table 5-5. Note that even in  areas of expected
high growth  (e.g., the area around Atlanta in the ACF basin), new development by 2050 is
expected to constitute only a small fraction of the total watershed area at the scale of the study
areas in this project.  The highest rate of land-use change in the studied watersheds is Coastal
Southern California, at 11.7%.  (Note that the ICLUS project does not cover the Cook Inlet
watershed in Alaska. Urban and residential development scenarios were thus not evaluated at
this study area.)
                                         5-15

-------
       Table 5-5. ICLUS projected changes in developed land area within different
       imperviousness classes by 2050
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
Change,
<20%
impervious
class (km2)
+665.2
+92.1
ND
+873.9
+353.5
+152.1
+307.2
+71.3
+8.9
+238.6
+1.3
+139.0
+103.6
+162.0
+329.4
+211.1
+492.4
+978.9
+56.9
+75.8
Change,
20-49%
impervious
class (km2)
+809.7
+87.0
ND
+776.1
+1,506.6
+204.8
+308.3
+142.9
+18.7
+327.2
+0.5
+228.8
+58.1
+1,001.0
+1,364.6
196.2
+306.6
+1,896.7
+168.1
+193.4
Change,
50-79%
impervious
class (km2)
+212.3
+16.0
ND
+361.5
+447.5
+51.0
+91.4
+60.9
+4.1
+215.5
+0.1
+57.1
+29.5
+1,089.1
+473.5
+69.6
+107.4
+891.1
+66.3
+95.0
Change,
>80%
impervious
class (km2)
+90.8
+1.3
ND
+102.2
+116.2
+15.6
+23.4
+18.5
+1.6
+59.2
0.0
+7.4
+8.2
+114.1
+83.6
+25.6
+29.2
+304.3
+8.3
+33.3
Total change in
developed land
(km2)
+1,778.0
+196.4
ND
+2,113.8
+2,424.0
+423.4
+730.1
+293.5
+33.2
+840.4
+1.9
+432.4
+199.3
+2,466.2
+2,251.1
+502.5
+935.6
+4,071.0
+299.6
+397.6
Increase as
percent of
study area (%)
+3.56
+0.51
ND
+4.65
+5.50
+1.40
+4.82
+0.67
+0.06
+3.13
+0.00
+0.88
+0.93
+11.72
+5.93
+0.71
+3.66
+8.76
+0.65
+1.37
Note: The ICLUS project does not cover the Cook Inlet watershed. Results shown are total new developed area,
including pervious and impervious fractions.
                                            5-16

-------
     6. STREAMFLOW AND WATER QUALITY SENSITIVITY TO DIFFERENT
       METHODOLOGICAL CHOICES: ANALYSIS IN THE FIVE PILOT STUDY
                                         AREAS
One goal of this study was to assess the implications of different methodological choices for
conducting climate change impacts assessments on the variability of simulation results.
Sensitivity studies in the five pilot study areas allow assessment of the variability resulting from
the use of different watershed models, and variability resulting from use of climate change
scenarios developed using different methods of downscaling GCM output.  The five pilot study
areas are the Minnesota River, ACF, Susquehanna, Willamette, and Salt/Verde/San Pedro
Rivers.  In each of these sites, independent simulations were conducted using the SWAT and
HSPF watershed models, and in addition to the six dynamically downscaled NARCCAP
scenarios, an additional set of climate change scenarios was evaluated, four based on the BCSD
statistically downscaled data set, and four based directly on GCMs with no downscaling.  This
section presents a summary of these results.

6.1. COMPARISON OF WATERSHED MODELS
The magnitude of the additional variability introduced by choice of a hydrologic model is of
interest when simulating hydrologic responses to climate change and urban development. Two
different watershed models,  SWAT and HSPF, were calibrated and applied to the five pilot study
areas. Evaluation of different watershed models can be considered an extension of the
scenario-based,  ensemble approach commonly used in  climate change studies. Detailed
examination of the calibration of each model in the five pilot study areas and the results of
change scenarios conducted with each model are presented in separate sections and the
appendices to this report.

HSPF and SWAT take different approaches to watershed simulation and have different structures
and algorithms,  resulting in different strengths and weaknesses. Most notably, the two models
differ in the way that they represent infiltration and plant-climate interactions. SWAT (in
standard application mode) simulates rainfall-runoff processes using a curve number approach,
operating at a daily time step.  The curve number approach first partitions incoming moisture
into direct runoff and a remainder that is available  for infiltration.  In contrast, HSPF simulates
rainfall-runoff processes using Green-Ampt infiltration, in which infiltration into the soil is
simulated first, with the remainder available for direct runoff or surface storage.

HSPF is typically run at a subdaily time  step, usually hourly for large watersheds, and has a more
sophisticated representation of runoff, infiltration,  and channel transport processes than does
SWAT.  SWAT's advantage is that it incorporates  a plant growth model (including
representation of changes in atmospheric CC>2 concentration) and can therefore simulate some of
the important feedbacks between plant growth and hydrologic response. Both models simulate
evapotranspiration of soil water stores, but HSPF does this using empirical monthly coefficients
relative to potential evapotranspiration, while SWAT incorporates a plant growth model that can,
in theory, dynamically represent plant transpiration of soil moisture.
                                          6-1

-------
6.1.1. Comparison of Model Calibration and Validation Performance
Models were calibrated and validated using multiple measures as summarized previously in this
report and described in detail in Appendices D-W. Calibration of both models was conducted in
accordance with the modeling QAPP (see Appendix B; Tetra Tech, 2008a) for each of the five
pilot study areas. Development and setup of the two watershed models proceeded from a
common basis, with both models using the same subbasin delineations, land use coverage (2001
NLCD), soils coverage (STATSGO), hydrography, digital elevation model, impervious area
fractions for developed land classes, and point source and dam representations. Other aspects of
model setup were designed to be similar, although it was not possible to be identical because of
differences in the way the two models conceptualize discretization of the land surface.  For
instance, hydrologic response units (the fundamental building blocks of the upland simulation)
were created as an overlay of land use and HSG for HSPF, while SWAT uses an overlay of land
use and STATSGO dominant soil, associating various other properties from the soil database in
addition to HSG with the model hydrologic response units. In addition, HSPF simulates
impervious surfaces as a separate land use, while SWAT assigns an impervious fraction to an
underlying land use.

Calibration/validation locations and observed data series were the same for both models.
Further, the calibration of both models was guided by prespecified statistical analyses that were
performed using identical spreadsheet setups obtained from a common template.  Despite these
commonalities, the scope of the modeling effort in this study required that models be developed
by different modeling teams, with inevitable differences in results. To reduce the likelihood of
bias, model calibration assignments were structured so that the same team did not apply both
HSPF and SWAT to a single study area, and each watershed model was implemented by at least
three different modeling teams for the pilot studies.

6.1.1.1. Streamflow Results
This section examines hydrologic simulations as compared to observed streamflow records based
on total volume error and the daily Nash-Sutcliffe coefficient of model  fit efficiency. Model
performance is first examined in terms of the quality of fit for the initial calibration watershed,
followed by similar analyses for the largest-scale downstream watershed.  Intercomparisons then
provide some insight into model performance relative to temporal change (calibration vs.
validation period) and relative to spatial change within each study area  (calibration watershed vs.
downstream watershed).

Summary results for percent error in total volume and  the Nash-Sutcliffe E coefficient for daily
streamflow are shown in Tables 6-1 and 6-2, respectively, for the initial calibration site along
with the calibration fit for the most downstream gage in the watershed.  In general, the quality of
model fit is good for both models. In most, but not all cases, the quality of model fit is slightly
better (smaller magnitude of percent error, larger E coefficient) for the HSPF simulations (e.g.,
see Figure 6-1 for the calibration period). This is likely due in large part to the use of daily
precipitation in SWAT versus hourly precipitation in HSPF, although the advantage accruing to
HSPF is muted by the fact that many of the "hourly" precipitation input series used are actually
disaggregated from daily totals. Monthly values of Nash-Sutcliffe E are higher for both models,
but attention is called to the daily scale because it better reflects the models' ability to separate
surface and subsurface flow pathways.  Note that E is low for the Arizona initial site on the
                                           6-2

-------
Verde River because streamflow is dominated by relatively constant deep groundwater
discharges.
       Table 6-1. Percent error in simulated total streamflow volume for 10-year
       calibration and validation periods at initial and downstream calibration
       gages
Study area
Apalachicola-
Chattahoochee-Flint (ACF)
Salt/Verde/San Pedro (Ariz)
Minnesota River (Minn)
Susquehanna (Susq)
Willamette (Willa)
Model
HSPF
SWAT
HSPF
SWAT
HSPF
SWAT
HSPF
SWAT
HSPF
SWAT
Initial site calibration
5.50
7.28
2.43
-2.46
1.61
-5.41
-0.16
-5.41
-3.92
-4.76
Initial site validation
5.79
3.33
6.31
5.68
14.78
-0.84
-8.00
-16.30
-9.80
12.10
Downstream calibration
16.79
16.53
4.48
9.43
-4.25
7.89
1.79
-9.74
2.58
-4.96
       Table 6-2. Nash-Sutcliffe coefficient of model fit efficiency (E) for daily
       streamflow predictions, 10-year calibration and validation periods at initial
       and downstream calibration gages
Study area
Apalachicola-Chattahoochee-
Flint (ACF)
Salt/Verde/San Pedro (Ariz)
Minnesota River (Minn)
Susquehanna (Susq)
Willamette (Willa)
Model
HSPF
SWAT
HSPF
SWAT
HSPF
SWAT
HSPF
SWAT
HSPF
SWAT
Initial site calibration
0.71
0.62
0.48
0.03
0.75
0.79
0.70
0.29
0.80
0.49
Initial site validation
0.65
0.56
0.45
-1.00
0.78
0.74
0.55
0.42
0.81
0.39
Downstream calibration
0.72
0.64
0.53
0.22
0.92
0.63
0.77
0.45
0.88
0.67
                                         6-3

-------
                                        • HSPF

                                        • SWAT
              Ariz
                           Susq
                                 Willa
• HSPF

• SWAT
                                                      ACF   CeAZ   MNRiver  Susq   Willm
       Figure 6-1. Comparison of model calibration fit to streamflow for the
       calibration initial site.

Note: Figures compare calibration results for HSPF and SWAT. Total volume error is converted to its absolute value.
The ability of the model to assess relative changes in response to altered climate forcing is of
paramount importance in this project.  Some insight on this topic can be gained by looking at the
sensitivity of model fit to temporal and spatial changes in application. Figure 6-2 summarizes
the sensitivity to temporal changes by examining the percent error in the calibration period and
the validation test. It is interesting to observe that for both the ACF and the Minnesota River, the
SWAT model achieved an improvement in total volume error during the validation period.
These are the two study areas with the greatest amount of row crop agriculture, and the results
may reflect SWAT's ability to reflect changing responses of crops to changes in climate over the
last 20 years.

Figure 6-3 examines model sensitivity to spatial scale, comparing performance during the
calibration period for the initial calibration target gage (HUC-8 spatial scale) and the most
downstream gage in the model (approximately HUC-4 spatial scale). The left panel shows the
change in the absolute magnitude of percent error, while the right panel shows the change in E.
A smaller magnitude of change in total volume error or a larger increase in E represents better
performance. The changes in total volume errors are generally small, regardless of whether
detailed spatial  calibration was pursued. In most cases, the models achieved an improvement in
E in going from the smaller to the larger scale.

6.1.1.2. Water Quality Results
The water quality calibration compared simulated monthly loads to monthly load estimates
obtained from a stratified regression on (typically sparse) observed data.  To compare these
results between models, the baseline adjusted E\' coefficient of model fit efficiency is most
appropriate. Results are summarized graphically for the calibration period at the calibration
initial site and downstream site in Figures 6-4 through 6-6. For suspended solids and total
phosphorus, the performances of the two models are similar, while HSPF appears to provide a
somewhat better fit for total nitrogen.
                                           6-4

-------
                                                                     HSPF-Cal
                                                                     HSPF-Val
                                                                     SWAT-Cal
                                                                     SWAT-Val
      -20
               ACF       Ariz      Minn      Susq
                                Study Area
                                           Willa
       Figure 6-2.  Sensitivity of model fit for total streamflow volume to temporal
       change.
    u
   I1"
   I8
   r
   = 3
   I.
J.J
         ACf    Aril    Minn
• HSPT
• SWAT
O.K
 0.2
O.lb
 0.1
0.0-j
  0
•o.os
-0.1
-0.15
 0.2
JUJ
             Acr

                              Susq
                                   Willj
• HSPf
• SWAT
       Figure 6-3.  Sensitivity of model fit for streamflow to spatial change.
Note: Change in percent total volume error represents the difference in the absolute value of percent error in going from the
initial calibration site to a larger scale, typically the furthest downstream site. Change in E represents the difference in the Nash-
Sutcliffe E coefficient in going from the calibration site to the larger-scale site.
                                         6-5

-------
                                 • HSPF

                                 • SWAT
  ACF     Ariz    Minn    Susq    Willa
• HSPF

• SWAT
                                               ACF    Ariz    Minn    Susq    Willa
Figure 6-4. Comparison of baseline adjusted model fit efficiency for total
suspended solids monthly loads for calibration site (left) and downstream site
(right).
                                 • HSPF

                                 • SWAT
              Minn    Susq    Willa
                                                     Ariz     Minn    Susq    Willa
• HSPF

• SWAT
Figure 6-5. Comparison of baseline adjusted model fit efficiency for total
phosphorus monthly loads for calibration site (left) and downstream site
(right).
                                 • HSPF

                                 • SWAT
              Minn    Susq    Willa
                                                          Minn    Susq    Willa
• HSPF

• SWAT
Figure 6-6. Comparison of baseline adjusted model fit efficiency for total
nitrogen monthly loads for calibration site (left) and downstream site (right).
                                     6-6

-------
6.1.1.3. Summary of Relative Model Performance
In general, the HSPF model provides a somewhat better fit to observed streamflow and water
quality data for the calibration periods. The effect is most noticeable in the coefficient of model
fit efficiency (£) for daily streamflow, where the HSPF approach of applying Philip infiltration
using hourly precipitation appears to yield an advantage over the SWAT daily curve number
method. However, relative performance of the two models is more similar as the analysis moves
to the validation period or to other sites for which detailed calibration has not been undertaken.
Most importantly, both models appear to be capable of performing adequately.

6.1.2.  Comparison of Simulated Changes Using SWAT and HSPF
Figure  6-7 compares HSPF and SWAT simulated changes in mean annual streamflow at the
downstream station of each of the five pilot watersheds for all 28 combinations of climate and
land-use change scenarios (expressed as a percent of the baseline conditions, representing
approximately 1970-2000). In general, the mean annual streamflow results provided by the two
models are similar, as is shown quantitatively below. One notable difference is for the
Minnesota River where SWAT projects higher flows relative to HSPF under future climate
conditions—an issue that is explored further in Section 6.1.3. Note that points plotting close to
or on top of each other for a given study site in Figure 6-7 are scenarios representing the same
climate change scenario with and without changes in urban development.
                200%
                                                                     + ACF

                                                                     • Ariz

                                                                     AMinn

                                                                     XSusq

                                                                     XWilla
                               50%        100%       150%
                          HSPF Total Flow Volume - Percent of Baseline
200%
       Figure 6-7. SWAT and HSPF simulated changes in total streamflow in pilot
       watersheds (expressed relative to current conditions).
                                          6-7

-------
Table 6-3 provides a statistical comparison of the HSPF and SWAT results at the downstream
station.  Three types of tests are summarized. The first is a t-test on the series of paired means
(HSPF and SWAT for each climate and land use scenario), which has a null hypothesis that the
mean of the differences between the series is not significantly different from zero. The second
test is a two-way analysis of variance (ANOVA) that looks at choice of watershed model (HSPF
or SWAT) as blocks and climate scenario as treatment. The null hypotheses for this test are that
the difference between series for a given source of variance is zero. The third test is a linear
regression on SWAT results as a function of HSPF results. Where the models are in full
agreement, the intercept of such a regression should not be significantly different from zero and
the slope should not be significantly different from unity.

For mean annual streamflow, both models produce similar results with a high Pearson correlation
coefficient. The null hypothesis from the Mest that the mean difference is zero cannot be
rejected. However, the two-way ANOVA shows that both the choice of watershed model and
the climate scenario are significant sources of variability in streamflow, with probability values
(p-value) well less than 0.1. Together these results suggest that the SWAT and HSPF results  are
similar in the aggregate, but may contain an underlying systematic shift.  A regression analysis
shows that the slope coefficient for SWAT and HSPF is 0.93, with a 95% confidence interval
that does not overlap 1.0,  and an intercept of 1,262  that also does not overlap zero. Thus, SWAT
projects a somewhat smaller response to increased rainfall, but results in higher baseflow
estimates (likely due to the effects of increased CC>2 on evapotranspiration, as explained further
below).
       Table 6-3.  Statistical comparison of HSPF and SWAT outputs at
       downstream station for the five pilot sites across all climate scenarios
Measure
Mean annual flow (cfs)
TSS load (t/yr)
TP load (t/yr)
TN load (t/yr)
Paired f-test on sample means
HSPF Mean
SWAT Mean
Pearson Correlation
t- statistic
p (two-tail)
20,546
20.435
0.989
0.616
0.539
2,398,714
2,865,178
0.733
-3.123
0.002
2,748
3,344
0.644
-4.783
O.001
35,346
43,275
0.948
-7.385
O.001
Two-way ANOVA on watershed model and climate scenario
p value — Model
p value — Climate
0.001
O.001
0.071
0.960
0.006
0.999
0.044
1.000
Linear regression; SWAT result as a function of HSPF result
Intercept
Intercept, 95%
confidence
Coefficient (slope)
Coefficient (slope)
95% confidence
1,261.7
695-1,828
0.933
0.911-0.956
141,717
-363,064-646,498
1.136
0.964-1.307
954.0
431-1,477
0.870
0.702-1.038
-1,173.1
-4,194-1,848
1.257
1.189-1.326
                                           6-8

-------
The comparison for total suspended solids is obscured by the extremely large projected increases
under certain scenarios for the Arizona basins (Verde River, in this case). Those increases are
mostly due to simulated channel erosion, for which both models are likely to be highly uncertain
because future simulated peak flows are outside the range of calibration data.  Figure 6-8 shows
the simulated total suspended solids results but with the x-axis truncated to exclude these
extreme results for the Verde River. Results for the other four pilot sites appear generally
consistent between models, although simulated increases from SWAT are generally less than
those from HSPF for the ACF, Susquehanna, and Willamette. In part this is due to differences in
the baseline simulation. For example, HSPF simulations show less channel transport and much
smaller total suspended solids loads at the mouth  of the Susquehanna than does SWAT for the
baseline scenario, resulting in a larger relative change with increased future streamflow. The
difference between results for SWAT and HSPF may also reflect the effects of increased
atmospheric CO2 concentration and longer growing periods simulated  by SWAT, leading to
more litter cover and reduced soil erosion.
                400%
                                                                       • Ariz

                                                                       AMinn

                                                                       XSusq

                                                                       XWilla
                 0%
                    0%       100%       200%       300%
                             HSPF TSS Load - Percent of Baseline
400%
       Figure 6-8.  SWAT and HSPF simulated changes in TSS at downstream
       station in pilot watersheds (expressed relative to current conditions).

Note: HSPF simulation for climate scenarios 9 (GFDL, nondownscaled GCM), 10 (CCSM, nondownscaled GCM), 12 (HadCMS,
BCSD), and 13 (CCSM, BCSD) yield increases in simulated total suspended solids load of greater than 400% and are omitted
from this plot.
For total suspended solids, the baseline load is higher in SWAT than in HSPF for three of the
five watersheds; thus the statistical comparison (see Table 6-3) shows a higher mean load from
                                           6-9

-------
SWAT, even though the percentage increases are often smaller. The Mest on means shows that
this difference is highly significant. However, the ANOVA show that neither the model choice
nor the climate scenario is a significant explanatory variable for the variance at the 95%
confidence level.  The regression analysis shows that the intercept is large, but not significantly
different from zero, while the slope is not significantly different from 1. Together these statistics
indicate that the total suspended solids simulation is subject to considerable uncertainty and that
differences between sites are more important than other factors.

Results for total phosphorus are generally similar to those seen for total suspended solids, with
much more extreme increases projected by both models for the Verde River (Ariz; see Figure
6-9).  HSPF simulations are especially high due to an assumption of phosphorus concentrations
in scoured channel sediment.  SWAT tends to simulate higher rates of increases for total nitrogen
(see Figure 6-10) than does HSPF (likely due to more rapid cycling of organic matter), with the
notable exception of the ACF study area. However, it appears that projections of total nitrogen
at the downstream end  of the ACF may be significantly  underestimated in the calibrated SWAT
model. Total nitrogen varies little in the Susquehanna model due to small changes in streamflow
and significant point source contributions.

For both total nitrogen  and total phosphorus the choice of model is a significant factor in the
ANOVA and higher mean loads are produced by SWAT.  The slope of a regression of SWAT on
HSPF is not significantly different from  1 for total phosphorus, consistent with the solids
simulation, but the intercept is significantly different from zero, indicating differences in the
baseflow simulation of total nitrogen. For total nitrogen, the intercept is not significantly
different from zero, but the slope is significantly greater than 1, suggesting that SWAT projects a
greater increase in total nitrogen loads under future climate conditions.

In sum, the comparison of relative response to change scenarios indicates that the two models
provide generally consistent results for hydrology, with  differences that may be in part due to the
inclusion of explicit representation of several processes  in SWAT (increased atmospheric CC>2,
changes in planting time, changes in crop growth and litter production, and changes in nutrient
recycling rates) that are not automatically included in HSPF.  Water quality results exhibit
greater variability between the models, due in large part to the uncertainty inherent in model
calibration.

An additional contributing cause to differences in results from the two models is the extent to
which spatial calibration of the model was pursued, which was left to modeler judgment.  In all
study areas, initial calibration and validation was pursued at an "initial calibration" gage and
monitoring station at an HUC-8 spatial scale. The calibration results were then carried to the
larger study area.  At this point, individual modeler preferences introduced some variability into
results.  Some modelers undertook detailed spatial adjustments to parameters; others extended
the initial parameter set with only minor modifications.  With more spatial adjustments a higher
degree of fit is generally to be expected for model calibration—although this does not necessarily
result in better performance in model validation.  In general, only limited spatial calibration
adjustments beyond the initial parameter set was carried out for the Minnesota River,
Susquehanna, and Willamette SWAT models and also for the Susquehanna HSPF model.
                                           6-10

-------
                200%
                                                                      *ACF

                                                                      • Ariz

                                                                      AMinn

                                                                      XSusq

                                                                      XWilla
                 0%
                    0%        50%       100%       150%       200%
                        HSPF Total Phosphorus Load - Percent of Baseline
       Figure 6-9.  SWAT and HSPF simulated changes in total phosphorus load in
       pilot watersheds (expressed relative to current conditions).

Note: 22 HSPF simulations for Ariz ranging from 200 to 875% are omitted.
Due to the potential influence of modeler choice and skill, it is cautioned that the results should
not be interpreted as a true head-to-head comparison of the two models, as the results for any
given watershed may be skewed by exogenous factors such as modeler calibration strategy.
Instead, it is most relevant to examine relative performance and potential inconsistencies
between simulations using the two models.
6.1.3.  Sensitivity to Increased Atmospheric

A key difference between HSPF and SWAT is that SWAT has a dynamic plant growth module
with ability to represent changes in atmospheric CC>2 on plant growth and water loss to ET. We
performed paired sets of SWAT simulations with and without increased CC>2 for all five pilot
sites to assess the sensitivity of streamflow and water quality endpoints to the effects of
increased atmospheric CO2 concentrations.
                                          6-11

-------
                200%
              cu
              c
              ro
              CO
              M—
              2 150%
              c
              cu
              £
              cu
              a.

              g 100%
              c
              cu
              -  50%
              to
                  0%
• Ariz

AMinn

XSusq

XWilla
                     0%        50%       100%       150%      200%
                         HSPF Total Nitrogen Load - Percent of Baseline
       Figure 6-10.  SWAT and HSPF simulated changes in total nitrogen load in
       pilot watersheds (expressed relative to current conditions).
IPCC estimates of future atmospheric CO2 concentrations under the assumptions of the A2
emissions scenario (the basis of climate and land-use change scenarios in this study) call for an
increase from 369 ppmv CC>2 in 2000 to about 532 ppmv (using the ISAM model reference run)
or 522 ppmv (using the Bern-CC model reference run) in 2050 (Appendix II in IPCC, 2001).
Plants require CC>2 from the atmosphere for photosynthesis. An important effect of increased
atmospheric CC>2 is a reduction in the time plant leaf stomata must be open to obtain the CC>2
needed for growth, resulting in reduced water loss as transpiration (Leakey et al., 2009; Cao et
al., 2010; Ainsworth and Rogers, 2007).  This effect can potentially counterbalance projected
increases in transpiration associated with increased air temperatures. It may also reduce water
stress on plants, resulting in greater biomass and litter production, which in turn will influence
pollutant loads.

In the past it has been argued that these effects, long documented at the leaf and organism level,
might not translate to true ecosystem effects. However, recent research, particularly results from
the Free-Air CC>2 Enrichment (FACE) experiments (Leakey et al., 2009) suggests that significant
reductions in evapotranspiration do occur at the ecosystem level with increased atmospheric CC>2
concentrations. Although there are differences in responses among plant species, with lesser
effects with C^ photosynthesis, the magnitude of the response to CC>2 levels projected by the
mid-21st century appears to be on the order of a 10% reduction in evapotranspiration response
(e.g., Bernacchi et al., 2007). Further, a recent study by Cao et al. (2010) suggests that up to
                                          6-12

-------
25% of the temperature increase projected for North America could result directly from
decreased plant evapotranspiration under increased CC>2 concentrations.

SWAT includes a plant growth module that accounts for the effects of changes in atmospheric
CC>2 concentration on stomatal conductance using the equation developed by Easterling et al.
(1992).  Using this approach, increased CC>2 leads to decreased leaf conductance, which in turn
results in an increase in the canopy resistance term in the PET calculation. The model also
simulates the change in radiation use efficiency of plants as a function of CO2 concentration
using the method developed by Stockle et al. (1992). Figure  6-11 shows the differences between
projected mid-21st century streamflow and water quality endpoints in the five pilot sites
simulated using SWAT with and without representation of the effects  of increased atmospheric
CC>2 concentrations (SWAT projections for the six NARCCAP climate scenarios incorporating
the ICLUS future  land use for each watershed). These simulations suggest increases in mean
annual streamflow from 3 to 38% due to increased CO2, with a median of 11%, in the same
range as the results summarized by Leakey et al. (2009). Simulations also suggest increased
atmospheric CC>2 results in increased pollutant loads. Total suspended solids loads show
increases from 3 to 57%, with a median of 15%. Total phosphorus loads increase from 0 to 29%,
with a median  of 6%.  Total nitrogen loads increase from zero to 34%, with a median of 6%.
The large increases in total suspended solids loads indicate that the effects of higher runoff under
increased atmospheric CO2 (largely due to greater soil moisture prior to rainfall events) may
outweigh benefits associated with greater ground cover—a finding that could have important
land management implications in the midwestern watersheds, including many of the Great Lakes
drainages. For the nutrients, the simulated load increases are less than for streamflow and total
suspended solids increases.  This presumably is due to the fact that increased atmospheric CO2
concentrations allow greater plant growth per unit of water, resulting in greater uptake and
sequestration of nutrients, and thus smaller increases in nutrient loads  relative to  streamflow and
total suspended solids.

The response to increased atmospheric CO2 concentration varies greatly by study area, with the
greatest effect  simulated by SWAT for the Minnesota River basin and the smallest effect for the
Willamette basin.  The large effect in the Minnesota River basin apparently occurs because the
land in this basin is predominantly in high-biomass corn-soybean rotation agricultural cropland
with precipitation and evapotranspiration in approximate balance. In contrast, the Willamette
basin is dominated by evergreen forest and has a moisture surplus for  much of the year.

Ficklin et al. (2009), working with the SWAT model in the San Joaquin watershed in California,
also  showed that increased atmospheric CC>2 could cause a significant relative decrease in
simulated evapotranspiration and a corresponding increase in water yield relative to simulations
that did not account for increased CC>2. However, Luo et al. (2013) recently suggested that the
approach used in SWAT to estimate the effects of CO2 on evapotranspiration is appropriate only
for arable land and may overestimate C (^-associated reductions from forest, pasture, and range
land. This remains an important topic for further investigation.
                                          6-13

-------
                 ACF        Central     Minnesota  Susquehanna  Willamette
                            Arizona        River
       Figure 6-11.  Differences between SWAT projections of mid-21st century
       streamflow and water quality (median across six NARCCAP scenarios) with
       and without representation of increased atmospheric
Note: Figure shows model simulation with increased CO2 minus projection with CO2 assumed constant at current levels.
Several important feedback loops other than the CC>2 effect on stomatal conductance are also
included in the SWAT plant growth model. First, planting, tillage, fertilization, and harvest
timing for crops (and start and end of growth for native plants) is represented by heat unit
scheduling relative to existing climate normals, allowing automatic adjustment in timing under a
changed temperature regime. Evapotranspiration is also simulated with the full
Penman-Monteith method, allowing dynamic simulation of leaf area development and crop
height, both of which impact ET. Finally, organic matter residue accumulation and degradation
on the land surface are dynamically simulated as a function of plant growth, and the effects of
altered cover on land surface erosion are represented.

All these factors are of potential importance in examining response to climate change.  In
contrast to SWAT, HSPF does not automatically compute these adjustments.  Instead, the user
would need to estimate changes in monthly parameters such as the lower zone evapotranspiration
coefficient (LZETP) and erosion cover externally and bring them into the model. While not well
understood, use of calibrated parameters in HSPF without these modifications could introduce
error to simulations under climatic conditions different from those  during the calibration period.
                                          6-14

-------
6.2.  SENSITIVITY TO DIFFERENT METHODS OF DOWNSCALING GCM OUTPUT
A variety of methods for downscaling large-scale GCM output to local scale projections are
available.  Both the selection of an underlying GCM and the choice of downscaling method have
a significant influence on the streamflow and water quality simulations. Indeed, in some basins
(e.g., Minnesota River, ACF) the difference among watershed model simulations as driven by
the six NARCCAP dynamically downscaled scenarios appears to be noticeably greater than the
range of model simulations driven by BCSD statistically downscaled or nondownscaled GCM
scenarios.  The results of the larger ensemble leads to the observation that incorporating
additional  information, either from dynamic RCMs or via statistical methods, can increase the
range of variability of simulated changes.

6.2.1. Climate Model Energy Inputs and PET Estimates
PET is calculated using the Penman-Monteith PET energy balance approach. The BCSD and
nondownscaled GCM scenarios do not provide all the required meteorological time series (see
Table 5-2 in Section 5.2.1.). As a result, PET for these scenarios was estimated using current
climate statistics for solar radiation, dew point, and wind time series. Comparisons presented in
Appendix Z suggest that PET estimates for the and GCM scenarios (scenarios 7-14) that do not
include solar radiation, dew point, and wind time series that are consistent with the simulated
precipitation and temperature are noticeably higher than estimates of PET derived from the
dynamically downscaled models that do provide these time series.

A comparison of the effects of data availability on PET calculations can be done through
comparison of scenarios that are based on the identical underlying GCM runs for CGCM3 and
GFDL that were each dynamically downscaled with two different RCMs  (as discussed in
Section 5.2.2.). Annual average PET estimates from these pairs are generally close to one
another, but may differ by up to 4.5% from their mean (see Table 5-4). For the CGCM3 model,
PET generated from the nondownscaled GCM is similar to that from the dynamically
downscaled scenarios, but PET calculated from the statistically downscaled scenario is from 2 to
19% higher. This  appears to be due to the fact that dew point temperature, which has an
important impact on PET,  is provided with the CGCM3  GCM but is not available from the
BCSD scenarios (see Table 5-2 above).  The difference is smallest for the Salt/Verde/San Pedro
River basins in Arizona, where dew point temperature is very low and not expected to change
much under future climates. In contrast, the  GFDL model does not provide dew point
temperature from the nondownscaled GCM.  For that model, both the nondownscaled and
statistically downscaled climate change scenarios produce higher PET estimates than the
NARCCAP dynamically downscaled scenarios. As with CGCM3, the smallest effect is seen in
the Salt/Verde/San Pedro River basins in Arizona, and the largest effect in the Susquehanna
basin, where a greater change in dew point temperature and relative humidity is projected. The
observed sensitivity  of PET estimates to climate variables other than air temperature and
precipitation suggests that simulation of future climates that does not account for changes in the
full suite of variables that influence PET could thus introduce significant biases into the
simulated water balance. Further investigation of this phenomenon was pursued through use of
"degraded" NARCCAP climate scenarios, as described below.
                                         6-15

-------
6.2.2. "Degraded" NARCCAP Climate Scenarios
To provide a consistent basis for comparison, all scenarios were created with a common
minimum set of variables.  Specifically, NARCCAP provided data on changes in precipitation
intensity (bin data), solar radiation, wind, and humidity that were not available in the GCM and
BCSD based scenarios. The following steps were taken to develop a consistent set of climate
scenario input series that differ only in the underlying climate model and downscaling technique:
   •   Representation of intensification in each of the NARCCAP dynamically downscaled
       scenarios was based on Approach 2 in Section 5.2.2., which assumes that all increases in
       precipitation occur in the top 30% of events, rather than using the direct analysis of
       intensity changes provided by NARCCAP.

   •   Complete information on changes in weather generator statistics for dew point
       temperature, solar radiation, and wind speed was removed for the NARCCAP
       dynamically downscaled scenarios, consistent with the information available for the
       BCSD scenarios. Incomplete information on these variables provided by the
       nondownscaled GCMs was also removed. (For the nondownscaled GCMs this affects
       weather scenarios 7, 9, and 10—see Table 5-2 above).

   •   Penman-Monteith PET was recalculated with the revised set of climate variables.

   •   Simulations use current land use to remove land-use change effects.
Note that these simplified or "degraded" NARCCAP scenarios are used only for the comparisons
presented in this section. Results presented in subsequent sections of this report use the
scenarios that contain all available meteorological information.

Comparison of the PET series generated with full climatological data to the degraded series in
which only precipitation and temperature are updated illustrates the effect of including these
additional variables (see Table 6-4). Further, the effect of individual meteorological time series
is discernible because the original set lacked solar radiation for Scenario 5, dewpoint temperature
for Scenario 9, and wind speed for Scenario 10 (see Table 5-2). Dewpoint temperature (which
tends to increase in future, warmer climates) has the biggest impact.  Including a climate
model-simulated dewpoint that is consistent with the scenario temperature and precipitation
regime results in a reduction in estimated annual PET of about 11% across all the meteorological
stations used for the five pilot watersheds.  The effect appears to be greater at higher latitudes.
The reduction in PET from including simulated dewpoint is around 10-20% for the Minnesota,
New York, Oregon, and Pennsylvania stations, but only 3-10% for the Alabama, Arizona,
Florida, and Georgia stations. In contrast, for  Scenario 9  (for which dewpoint temperature was
not available), the original PET series were on average 1.9% higher than the degraded series.
Omission of solar radiation or wind speed results from the climate scenario appears to have at
most a minor impact on the estimated PET.

In retrospect, these results suggest that a better approach to simulation of PET in cases where the
climate models do not provide dewpoint would be to assume that relative humidity remains
                                          6-16

-------
constant and recalculate a new dewpoint based on the relative humidity and climate-modified air
temperature, thus providing a more physically realistic estimate of vapor pressure deficit.
       Table 6-4. Effects of omitting simulated auxiliary meteorological time series
       on Penman-Monteith reference crop PET estimates for "degraded" climate
       scenarios
State
AL
AZ
FL
GA
MN
NY
OR
PA
All (%)
All
(in/yr)
Climate Scenario (GCM/RCM)
1
CGCM3/
CRCM
-4.87%
-2.38%
-7.14%
-9.30%
-14.68%
-23.27%
-15.82%
-17.62%
-12.53%
-6.36
2
HadCM3/
HRM3
-4.44%
-3.01%
-8.48%
-7.21%
-10.30%
-16.99%
-14.28%
-12.54%
-9.93%
-5.27
3
GFDL/
RCM3
-5.21%
-4.12%
-7.45%
-7.79%
-13.73%
-17.68%
-7.75%
-14.77%
-9.97%
-5.16
4
GFDL/
GFDL hi
res
-10.90%
-3.59%
-16.69%
-18.01%
-10.30%
-20.62%
-12.90%
-18.93%
-12.62%
-6.48
5
CGCM3/
RCM3
-5.76%
-2.97%
-9.04%
-10.15%
-16.46%
-22.95%
-13.67%
-18.59%
-12.86%
-6.42
6
CCSM/
WRFP
-4.47%
-3.08%
-9.02%
-7.27%
-21.16%
-18.30%
-13.29%
-13.40%
-12.48%
-6.31
7
CGCM3
(not down-
scaled)
-4.89%
-0.99%
-7.35%
-8.71%
-13.83%
-23.01%
-12.73%
-17.96%
-11.37%
-5.63
9
GFDL
(not down-
scaled)
2.66%
2.69%
2.92%
1.79%
1.68%
-1.29%
0.11%
0.28%
1.19%
0.90
10
CCSM (not
down-
scaled)
-7.11%
-3.02%
-10.91%
-14.04%
-16.46%
-20.48%
-10.17%
-17.28%
-12.39%
-6.55
Note: Auxiliary time series are solar radiation, dewpoint temperature, and wind. Scenario 5 did not have a solar radiation time
series; Scenario 9 did not have a dewpoint temperature time series; Scenario 10 did not have a wind time series. Results are
averages across entire study area. See Table 5-1 for details of the climate scenarios.
These results suggest that downscaling approaches that omit dewpoint temperature can introduce
significant biases.  Specifically, simulation without adjusting for future changes in dewpoint
temperature is likely to overestimate PET, leading to an underestimation of soil moisture and
streamflow.

6.2.3. Sensitivity of Flow and Water Quality to Approaches for Downscaling GCM
       Projections
The effect of downscaling approach on the variability of watershed model simulations can be
investigated quantitatively by comparing the results from simulations based on degraded
NARCCAP, GCM, and BCSD scenarios.  Table 6-5 presents results obtained with current land
use and the SWAT watershed model (with increased atmospheric CO2) at the most downstream
gage in  each study area. Table 6-6 presents detailed results for multiple streamflow and water
quality parameters in the Minnesota River study area. Differences among results with different
downscaling methods are qualitatively similar for HSPF output (not shown).
                                           6-17

-------
       Table 6-5. Summary of SWAT-simulated total streamflow in the five pilot
       study areas for scenarios representing different methods of downscaling
Study area
Apalachicola-Chattahoochee-
Flint (ACF)
Salt/Verde/San Pedro (Ariz)
Minnesota River (Minn)
Susquehanna (Susq)
Willamette (Willa)
Downscaling
method
NARCCAP
BCSD
GCM
NARCCAP
BCSD
GCM
NARCCAP
BCSD
GCM
NARCCAP
BCSD
GCM
NARCCAP
BCSD
GCM
Number of
scenarios
6
4
4
6
4
4
6
4
4
6
4
4
6
4
4
Median
(cms)
710.4
675.5
655.0
19.4
24.0
26.0
229.5
236.8
238.3
834.8
935.7
868.7
878.8
833.0
843.3
Maximum
(cms)
818.8
722.0
750.7
24.5
28.4
27.0
274.3
286.3
277.0
855.5
948.4
1,017.1
951.8
1,003.7
970.7
Minimum
(cms)
478.6
655.3
581.3
12.9
21.3
19.9
149.4
209.7
124.4
705.6
879.2
807.0
763.6
800.3
810.6
CV
0.208
0.042
0.105
0.233
0.122
0.131
0.230
0.153
0.301
0.068
0.035
0.106
0.086
0.108
0.082
Notes: Results shown are for most downstream station in each study area; coefficient of variation (CV) = standard deviation
divided by the mean. Climate scenarios are degraded to a common basis of scenario precipitation and air temperature
information only.
Results show considerable variability among climate models and downscaling techniques in
different basins and for different streamflow and water quality endpoints. No consistent pattern
attributable to downscaling method is evident for the case in which all climate model outputs are
evaluated using a common basis of precipitation and air temperature only. As was discussed in
Section 6.1.3., the additional information on other meteorological variables can have a profound
effect on PET and watershed responses.

It is noteworthy that the dynamically downscaled results may differ significantly from the
statistically downscaled results from the same GCM, and that the results may also be quite
different when the same GCM is downscaled with a different RCM (e.g., refer to Table 5-1 and
compare climate scenarios 1 and 5 for CGCM3, also 3 and 4 for the GFDL).  As noted in Section
5.2., direct comparison between NARCCAP and BCSD downscaling of a single GCM can only
be reliably undertaken for the GFDL and CGCM3 models, because slightly different GCM runs
were used to produce NARCCAP and BCSD results for other GCMs.
                                           6-18

-------
       Table 6-6.  Summary of SWAT-simulated streamflow and water quality in
       the Minnesota River study area for scenarios representing different methods
       of downscaling
Endpoint
Total Streamflow (cms)
100-Yr High Flow (cms)
7 Day Average Low
Flow (cms)
Total Suspended
Solids (MT/yr)
Total Phosphorus
(MT/yr)
Total Nitrogen (MT/yr)
Downscaling
method
NARCCAP
BCSD
GCM
NARCCAP
BCSD
GCM
NARCCAP
BCSD
GCM
NARCCAP
BCSD
GCM
NARCCAP
BCSD
GCM
NARCCAP
BCSD
GCM
Number of
scenarios
6
4
4
6
4
4
6
4
4
6
4
4
6
4
4
6
4
4
Median
229.5
236.8
238.3
3,415.4
3,960.2
3,565.7
27.7
25.8
28.2
1,926,166
2,002,421
1,914,800
36,304
40,579
38,747
2,700
3,073
2,889
Maximum
274.3
286.3
277.0
3,700.2
5,055.0
4,432.3
38.5
37.9
37.0
2,520,444
2,428,565
2,557,634
42,119
44,936
42,087
3,283
3,453
3,162
Minimum
149.4
209.7
124.4
3,155.7
3,617.6
2,508.7
14.3
22.3
12.9
896,806
1,376,608
633,793
25,843
32,451
21,538
2,007
2,356
1,489
CV
0.230
0.153
0.301
0.058
0.153
0.227
0.353
0.247
0.395
0.385
0.265
0.460
0.191
0.150
0.264
0.194
0.183
0.292
MT = metric ton

Notes: Results shown are for most downstream station in each study area; coefficient of variation (CV) = standard deviation
divided by the mean. Climate scenarios are degraded to a common basis of scenario precipitation and air temperature
information only.
Both the GFDL and CGCM3 A2 scenario runs for 2041-2070 were downscaled with two
different NARCCAP RCMs—with one RCM (RCM3) in common between the two. A
comparison in terms of the ratio of simulated future mean annual streamflow to simulated current
mean annual streamflow, using SWAT, is made in Figure 6-12 for the GFDL and in Figure 6-13
for the CGCM3 model. For both GCMs, the NARCCAP downscaling, BCSD downscaling, and
nondownscaled GCM output produce relatively consistent results for the Willamette and
Susquehanna basins, but diverge for the Minnesota River. For the Arizona basin, the two
different downscaling approaches diverge for the GFDL but not the CGCM3 GCM. Elevated
coefficients of variation (CVs) on mean annual streamflow in both the Minnesota River and
Arizona basins appear to be largely due to the difference in downscaling results obtained with
                                          6-19

-------
the GFDL high-resolution regional model, which suggests lower flow than other dynamically
downscaled interpretations of the GFDL GCM.
9nn°/


cu
c
"3
in
ro
CO
M-
o

cu
u
*- 1 nn°/
Q.
E
_3
o
5
_o
u_
1
no/






I
V
X
*
* 1
• • •



ACF Ariz Minn SL



KLMo
• High Res
A BCSD
X Raw GFDL


i 2




jsq Willa




























       Figure 6-12.  Consistency in SWAT model projections of mean annual
       streamflow at downstream stations with downscaled (NARCCAP, BCSD)
       and nondownscaled GCM projections of the GFDL GCM.

Note: The climate change scenarios used in this analysis are simplified to include changes only in air temperature and
precipitation (variables common to the NARCCAP, BCSD, and GCM data sets) to provide a common basis for comparison.
Figures 6-12 and 6-13 demonstrate that a single GCM may yield rather different results
depending on the RCM used for dynamical downscaling. In the current state of the science it
does not appear that the use of dynamical downscaling reduces uncertainty; however, use of
multiple downscaling approaches helps to inform the potential range of climate futures.

To date, relatively few comparisons of RCM model performance in the NARCCAP data sets
have been undertaken. An exception is the study of Wang et al. (2009) for the Intermountain
Region of the Western United States.  Significant orographic effects in this area lead to a
complex combination of precipitation annual and semiannual cycles that form four major climate
regimes in this area. Wang et al. compared results from six RCMs over this region to the North
American Regional Reanalysis  (NARR) precipitation study (Mesinger et al., 2006) and found
that each model produces its own systematic bias in the central Intermountain Region where the
four different climate regimes meet. All six of the RCMs appeared to produce simulated annual
cycles that are too strong and winter precipitation that is too high under current conditions.  The
BCSD statistical approach  can correct this for current conditions; however, the  statistical
approach would not account for any future large-scale changes in the interaction of the major
climate regimes.
                                          6-20

-------
               cu
               c
               ro
               CO
               c
               cu
               H
               cu
               E
               _3
               o

               I
                    0%




•
1









•
+ RCM3
• CRCM
A BCSD
XRawCGCMB



A
* *
...» 	 A 	




                            ACF
Ariz
Minn
Susq
Willa
       Figure 6-13. Consistency in SWAT model projections of mean annual
       streamflow at downstream stations with downscaled (NARCCAP, BCSD)
       and nondownscaled GCM projections of the CGCM3 GCM.

Note: The climate change scenarios used in this analysis are simplified to include changes only in air temperature and
precipitation (variables common to the NARCCAP, BCSD, and GCM data sets) to provide a common basis for comparison.
Wang et al. (2009) also demonstrate that the different RCMs are largely consistent in the
Cascade Range (OR, WA), where the dominant upper level flow first encounters land, which fits
with the reduced level of variability between downscaling methods noted for the Willamette
study area. The differences among RCMs reported by Wang et al., and the difference from
NARR, are greatest on the windward side of the Rocky Mountains in Colorado and remain large
into Arizona. Interestingly, the apparent wet bias of the CRCM and dry bias of most other
RCMs relative to NARR in Arizona reported by Wang et al. does not appear to carry through
into the future scenarios reported here, suggesting that the RCMs may be providing different
simulated solutions to the future interaction of large-scale climate regimes in this area.

In addition to uncertainties in representing climate forcing at the watershed level, as discussed in
this section, previous sections have shown that the results are sensitive to the selection of a
watershed model, and to modeler skill  in calibrating the model. Furthermore, the results are
undoubtedly also sensitive to feedback loops that are not incorporated into the models. Results
produced in this study thus likely do not span the full range of potential future impacts (even
conditional on the A2 storyline) for the reasons given above, among others.  Nonetheless, the
range of uncertainty is considerable, and generally covers the zero point, as is summarized at
selected downstream analysis points shown in Table 6-7.
                                           6-21

-------
       Table 6-7. Range of SWAT-projected changes in annual streamflow and
       pollutant loads for combined mid-21st century NARCCAP climate change
       and ICLUS urban and residential development scenarios
Downstream location
ACF: Apalachicola River Outlet
Ariz: Verde River ab Tangle Creek
Susq: Susquehanna River Outlet
Minn: Minnesota River Outlet
Willa: Willamette River Outlet
Change in flow
(%)
-26. 9 to +23. 6
-29.4 to +26.7
-10.0 to +11.0
-14.3 to +62.1
-8.4 to +15. 9
Change in total
solids load
(%)
-47.2 to +6.1
-52. 6 to +118.4
-15.6 to +17.8
-22.9 to +122.9
-10. 3 to +24. 5
Change in total
nitrogen load
(%)
-4.6 to +25. 6
-7.2 to +46. 6
+32.1 to +61. 9
+4. 9 to +71.0
-10. 9 to +3. 3
Change in total
phosphorus load
(%)
-6. 6 to +73.1
-32.8 to +63.4
+6. 3 to +28.1
-6.3 to +59.5
-13. 3 to +4.2
The ranges shown in Table 6-7 suggest that for 2041-2070 conditions it is not possible in most
cases to even state the sign of change in watershed response with a high degree of assurance
unless one is willing to assert that one of the RCMs is more reliable than another. Rather, the
results tell us that the range of potential responses is large.

Based on the analysis presented here, however, the differences in simulation results in our study
are largely a result of combined differences in the underlying GCM and the downscaling
approach used, and more specifically, largely a result of heterogeneity in simulated precipitation
amounts and patterns. For the 2041-2070 timeframe, these warming-induced increases in
simulated PET are generally insufficient to overcome this range of variability in projected
precipitation.  This may not be the case, however, for more distant future simulation
periods—given continually increasing temperature and PET, evapotranspiration increases are
likely to ultimately exceed the range of variability in projected precipitation in many basins,
resulting in more uniform decreases in runoff.
                                          6-22

-------
   7. REGIONAL SENSITIVITY OF STREAMFLOW AND WATER QUALITY TO
          CLIMATE CHANGE AND LAND DEVELOPMENT: RESULTS IN ALL
                                    20 WATERSHEDS
This section presents simulation results in all 20 study areas using SWAT. Model simulations
evaluate the effects of mid-21st century climate change alone (see Section 7.1.), urban and
residential development alone (see Section 7.2.), and the combined effects of climate change and
urban development (see Section 7.4) on streamflow, TN, TP, and TSS. Scenarios also assume
future increases in atmospheric CC>2.  Results are presented for a single representative analysis
point in each study area (see Table 7-1). For study areas composed of a single watershed, this is
the outlet (pour point) of the entire study area. For study areas composed of multiple, adjacent
watersheds draining to the coast, the analysis point reported here is at or near the outlet of the
largest river within the study area. Results for additional locations within each study area are
presented in Appendix X for the five pilot study areas  and in Appendix Y for the other 15 study
areas.
       Table 7-1. Downstream stations within each study area where simulation
       results are presented
Study area
Apalachicola-Chattahoochee-Flint Basins (ACF)
Southern California Coastal (SoCal)
Cook Inlet Basin (Cook)
Georgia-Florida Coastal Plain (GaFla)
Illinois River Basin (Illin)
Lake Erie Drainages (LErie)
Lake Pontchartrain Drainage (LPont)
Nebraska: Loup and Elkhorn River Basin (Neb)
Minnesota River Basin (Minn)
Tar and Neuse River Basins (TarNeu)
New England Coastal Basins (NewEng)
Powder and Tongue River Basin (PowTon)
Rio Grande Valley (RioGra)
Sacramento River Basin (Sac)
Arizona: Salt, Verde, and San Pedro (Ariz)
South Platte River Basin (SoPlat)
Susquehanna River Basin (Susq)
Trinity River Basin (Trin)
Upper Colorado River Basin (UppCol)
Willamette River Basin (Willa)
Location presenting results
Apalachicola R at outlet
Los Angeles R at outlet
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Elkhom R at outlet
Minnesota R at outlet
Neuse R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Salt River near Roosevelt
S. Platte R at outlet
Susquehanna R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
                                          7-1

-------
7.1.  SELECTION OF WATERSHED MODEL FOR USE IN ALL STUDY AREAS
Resource limitations for this study precluded the application of SWAT and HSPF in all 20 study
areas. Analyses at Pilot sites were used to select a single model for application in all 20 study
areas. Analyses in the Pilot sites show HSPF and  SWAT are each capable of providing a good
fit to streamflow and pollutant loads for existing conditions.  The quality of fit depends in part on
the strategy and skill of the individual modeler. In this study, the quality of fit was also
influenced by the availability in certain areas of preexisting, calibrated models which were
adapted for use as compared to locations where new models were developed and calibration
subject to resource limitations.

For the purposes of this study, the SWAT model was considered to have a technical advantage
because it can account for the influence of changes in atmospheric CO2 concentration and other
feedback responses of plant growth to climate change. HSPF does not automatically account for
these effects.  While it uncertain how well SWAT is able to represent the complex processes
affecting plant growth, nutrient dynamics, and water budgets under changing climate (see Luo et
al., 2013), it was considered important to include some representation of these processes  to better
understand potential watershed sensitivity to a wide range of conditions. In addition, there  are
also practical advantages to the choice of SWAT,  as the model is somewhat easier to set up and
calibrate than is HSPF.

Conversely, the HSPF model proved generally better able to replicate observations during
calibration, as shown in Section 6.1.1., although the difference between HSPF and SWAT model
performance was small for the selected response variables. HSPF is often  able to provide a
better fit to streamflow after calibration due to the use of hourly precipitation and a more
sophisticated algorithm compared to SWAT's daily curve number approach—although this
advantage is diminished by the need to use disaggregated daily total rainfall to drive the models
in many areas. Increased accuracy in hydrology—especially the accurate partitioning between
surface and subsurface runoff—should also provide increased accuracy in the simulation of
sediment yield and the transport of sediment-associated nutrients. However, at the larger
watershed scales studied here (HUC-8  and greater), such advantages will tend to diminish as
observations reflect the integration of flows and loads from multiple subwatersheds driven by
multiple weather stations. Further, SWAT is generally considered to perform better under
limited calibration and thus may have an advantage for extension to changed conditions of land
use and climate (Gassman et al., 2007).

The file structure of the HSPF model is also considerably more efficient for implementing and
running multiple scenarios.  SWAT's use of the curve number approach to hydrology and a daily
time step can also cause difficulties in representing the full hydrograph and introduces
uncertainties into the simulation of erosion and pollutant loading as a function of surface flow
(Garen and Moore, 2005).  This is  a concern in particular for the simulation of urban hydrology
at small spatial scales; however, these concerns are of lesser importance at the larger spatial
scales that are the focus of this  study.

Given that both models were capable of performing adequately, the SWAT model was selected
for use in the 15 nonpilot watersheds due to its integrated plant growth model and practical
advantages of ease of calibration.
                                           7-2

-------
It should be recognized that there are other feedback cycles that are not incorporated in either
model, such as the potential for any increased rate of catastrophic forest fires (Westerling et al.,
2006), changes to vegetative communities as a result of pests and disease (Berg et al., 2006), and
human adaptations such as shifts to different crops and agricultural management strategies
(Polsky and Easterling, 2001).

7.2.  SENSITIVITY TO CLIMATE CHANGE SCENARIOS
This section presents the results of SWAT simulations in all 20 study areas for climate change
scenarios alone (that is, with land use held constant at existing conditions). In general, the
different climate scenarios provide a consistent picture of temperature increases by mid-century
(on the order of 2 to 3°C or 3 to 6°F), although there do appear to be systematic differences
between the scenarios (for example, the NARCCAP scenario using the GFDL model downscaled
with RCM3 typically is the coolest scenario for the  watersheds studied here). In contrast,
changes in precipitation between the historical and future periods differ widely across climate
change scenarios, with some producing increases and some decreases in total precipitation.

Projected mid-21st century precipitation, air temperature, PET, and simulated AET (from SWAT)
for each of the six NARCCAP climate change scenarios in each study area are shown in Tables
7-2 through 7-5.  For Cook Inlet (Alaska) results are shown only for the three NARCCAP
scenarios that provide climate projections for this portion of Alaska. The projected future
climate annual average as a percent of baseline resulting from each of the six NARCCAP
scenarios is shown for precipitation, PET, and AET; absolute change is shown for the annual
average temperature. It should be noted that while the projected future average annual
temperature increases in  all cases, PET does not always increase. This is particularly noticeable
in some of the southwestern study areas  (e.g., Rio Grande Valley) where at least some future
climate scenarios project increases in humidity  and  cloudiness that offset the temperature impact
on PET. While shown here for comparison to PET, AET is a model input, not a model output.
AET is driven by PET, but can also be limited by lack of soil moisture and is affected by changes
in the seasonal timing of both precipitation and plant growth.
                                          7-3

-------
Table 7-2. Average annual precipitation (in/yr and percent of baseline) for current conditions and mid-21s
century climate scenarios
Study area
ACF — Apalachicola-Chattahoochee-Flint Basins
Ariz — Arizona: Salt, Verde, and San Pedro
Cook — Cook Inlet Basin
GaFla — Georgia-Florida Coastal Plain
Illin — Illinois River Basin
LErie — Lake Erie Drainages
LPont — Lake Pontchartrain Drainage
Minn — Minnesota River Basin
Neb — Nebraska: Loup and Elkhorn River Basins
NewEng — New England Coastal Basins
PowTon — Powder and Tongue River Basins
RioGra — Rio Grande Valley
Sac — Sacramento River Basin
SoCal — Southern California Coastal Basins
SoPlat— South Platte River Basin
Susq — Susquehanna River Basin
TarNeu — Tar and Neuse River Basins
Trin — Trinity River Basin
UppCol — Upper Colorado River Basin
Willa — Willamette River Basin
Current
conditions
52.14
19.38
24.22
52.98
37.63
36.88
64.76
27.61
24.43
46.42
13.85
12.20
35.81
19.62
15.93
39.73
48.90
42.83
15.88
55.43
CRCM_
cgcm3 (%)
105.1
87.4
NO
101.3
101.5
102.4
96.0
102.3
99.5
106.1
99.1
89.1
102.3
96.2
95.4
106.6
99.5
94.8
90.3
106.5
HRM3
hadcm3 (%)
114.3
94.3
118.3
117.3
114.2
114.2
109.2
106.7
103.4
113.2
100.2
91.1
88.6
117.1
92.2
109.2
122.3
110.4
97.3
101.1
RCM3
gfdl (%)
106.2
110.4
NO
106.5
103.9
104.9
106.4
110.3
103.4
107.7
104.8
106.5
95.8
97.0
97.5
103.6
112.6
98.6
108.3
97.6
GFDL
slice (%)
97.2
85.9
113.9
95.3
104.1
109.0
92.5
97.8
86.2
107.4
86.5
90.6
99.6
95.5
87.1
105.4
103.2
83.4
95.7
88.4
RCM3 cgcm3
(%)
111.2
98.5
NO
112.0
105.3
104.0
100.9
110.7
106.3
104.7
105.6
88.3
99.1
99.4
98.9
105.7
108.0
101.8
94.8
105.1
WRFP ccsm
(%)
90.4
87.9
122.6
85.1
93.3
91.7
87.8
112.1
104.8
98.1
120.0
99.4
96.3
87.6
101.2
97.9
92.4
105.9
95.2
94.5
Median ratio
(%)
105.6
91.1
118.3
103.9
104.0
104.5
98.5
108.5
103.4
106.7
102.5
90.8
97.7
96.6
96.5
105.6
105.6
100.2
95.4
99.4

-------
Table 7-3. Average annual temperature (°F and change from baseline) for current conditions and mid-21st
century climate scenarios
Study area
ACF — Apalachicola-Chattahoochee-Flint Basins
Ariz — Arizona: Salt, Verde, and San Pedro
Cook — Cook Inlet Basin
GaFla — Georgia-Florida Coastal Plain
Illin — Illinois River Basin
LErie — Lake Erie Drainages
LPont — Lake Pontchartrain Drainage
Minn — Minnesota River Basin
Neb — Nebraska: Loup and Elkhorn River Basins
NewEng — New England Coastal Basins
PowTon — Powder and Tongue River Basins
RioGra — Rio Grande Valley
Sac — Sacramento River Basin
SoCal — Southern California Coastal Basins
SoPlat— South Platte River Basin
Susq — Susquehanna River Basin
TarNeu — Tar and Neuse River Basins
Trin — Trinity River Basin
UppCol — Upper Colorado River Basin
Willa — Willamette River Basin
Current
conditions
64.33
56.41
33.13
68.29
49.57
49.13
66.48
44.18
47.94
46.32
44.84
44.72
58.23
61.38
45.06
48.18
59.93
64.91
40.80
51.48
CRCM_
cgcm3
+3.81
+4.93
NO
+3.56
+5.36
+5.19
+3.77
+5.61
+5.20
+4.97
+4.77
+5.13
+4.16
+3.58
+4.98
+4.98
+4.28
+4.35
+5.20
+3.79
HRM3_
hadcm3
+4.16
+5.19
+5.20
+3.99
+4.66
+4.65
+4.53
+5.29
+5.10
+4.81
+4.97
+5.37
+4.76
+3.97
+5.20
+4.98
+4.51
+4.66
+5.14
+4.37
RCM3_ gfdl
+3.62
+4.35
NO
+3.45
+4.38
+4.29
+3.61
+4.01
+3.88
+4.07
+3.81
+4.20
+3.75
+3.72
+4.14
+4.16
+3.83
+3.97
+4.13
+2.80
GFDL_ slice
+4.49
+4.96
+3.99
+4.36
+4.84
+4.75
+4.13
+5.02
+5.09
+4.12
+4.71
+5.84
+3.47
+3.27
+5.51
+4.72
+4.18
+4.45
+5.53
+3.03
RCM3_
cgcm3
+3.45
+4.75
NO
+3.32
+4.75
+4.67
+3.41
+4.60
+4.53
+4.67
+4.50
+5.02
+3.94
+3.98
+4.93
+4.59
+3.70
+3.79
+4.90
+3.59
WRFP_
ccsm
+4.35
+4.62
+5.30
+3.68
+5.36
+5.11
+3.79
+4.90
+4.65
+4.50
+4.27
+4.74
+4.06
+3.57
+4.77
+4.60
+4.14
+4.38
+5.04
+3.57
Median
change
+3.98
+4.84
+5.20
+3.62
+4.80
+4.71
+3.78
+4.96
+4.87
+4.58
+4.61
+5.08
+4.00
+3.65
+4.96
+4.66
+4.16
+4.36
+5.09
+3.58

-------
Table 7-4. Average annual PET (in/yr and percent of baseline) for current conditions and mid-21st century
climate scenarios
Study area
ACF — Apalachicola-Chattahoochee-Flint Basins
Ariz — Arizona: Salt, Verde, and San Pedro
Cook — Cook Inlet Basin
GaFla — Georgia-Florida Coastal Plain
Illin — Illinois River Basin
LErie — Lake Erie Drainages
LPont — Lake Pontchartrain Drainage
Minn — Minnesota River Basin
Neb — Nebraska: Loup and Elkhorn River Basins
NewEng — New England Coastal Basins
PowTon — Powder and Tongue River Basins
RioGra — Rio Grande Valley
Sac — Sacramento River Basin
SoCal — Southern California Coastal Basins
SoPlat— South Platte River Basin
Susq — Susquehanna River Basin
TarNeu — Tar and Neuse River Basins
Trin — Trinity River Basin
UppCol — Upper Colorado River Basin
Willa— Willamette River Basin
Current
conditions
62.04
81.27
16.56
65.82
42.91
45.27
59.19
49.36
61.94
43.22
55.39
54.48
66.77
64.41
53.25
43.81
56.38
77.27
38.14
43.64
CRCM
cgcm3 (%)
101.2
103.6
NO
99.9
112.3
102.0
101.4
106.3
100.4
103.3
101.3
94.4
99.2
99.2
102.0
102.9
100.5
99.5
106.8
97.4
HRM3
hadcm3 (%)
103.8
103.8
106.2
101.1
110.5
100.7
106.9
110.4
100.5
105.9
102.7
100.1
103.0
100.2
103.4
107.6
100.9
100.1
107.7
102.0
RCM3 gfdl
(%)
101.6
100.3
NO
99.6
109.3
99.6
103.0
98.6
97.0
100.3
99.0
90.4
102.5
99.9
100.3
101.2
99.2
99.0
103.9
100.5
GFDL
slice (%)
97.5
103.0
99.7
100.3
111.2
101.1
103.9
110.5
101.4
100.6
102.3
99.9
98.8
99.0
104.1
101.4
100.1
99.8
108.8
98.9
RCM3
cgcm3 (%)
98.3
102.6
NO
98.6
110.0
100.2
99.3
99.7
98.1
100.6
100.3
95.9
98.8
100.0
101.9
99.5
99.0
98.1
105.8
98.6
WRFP
ccsm (%)
105.2
106.4
104.1
100.6
111.2
101.3
101.2
94.2
97.9
101.3
99.0
92.6
101.6
99.2
101.4
104.9
100.1
98.7
106.2
98.7
Median
ratio (%)
101.4
103.3
104.1
100.1
110.9
100.9
102.2
103.0
99.2
100.9
100.8
95.1
100.4
99.6
102.0
102.1
100.1
99.3
106.5
98.8

-------
Table 7-5. Average annual SWAT-simulated actual ET (in/yr and percent of baseline) for current conditions
and mid-21st century climate scenarios
Study area
ACF — Apalachicola-Chattahoochee-Flint Basins
Ariz — Arizona: Salt, Verde, and San Pedro
Cook — Cook Inlet Basin
GaFla — Georgia-Florida Coastal Plain
Illin — Illinois River Basin
LErie — Lake Erie Drainages
LPont — Lake Pontchartrain Drainage
Minn — Minnesota River Basin
Neb — Nebraska: Loup and Elkhorn River Basins
NewEng — New England Coastal Basins
PowTon — Powder and Tongue River Basins
RioGra — Rio Grande Valley
Sac — Sacramento River Basin
SoCal — Southern California Coastal Basins
SoPlat— South Platte River Basin
Susq — Susquehanna River Basin
TarNeu — Tar and Neuse River Basins
Trin — Trinity River Basin
UppCol — Upper Colorado River Basin
Willa— Willamette River Basin
Current
conditions
32.22
14.47
7.95
30.86
22.90
22.75
29.83
21.64
18.00
23.31
16.83
10.32
15.26
8.75
13.06
23.73
29.48
27.58
13.13
19.84
CRCM
cgcm3 (%)
106.1
86.8
NO
98.5
101.5
94.4
100.1
96.1
97.8
103.3
93.3
84.2
99.2
97.2
96.1
104.8
97.2
95.0
91.7
87.9
HRM3
hadcm3 (%)
110.6
94.8
109.1
101.1
103.4
95.3
107.0
99.9
101.8
110.4
94.9
87.9
97.6
102.7
94.0
108.4
99.9
99.9
98.0
92.9
RCM3 gfdl
(%)
106.6
102.8
NO
99.7
101.3
96.0
101.3
94.9
100.6
103.5
96.7
98.0
94.7
92.9
96.0
102.6
98.0
97.0
101.2
85.4
GFDL
slice (%)
106.2
86.3
103.6
98.9
101.0
97.2
103.4
97.2
94.3
102.5
83.7
88.7
97.0
93.7
90.4
103.7
97.8
90.1
98.6
82.9
RCM3
cgcm3 (%)
104.9
97.3
NO
99.1
101.7
93.3
99.0
95.7
98.9
104.3
97.2
85.2
95.9
96.8
97.1
102.0
97.4
96.4
94.0
88.7
WRFP ccsm
(%)
102.8
89.8
108.6
95.0
98.9
90.8
98.2
92.7
97.9
104.8
105.4
94.7
94.9
94.1
98.5
104.8
95.6
97.8
95.5
88.3
Median
ratio (%)
106.2
92.3
108.6
99.0
101.4
94.9
100.7
95.9
98.4
103.9
95.8
88.3
96.5
95.5
96.0
104.2
97.6
96.7
96.7
88.1

-------
In addition to changes in precipitation amount, this study considers the impacts of changes in
precipitation intensity, which may have significant effects on the partitioning between surface
and subsurface flows and associated generation of pollutant loads. As described in Section 5.2.,
a change factor approach was used to modify historical meteorological time series to represent
mid-21st century climate futures projected by a variety of downscaled (and nondownscaled)
GCM projections. Potential intensification of precipitation is represented by reapportioning the
net change in precipitation volume according to GCM forecasts of the distribution of event
intensities above and below the 70*  percentile of the distribution of current (1971-2000) rainfall
events.  Under current conditions, the fraction of rainfall volume occurring in events above the
70th percentile ranges from a low of 61% (Cook Inlet and Willamette) to a high of 93%
(Southern California Coastal). Projected mid-21st century changes in precipitation intensity  from
the six NARCCAP scenarios, shown in Table 7-6, are mixed.  Across all study areas there is an
average increase in the fraction of total volume above the 70*  percentile of the current
distribution of 1.19 percentage points.  However, for most study areas the six NARCCAP
scenarios are not in full agreement as to whether intensification of precipitation (as defined
relative  to the 70* percentile event) will increase.  An increase in the volume in high-intensity
events is consistently projected across all six of the NARCCAP mid-21st century projections in
only six of the 20 study areas (Susq, Minn,  Cook,  LErie, Illin, and NewEng).  Two RCM/GCM
combinations (HRM3_hadcm3 = Scenario 2) and (RCM3_cgcm3 = Scenario 5) project increases
in intensity in all study areas. No study area is expected to have a decrease in precipitation
volume  in high intensity events across all NARCCAP scenarios, while six study areas (Cook,
Illin, LErie, Minn, NewEng, and  Susq) are projected to have an increase in high-intensity events
across all six NARCCAP scenarios. By far the largest increases in high-intensity events are
projected for the Cook Inlet watershed  in Alaska, followed by the Upper Colorado basin.

The simulated watershed responses  to mid-21st century climate change scenarios are shown in
Tables 7-7 through 7-14.  For endpoints other than days to streamflow centroid, the results are
displayed as a percentage relative to the current baseline (generally, 1972-2003), allowing
comparison across multiple basins with different magnitudes of streamflow and pollutant  loads.
For Cook Inlet (Alaska), the results are shown only for the three NARCCAP scenarios that
provide  climate projections for this portion  of Alaska.

Table 7-7 summarizes results for total average annual streamflow volume, with results ranging
from 62% to 240% of current average flows. Results for 7-day low streamflow and 100-year
peak flows (estimated with log-Pearson III fit) are shown in Tables 7-8 and 7-9, respectively.
The Kenai River has by far the greatest increase in 7-day low flows because warmer
temperatures alter the snow/ice melt regime, while the largest increases in 100-year peak flows
are for the Neuse River on the east coast.

Table 7-10 summarizes the estimated change in days to streamflow centroid relative to the start
of the water year. Many stations  show negative shifts, indicating earlier snowmelt resulting in an
earlier center of streamflow mass. In contrast, several stations show positive shifts due to
increased summer precipitation.

Results  for the Richards-Baker flashiness index (see Table 7-11) show generally small
percentage changes, with a few exceptions. Baker et al. (2004) suggest that changes on the order

-------
of 10% or more may be statistically significant.  It is likely, however, that the focus on larger
watersheds reduces the observed flashiness response.
       Table 7-6.  Changes in precipitation intensity for NARCCAP mid-21st
       century climate scenarios

ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM cgcm3
(%)
3.28
-0.19
ND
2.74
1.95
2.25
2.49
1.92
1.08
2.55
1.67
-0.04
1.87
0.24
-0.20
3.28
2.59
1.32
-0.10
2.50
HRM3 hadcm3
(%)
1.35
0.53
7.51
1.60
1.56
2.01
0.71
1.22
1.56
1.51
1.54
0.96
0.47
1.83
0.85
1.58
1.38
1.24
1.95
2.59
RCM3 gfdl
(%)
1.87
0.03
ND
2.33
1.09
1.64
2.50
1.94
1.61
1.74
1.66
1.64
0.00
-0.76
1.13
2.08
1.55
0.61
2.36
0.46
GFDL slice
(%)
-0.68
0.36
3.71
-1.23
0.87
1.81
-0.48
0.09
-0.12
0.36
-1.09
0.37
-1.06
-0.28
-0.15
0.41
-0.42
-0.66
0.79
-2.63
RCM3 cgcm3
(%)
2.43
0.79
ND
2.99
1.25
1.12
1.61
1.38
0.99
1.21
1.59
0.91
2.28
0.29
1.27
1.77
1.21
0.14
2.06
2.71
WRFG ccsm
(%)
-0.12
-0.43
4.62
-0.89
0.20
0.27
-0.87
0.43
0.14
0.15
0.97
0.52
1.42
-0.15
0.51
0.41
-0.02
0.46
0.71
0.97
Note: Potential change in precipitation intensity is shown as the change total volume of precipitation event above the 70th
percentile of the current (1971-2000) distribution of rainfall event volumes.
Simulated changes in pollutant loads (TN, TP, TSS) are summarized in Tables 7-12 through
7-14. The patterns are generally similar to changes in streamflow.  Increases in pollutant loads
are suggested for many watersheds, but there are also basins where loads decline, mostly due to
reduced flows.
                                             7-9

-------
Table 7-7. Simulated total streamflow volume (climate scenarios only; percent relative to current conditions) for
selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhom R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM cgcm3
(%)
107
80
ND
114
94
116
96
109
117
108
101
72
104
92
90
109
103
98
86
116
HRM3 hadcm3
(%)
122
80
154
153
125
150
110
113
125
115
85
69
89
138
74
106
158
146
95
106
RCM3 gfdl
(%)
108
149
ND
128
101
120
115
147
137
111
140
112
98
102
90
106
137
106
116
105
GFDL slice
(%)
88
75
132
92
102
136
84
86
68
111
70
66
98
103
65
108
110
62
89
92
RCM3 cgcm3
(%)
124
94
ND
156
105
122
106
146
138
106
130
69
100
106
107
111
125
118
92
114
WRFP ccsm
(%)
73
73
167
75
78
88
77
162
143
94
240
84
99
84
119
90
86
134
91
98
Median
(%)
107
80
154
121
101
121
101
130
131
109
115
71
99
103
90
107
118
112
91
105

-------
Table 7-8. Simulated 7-day low flow (climate scenarios only; percent relative to current conditions) for selected
downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhom R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM cgcm3
(%)
97
58
ND
104
85
104
73
115
119
110
102
81
101
96
93
91
94
26
85
131
HRM3 hadcm3
(%)
120
77
267
141
123
184
106
136
133
140
92
64
91
114
87
120
170
167
94
113
RCM3 gfdl
(%)
105
130
ND
121
97
126
88
201
151
130
145
120
95
98
97
104
135
64
121
108
GFDL slice
(%)
85
87
280
95
91
132
74
81
48
118
67
62
96
98
74
89
113
23
85
83
RCM3 cgcm3
(%)
113
79
ND
136
100
128
89
182
148
124
127
74
99
100
102
107
125
70
91
127
WRFP ccsm
(%)
64
90
401
78
70
58
62
228
154
120
235
86
93
92
113
86
70
85
90
102
Median
(%)
101
83
280
113
94
127
81
159
140
122
115
77
95
98
95
98
119
67
91
111

-------
           Table 7-9. Simulated 100-year peak flow (log-Pearson III; climate scenarios only; percent relative to current
           conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhom R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM cgcm3
(%)
119
119
ND
130
120
96
105
84
126
114
118
90
105
83
132
107
71
97
78
116
HRM3 hadcm3
(%)
144
101
132
145
153
106
150
83
117
130
113
77
98
89
127
130
292
106
84
130
RCM3 gfdl
(%)
110
104
ND
129
107
87
108
96
109
111
133
108
125
161
98
106
161
107
97
114
GFDL slice
(%)
90
68
125
94
99
93
99
88
92
138
82
66
117
95
126
128
111
60
91
79
RCM3 cgcm3
(%)
128
120
ND
157
128
93
105
90
139
89
121
72
102
127
151
172
224
86
94
116
WRFP ccsm
(%)
94
66
132
107
97
92
65
96
103
80
146
92
131
77
150
100
63
106
84
95
Median
(%)
114
102
132
130
114
93
105
89
113
112
119
83
111
92
129
118
136
102
87
115
to

-------
Table 7-10. Simulated changes in the number of days to streamflow centroid (climate scenarios only; relative to
current conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhom R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_ cgcm3
-2
-18
ND
-3
-12
-2
-14
-13
-12
-17
-6
25
-4
5
-12
-18
-14
16
-11
3
HRM3_ hadcm3
-2
41
-3
17
6
-4
13
-19
6
-14
-3
6
-7
48
-20
16
23
21
-14
-8
RCM3_ gfdl
1
28
ND
25
-3
1
-24
-6
1
-19
1
3
-4
-3
-14
-6
30
30
-7
-1
GFDL_ slice
8
17
-5
-8
-12
0
-7
-15
-15
-13
-16
11
-1
10
-19
-12
-12
3
-10
3
RCM3_ cgcm3
-6
-6
ND
-5
-2
10
-6
-3
-6
-9
-4
14
-3
-3
-3
-6
10
6
-8
1
WRFP_ ccsm
1
53
-1
11
-15
-8
-11
2
2
-18
7
17
-8
1
-12
0
-5
37
-10
8
Median
-1
22
-3
4
-7
-1
-9
-10
-2
-16
-3
13
-4
3
-13
-6
2
18
-10
2

-------
Table 7-11.  Simulated Richards-Baker flashiness index (climate scenarios only; percent relative to current
conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhom R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM cgcm3
(%)
106
81
ND
93
106
99
105
104
95
101
102
109
124
103
99
107
96
71
101
101
HRM3 hadcm3
(%)
125
102
94
62
104
101
105
112
98
103
108
117
103
119
91
111
113
68
107
105
RCM3 gfdl
(%)
109
121
ND
76
103
99
106
107
94
99
104
95
112
100
101
107
115
72
111
100
GFDL slice
(%)
94
98
102
117
106
100
104
100
95
101
100
119
109
105
87
110
98
73
105
97
RCM3 cgcm3
(%)
125
103
ND
59
105
100
104
109
96
98
103
103
116
105
108
112
103
69
104
101
WRFP ccsm
(%)
90
119
96
187
104
96
102
108
94
93
109
106
123
99
106
103
91
68
101
102
Median
(%)
108
102
96
84
105
100
104
108
95
100
104
108
114
104
100
109
101
70
105
101

-------
Table 7-12.  Simulated total suspended solids load (climate scenarios only; percent relative to current conditions)
for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_
cgcm3 (%)
125
89
ND
121
116
123
100
107
122
118
108
60
139
71
91
117
106
63
80
124
HRM3
hadcm3 (%)
146
79
234
176
142
169
115
119
131
128
84
53
94
111
87
108
199
124
90
111
RCM3
gfdl (%)
129
184
ND
138
115
126
128
187
147
117
169
114
122
81
94
108
162
62
124
109
GFDL_
slice (%)
93
66
196
90
128
153
83
77
60
122
66
49
118
81
80
115
115
27
82
90
RCM3_
cgcm3 (%)
144
106
ND
181
120
129
111
197
162
111
153
59
99
84
100
118
143
83
89
121
WRFP ccsm
(%)
53
74
244
74
90
86
71
225
162
85
351
71
108
65
104
84
82
113
85
97
Median
(%)
127
84
234
130
118
128
106
153
139
118
131
59
113
81
93
112
129
73
87
110

-------
Table 7-13.  Simulated total phosphorus load (climate scenarios only; percent relative to current conditions) for
selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM cgcm3
(%)
138
82
ND
115
107
118
113
97
118
111
107
54
100
53
90
128
112
124
79
100
HRM3 hadcm3
(%)
152
83
89
171
112
150
131
115
124
118
86
43
86
88
78
106
230
163
88
98
RCM3 gfdl
(%)
134
155
ND
135
107
132
135
151
138
111
163
127
104
71
99
111
169
130
119
96
GFDL slice
(%)
118
70
90
89
113
148
94
97
65
115
67
51
115
60
72
127
120
83
81
94
RCM3_
cgcm3 (%)
148
106
ND
173
108
117
115
138
145
106
148
41
95
62
108
115
166
135
84
100
WRFP_
ccsm (%)
106
88
113
76
99
88
83
160
147
94
324
67
108
54
111
109
94
160
83
96
Median
(%)
136
86
90
125
108
125
114
126
131
111
127
53
102
61
95
113
143
132
84
97

-------
Table 7-14.  Simulated total nitrogen load (climate scenarios only; percent relative to current conditions) for
selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM cgcm3
(%)
116
90
ND
127
103
128
123
126
93
119
109
49
99
93
86
162
111
121
73
104
HRM3 hadcm3
(%)
125
91
200
160
118
158
141
130
97
128
91
38
89
140
70
147
189
165
82
97
RCM3 gfdl
(%)
115
142
ND
135
106
162
143
163
145
117
165
125
100
131
91
147
154
125
110
95
GFDL slice
(%)
106
86
175
112
110
191
106
105
88
121
71
47
110
98
63
156
118
80
76
89
RCM3_
cgcm3 (%)
122
105
ND
166
108
125
120
158
104
114
148
37
98
90
109
150
144
136
80
103
WRFP_
ccsm (%)
95
84
223
85
93
94
91
171
107
101
320
64
107
101
116
132
99
164
79
93
Median
(%)
116
90
200
131
107
143
121
144
101
118
128
48
100
100
89
149
131
130
80
96

-------
For most measures in most watersheds, there is a substantial amount of variability between
scenario projections based on different methods of downscaling GCM outputs.  This reflects our
uncertainty in predicting future climate, especially the future joint distribution of precipitation
and potential evapotranspiration that is fundamental to watershed response, and reinforces the
need for an ensemble approach for evaluating the range of potential responses.

Climate change could also alter the seasonal dynamics of streamflow and nutrient loading.
Seasonal effects are investigated here in summary form through calculation of the ratio of winter
(January-March) to summer (July-September) runoff volume averaged over all HUC-8s in a
study area. More detailed results showing simulated changes in streamflow by month are
presented in Appendices X and Y. The different study areas have very different seasonal runoff
volume ratios under current conditions, ranging from a winter:summer low of 0.11 in the Cook
Inlet basin to a high of 11 in the Willamette River basin.  The average ratios under the mid-21st
century NARCCAP climate change scenarios are shown relative to the current ratio in Figure
7-1.  In most cases, the future climate scenarios span the current ratio; however, in the case of the
South Platte and Upper Colorado study areas, currently dominated by snowmelt runoff from the
Rocky Mountains, all future climate scenarios project an increase in the ratio. In some basins the
range of future projected seasonal runoff ratios is quite large. For the Salt, Verde, and San Pedro
River basins (Ariz) the average future ratios by climate scenario range from 0.8 to 5.4, depending
on whether the climate scenario projects greater increases in the summer monsoon or winter
rainy period, while in the Lake Erie drainages (LErie) the range is from 1.5 to 7.8.  The
distribution for each of the six NARCCAP  climate scenarios is summarized in Figure 7-2. There
are clear differences between the different scenarios, with some projecting a much greater
increase in the winter: summer runoff ratio than others.

7.3.  SENSITIVITY TO URBAN AND RESIDENTIAL DEVELOPMENT SCENARIOS
This section presents the results of SWAT simulations in all 20 study areas for mid-21st century
urban and residential development alone (that is, with climate held constant at existing
conditions).  Results in the pilot study areas (see Section 6.) suggested that effects of urban and
residential development by 2050 on streamflow and pollutant loads is likely to be comparatively
small relative to the potential range of impacts associated with climate change.  This is largely a
reflection of the scale of the analysis:  at the scale of large (HUC-4 to HUC-8) watersheds,
developed land is rarely a large portion of the total land area. Significant effects may occur in
smaller subbasins where extensive new land development occurs.

Over the full extent of individual study areas, current impervious surface area ranges from near
zero to 13.8% of the total area, while projected changes (increases) in impervious cover area
range from 0 to 5.3% of the total area (see Table 7-15). While several fast-growing metropolitan
areas are included within the study areas, the impact of these areas is diminished at larger spatial
scales. At the HUC-8 and larger scale, it is not surprising that projected changes in urban and
residential development have only a relatively small effect compared to climate change, which
affects all portions of a watershed. The largest response of total streamflow volume to land-use
change at the full-basin scale is simulated for the Trinity River in Texas, where total flow
increased by 6%, while the estimated  100-year peak flow decreased and days to streamflow
centroid increased (i.e., later runoff).  This reflects increases in development upstream in  the
Dallas-Fort Worth metropolitan area.  A stronger response to land development is seen at smaller
                                          7-18

-------
spatial scales where development can account for a larger fraction of watershed area.
Development effects are also more likely be reflected in high or low streamflow statistics. For
example, in the Los Angeles River projected changes in urban and residential development result
in little change in model-simulated total streamflow volume, but the 100-year peak flow
increases by nearly 25%.
1 nn nn
J.UU.UU






J^
'S
c

cc
>- 1 n nn .
QJ 1U.UU
E
of Winter to Sum
i-*
b
o
g
+j
re
QC






01 n
. J.U

















£










LL.
U
















^
*
A
*










M
<
















__.








' '
* '

O
o
U
















w











03
ID

I
I
1
1
1
1
1
1
1
1
1
1
1
1

1 o
1
1
':•
-li—
i
i
i
i
i
i
i
i
i
i
i
i
i

C _OJ
= LU
















1










•4-J
C
o
















X.










c
c
















X
- >











OJ
















i
.__










QD
NewEn
















X
->*'-
I










c
o
o

















*









O3
(5
O
















ft
__.










U
TO
l/l
















s
,__











U
o
LO
















	










•4-J
o
LO
















i










O"
tn
LO
















I
-^
H










(U
L_
O3
















__,










C
H
















.__


n








o
U
Q.
Q.












^^
• •
"^J"
^
.__










O3
i













/- j.
^Current

• 2
A3
X4
X5

• 6









       Figure 7-1.  Ratio of winter (January-March) to summer (July-September)
       runoff volume under current and mid-21st century NARCCAP climate
       scenarios.

Notes: Results are averages over all HUC-8s simulated within a study area. Climate scenarios are (RCM and GCM): (1)
CRCM_cgcm3, (2) HRM3_hadcm3, (3) RCM3_gfdl, (4) GFDL High Res_gfdl, (5) RCM3_cgcm3, and (6) WRFP_ccsm.
The simulated watershed responses to projected mid-21st century urban and residential
development are shown in (see Table 7-16). Results across all 20 watersheds are small, as would
be expected given the small changes in developed lands, when expressed as a fraction of total
watershed area, at the scale of modeling in this study. Larger effects are likely in smaller
subbasins within the study areas where urban and residential development is concentrated. Note
that results are not available for the Kenai River (Cook Inlet, AK study area) because ICLUS
projections do not include Alaska.
                                           7-19

-------

|3.0
CL
O
CD ^-°
"ro
CD
"ro
or
1_
CD
1 ^ "
W
i_
CD
"£ 1 n
C I .U
§
£ 0.5

n n









X

X




X


I
T |

I ^ J

X

1 2 3































J

L





















I

I




























































X




4

5








6
Climate Scenario
       Figure 7-2. Box plots of the distribution of the ratio of winter
       (January-March) to summer (July-September) runoff volume normalized to
       the ratio under current conditions.

Notes: The box shows the interquartile range, with median indicated by a horizontal line, and the whiskers extend 1.5 times the
interquartile range. Outliers beyond the whiskers are shown by individual points.  The data are averages over all HUC-8s
simulated within a study area. Climate scenarios are (RCM and GCM): (1) CRCM_cgcm3, (2) HRM3_hadcm3, (3) RCM3_gfdl,
(4) GFDL High Res_gfdl, (5) RCM3_cgcm3, and (6) WRFP_ccsm.
7.4.  RELATIVE EFFECTS OF CLIMATE CHANGE AND URBAN DEVELOPMENT
      SCENARIOS
The changes in urban and residential development projected by ICLUS for 2050 suggest changes
may be large locally but are small relative to the area of basins modeled in this study (see Table
5-5).  Urban and residential development has long been recognized as a source of hydrologic
changes and water quality degradation at local scales in developing areas (e.g., U.S. EPA, 1984).
The cumulative impacts of development, however, tend to be relatively small at the larger basin
scale evaluated in this study simply because only a small fraction of most HUC-4 scale
watersheds is developed or projected to be developed by 2050.
                                           7-20

-------
       Table 7-15. Projected mid-21s century impervious cover changes in study
       areas from ICLUS for A2 emissions storyline
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
Current (2001)
impervious cover (%)
2.04
0.19
0.24
2.50
6.19
3.48
3.24
1.06
0.38
5.59
0.08
0.55
0.73
13.80
2.06
1.50
1.70
4.17
0.37
2.51
Projected mid-21s* century impervious
cover (%)
3.06
0.30
ND
3.86
8.22
3.88
4.56
1.28
0.39
6.74
0.08
0.81
0.95
19.11
4.27
1.69
2.55
7.37
0.61
3.06
Change in
impervious cover (%)
1.02
0.11
ND
1.36
2.03
0.40
1.32
0.22
0.01
1.15
0.00
0.26
0.22
5.31
2.21
0.19
0.85
3.20
0.24
0.55
The relative magnitude of effects from urban development versus climate change in our
simulations can be examined by looking at changes in mean annual streamflow. Figure 7-3
compares the HSPF simulated change in mean annual streamflow in the pilot study areas for
mid-21st century urban and residential development compared to the six NARCCAP climate
change scenarios.  The results summarize the range of responses across selected HUC-8
subbasins and calibration locations contained within each study area.  Table 7-17 compares the
range of SWAT simulated changes in mean annual streamflow in all study locations for mid-21st
century urban and residential development and the six NARCCAP climate change scenarios.
Results summarize the ranges at the HUC-8 and larger scale within the study areas.
                                         7-21

-------
          Table 7-16. Simulated response to projected 2050 changes in urban and residential development (percent or

          days relative to current conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study
area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
Total flow
(%)
100.3
100.1
ND
100.3
102.4
100.5
100.8
100.2
100.3
100.4
100.0
100.1
100.1
101.4
102.8
100.2
101.7
106.4
100.1
99.9
7-day low
flow (%)
100.4
100.0
ND
99.9
104.0
100.8
102.6
100.3
100.3
100.5
100.0
100.1
100.1
101.3
100.7
100.7
105.2
188.1
100.6
100.0
100-yr peak
flow (%)
100.3
100.2
ND
100.6
102.1
101.4
101.6
99.9
101.5
101.4
100.0
100.4
99.9
114.4
101.1
99.7
102.1
74.2
100.3
100.1
Days to flow
centroid
-0.1
0.1
ND
0.3
1.0
0.2
0.2
0.3
0.0
0.0
0.0
0.0
-0.1
0.0
0.9
0.1
0.7
3.7
-0.1
0.0
Richards-Baker
flashiness (%)
100.0
100.3
ND
99.5
98.4
100.9
100.4
100.1
102.8
101.3
100.0
100.2
100.4
103.9
103.9
100.1
99.1
68.8
99.8
100.7
TSS load
(%)
100.6
100.2
ND
100.4
100.5
100.6
98.7
98.0
100.1
101.2
100.0
101.1
99.7
106.6
103.9
100.2
102.3
61.9
100.0
99.7
TP load
(%)
101.1
100.4
ND
108.9
100.2
101.3
106.8
99.3
100.1
103.8
100.0
95.4
102.1
138.2
104.0
99.7
106.7
110.0
100.8
99.9
TN load
(%)
100.5
100.2
ND
102.5
99.2
99.6
103.9
99.5
99.8
102.0
100.0
99.6
104.7
111.1
103.4
99.2
103.3
106.2
100.2
102.5
to
to

-------
Simulations using both HSPF and SWAT show a smaller range of response to projected future
changes in urban development than to projected climate change. As discussed previously, at the
spatial scale of these simulations projected future changes in developed land were a relatively
small fraction of total watershed area. At smaller spatial scales, however, the effects of urban
and residential development could be greater. Results for pollutant loads are similar to those for
streamflow.

The simulated response to land-use change is also sensitive to model choice—or, more precisely,
an interaction between the model and the way in which the ICLUS is interpreted. In the SWAT
setup, there are representations of both directly connected (effective) and disconnected
impervious area. New developed land use implied by ICLUS is identified to the model as a total
area in a given development density class, then subdivided by the model into pervious and
impervious fractions using basin-specific estimates of total and effective impervious area. The
effective impervious fraction for a given development category is calculated from the 2000
NLCD and assumed invariant. The model then assumes that the effective impervious area has a
curve number of 98, while the remaining disconnected impervious area provides a small
modification to the curve number assigned to the pervious fraction of the HRU.

In contrast, HSPF has pervious (PERLND) and impervious (IMPLND) land uses, but does not
distinguish a separate disconnected impervious class. For HSPF, the new developed area in
ICLUS is assigned to the relevant pervious and impervious land-use fractions based on the
basin-specific percent imperviousness for the land-use class. In essence, this means that
somewhat greater future connected imperviousness is being specified to the HSPF model than is
specified to the SWAT model. While the two approaches are rather different, they are consistent
with typical modeling practice for the two models.

Several other details of the SWAT modeling process adopted in this study affect results.  The
approach to implementing changes in urban development in SWAT was to remove land from
existing undeveloped and nonexempt land uses and reassign it to new developed classes that
have the parameters of the most dominant soil and lowest HRU slope in the subbasin. In some
cases (particularly when a subbasin is already largely developed) the dominant soil in the
watershed may have characteristics different from the soils and slopes of the remaining
undeveloped land.  For HSPF, the urban land uses are not associated  with a specific soil or HSG.

In addition, a special circumstance occurs in the Willamette SWAT model. In that model, new
developed land primarily comes from dense forest cover. The model tends to simulate greater
evapotranspiration for urban grass than for intact evergreen forest, which appears to offset
increases in total streamflow volume due to increased impervious area.

The effects of land-use change on simulated streamflow  extremes can be more dramatic in basins
where strong growth is expected, but also tend to be smaller than the range of simulated climate
responses.  For example, in the ACF basin, land-use change alone can increase the simulated
100-year flood peak by up to 27%, but the range of responses to the six NARCCAP climate
scenarios is from 17 to 66%.
                                         7-23

-------
         100%
           80o/0 |-~r— -c-h_ange_due_tp_
                             urbanization
           60% + -/
      o
_   40%  +-
CO
3
i   20%  +-
<
S    o%
o>
1  -20%

|>  -40%
CD
u  -60%

    -80%

   -100%  --
                                                       Change due to
                                                       climate
       Figure 7-3. Comparison of simulated responses of mean annual streamflow
       to urban development and climate change scenarios—HSPF model.

Note: The blue area represents the range of responses to the six NARCCAP RCM-downscaled 2050 climate scenarios across the
different HUC-8 scale reporting sites (with no change in land use). The red bars represent the maximum response to land-use
change among the reporting sites (with no change in climate). Results are shown for Apalachicola River at outlet (ACF), Sat
River near Roosevelt (Ariz), Minnesota River at outlet (Minn), Susquehanna River at outlet (Susq), and Willamette River at
outlet (Willa).
7.5.  SENSITIVITY TO COMBINED CLIMATE CHANGE AND URBAN
      DEVELOPMENT SCENARIOS
This section presents the results of SWAT simulations in all 20 study areas for the combined
effects of mid-21st century climate change and urban and residential development scenarios.
Simulation results are generally consistent with results for climate scenarios alone (presented in
Section 7.2.) given the relatively small response to projected urban and residential development
at the spatial scale of modeling in this study. Results are presented for selected locations in each
study area in Tables 7-18 through 7-27.  For study sites comprised of a single watershed, results
are shown for a downstream outlet. For study sites comprised of multiple adjacent basins results
are shown for a single representative basin, typically the largest.  These same results for each
study area are also shown as scatterplots in Figures 7-4 through 7-24, followed by maps showing
the simulated median values for the six NARCCAP scenarios at the HUC-8 scale within study
areas.  It should be noted that use of the median values alone without taking into account the full
range of simulated responses to all scenarios is potentially misleading. Median values are
presented here only as an indicator of variability between study areas and should not alone be
considered indicative of broad regional trends. It  should also be noted that simulation results for
                                           7-24

-------
Kenai River in the Cook Inlet basin do not include urban and residential development scenarios.
ICLUS projections are not available for the Alaska study area, but are anticipated to be small.
       Table 7-17. Simulated range of responses of mean annual streamflow to mid-
       21st century climate and land-use change at the HUC-8 and larger spatial
       scale

ACF
Ariz
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
Climate Change Response
Minimum (%)
-45.73
-35.29
-39.73
-22.20
-22.89
-24.75
-23.39
-79.14
-12.55
-42.49
-45.38
-20.79
-26.91
-53.04
-23.80
-13.65
-60.57
-20.21
-17.51
Maximum (%)
24.84
152.52
69.85
34.00
72.13
21.82
85.38
72.64
19.80
206.01
19.86
10.29
62.19
59.23
25.79
61.60
125.65
22.93
23.21
Land-Use Change Response
Minimum (%)
0.00
0.00
0.01
0.00
0.00
0.00
0.00
0.00
0.02
0.00
-0.07
-0.03
-3.60
-1.00
0.00
0.28
7.09
-0.38
-1.18
Maximum (%)
0.68
1.48
7.36
11.90
1.84
1.24
0.19
0.27
0.76
0.00
0.13
0.47
6.36
2.82
0.23
4.31
34.91
0.47
0.00
Note: Cook Inlet basin is not shown because ICLUS land-use change information is not available. Results based on SWAT
simulations for the six NARCCAP climate change scenarios and ICLUS 2050 projected changes in developed land.
The simulated ranges of total streamflow volume changes shown in Figure 7-4 suggest several
observations.  The first is that increases in streamflow volume for the Kenai River (Cook Inlet
basin) are on average larger than for other basins. Perhaps more importantly, for a majority of
the basins the different downscaled models do not provide a consistent sign for changes in
streamflow for the 2041-2070 period, with some simulating increases and some decreases.  The
models are in complete agreement as to the sign of change only for Kenai River (increase).  It is
                                          7-25

-------
also worth noting that the Weather Research and Forecasting Model (WRFP) downscaling of the
CCSM GCM often seems to be an outlier relative to the other models.

Figure 7-5 shows the median simulated annual streamflow volume (as the median over the six
NARCCAP scenarios; expressed as percent of baseline conditions) at the HUC-8 spatial scale for
each study area.  On this map, a neutral gray tone represents no change from current conditions
(100% of current conditions).  Browns indicate streamflow volumes less than current, with
greater color intensity reflecting lower streamflow; blues represent flow volumes greater than
current, with greater intensity reflecting higher flows.  Simulated median values suggest a
general trend of decreasing streamflow volume in the central Rockies, accompanied by increases
in streamflow in the northern plains. Only moderate changes are seen for the west coast and
Mississippi Valley, while streamflow volume generally increases on the  east coast.

In addition to streamflow volume, changes in the timing and rate of streamflow can also affected
by climate change. At a national scale, the number of days to the streamflow centroid—the point
at which half the streamflow volume of an average year is achieved (calculated from the
October 1 start of the water year)—is a useful measure of changes in the seasonal distribution of
streamflow.  Figure 7-11 shows that the centroid of streamflow comes earlier in the year in
model-simulated response to warmer temperatures for many of the snow-melt dominated basins,
particularly Cook Inlet in Alaska and higher elevations in the Rockies, but also for many basins
in the southeast.  The latter result reflects changes in precipitation timing, with increased winter
precipitation and decreased summer precipitation.   Several of the western basins have later dates
for the streamflow centroid due to a substantial increase in model-simulated spring or summer
precipitation relative to winter snowpack that counteracts the effects of earlier snowmelt.
Appendices X and Y provide more detailed information about seasonal shifts in streamflow
timing in the study areas.

The geographic distribution of 100-year peak flows (Log-Pearson III) fit is displayed in Figure
7-9 and shows considerably more heterogeneity. Simulated peak flows increase in many basins,
but show less of a clear pattern (see Figure 7-8). Peak flows tend to decline in the area of the
Southwest where total streamflow volumes decline, while the greatest increases are seen in
Alaska and the populated areas of the east and west coast. The increase in 100-year peak flows
is generally greater (or, in some instances, the reduction less) than the change in total streamflow
volume, consistent with the findings of Taner et al. (2011) for Lake Onondaga.

Results also suggest a large (factor of 5) increase in low flows for the Kenai River (see Figure
7-6).  This reflects greater dry season melt rates of ice under a warmer climate in Alaska. The
models also consistently show large declines in low flows for the Rio Grande Valley.
                                          7-26

-------
          Table 7-18.  Simulated total streamflow volume (climate and land-use change scenarios; percent relative to
          current conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_
cgcm3 (%)
107
80
ND
115
96
117
96
110
117
108
101
73
104
92
92
109
104
102
86
116
HRM3
hadcm3 (%)
122
80
154
154
126
151
111
113
126
116
85
69
89
140
76
107
160
150
95
106
RCM3
gfdl (%)
108
149
ND
128
103
120
116
147
137
111
140
112
98
104
92
106
138
110
116
104
GFDL slice
(%)
89
75
132
93
104
136
85
86
68
112
70
66
98
103
67
108
111
66
89
92
RCM3_
cgcm3 (%)
124
94
ND
157
106
123
107
146
138
106
130
69
100
107
110
111
127
122
92
114
WRFP_
ccsm (%)
73
73
167
75
79
89
78
162
143
94
240
84
99
85
121
90
88
138
91
98
Median
(%)
108
80
154
122
103
122
102
130
131
110
115
71
99
103
92.27
108
119
116
91
105
to

-------
to

-------
VO
                   0  200 400    800
                                  Kilometers
           Figure 7-5. Median simulated percent changes in total future streamflow volume for six NARCCAP scenarios

           relative to current conditions by HUC-8 (median of NARCCAP climate scenarios with urban development).

           Note: Cook Inlet results do not include land-use change.

-------
Table 7-19. Simulated 7-day low flow (climate and land-use change scenarios; percent relative to current
conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_
cgcm3 (%)
98
58
ND
105
88
105
76
115
119
112
102
81
101
98
94
92
100
33
85
131
HRM3
hadcm3 (%)
120
77
267
141
126
184
108
137
133
141
92
64
91
115
88
121
175
199
94
113
RCM3
gfdl (%)
105
131
ND
121
100
127
91
202
152
131
145
120
95
99
98
105
139
87
122
108
GFDL
slice (%)
86
87
280
95
94
133
77
82
48
119
67
62
96
100
75
90
118
36
86
82
RCM3_
cgcm3 (%)
113
79
ND
136
103
129
92
182
148
125
127
74
99
101
103
108
129
93
92
127
WRFP_
ccsm (%)
64
90
401
78
73
59
64
228
154
121
235
86
93
93
114
87
74
102
91
102
Median
(%)
101
83
280
113
97
128
84
159
141
123
115
77
95
99
96
98
123
90
91
111

-------

ylfVW
4L/U/0


_C
m
(0 3 r\r\ O/
fQ J(JL)/o
•5
•N
§ 250%
p
o_
> T rmo/
5 ZUU/o
^2
^L
J£
O -] Rf)Q/
s
i nn^
lUU/o
rpiO/
JU/O "

nQ£
u/o























	



I
_
X



j;
zl
O

15
a:
_ro

7;
LL.
1











	


	



A

<^k



•K

01
ie: Maume
LLJ
^J











	


	



^H
•




JZ

0
E
ro
a:
OJ
.t;
0
Q.
^J










	

	
•
A
m
X

^
•

X



j-

0
E
re
a;
ro
i: Minnesot
c
I





















A


JZ
zt
0
E
15
TV
C.
0
UJ
01




















1





JZ
ZI

E
ro
OL
^
: Merrimac
00
c
LLJ
3
(U











*





A

•
X



JZ



ro
OL
O)
00
O
Lz
o
J
o
0.


















A

I



HI
Z
2^
UJ
ZI
cr
n
5
Grande at ,
0
'cc

o
be








	








T



j;
^
o

ra
u.
i
ramento Ri
u
re
1/1
U
re
LO













	



•
i.

	


JZ
ZI
o

In
cc

"3
00
c
10
3
ro
u
0











—


	



•

^•^
f
A
	


c
o
1_
HI
c
OJ
X
•r^
(D
C£
HI
IS
0.

ro
0.
o
1/1










	


	



i
*
V
	


JZ
^
0
E
ro
Q^
ro
usque hann
i/>
1
t/)













	



A
A
V
	


az
D
0
E
15
0£
HI
ra
ZI
HI
zi
HI
ro
i —













_
^p




^
*

¥

JZ
Z)
o

15
oe

±i
1—
t:
1—
















	



A

T
	


0)
c
—
a
ra
ro
K

c
Colorado R

o
u
a.
Q.
"









	


	



^.
If
••
™"

~~
X



JZ
D

E
ro
ae
OJ
: Willamett
_ro
i








Climate Scenario

* l-CRCM_cgcm3
• 2-HRM3_hadcm3
4 3-RCM3_gfdl
,1 /trpil cliro
H-OrUL bMCc

t 5-RCM3_cgcm3

*6-WRFP_ccsm
— Median


















Figure 7-6. Simulated 7-day low flow relative to current conditions (NARCCAP climate scenarios with urban
development) for selected downstream stations.

-------
to
                    0  200 400    800
                                  Kilometers
                       7-day Low Flow

                       Median Change (%)
           Figure 7-7. Median simulated percent changes in 7-day average low flow volume for six NARCCAP scenarios

           relative to current conditions by HUC-8 (median of NARCCAP climate scenarios with urban development).


           Note: Cook Inlet results do not include land-use change.

-------
Table 7-20. Simulated 100-year peak flow (log-Pearson III; climate and land-use change scenarios; percent
relative to current conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_
cgcm3 (%)
117
119
ND
131
121
96
107
84
128
116
118
90
105
100
132
108
71
97
78
116
HRM3
hadcm3 (%)
145
101
132
145
155
107
152
83
117
134
113
77
98
112
126
130
294
107
83
131
RCM3 gfdl
(%)
110
104
ND
130
109
88
110
96
110
113
133
108
122
194
101
107
163
108
97
114
GFDL
slice (%)
90
68
125
95
103
94
100
87
93
141
82
66
117
124
129
129
113
60
91
79
RCM3_
cgcm3 (%)
128
121
ND
158
129
94
107
89
139
90
121
72
102
158
163
173
227
87
93
116
WRFP_
ccsm (%)
94
66
132
107
98
93
66
96
102
82
146
92
131
93
152
101
64
107
84
95
Median
(%)
114
102
132
130
115
94
107
88
114
115
119
83
111
118
131
118
138
102
87
115

-------
          250%
        a
          200% -
        41
        O.

        0
        o
          150%
          100%
            0%
  Climate Scenario


•» l-CRCM_cgcm3


• 2-HRM3_hadcm3


- 3-RCM3jgfdl


  4-GFDL_slice


i  5-RCM3_cgcm3


  6-WRFP_ccsm


— Median
Figure 7-8. Simulated 100-year peak flow relative to current conditions (NARCCAP climate scenarios with

urban development) for selected downstream stations.

-------
        0  200 400    800
                       Kilometers
Figure 7-9. Median simulated percent changes in 100-year peak flow for six NARCCAP scenarios relative to

current conditions by HUC-8 (median of NARCCAP climate scenarios with urban development).


Note: Cook Inlet results do not include land-use change.

-------
Table 7-21. Simulated change in the number of days to streamflow centroid (climate and land-use change
scenarios; relative to current conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_
cgcm3
-2
-18

-3
-11
-2
-14
-13
-12
-17
-6
25
-4
6
-11
-18
-13
17
-11
o
6
HRM3_
hadcm3
-2
41
-3
17
6
-4
14
-19
6
-14
-3
6
-7
48
-19
16
23
23
-14
-8
RCM3_ gfdl
1
28

25
-2
1
-23
-6
1
-19
1
3
-4
-3
-13
-6
31
31
-7
-1
GFDL_ slice
8
17
-5
-8
-12
0
-7
-15
-15
-13
-16
11
-1
10
-18
-12
-11
4
-10
o
6
RCM3_
cgcm3
-6
-5

-5
-1
10
-5
o
— J
-6
-9
-4
14
o
— J
-3
-2
-5
11
7
-8
1
WRFP_
ccsm
1
54
-1
11
-14
-8
-11
2
2
-18
7
17
-8
0
-11
0
-5
25
-11
8
Median
-1
22
-3
4
-6
-1
-9
-9
-2
-16
-3
13
-4
3
-12
-6
3
20
-10
2

-------
fin




•o
0
£ ™
c 30 -
a

u. 2U
a

« m

Q
c
O
C
n
5

~f-\J


























X


ff
X



—
^
0
E
*-*
ro
DC
IT!
Apalachicol

'•^




m

.


A

X






.



^

Ol
Si
o
0
DC

C
i
DC
1




















f



_c
D
0
E
to
DC

C
QJ
O
0
u











A








^rv
\/



_c
^
0
E
fi
cc:
(U
la: Suwane
re

















H

i
t


c
1
o


Ol
CO
ra
oc
'o
c
'^













^




I



j=
^
0
E
TO
QC
Ol
E
ro
.
±1
c
I—
iz
£





















f


OJ

— '
Ol
m
—
j_
C
Colorado R

o
u
a.
Q.
—*













^






.n
^
Q
E
ro
DC
01
: Willamett
JS
•—
*







Climate Scenario
* l-CRCM_cgcm3
• 2-HRM3 hadcm3

* 3-RCM3_gfdl

4-GFDL slice


i 5-RCM3 cgcm3
£. WRPP rrtm
— Median
















Figure 7-10. Simulated change in days to streamflow centroid relative to current conditions (NARCCAP climate
scenarios with urban development) for selected downstream stations.

-------
oo
                    0  200 400    800
                                  Kilometers
                        Change in Days
                        To Flow Centroid
           Figure 7-11. Median simulated change in the number of days to streamflow centroid for six NARCCAP
           scenarios relative to current conditions by HUC-8 (median of NARCCAP climate scenarios with urban
           development).

           Note: Cook Inlet results do not include land-use change.

-------
           Table 7-22. Simulated Richards-Baker flashiness index (climate and land-use change scenarios; percent relative
           to current conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_
cgcm3 (%)
106
81
ND
93
105
100
105
105
97
102
102
109
124
104
103
107
95
71
101
102
HRM3
hadcm3 (%)
125
103
94
62
103
102
105
112
101
104
108
117
103
125
95
111
112
69
107
105
RCM3 gfdl
(%)
109
121
ND
76
102
100
106
108
97
100
104
95
113
103
105
107
114
72
111
100
GFDL
slice (%)
94
98
102
116
106
101
104
101
96
102
100
120
109
105
91
110
97
73
105
98
RCM3_
cgcm3 (%)
126
103
ND
59
105
100
104
109
98
99
103
103
117
108
113
112
102
70
103
101
WRFP_
ccsm (%)
90
119
96
185
103
97
102
108
97
94
109
106
124
104
110
103
90
68
101
102
Median
(%)
107
103
96
84
104
100
104
108
97
101
104
108
115
105
104
109
100
70
104
102
VO

-------





g
1
u
•5
c
5
a.
X
c
m
EC
snai


60%
/i no/





















H








.c
^
0
E
ts
en
re
': Apalachicol













.
*
X

^^




a

a)
;/>
O
0

1=
§
be:
•H*
M














X
*






.c
^
0
E
ts
L-.

C
(U
o
0
u














X

^M

A
•
I

^
D
0
E
ts
o:
01
aFIa: Suwane
a













i







c
1
o

ro
fl
CO
ts
EC
l/l
'o
c
=













A







_z

0
E
—
OS
CY"
U
Erie: Maume














t



	 ,

	 j

_z

0
E
ts
L-

acramento Ri
u
m











•
i







.c
^
0
E
ts
rf
v>
•il: LosAngele
u
0













i
s






c

1_
^
c
111
X
—
cc
1)
t;
ro
0.
t/i
ro
0.
L/1












1







.c
^
0
E
ns
rf*
ro
Susquehann
^
1/1




















_n
D
0
E
ts
(Y
OJ
VI
a
01
id
m
•z.
i™


















£



jr
^
0
E
ts
C£

.t!
C
H
c
H














i







0)

— '
01
m
1_
C
1: Colorado R
o
u
a.
a.




















j;
D

E
ts
LT
0)
la: Willamett
g








Climate Scenario
* 1-CRCM cgcm3

• 2-HRM3_hadcm3
3-RCM3_gfdl
4-GFDL_slice
1 5-RCM3_cgcm3
• 6-WRFP_ccsm
— Median

















Figure 7-12. Simulated Richards-Baker flashiness index relative to current conditions (NARCCAP climate
scenarios with urban development) for selected downstream stations.

-------
         0  200 400     800
                       Kilometers
            Flasniness Index

            Absolute Change
Figure 7-13. Simulated absolute changes in the Richards-Baker flashiness index for six NARCCAP scenarios

relative to current conditions by HUC-8 (median of NARCCAP climate scenarios with urban development).


Note: Cook Inlet results do not include land-use change.

-------
Regional differences also occur in the degree of agreement among simulated watershed
responses to climate change scenarios. Table 7-23 shows the CV (standard deviation divided by
the mean) for SWAT-simulated percentage changes in different streamflow endpoints at the
downstream location of each study site for the six NARCCAP scenarios (calculated without
land-use change to isolate the impacts of climate). The CV for total streamflow is large at some
stations, such as Salt River and Tongue River, indicating poor model agreement on the
magnitude of change.  Note that CVs on total streamflow are artificially reduced at some stations
(e.g., Colorado River,  Sacramento River) due to the presence of constant upstream boundary
conditions (representing interbasin transfers for the Colorado and releases from an upstream dam
on the Sacramento River). The largest divergences among simulated high flows are seen at
different stations than  the largest divergences among total streamflow volume estimates.

CVs were also calculated reflecting the variability in response across the selected downstream
stations for all study areas for each NARCCAP climate change scenario. Table 7-24 shows these
values along with the average absolute difference from the median of all scenarios for each
NARCCAP scenario.  For total streamflow volume, the CCSM downscaled with WRFP has both
the greatest station-to-station variability (highest CV) and largest average absolute difference
from the median of all six simulations.

Simulated changes in pollutant loads are shown in Tables 7-25 through 7-27, and Figures 7-14
through 7-19.  Changes in projected pollutant loads are qualitatively similar to those seen for
response to climate change only, but further increased in areas with significant new urban
development.  In general, projected changes in pollutant loads follow a pattern similar to the
changes in total streamflow volume. Total  suspended solids loads (see Figure 7-15) increase in
most basins, except for declines in the Rocky Mountain and Southwest study areas where overall
streamflow decreases.  The large increases in solids loads for some basins (especially sand bed
rivers in the west) are mostly driven by channel scour. These results should be considered highly
uncertain given the simplified approach to channel scour included in SWAT version 2005 and
the differences among individual models in calibration to channel scour. The regional pattern for
total phosphorus loads is similar, as much of the total phosphorus load is driven by erosion (see
Figure 7-17), with the  notable exception of the Cook Inlet basin in Alaska.  The regional pattern
for total nitrogen loads is also generally similar, with  some additional variability associated with
the interactions of plant growth and erosion (see Figure 7-19).

Changes in the timing of nutrient load delivery may be even more significant for ecological
impacts (c.f, Tu, 2009; Wilson and Weng, 2011; Tong et al., 2011; Marshall and Randhir, 2008).
Potential ecological impacts of changes in timing of pollutant delivery simulated in the rich data
set generated by this study remain to be evaluated.

7.6. WATER BALANCE INDICATORS
Several  additional endpoints—identified here as water balance indicators—were calculated for
each study area. Water balance indicators are defined in Section 4.3. This section presents
results describing the SWAT-simulated changes in these indicators in response to the six mid-
21st century NARCCAP climate change and urban development scenarios.

Table 7-28 provides a  summary of water balance indicators for each study area.  Figures 7-20
through 7-24 show the median values for changes in water balance metrics for simulations using
                                          7-42

-------
the six NARCCAP climate change scenarios at each study location. As stated previously,
median values are presented here only as an indicator of variability within and among study
areas and should not alone be considered indicative of broad regional trends.  Appendices X and
Y provide more detailed results for changes in water balance indicators including analysis at
additional locations in each study area.
       Table 7-23. Coefficient of variation of SWAT-simulated changes in
       streamflow by study area in response to the six NARCCAP climate change
       scenarios for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
Total flow
0.038
0.091
0.021
0.089
0.023
0.035
0.023
0.066
0.064
0.005
0.293
0.039
0.003
0.032
0.044
0.005
0.055
0.079
0.013
0.008
100-yr peak
0.037
0.060
0.001
0.043
0.039
0.004
0.070
0.004
0.024
0.046
0.039
0.028
0.016
0.100
0.029
0.057
0.534
0.036
0.006
0.030
7-day low flow
0.043
0.067
0.172
0.053
0.033
0.137
0.029
0.198
0.128
0.009
0.273
0.056
0.001
0.005
0.019
0.017
0.101
0.378
0.020
0.028
                                         7-43

-------
Table 7-24. Coefficient of variation of SWAT-simulated changes in
streamflow by NARCCAP climate scenario for selected downstream stations
RCM/GCM
CRCM_cgcm3
HRM3_hadcm3
RCM3_gfdl
GFDL_slice
RCM3_cgcm3
WRFP_ccsm
Total flow
CV
0.016
0.068
0.026
0.049
0.036
0.167
Average absolute
difference from
median (%)
14.66
15.38
19.54
18.37
16.06
25.25
100-yr peak flow
CV
0.032
0.166
0.035
0.048
0.108
0.063
Average absolute
difference from
median (%)
14.97
19.76
18.52
17.56
25.00
19.83
7-day low flow
CV
0.058
0.163
0.073
0.264
0.068
0.571
Average absolute
difference from
median (%)
27.95
23.45
27.32
20.10
21.52
31.89
                                7-44

-------
Table 7-25. Simulated total suspended solids load (climate and land-use change scenarios; percent relative to
current conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_
cgcm3 (%)
126
89
ND
121
117
123
99
104
122
119
108
61
138
75
95
118
108
64
80
124
HRM3
hadcm3 (%)
147
79
234
177
142
170
113
117
131
129
84
54
94
121
91
108
201
126
90
111
RCM3
gfdl (%)
128
184
ND
139
115
127
125
183
147
119
169
115
121
86
98
109
164
64
124
108
GFDL
slice (%)
93
66
196
90
128
154
82
76
60
123
66
50
118
85
84
116
117
28
82
89
RCM3_
cgcm3 (%)
145
106
ND
182
121
130
110
192
162
112
153
60
99
90
104
118
145
85
89
121
WRFP_
ccsm (%)
53
74
244
74
91
87
70
219
162
86
351
72
108
69
108
85
84
115
85
97
Median
(%)
127
84
234
130
119
128
104
150
139
119
131
60
113
86
97
112
131
74
87
110

-------






^ nn°/
ttj
"HJ
m T C-O0/,
CO
|
c
H
£
•o
O
i
crw































X


.E
^
0
E
ro
LL.
fD
^palachicol

|
<















A

X
t


^

(D
y,
o
0
CC

c
§
en
'>_
^f














t

X





j:
^
0
E
ro
ry
—
c
O
0
u


















i

T
X


.c
^
0
E
-
C£.
su
la: Suwane
to
In


















I
*


C
1
o
t>
ra
u
DC
ce
w
'o
c
_
—















u
X

•


X
^
0
E
ro
CC
a>
ie: Maume
LLJ
—^


















i
. -_.J


j:
D
0
E
re
fY*
-•
.ti
0
D.


















I

5
X


.n
D
0
E
-
CC
re
: Minnesot
c

Z-

















t
x


j:
^
0
E
15
OL
C
0
.i.-
LJJ
01




















i
•JB


.n
^


TO
c;
_i/
: Merrimac
c
LLJ
=J

*


^
D
0
E
TO
ff-
HJ
v\
a
±
b
aj
2
I™



















1
-.

X
j:
^
0
E
T5
ry
-^
,t!
H
c
™




















A



0)


(U
m
j_
C
Colorado R

o
u
Q.
















i
*


x
^
Q
E
%
ry
OJ
: Willamett
JS
•^
5













• l-CRCM_cgcm3
• 2-HRM3 hadcm3
. 3-RCM3_gfdl
4-GFDL slice
1, 5-RCM3_cgcm3
*6-WRFP ccsm
— Median
















Figure 7-14. Simulated total suspended solids load relative to current conditions (NARCCAP climate scenarios
with urban development) for selected downstream stations.

-------
        0  200 400    800
                      Kilometers
Figure 7-15.  Median simulated percent changes in total suspended solids loads for six NARCCAP scenarios
relative to current conditions by HUC-8 (median of NARCCAP climate scenarios with urban development) for
selected downstream stations.

Note: Cook Inlet results do not include land-use change.

-------
          Table 7-26. Simulated total phosphorus load (climate and land-use change scenarios; percent relative to current
          conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_
cgcm3 (%)
139
82
ND
125
107
121
123
97
118
116
107
51
102
78
93
128
123
148
80
100
HRM3
hadcm3 (%)
153
84
89
190
112
155
144
115
124
125
86
40
88
128
81
106
259
188
88
98
RCM3
gfdl (%)
136
156
ND
149
107
136
147
151
138
116
163
125
106
102
104
110
184
153
120
97
GFDL_
slice (%)
119
70
90
96
113
151
103
97
65
120
67
49
117
83
75
127
134
98
82
94
RCM3_
cgcm3 (%)
150
107
ND
189
108
120
125
138
145
111
148
37
97
89
113
114
183
155
84
100
WRFP_
ccsm (%)
107
88
113
82
99
89
89
160
148
97
324
64
110
71
115
108
103
187
84
96
Median
(%)
138
86
90
137
107
128
124
126
131
116
127
50
104
86
99
112
158
154
84
97
oo

-------
VO



,nno.
jsUU/o



a /!jU/o
M
(U
00
•!•• T Ar\0/
^ 200%
•g
u
Cu 1 cn0/
i IjU/o
•a
F 1 nn°/
1 UU /o

n%
u/o




























X




.E
ZJ
0
S
a:
IB
0
.y
jz
u
ro
a
0.
~
§


















A



>

*i

be
o
a
E
ro
u
ro
LO
(J
ro


















•
A
IK
f

j:
^
0
ts
ce
in
W>
C
<£
w
"ro
U
0



















A
T
A
^•^^


c
o
£
u
c
u
I
<-
IB
££
OD
a.
^
00
a.
o
in





-.. .








	



f
•



JZ
^
0
It!
rv
ro
c
c
ro
_r
S
cr
Ul
ZJ

1
[/I















A

—
^fc
^



j=
^
0
s
ae

on
r
ai
zi
HI
IB
| 	

















B

*

*. x
JR^


JZ
"^
0
ro
ce

±i
i—
iz
K





















•


T

01
.^
09
m
ro
t/i

c
ce
o
•o
"o
i,
o
u
a.




















^
^^•P


j;
"D
g
ro
ce
u
•*-*
E
i
JB
i











Climate Scenario
* l-CRCM_cgcm3

• 2-HRM3_hadcm3

4 3-RCM3_gfdl

4-QFDL_slice
••1 5-RCM3_cgcm3
6-WRFP_ccsm

















          Figure 7-16.  Simulated total phosphorus load relative to current conditions (NARCCAP climate scenarios with
          urban development) for selected downstream stations.

-------
         0  200 400    800
                       Kilometers
            Total Phosphorus

            Median Change (%)
Figure 7-17. Median simulated percent changes in total phosphorus loads for six NARCCAP scenarios relative

to current conditions by HUC-8 (median of NARCCAP climate scenarios with urban development).


Note: Cook Inlet results do not include land-use change.

-------
Table 7-27. Simulated total nitrogen load (climate and land-use change scenarios; percent relative to current
conditions) for selected downstream stations
Station
Apalachicola R at outlet
Salt River near Roosevelt
Kenai R at Soldotna
Suwanee R at outlet
Illinois R at Marseilles, IL
Maumee R at outlet
Amite R at outlet
Minnesota R at outlet
Elkhorn R at outlet
Merrimack R at outlet
Tongue R at outlet
Rio Grande R below Albuquerque
Sacramento R at outlet
Los Angeles R at outlet
S. Platte R at outlet
Susquehanna R at outlet
Neuse R at outlet
Trinity R at outlet
Colorado R near State Line
Willamette R at outlet
Study area
ACF
Ariz
Cook
GaFla
Illin
LErie
LPont
Minn
Neb
NewEng
PowTon
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
CRCM_
cgcm3 (%)
117
90
ND
129
103
127
130
126
93
123
109
50
104
125
89
161
120
140
73
106
HRM3
hadcm3 (%)
126
91
200
167
117
158
152
130
97
131
91
38
94
159
72
146
207
187
82
98
RCM3 gfdl
(%)
116
142
ND
139
105
161
153
163
145
121
165
127
105
154
95
146
166
142
111
97
GFDL slice
(%)
107
87
175
113
109
190
113
104
88
124
71
48
113
102
65
155
125
93
76
91
RCM3_
cgcm3 (%)
123
105
ND
171
107
125
127
158
104
116
148
37
103
96
112
149
155
153
80
105
WRFP_
ccsm (%)
96
85
223
86
93
94
95
170
107
103
320
65
111
101
120
131
105
186
79
95
Median
(%)
117
91
200
134
106
142
128
144
100
122
128
49
104
113
92
147
140
148
80
97

-------
to




Q> ZjU/o
"3
U)
re
CO
NM
v
U
IH
^ 150% -
•o
1
fc i nn°/s

soy





















f
•


.E
13
0
15
Of
ra
^palachicol














A
X
t


*i
s
o
0
cc

c
1
'u-
"*












X



.E
13
0
— •
ro
[Y1
—
C
.ii
O
0
U














i
„.
X
•


.E
0
ro
C£
111
la: Suwane
ro
<2












^B
£
AV
•F


c
o
ra
CJ
CD
CSL
i/i
'o
C
=











X
.A.
*


.c
0
ro
at
03
ie: Maume
—













M
X
A
V


^
0
15
a:

.*;
0
a.














I
9
x



J5
0
ro
a:
ro
: Minnesot

5











_ . .
"A"
-
f


j:
13
0
15
p^
n
0
UJ
01














*
9



^
1
15
nr-

: Merrimac
DC
C
LLJ
3
z











A
X
X


.E
13
ro
cr
0)
ttO
o
F-
o
5
a.












A



U
E"
01
13
f-i
ej|
HI
C
5
.0
ro
(D
Q
C£










	
^^^
1


j:
0
15
^
>
rarnento Ri
U
ro
to
U
ro
ui











*
1


.c
13
0
ts
cc

Los Angele
ro
U
0












^
*


c
ELI
C
(U
X
—
UJ
i
s
uo

ro
0.
1/1











4



j:
0
ro
fY
ro
usquehann
m
1
uo









B

..
¥
•



_n
0
15
oc

on
3
(U
z
13
m
m












H
\v-
X


.E
0
ro
cc:

.ti
l—
if
K















f



-------
         0  200 400    800
                       Kilometers
Figure 7-19. Median simulated percent changes in total nitrogen loads for six NARCCAP scenarios relative to

current conditions by HUC-8 (median of NARCCAP climate scenarios with urban development).


Note: Cook Inlet results do not include land-use change.

-------
Table 7-28. Simulated percent changes in water balance statistics for study
areas (NARCCAP climate with land-use change scenarios; median percent
change relative to current conditions)
Study Area
ACF
Ariz-Salt
Ariz-San Pedro
Ariz-Verde
Cook
GaFla-North
GaFla-Tampa
Illin
LErie
LPont
Minn
Neb-Elkhorn
Neb-Loup
NewEng
PowTon-Powder
PowTon-Tongue
RioGra
Sac
SoCal
SoPlat
Susq
TarNeu
Trin
UppCol
Willa
Dryness Ratio
(fraction of
precipitation lost to
ET) (%)
0
1
-1
-2
-8
-10
-6
-1
-3
-10
-5
0
o
— 3
-i
-7
-6
2
0
-2
-1
0
-8
-4
1
-11
Low Flow Sensitivity
(baseflow generation,
cfs/mi2) (%)
-16
-10
-7
-3
22
47
8
-7
22
59
28
3
12
-6
18
5
-28
-4
-5
-6
-6
15
-1
-8
5
Surface Runoff
Fraction of
Flow (%)
22
-5
23
7
4
-8
11
15
-4
-14
49
16
-1
1
-1
6
o
J
4
7
1
16
5
2
-4
1
Snowmelt
Fraction of
Flow (%)
-57
-46
-52
-50
-12
-32
-72
-39
-32
-22
-24
-24
-33
-82
-18
-17
-1
-45
-54
-17
-31
-49
-43
-15
-68
Deep Recharge
Rate
(depth) (%)
-14
-15
-12
4
-43
39
7
-6
20
47
24
1
13
-5
NA
-8
-28
-6
1
NA
-5
15
0
-16
6
                                 7-54

-------
         0 200 400    800
                        Kilometers
Figure 7-20.  Median simulated percent changes in watershed Dryness Ratio for six NARCCAP scenarios

relative to current conditions (median of NARCCAP climate scenarios with urban development).

Note: Dryness ratio is the fraction of input precipitation lost to ET. Cook Inlet results do not include land-use change.

-------
                            *^-v

                         Ar'T
                      ^k    ex
         0  200 400     800
                       Kilometers
Figure 7-21. Median simulated percent changes in watershed Low Flow Sensitivity for six NARCCAP scenarios

relative to current conditions (median of NARCCAP climate scenarios with urban development).


Note: Low Flow Sensitivity is the rate of streamflow generation by baseflow (cfs/mi2). Cook Inlet results do not include land-use change.

-------
         0  200 400     800
                        Kilometers
         Surface Runoff Fraction of Flow

         Median Change (%)
Figure 7-22. Median simulated percent changes in watershed Surface Runoff Fraction for six NARCCAP

scenarios relative to current conditions (median of NARCCAP climate scenarios with urban development).


Note: Surface Runoff Fraction is the fraction of streamflow contributed by overland flow pathways. Cook Inlet results do not include land-use change.

-------
oo
                     0  200 400     800
                                    Kilometers
                       Snow Fraction

                       Median Change (%)
            Figure 7-23. Median simulated percent changes in watershed Snowmelt Fraction for six NARCCAP scenarios

            relative to current conditions (median of NARCCAP climate scenarios with urban  development).


            Note: Snowmelt Fraction is the fraction of streamflow contributed by snowmelt.  Cook Inlet results do not include land-use change.

-------
VO
                    0  200 400     800
                                   Kilometers
            Figure 7-24. Median simulated percent changes in watershed Deep Recharge for six NARCCAP scenarios

            relative to current conditions (median of NARCCAP climate scenarios with urban development).


            Note: Deep Recharge is the depth of water recharging deep aquifers per unit time. Cook Inlet results do not include land-use change. Areas shown in black have no deep
            recharge simulated.

-------
The water balance summaries are presented as averages over whole watersheds. These are
generally consistent with the project study areas, except that several study areas (e.g., Central
Nebraska) were simulated using more than one SWAT model and thus show multiple results.
Figure 7-20 shows the change in the Dryness Ratio, expressed as the ratio of ET to precipitation.
The central tendency of the Dryness Ratio is estimated to increase in the southern Rocky
Mountains and adjacent parts of Arizona, consistent with median decreases in simulated mean
annual streamflow (see Figure 7-4).

Another aspect of low flows is shown by the Low Flow Sensitivity metric—the average rate of
baseflow generation per square mile of watershed area.  This metric (see Figure 7-21) decreases
in areas for which the Dryness Ratio increases. However, it also decreases in various other
watersheds (such as SoCal and ACF) for which there is little change in the Dryness Ratio. Areas
where the Low Flow Sensitivity metric decreases may be expected to experience difficulties in
maintaining minimum streamflow for aquatic life support or for meeting wasteload dilution
expectations.

The Surface Runoff Fraction (the fraction of streamflow contributed by overland flow pathways)
increases strongly for various study areas on the east coast and some other areas, mostly due to
intensification of rainfall events in climate models (see Figure 7-22). Study areas for which the
Surface Runoff Fraction strongly increases, such as ACF and Ariz-San Pedro, are those where
the Low Flow Sensitivity decreases despite relatively small changes in the Dryness Ratio.

Snowmelt Fraction, the fraction of runoff that is due to melting snow (see Figure 7-23) declines
in all watersheds. The strongest percentage declines (in southern and coastal areas) are
somewhat misleading, as these watersheds generally have small amounts of snow. The lesser
percentage declines throughout the Rockies are of greater concern to water management in the
west.

The combination of a greater fraction of surface runoff in many watersheds coupled with
increased dryness and reduced total streamflow in many western watersheds leads to a reduction
in projected Deep Recharge (rates of recharge to deep aquifers) in many study areas (see Figure
7-24).  The risks are estimated to be particularly acute in the Rockies and the ACF basins. In
other areas, increased precipitation in the models counteracts other forces through mid-century,
including the  critical recharge areas in central Nebraska.

7.7. MODELING ASSUMPTIONS AND LIMITATIONS
Model simualtions in the study  provide an improved understanding of streamflow and water
quality sensitivity in different regions  of the United States to a range of plausible mid-21st
century climate change and urban development scnearios. The study also illustrates certain
challenges associated with the use of watershed models for conducting scenario-based studies of
climate change impacts. In the process, this study adds to our knowledge of how to implement
such investigations.

A number of sources of uncertainty must be considered in interpreting results from watershed
hydrologic and water quality simulations of response to climate change—including uncertainty
in the emissions scenario, uncertainty in the GCM simulations of future climate, uncertainty in
the downscaling of these GCM outputs to the  local scale, and uncertainty in the watershed
                                          7-60

-------
models used to translate potential changes in local climate to watershed response. The strong
dependence of streamflow and water quality on climate drivers (e.g., temperature, precipitation,
etc.) means that accurate weather data is necessary to generate accurate estimates of future flow
and water quality conditions.  Inherent in the scenario approach to modeling climate futures is
uncertainty in knowledge of future climate conditions. It is therefore necessary to choose a range
of scenarios that reflect the full, plausible set of future conditions.

Simulation results showed a wide range of watershed responses to differences in climatic
forcing. Results suggest the variability resulting from scenarios based on different methods of
downscaling with a single GCM can be of the same order of magnitude as the variability among
GCMs.  In many cases, simulations for scenarios based on different downscaling approaches
with a single GCM do not agree even in the direction of projected changes relative to current
values.  In part, this issue reflects the skill associated with RCM simulations. A recent study by
Racherla et al. (2012) investigated the value added by using an RCM (the Weather Research and
Forecasting or WRF model) with the GCM GISS-ModelE2 and concluded that the RCM does
not achieve holistic improvement in the simulation of seasonally and regionally averaged surface
temperature or precipitation for historical data.  They further suggested that no strong
relationship exists between skill in capturing climatological means and skill  in capturing climate
change. If RCMs do not add considerable value to the global simulation, the underlying
uncertainties can only be reduced by improving the global-scale climate simulations.

As with any study of this type, simulation results are conditional upon the specific methods,
models, and scenarios used. The simulated range of response in this study is limited by the
particular set of climate model projections available in the NARCCAP archives (the subset of
BCSD projections was selected to match those in the NARCCAP set). For example, all climate
change scenarios evaluated in this study are based on the IPCC A2 greenhouse gas emissions
storyline.  While simulations in this study represent a credible set of plausible future climatic
conditions, the scenarios evaluated should not be considered comprehensive of all possible
futures. A recent summary by Mote et al.  (2011) concludes that ensemble scenarios with a
limited number of projections taken from the full set of available climate models yields results
that differ little from those achieved from larger sets given the current state of science;
furthermore, attempting to preselect the "best" models based on measures of model skill does
little to refine the estimate of central tendency of projected change. Mote et al. recommend a
sample size of approximately ten climate scenarios, which is greater than the six used  in this
study.  Inclusion of additional sources and types of scenarios could alter the  ranges of change
simulated in this study. Similarly, alternative urban and residential development scenarios would
also expand the ensemble range of future responses.

Watershed model simulations developed here also do not consider feedback effects of human
and ecological adaptation to change. In essence, the climate-land use-watershed system is
considered independent of management and adaptation in this study. At the most direct level,
various aspects of human water management such as operation of dams, water use,
transboundary water inputs, and point source discharges are considered fixed at present levels.
In fact, we know these will change. For instance, a warmer climate is likely to result in
increased irrigation withdrawals for crops, while more intense precipitation is likely to result in
changes in operating  rules for dams. In some cases, the models are driven by fixed upstream
boundary conditions (e.g., the Sacramento River model).  There was, however, insufficient
                                          7-61

-------
knowledge of these changes to incorporate them into the scenarios.  The analyses thus provide an
increased understanding of the marginal changes in watershed responses due to potential changes
in climate and urban and residential development, but do not account for the net changes from all
factors, including human use and management of water.

At a more sophisticated level, both natural and human communities are likely to adapt to climate
changes, influencing the watershed response.  The SWAT plant growth model takes into account
the effects of changed climate on plant growth as a function of CO2, temperature, water stress,
and nutrient availability. However, it does not take into account changes in the type of land
cover that may occur as a result of such stresses—either slowly, as through a gradual shifting of
ecological niches, or catastrophically, as might occur through drought-induced forest fires.
Human adaptations that affect watershed processes will also occur.  For example, crop types (or
total area in crops) are likely to change as producers respond to changes in growing season
length and water availability (e.g., Polsky and Easterling, 2001).  Simulation models are not yet
available to provide a credible analysis of such feedback loops at the scale necessary for
evaulating watershed responses.

In addition, many of the modeled study areas are highly managed systems influenced by dams,
water transfers and withdrawals, and point and nonpoint pollution sources.  Given the difficulty
inherent in modeling watershed response at the large spatial scale used in this study, detailed
representation of all management and operational activities was not possible. Results therefore
represent the potential response of watersheds to different change scenarios, but should  not be
considered quantitative forecasts of future conditions.

7.7.1. Model Calibration
Reliably reproducing the baseline period is important for any study of watershed response to
climate change because any biases present in the model calibration are likely to also affect the
future simulations of streamflow (Prudhomme and Davies, 2009), possibly with nonlinear
amplification. The experiences of this project emphasize the importance (and challenges) of
calibration and validation for watershed models. Water quality calibration is particularly
challenging due to limited amounts  of readily available monitoring data. Additional efforts
similar to the one presented here should either focus on watersheds for which well-calibrated
models already exist (and the effort of assembling water quality input and monitoring data from
multiple sources has already been completed) or allocate sufficient time and budget to conduct
detailed, site-specific calibration.

The calibration process can introduce modeler bias, which could be mitigated through use of an
automated model calibration scheme. We avoided this option based on past experience  with the
SWAT and HSPF models in which automated calibration often converges to physically
unrealistic model parameter sets.  It may, however, be advisable to pursue stepwise, guided
model calibration with carefully specified parameter constraints to avoid the effects of user bias,
as was done, for example, in recent USGS simulations of watershed-scale streamflow response
to climate change using the PRMS model (Hay et al., 2011). PRMS, however, only addresses
streamflow and has a much more parsimonious data set than does SWAT or HSPF. Nonetheless,
the advantages of controlling for modeler bias may make use of a semiautomated calibration
procedure desirable.
                                          7-62

-------
The significance of calibration bias is mitigated by focusing on projected changes relative to
baseline conditions as compared to actual future values.  If biases are consistent and linear
between the baseline and future condition, the effect of such biases will tend to cancel out when
relative change is calculated.  There is, however, no guarantee that biases will be linear.  Further
testing to evaluate the effects of alternative model calibrations on the simulated response of
different study areas would be desirable.

7.7.2. Watershed  Model Selection
Simulation results are sensitive to the watershed model applied. In the pilot studies, both HSPF
and SWAT appeared capable of providing similar quality of fit to observed streamflow at the
large basin scale and to pollutant loads at the monthly scale, while HSPF, using a shorter time
step, was better able to resolve streamflow at smaller spatial scales and better able to match
observed concentrations when fully calibrated.  An important result of model comparisons
conducted in this study is the significant effect that increased atmospheric CO2 concentrations
(effects of reduced stomatal conductance that decrease ET) appeared to have on the water
balance.  SWAT's  integrated plant growth model takes this effect into account, whereas HSPF
does not.

It is unclear, however, how well SWAT is able to represent the complex processes affecting
plant growth, nutrient dynamics, and water budgets under changing climate. For example, as
CC>2 levels increase, leaf level reductions in stomatal  conductance and evapotranspiration may be
offset by increased plant growth and leaf area. The effects of CC>2 on plant growth may also be
altered over time due to nutrient limitation (Reich et al., 2006). Further study is required to
better understand how climate change will affect these processes. It should also be noted that
SWAT (as implemented here, using version SWAT2005) has limitations in its representation of a
number of important watershed processes, including simplified simulation of direct runoff using
a curve number approach, erosion prediction with MUSLE that does not fully incorporate
changes in energy that may occur with altered precipitation regimes, and a simplistic
representation of channel erosion processes that appears unlikely to provide a firm foundation for
simulating channel stability responses to climate change.  More recent versions of SWAT
considerably expand the options for simulating  channel erosion, but do not appear to be fully
validated at this time and are limited by the model's use of a daily time step for hydrology.

These considerations suggest that a more sophisticated watershed model formulation, combining
a plant growth model (as in SWAT) with a more detailed hydrologic simulation would be
preferable for evaluating watershed responses to climate change.  However, even if such a model
was available, fully validated, and ready for use, it would likely require a significantly higher
level of effort for model implementation and calibration.

Comparison of change scenarios using HSPF and SWAT suggests one must proceed with
caution when attempting to estimate even relative aggregate impacts at a national scale through
use of watershed models with different underlying formulations. For example, a national
synthesis that drew conclusions from a mix of models, some of which did and  others of which
did not include explicit simulation of effects of increased CC>2 on evapotranspiration, could reach
erroneous conclusions regarding the relative intensity of impacts in different geographical areas.
                                          7-63

-------
                         8.  SUMMARY AND CONCLUSIONS


This report describes watershed modeling in 20 large, U.S. drainage basins (6,000-27,000 mi2 or
15,000-60,000 km2) to characterize the sensitivity of streamflow, nutrient (nitrogen and
phosphorus) loading, and sediment loading to a range of potential mid-21st century climate
futures, to assess the potential interaction of climate change and urbanization in these basins, and
to improve our understanding of methodological  challenges associated with integrating existing
tools (e.g., climate models, downscaling approaches, and watershed models) and data sets to
address these scientific questions.  Study areas were selected to represent a range of geographic,
hydroclimatic, physiographic, land use, and other watershed attributes. Other important criteria
used in site selection included the availability of necessary data for calibration and validation of
watershed models, and opportunities for leveraging the availability of preexisting watershed
models.

Models were configured by subdividing study areas into modeling units, followed by continuous
simulation of streamflow and water quality for these units using meteorological, land use, soil,
and stream data.  A unique feature of this study is the use of a consistent watershed modeling
methodology and a common set of climate and land-use change scenarios in multiple locations
across the nation. Models in each  study area are  developed for current (1971-2000) observed
conditions, and then used to simulate results under a range of potential mid-21st century
(2041-2070) climate change and urban development scenarios. Watershed modeling was
conducted at each study location using the SWAT model and six climate change  scenarios based
on dynamically downscaled (50 x  50 km2) output from four of the GCMs used in the IPCC 4
Assessment Report for the period 2041-2070 archived by the NARCCAP.  Scenarios were
created by adjusting historical weather series to represent projected changes in climate using a
change factor approach. To explore the potential interaction of climate change and urbanization,
simulations also include urban and residential development scenarios for each of the 20 study
watersheds. Urban and residential development scenarios were acquired from EPA's national-
scale ICLUS project.

In a subset of five study areas (the Minnesota River, the Susquehanna River, the
Apalachicola-Chattahoochee-Flint, the Salt/Verde/San Pedro, and the Willamette River Basins),
additional simulations were conducted to assess the variability in simulated watershed response
resulting from use of different watershed models and different approaches for downscaling GCM
climate change scenarios. In these study areas, watershed simulations were also run with eight
additional scenarios derived from the same four GCMs used in NARCCAP: four scenarios
interpolated to station  locations directly from the GCM output, and four scenarios based on
BCSD statistically downscaled climate projections described by Maurer et al. (2007).  In
addition, in these five  study areas,  all scenario simulations were run independently with a second
watershed simulation model, the HSPF.

Given the large size of study areas, calibration and validation of all models was completed by
first focusing on  a single HUC-8 within the larger study area (preferably one with a good record
of streamflow gaging and water quality monitoring data), and then extending the calibration to
adjacent areas with modifications as  needed to achieve a reasonable fit at multiple spatial scales.
                                           3-1

-------
Large-scale GCM projections are generally consistent in showing a continued warming trend
over the next century (although with sometimes significant regional-scale disagreements in the
magnitude of this warming), but offer a much wider range of plausible outcomes in other aspects
of local climate—particularly the timing and intensity of precipitation and the energy inputs (in
addition to air temperature) that determine potential evapotranspiration—that interact to create
watershed responses.

The simulated watershed responses to these changes provide an improved understanding of
system sensitivity to potential climate change and urban development scenarios in different
regions of the country and provide a range of plausible future hydrologic and water quality
change scenarios that can be applied in various planning and scoping frameworks. The results
illustrate a high degree of regional variability in the response of different streamflow and water
quality endpoints to a range of potential mid-21st century climatic conditions in different regions
of the nation. Watershed hydrologic response is determined by the interaction of precipitation
and evapotranspiration, while water quality response is largely dependent on hydrology.
Comparison of simulations in all 20 study areas for the 2041-2070 time horizon suggest
potential streamflow volume decrease in the Rockies and interior southwest, and increases in the
east and southeast coasts.  Wetter winters  and earlier snowmelt are likely in many of the northern
and higher elevation watersheds.  Higher peak flows will also increase erosion and sediment
transport; nitrogen and phosphorus loads are also likely to increase in many watersheds.

Both the selection of an underlying GCM and the choice of downscaling method have a
significant influence on the streamflow and water quality simulations. In many cases, the range
of simulated responses across the different climate models and downscaling methodologies do
not agree in direction.  The ultimate significance of any given simulation of future change will
depend on local context, including the historical range of variability, thresholds and management
targets, management options,  and interaction with other stressors. The simulation results in this
study do, however, clearly illustrate that the potential streamflow and water quality response in
many areas could be large.

Watershed simulations were run in all study areas with and without projected mid-21st century
changes in urban and residential development. These results suggest that at the HUC-8 spatial
scale evaluated in this study, watershed sensitivity to projected urban and residential
development will be small relative to the changes resulting from climate change. It is important,
however, to qualify  this result. The finest spatial scale reported in this study is that of an 8-digit
HUC, and most urbanized areas are located on larger rivers downstream of multiple 8-digit
HUCs. Over the whole of individual study areas, urban and residential growth scenarios
represented changes in the amount of developed land on the order of <1 to about 12% of total
watershed area and increases in impervious surfaces on the order of 0 to 5% of total watershed
area.  The effects of urban development on adjacent water bodies at higher levels of development
are well documented.  It is thus likely that at smaller spatial scales within study areas where the
relative fraction of developed land is greater, the effects of urbanization will be greater.
Identifying the scale at which urbanization effects become comparable to the effects of a
changing climate is  an important topic for future research.

The simulation results also illustrate a number of methodological issues related to impacts
assessment modeling.  These include the sensitivities and uncertainties associated with use  of
                                           8-2

-------
different watershed models, different approaches for downscaling climate change simulations
from global models, and the interaction between climate change and other forcing factors, such
as urbanization and the effects of changes in atmospheric CO2 concentrations on
evapotranspiration.  Uncertainty associated with differences in emission scenarios and climate
model sensitivities is well known and widely discussed in previous assessments of climate
change impacts on water (e.g., IPCC, 2007; Karl et al., 2009). This study illustrates a potentially
significant additional sensitivity of watershed simulations to the method selected for
downscaling GCM model output.  Results of the intercomparison of climate change data sets
suggest that the variability between downscaling of a single GCM with different RCMs can be of
the same order of magnitude as the ensemble variability between GCMs.

This study also suggests potentially important sensitivity  of results to the use of different
hydrologic models (HSPF and SWAT in this study), associated with differences in process
representation, such as accounting for the influence of increased atmospheric CO2 on
evapotranspiration.  One notable insight from these results is that, in many watersheds, climate
change (when precipitation amount and/or intensity is altered), increasing urbanization, and
increasing atmospheric CC>2 can have similar or additive effects  on  streamflow and pollutant
loading (e.g., a more flashy runoff response with higher high flows and lower low flows). The
results, while useful as guidance for designing and conducting similar impacts assessment
studies, are only a first step in understanding what are likely highly complex and context-
dependent relationships. Further study and evaluation of the implications of these and other
questions  is necessary for improving the plausibility and relevance of coupled climate-hydrology
simulations, and ultimately for informing resource managers and climate change adaptation
strategies.

The model simulations in this study contribute to a growing understanding of the complex and
context-dependent relationships between climate change,  land development, and water in
different regions of the nation.  As a first order conclusion, results indicate that in many locations
future conditions are likely to be different from past experience.  In the context of decision
making, being aware and planning for this uncertainty is preferable to accepting a position that
later turns out to be incorrect. Results also provide a plausible envelope on the range of
streamflow and water quality responses to mid-21st century climate change and urban
development in different regions of the nation. In addition, in many study areas the simulations
suggest a likely direction of change of streamflow and water quality endpoints.  This information
can be useful in planning for anticipated but uncertain future conditions.  The sensitivity studies
evaluating different methodological choices help to improve the scientific foundation for
conducting climate change impacts assessments, thus building the capacity of the water
management community to understand and respond to climate change.

Understanding and responding to climate change is complex, and this study is only an
incremental step towards fully addressing these questions. It must be stressed that results are
conditional upon the methods, models, and scenarios used in this study.  Scenarios represent a
plausible range but are not comprehensive of all possible  futures. Several of the study areas are
also complex, highly managed systems; all infrastructure  and operational aspects of water
management are not represented in full detail.  Finally, changes in agricultural practices, water
demand, other human responses, and natural ecosystem changes such as the prevalence of forest
fire (e.g., Westerling et al., 2006) or plant disease that will influence streamflow and water
                                           8-3

-------
quality are not considered in this study. Further study is required to continue to build the
scientific foundation for assessing these and other questions relevant to the scientific and
watershed management communities.

Successful climate change adaptation strategies will need to encompass practices and decisions
to reduce vulnerabilities across a wide range of plausible future climatic conditions. Where
system thresholds are known, knowledge of the range of potential changes can help to identify
the need to consider future climate change in water planning. Many of these strategies might
also help reduce the impacts of other existing stressors.  It is the ultimate goal of this study to
build awareness of the potential range of future watershed response so that where simulations
suggest large and potentially disruptive changes, the management community will respond to
build climate resiliency.
                                           8-4

-------
                                    REFERENCES
Ainsworth, EA; Rogers, A. (2007) The response of photosynthesis and stomatal conductance to
       rising [CC>2]: mechanisms and environmental interactions.  Plant Cell Environ
       30:258-270.

Alexander, LV; Zhang, X; Peterson, TC; et al. (2006) Global observed changes in daily climate
       extremes of temperature and precipitation. J Geophys Res 111(D05109). doi:
       10:1029/2005JD006290.

Alexander, RB; Smith, RA; Schwarz, GE; et al. (2008) Differences in phosphorus and nitrogen
       delivery to the Gulf of Mexico from the Mississippi River Basin. Environ Sci Technol
       42(3):822-830.

Allan, RP; Soden, BJ. (2008) Atmospheric warming and the amplification of precipitation
       extremes.  Science 321(5895): 1481-1484.

Allan, RJ. (1986) The role of particulate matter in the fate of contaminants in aquatic
       ecosystems.  Scientific Series 142. Inland Waters Directorate, Environment Canada,
       Ottawa.

Allen, RG; Pereira, LS; Raes, D; et al. (1998) Crop evapotranspiration: guidelines for computing
       crop water requirements.  Irrigation and Drainage Paper 56. Food and Agriculture
       Organization of the United Nations, Rome. Available online at http://www.engr.scu.edu/
       ~emaurer/classes/ceng!40_watres/handouts/F AO_56_Evapotranspiration.pdf

Allen, RG; Walter, IA; Elliott, RL; et al. (2005) The ASCE standardized reference
       evapotranspiration equation. Reston, VA: American Society of Civil Engineers.

Anandhi, A; Frei, A; Pierson DC; et al. (2011) Examination of change factor methodologies for
       climate change impact assessment. Water Resour Res
       47(W03501):doi:10.1029/2010WR009104.

Baker, DB; Richards, P; Loftus, TT; et al. (2004) A new flashiness index: characteristics and
       applications to Midwestern rivers and streams. J Am Water Resour Assoc
       40(2):503-522.

Berg, EE; Henry, JD; Fastie, CL; et al. (2006) Spruce beetle outbreaks on the Kenai Peninsula,
       Alaska, and Kluane National Park and Reserve, Yukon Territory: relationship to summer
       temperatures and regional differences in disturbance regimes.  For Ecol Manage
       227:219-232.

Bernacchi, CJ; Kimball, BA;  Quarles, DR; et al.  (2007) Decreases in stomatal conductance of
       soybean under open-air elevation of [CO2] are closely coupled with decreases in
       ecosystem evapotranspiration.  Plant Physiol 143:134-144.
                                          R-l

-------
Bicknell, BR; Imhoff, JC; Kittle JL, Jr; et al. (2001) Hydrological simulation program—Fortran
       (HSPF). User's manual for release  12. U.S. Environmental Protection Agency, National
       Exposure Research Laboratory, Athens, GA, in cooperation with U.S. Geological Survey,
       Water Resources Division, Reston,  VA.

Bicknell, BR; Imhoff, JC; Kittle JL, Jr; et al. (2005) HSPF version 12.2 user's manual.  U.S.
       Environmental Protection Agency, National Exposure Research Laboratory, Office of
       Research and Development, Athens, GA.

Bosch, NS; Allan, JD; Dolan, DM; et al.  (2011) Application of the Soil and Water Assessment
       Tool for six watersheds of Lake Erie: model parameterization and calibration. J Great
       Lakes Res 37:263-271.

Bouraoui, F; Galbiati, L; Bidoglio,  G. (2002) Climate change impacts on nutrient loads in the
       Yorkshire Ouse catchment (UK).  Hydrol Earth Syst Sci 6:197-209.

Burwell, RE;  Timmons, DR; Holt, RF. (1975) Nutrient transport in surface runoff as influenced
       by soil cover and seasonal periods.  Soil Sci Soc Am J 39(3):523-528.

Cao, L; Bala,  G; Caldeira, K; Nemani, R; Ban-Weiss, G. (2010) Importance of carbon dioxide
       physiological forcing to future climate change. PNAS 107(21):9513-9518.

CECWG (Commission for Environmental Cooperation Working Group). (1997) Ecological
       regions of North America—toward  a common perspective. Commission for
       Environmental Cooperation, Montreal. Available online at http://www.calepa.ca.gov/
       education/eei/Curriculum/Gradel 1/1186/1186EnvGuide.pdf

Chang, H; Evans, BM; Easterling, DR. (2001) The effects of climate change  on streamflow and
       nutrient loading. J Am Water Resour Assoc 37(4):973-985.

Cox, P; Stephenson, D. (2007) A changing climate for prediction.  Science 317(5835):207-208.

Crawford, NH; Linsley, RK. (1966) Digital simulation in hydrology: Stanford watershed model
       IV.  Technical Report 39. Department of Civil Engineering, Stanford University, CA.
       Available online at http://www.hydrocomp.com/publications/StanfordModelIV.PDF

Dai, A. (2006) Precipitation characteristics in eighteen coupled climate models.  J Clim
       19:4605-4630.

Donigian, AS, Jr. (2000) HSPF training workshop handbook and CD. Lecture #19: calibration
       and  verification issues. U.S. Environmental Protection Agency Headquarters,
       Washington Information Center, 10-14 January, 2000. Prepared for U.S. Environmental
       Protection Agency, Office of Water, Office of Science and Technology, Washington, DC.

Duda, P; Kittle, J, Jr; Gray, M; et al. (2001) WinHSPF Version 2.0: an interactive Windows
       interface to HSPF (WinHSPF) user's manual.  AQUA TERRA Consultants, Decatur, GA.
       Available online at http://www.epa.gov/athens/research/ftable/toc_intr.pdf
                                          R-2

-------
Easterling, WE; Rosenberg, NJ; McKenney, MS; et al. (1992) Preparing the erosion productivity
       impact calculator (EPIC) model to simulate crop response to climate change and the
       direct effects of CO2.  Agric For Meteorol 59(1-2): 17-34.

Emori, S; Brown, SJ. (2005) Dynamic and thermodynamic changes in mean and extreme
       precipitation under changed climate. Geophys Res Lett 32(L17706)
       doi: 10.1029/2005GL023272

Ficklin, DL; Luo, Y; Luedeling, E; et al. (2009) Climate change sensitivity assessment of a
       highly agricultural watershed using SWAT. J Hydrol 374(1/2): 16-29.

Follett, RF. (1995) Fate and transport of nutrients: nitrogen. Working Paper No. 7. U.S.
       Department of Agriculture, Agricultural Research Service, Soil-Plant-Nutrient Research
       Unit, Fort Collins, CO. Available  online at http://www.nrcs.usda.gov/wps/portal/nrcs/
       detail/national/technical/? cid=nrcs!43_014202

Garen, DC; Moore, DS. (2005) Curve number hydrology in water quality modeling: uses,
       abuses, and future directions.  J Am Water Resour Assoc 41(2):377-388.

Garrick, M; Cunnane, C; Nash, JE. (1978) A criterion of efficiency for rainfall-runoff models.  J
       Hydrol 36(3-4):375-381.

Gassman, PW; Reyes, MR; Green, CH; et al. (2007) The Soil and Water Assessment Tool:
       historical development, applications, and future research directions.  Trans Am Soc Agri
       BiolEng 50(4): 1211-1250.

Gesch, D; Oimoen, M; Greenlee, S; Nelson, C;  Steuck, M; Tyler, D. (2002) The national
       elevation dataset. Photogramm Eng Remote Sens 68(1):5-11.

Gleckler, PJ; Taylor, KE; Doutriaux, C. (2008) Performance metrics for climate models. J
       Geophys Res 113(D06104): doi:10.1029/2007JD008972.

Groisman, PY; Knight, RW; Karl, TR. (2012) Changes in intense precipitation over the central
       United States. J Hydrometeorol 13:47-66.

Gutowski, WJ; Hegerl, GC; Holland,  GJ; et al. (2008) Causes of observed changes in extremes
       and projections of future changes.  In: Karl, TR; Meehl, GA; Miller, CD; et al; eds.
       Weather and climate extremes in a changing climate.  Regions of focus: North America,
       Hawaii, Caribbean,  and U.S. Pacific Islands. Report by the U.S. Climate Change Science
       Program. Global Change Research, Washington, DC; p. 81-116. Available online at
       http://www.agci.org/dB/PDFs/Publications/07Sl_USCCSP.pdf

Hawkins, E; Sutton, R. (2009) The potential to narrow uncertainty in regional climate
       predictions. Bull Am Meteor  Soc  90:1095-1107.

Hay, LE; Markstrom, SL; Ward-Garrison, C. (2011) Watershed-scale response to climate change
       through the twenty-first century for selected basins across the United States. Earth
       Interact 15:1-37.

-------
Homer, C; Huang, C; Yang, L; et al. (2004) Development of a 2001 national landcover database
      for the United States. Photogramm Eng Remote Sens 70(7):829-840.

Homer, C; Dewitz, J; Fry, J; et al. (2007) Completion of the 2001 national land cover database
      for the conterminous United States.  Photogramm Eng Remote Sens 73(4):337-341.
      Available online at http://www.epa.gov/mrlc/pdf/april-07-highlight.pdf

Kurd, B; Leary, N; Jones, R; et al. (1999) Relative regional vulnerability of water resources to
      climate change.  J Am Water Resour Assoc 35(6): 1399-1409.

IPCC (Intergovernmental Panel on Climate Change). (2001) Climate change 2001: the scientific
      basis.  In: Houghton, JT; Ding, Y; Griggs, DJ; et al.; eds. Intergovernmental panel on
      climate change: working group I. Cambridge, UK: Cambridge University Press.

IPCC (Intergovernmental Panel on Climate Change). (2007) Climate change 2007: synthesis
      report—summary for policymakers. Available online at:
      http://www.ipcc.ch/pdf/assessment-report/ar4/syr/ar4_syr_spm.pdf

Jensen, ME; Burman, RD; Allen, RG. (1990) Evapotranspiration and irrigation water
      requirements. ASCE Manuals and Reports on Engineering Practice No. 70. ASCE, NY.

Johnson, T; Weaver, C.  (2009) A framework for assessing climate change impacts on water and
      watershed systems. Environ Manage 43:118-134.

Karl, TR; Knight, RW. (1998) Secular trends of precipitation amount, frequency, and intensity in
      the United States. Bull Am Meteorol Soc 79(2):231-241.

Karl, TR; Melillo, JM; Peterson, TC; eds. (2009) Global climate change impacts in the United
      States.  Cambridge, UK: Cambridge University Press.

Kharin, VV; Zwiers, FW; Zhang, X; et al. (2013) Changes in temperature and precipitation
      extremes in the CMIP5 ensemble. Clim Change 119(2):345-357. DOI 10.1007/sl0584-
      013-0705-8

Kundzewicz, ZW; Mata, LJ; Arnell, NW; et al. (2007) Freshwater resources and their
      management. In: Parry, ML; Canziani, OF; Palutikof, JP; et al; eds. Climate change
      2007: impacts, adaptation and vulnerability. Contribution of Working Group II to the
      Fourth Assessment Report of the Intergovernmental Panel on Climate Change.
      Cambridge, UK: Cambridge University Press; pp. 173-210.

Leakey, ADB; Ainsworth, EA; Bernacchi, CJ; et al. (2009) Elevated CC>2 effects on plant
      carbon, nitrogen, and water relations: six important lessons from FACE.  J Exp Bot
      60(10):2859-2876.

Legates, DR; McCabe, GJ, Jr. (1999) Evaluating the use of "goodness-of-fit" measures in
      hydrologic and hydroclimatic model validation.  Water Resour Res 35(1):233-241.
                                         R-4

-------
Lempert, RJ; Groves, DG; Popper, SW; et al. (2006) A general, analytic method for generating
       robust strategies and narrative scenarios.  Manage Sci 52(4):514-528.

Lumb, AM; McCammon, KB; Kittle, JL, Jr. (1994) User's manual for an expert system
       (HSPEXP) for calibration of the hydrological simulation program-FORTRAN. U.S.
       Geological Survey Water Resources Investigation Report 94-4168.  U.S. Geological
       Survey, Reston, VA.

Luo, Y; Ficklin, DL; Liu, X; et al. (2013) Assessment of climate change impacts on hydrology
       and water quality with a watershed modeling approach.  Sci Total Environ
       450-451:72-82.

Marshall, E; Randhir, T. (2008) Effect of climate change on watershed system:  a regional
       analysis. Clim Change 89(3/4):263-280.

Maurer, EP; Brekke, L; Pruitt, T; et al. (2007) Fine-resolution climate projections enhance
       regional climate change impact studies. Eos Trans Am Geophys Union  88(47):504.

Mearns, L. (2009) The North American regional  climate change assessment program
       (NARCCAP): overview of phase II results.  Earth Environ Sci Trans 6(022007):
       doi: 10.1088/1755-1307/6/2/022007.

Mesinger, F; DiMego, G; Kalnay, E; et al. (2006) North American regional reanalysis. Bull Am
       Meteorol Soc 87(3):343-360.

Milly, PCD; Dunne, K; Vecchia, A. (2005) Global pattern of trends in streamflow and water
       availability in a changing climate. Nature 438:347-350.

Milly, PCD; Betancourt, J; Falkenmark, M; et al. (2008) Stationarity is dead: Whither water
       management? Science 319:573-574.

Monteith, DT; Evans, CD; Reynolds, B. (2000) Are temporal variations in the nitrate content of
       UK upland freshwaters linked to the North Atlantic Oscillation? Hydrol Processes
       14(10):1745-1749.

Monteith, JL. (1965) Evaporation  and the environment. In: The state and movement of water in
       living organisms. XlXth Symposium, Swansea. Soc Exp Biology, Vol 19. Cambridge,
       UK. Cambridge University Press; pp. 205-234.

Moriasi, DN;  Arnold, JG; Van Liew, MW; et al.  (2007) Model evaluation guidelines for
       systematic quantification of accuracy in watershed simulations.  Trans Am Soc Agri Biol
       Eng 50(3):885-900.

Mote, P; Brekke, L; Duffy, PB; et al.  (2011) Guidelines for constructing climate scenarios. Eos
       Trans Am Geophys Union 92(31):257-258.

Murdoch, PS; Baron, JS; Miller, TL. (2000) Potential effects of climate change  on surface-water
       quality in North America.  J Am Water Resour Assoc 36(2):347-366.
                                         R-5

-------
Nash, JE; Sutcliffe, JV. (1970) River flow forecasting through conceptual models part I—a
       discussion of principles. JHydrol 10(3):282-290.

Neitsch, SL; Arnold, JG; Kiniry, JR; et al. (2005) Soil and water assessment tool, theoretical
       documentation.  Grassland, Soil and Water Research Laboratory, U.S. Department of
       Agriculture, Agricultural Research Service, Temple, TX. Available online at
       http://swat.tamu.edu/media/1292/swat2005theory.pdf

Paul, MJ; Meyer, JL. (2001) Streams in the urban landscape. Annu Rev Ecol Syst 32:333-365.

Poff, NL; Tokar, S; Johnson, P. (1996) Stream hydrological and ecological responses to climate
       change assessed with an artificial neural network. Limnol Oceanogr 41(5): 857-863.

Polsky, C; Easterling, WE. (2001) Adaptation to climate variability and change in the U.S.  Great
       Plains: a multi-scale analysis of Ricardian climate sensitivities.  Agric Ecosyst Environ
       85(1-3):133-144.

Preston, SD; Bierman, VJ, Jr; Silliman, SE. (1989) An evaluation of methods for the estimation
       of tributary mass loads.  Water Resour Res 25(6): 1379-1389.

Prudhomme, C; Davies, H. (2009) Assessing uncertainties in climate change impact analyses on
       the river flow regimes in the UK. Part 1: baseline climate.  Clim Change
       93(1-2):177-195.

Pyke, C; Warren, M; Johnson, T; et al. (2011) Assessment of low impact development for
       managing stormwater with changing precipitation due to climate change. Landsc Urban
       Plan 103(2): 166-173.

Racherla, PN;  Shindell, DT; Faluvegi, GS. (2012) The added value to global model projections
       of climate change by dynamical downscaling: a case study over the continental U.S.
       using the GISS-ModelE2 and WRF models.  J Geophys Res-Atmos  117(D20118):
       doi:10.1029/2012JD018091.

Raisanen, J. (2007) How reliable are climate models? Tellus A 59:2-29.

Reich, PB; Hungate, BA; Luo, Y. (2006) Carbon-nitrogen interactions in terrestrial ecosystems
       in response to rising atmospheric carbon dioxide.  Annu Rev Ecol Evol Syst 37:611-636.

Sarewitz, D; Pielke,  RA, Jr; Byerly, R, Jr. (2000) Introduction: death, taxes, and environmental
       policy. In: Sarewitz, D; Pielke, RA, Jr; Byerly,  R, Jr; eds. Prediction: science, decision
       making, and the future of nature. Washington, DC: Island Press.

SCS (Soil Conservation Service). (1972) Hydrology guide for use in watershed planning.
       National engineering handbook, section 4: hydrology, supplement A. U.S. Department of
       Agriculture, Natural Resources Conservation Service, Washington, DC.

Seaber, PR; Kapinos, FP; Knapp, GL.  (1987) Hydrologic unit maps. Water-Supply Paper 2294.
       U.S. Geological Survey, Denver, CO.
                                          R-6

-------
Sharpley, AN; Williams, JR; eds. (1990) EPIC—erosion productivity impact calculator, 1.
       Model documentation.  U.S. Department of Agriculture, Agricultural Research Service.
       Tech Bull 1768.

Stainforth, DA; Allen, MR; Tredger, ER; et al. (2007) Confidence, uncertainty, and decision-
       support relevance in climate predictions. Phil Trans R Soc A 365:2145-2161.

Stockle, CO; Williams, JR; Rosenberg, NJ; et al. (1992) A method for estimating the direct and
       climatic effects of rising atmospheric  carbon dioxide on growth and yield of crops: part
       1—modification of the EPIC model for climate change analysis. Agric Syst 38:225-238.

Suddick, EC; Davidson, EA; eds. (2012) The role of nitrogen in climate change and the impacts
       of nitrogen-climate interactions on terrestrial and aquatic ecosystems, agriculture, and
       human health in the United States. Technical report submitted to the U.S. National
       Climate Assessment. North American Nitrogen Center of the International Nitrogen
       Initiative (NANC-INI), Woods Hole Research Center,  Falmouth, MA. Available online at
       http://www.whrc.org/resources/publications/pdf/SuddicketalWHRC.12.pdf

Sun, Y; Solomon, S; Dai, A; et al. (2006) How often does it rain? J Clim 19:916-934.

SWCS (Soil and Water Conservation Society). (2003) Conservation implications of climate
       change: soil erosion and runoff from cropland.  Ankeny, IA: Soil and Water Conservation
       Society.

TAMU (Texas A&M University). (2010) SWAT—soil and water assessment tool.  Available
       online at http://swatmodel.tamu.edu/software/arcswat.

Taner, MU; Carleton, JN; Wellman, M. (2011) Integrated model projections of climate change
       impacts on a North American lake.  Ecol Model 222(18):3380-3393.

Tetra Tech. (2008a) Quality assurance project plan for watershed modeling to evaluate potential
       impacts of climate and land-use change on the hydrology and water quality of major U.S.
       drainage basins. Prepared for the Office of Research and Development Global Change
       Research Program, U.S. Environmental Protection Agency, Washington, DC.

Tetra Tech. (2008b) Minnesota River basin turbidity TMDL and Lake Pepin excessive nutrient
       TMDL, model calibration and validation report. Prepared for Minnesota Pollution
       Control Agency, St. Paul, MN.

Tong, STY; Sun, Y; Ranatunga, T; et al. (2011) Predicting plausible impacts of sets of climate
       and land-use change scenarios on water resources.  Appl Geogr 32(2):477-489.

Trenberth, KE; Dai, A; Rasmussen, RM; et al. (2003)  The changing character of precipitation.
       Am Meteorol Soc 84:1205-1217.

Trenberth, KE; Jones, PD; Ambenje, P; et al.  (2007) Observations: surface and atmospheric
       climate change.  In: Solomon, S;  Qin, D; Manning, M; et al; eds. Climate change 2007:
       the physical science basis. Contribution of working group I to the fourth assessment
                                          R-7

-------
       report of the intergovernmental panel on climate change. Cambridge, UK and New York:
       Cambridge University Press.

Tu, J. (2009) Combined impact of climate and land-use changes on streamflow and water quality
       in eastern Massachusetts, USA. J Hydrol 379(3-4):268-283.

USDA (United States Department of Agriculture). (1991) State soil geographic (STATSGO) data
       base; data use information. Miscellaneous Publication 1492. National Soil Survey
       Center, Natural Resources Conservation Service, Fort Worth, TX.

U.S. EPA (Environmental Protection Agency). (1984) Report to congress: nonpoint source
       pollution in the U.S. Office of Water Program Operations, Water Planning Division,
       Washington, DC.

U.S. EPA (Environmental Protection Agency). (2000) Estimating hydrology and hydraulic
       parameters for HSPF.  BASINS Technical Note 6.  Office of Water, Washington, DC;
       EPA-823-ROO-012. Available online at http://water.epa.gov/scitech/datait/models/basins/
       upload/2000_08_14_BASINS_tecnote6.pdf

U.S. EPA (Environmental Protection Agency). (2001) BASINS version 3.0 user's manual.
       Office of Water, Washington, DC; EPA-823-B-01-001. Available online at
       http://water.epa.gov/scitech/datait/models/basins/upload/2009_04_03_BASINS_b3docs_
       usermanual.pdf

U.S. EPA (Environmental Protection Agency). (2002) Nitrogen: multiple and regional impacts.
       Clean Air Markets Division, Washington, DC; EPA-430-R-01-006. Available online at
       http ://www. epa.gov/airmarkets/resource/docs/nitrogen.pdf

U.S. EPA (Environmental Protection Agency). (2008) Using the BASINS meteorological
       database—version 2006. BASINS Technical Note 10. Office of Water, Washington,
       DC. Available online at http://water.epa.gov/scitech/datait/models/basins/upload/
       2009_04_13_BASINSs_tecnotel0.pdf

U.S. EPA (Environmental Protection Agency). (2009a) BASINS 4.0—fact sheet.  Office of
       Water, Wasthingtion, DC. Available online at
       http://www.epa.gov/waterscience/BASINS/fs-basins4.html (accessed January 27, 2010).

U.S. EPA (Environmental Protection Agency). (2009b) BASINS 4.0 climate assessment tool
       (CAT): supporting documentation and user's manual. Global Change Research Program,
       National Center for Environmental Assessment, Office of Research and Development,
       Washington, DC; EPA-600-R-08-088F. Available online at http://cfpub.epa.gov/ncea/
       cfm/recordisplay.cfm?deid=203460

U.S. EPA (Environmental Protection Agency). (2009c) ICLUS V1.2 user's manual: ArcGIS
       tools and datasets for modeling US housing density growth. Global Change Research
       Program, National Center for Environmental Assessment, Office of Research and
       Development, Washington, DC; EPA-600-R-09-143A.
                                         R-8

-------
U.S. EPA (Environmental Protection Agency). (2010) NHDPlus user guide. Office of Water,
       Washington, DC. Available online at ftp://ftp.horizon-systems.com/NHDPlus/
       NHDPlus V1 /documentation/NHDPLUS V1 JJserGuide. pdf

USGS (United States Geological Survey). (1982) Guidelines for determining flood flow
       frequency. Bulletin #17B of the Hydrology Subcommittee, Interagency Advisory
       Committee on Water Data, Reston, VA. Available online at http://water.usgs.gov/
       osw/bulletin!7b/dl_flow.pdf

Volkery, A; Ribeiro, T. (2009) Scenario planning in public policy: understanding use, impacts
       and the role of institutional context factors.  Technol Forecast Soc Change
       76(9): 1198-1207.

Walsh, CJ; Roy, A; Feminella, J; et al. (2005) The urban stream syndrome: current knowledge
       and the search for a cure. J N Am Benthol Soc 24(3):706-723

Wang, S-Y; Gillies, RR; Takle, ES; et al. (2009) Evaluation of precipitation in the Intermountain
       Region as simulated by the NARCCAP regional climate models. Geophys Res Lett
       36(L11704)doi:10:1029/2009GL037930.

Westerling, AL; Hidalgo, HG; Cayan, DR; et al. (2006) Warming and earlier spring increases
       western U.S. forest wildfire activity.  Science 313(5789):940-943.

Whitehead, PG; Wilby, RL; Battarbee, RW; et al. (2009) A review of the potential impacts of
       climate change on surface water quality. Hydrol Sci 54:101-123.

Wilcox, BP; Rawls, WJ; Brakensiek, DL; et al. (1990) Predicting runoff from rangeland
       catchments: a comparison of two models. Water Resour Res 26(10):2401-2410.

Williams, J.R.  (1975) Sediment-yield prediction with universal equation using runoff energy
       factor.  In: Present and prospective technology for predicting sediment yield and sources:
       proceedings of the sediment-yield workshop, November 28-30, 1972.  U.S. Department
       of Agriculture Sedimentation Lab, Oxford, MS. ARSS-40; pp. 244-252. .

Williams, MW; Losleben, M; Caine, N; et al. (1996) Changes in climate and hydrochemical
       responses in a high-elevation catchment in the Rocky Mountains, USA. Limnol
       Oceanogr 41(5):939-946.

Wilson, CO; Weng, Q. (2011) Simulating the impacts of future land use and climate changes on
       surface water quality in the Des Plaines River watershed, Chicago Metropolitan
       Statistical Area, Illinois.  Sci Total Environ 409(20):4387-4405.

Winchell, M; Srinivasan, R; DiLuzio, M; et al. (2008) ArcSWAT 2.1 interface for SWAT 2005,
       user's guide. U.S.  Department of Agriculture, Agricultural Research Service, Temple,
       TX. Available online at http://www.geology.wmich.edu/sultan/5350/Labs/
       ArcSWAT_Documentation.pdf
                                         R-9

-------
United States
Environmental Protection
Agency
National Center for Environmental Assessment (8601)
Office of Research and Development
Washington, DC 20460

-------
Appendix  A
Model  Setup  Process
  1   Overview
 This memorandum describes the protocol to ensure an efficient and consistent model setup process
 to implement the SWAT and HSPF models for the 20 Watersheds study areas.

 Modeling in this study addresses large study areas, with an emphasis on relative change for future
 conditions. A simplified approach is used for land use and soils coverages to the extent possible to make
 this efficient. SWAT setup and calibration starts from a common land use platform representing current
 (calibration) conditions.
 Simplifying principles include the following:

              Optimize for automated processing, taking advantage of features already built into
              ArcSWAT.
              Calibrate first for one 8-digit HUC, then extend to whole  study area. The calibration
              HUC should be selected in an area with the greatest availability of hydrology and water
              quality calibration data.  Avoid selecting a calibration HUC8 with complicating features,
              such as large reservoirs.
              SWAT is set up in the usual way, using an HRU overlay of land use and STATSGO
              soils.
              Use only weather stations already processed for BASINS 4 and supplied to the team.
              Account for sparse coverage by using elevation bands in areas of high relief.
              Existing impervious area is identified based on NLCD products.
              HSPF setup is based on the same spatial coverages as SWAT but requires additional
              processing in the WinHSPF interface.

              HSPF is developed only for the five pilot watersheds.
                                          A-l

-------
2   Processing  Steps
2.1  INITIAL SETUP
Initial GIS processing has been performed by Tt in ArcGIS for the whole study area. Processed GIS
inputs are then used in ArcSWAT (which runs as an extension in ArcGIS).

Watershed Boundaries and Reach Hydrography

Subbasin boundaries and reach hydrography have been created from NHDPlus. In general these
delineations should be sufficient for application and isolation of appropriate calibration points.  Note that
many river "basins" have multiple outlets, notably those adjacent to the Great Lakes or ocean shorelines.
DEM

Mosaic-ed DEMs are supplied for the full extent of the model watershed area. ArcSWAT will create a
slope grid during model setup. For  some of the study areas (notably those with significant shoreline
adjacent to an ocean or Great Lake), the DEM extent does not fully overlap the subbasin extent, due to
differences in shoreline representation in the parent spatial files. If the study area has shoreline, or if the
SWAT "subbasin delineation" process fails, we recommend the following ArcGIS procedure to fill the
DEM NoData "holes".

    •  In the Single Output Map Algebra tool, use the following statement:
       con(grid>0,grid,0)
       where grid is the name of the DEM.

       Before executing the tool, in the Environments section set the extent to the subbasin shapefile.

       The procedure will replace NoData cells with 0 inside the subbasin extent, but not beyond. The
       error associated with using 0 is likely to be minimal and not adversely affect model setup.
Land Cover and Soils

    1.  Clipped NLCD 2001 Land Cover is provided for the extent of the model watershed area. The
       grid files have been modified to include four reclassified cells in each subbasin. These cells and
       their values are used as placeholders for updating developed land cover in future scenarios. You
       must use the land cover grid ending in "_sw" for SWAT applications.

    2.  NLCD 2001 Urban Imperviousness bas been analyzed to calculate developed class impervious
       area. This is provided in an Excel spreadsheet.
    3.  STATSGO soils grids are provided with ArcSWAT.
                                             A-2

-------
Other Data Provided

    4.  BASINS4 weather stations in proximity of model watersheds.  Weather stations from this set
       must be used to enable climate updates for the scenarios. The locations are provided within the
       SWAT precipitation and air temperature input files on the project FTP site. ArcSWAT will
       create  shapefiles for the weather stations during model setup.
    5.  Locations of major point sources are provided in an Excel file.
    6.  Shapefiles with locations of long-term gaging stations and water quality monitoring stations.
    7.  Area-averaged nitrogen wet atmospheric deposition concentration (mg/L as N) for the study area.
    8.  Area-averaged percent impervious values for each of the NLCD developed classes. The values
       were developed from an analysis specific to each study area. Both total (FIMP) and directly
       connected (FCIMP) values are provided.
Other Setup Tasks (to be undertaken by modeler)

    1.  Identify locations and characteristics of any major reservoirs. Reservoirs included in the model
       should be kept to the essential minimum of those that are sufficiently significant to the water
       balance of the simulated area (at the HUC-8 scale or greater) to include explicitly.
    2.  Identify locations and characteristics of any major features of the watershed affecting water
       balance (e.g., diversions, upstream areas not modeled, reaches that lose flow to groundwater).
       Some of these features may best be represented with observed flow series as boundary conditions.
       Irrigation should be explicitly considered only where needed as a significant part of the basin-
       scale water balance, e.g. Rio Grande.
Special notes for the Cook Inlet study areas

           a.  The projection is different from the one used for the lower 48 states. Be sure that all input
               data used in building the model is in the same projection as the DEM and land cover grid.
           b.   The DEM is in meters, not cm.

           c.  No wet atmospheric deposition rates for nitrogen were available, so it can be omitted
               from the model.
           d.   The SWAT soils database does not include Alaska, so the user will need to obtain
               appropriate STATSGO data for the extent of the model area.
2.2  SWAT MODEL SETUP
SWAT model setup follows directly from the initial setup, using the ArcSWAT extension in ArcGIS.
The following items should be noted and/or followed during SWAT model setup:
Watershed Delineation
Use the Automatic Watershed Delineation option with the following steps:
    1.  Open the Watershed Delineation window.
    2.  Import the DEM for the watershed and choose the z unit as cm.
    3.  Select the option for using user defined watersheds.
    4.  Import the subbasin and reach shapefiles subsequently. Create outlets.
    5.  Click the calculate subbasin parameters button. Please check "Skip Longest flow path
       calculation" option before calculating subbasin parameters.

-------
    6.  Once the subbasin parameters have been calculated, add reservoirs, if any.  It is desirable to avoid
       inclusion of reservoirs where possible due to the difficulty in representing operating rules
       adequately. Therefore, only include significant/major reservoirs that have a major impact on
       flows. If a reservoir is at the terminus of the model area it may be ignored so that the model
       represents input to, rather than output from the terminal reservoir.

    7.  For point sources, only those identified as majors and supplied by Tt are included in the model.
       Create a GIS coverage of the point sources in the subbasin from the list of majors supplied by
       Tetra Tech. Bring in the point sources layer to aid manual addition of point sources using the
       ArcSWAT interface. Define the major point sources at this stage. For some watersheds it will
       also be necessary to define an upstream boundary condition "point source".
    8.  Run Watershed Topographic Report for later use.
HRU Analysis
Start with the Land Use/Soils/Slope Definition
    1.  Save the model often, and make complete backups as you finish major tasks.
    2.  Use  SWAT to classify the slopes into two categories with a breakpoint at 10%.
    3.  Bring in STATSGO soils with the  MUID option.

    4.  NLCD 2001 land use coverage that is supplied can be loaded directly into ArcSWAT without
       modification. The default NLCD class to SWAT class mapping is appropriate for most areas;
       however, there are added future urban land uses (codes 121, 122, 123, and 124).  Use the supplied
       luc.txt file to ensure correct mapping.  Adjustments to the land use assignment can be made
       during cover setup, or parameters for SWAT classes adjusted at a later time. The clipped grids
       distributed for a project include a nominal representation of all potential future developed classes
       in each subwatershed, with near-zero area (e.g., 0.001 ha). This will provide a basis for ready
       modification to address future land use scenarios.
    5.  Assign impervious percentage to developed land use classes in the SWAT urban database using
       the values provided  for the study area.  The same assumptions must be applied for the future
       developed land use classes UFRL,  UFRM, UFRH, and UFHI, i.e., the future classes will have
       the same total and connected impervious fractions as the  corresponding URLD, URMD, URHD,
       and UIDU urban land uses. Within ArcSWAT, the impervious values are saved to the main
       program geodatabase at C:\Program Files\SWAT\SWAT2005.mdb.  The corresponding
       urban.dat file is regenerated from the geodatabase each time the model is run from the ArcGIS
       interface.  This has another important implication: If a given model is ported to a different
       machine, the SWAT2005.mdb file must also be ported.
    6.  Proceed to the HRU Definition tab. Create HRUs by overlaying land use, soil, and slope at
       appropriate cutoff tolerance levels  (usually 5% for land use, 10% for soil, and 5% for slope).  BE
       SURE to EXCLUDE all 8 urban land use classes (URLD, URMD, URHD, UIDU, UFRL,
       UFRM, URFH, and UFHI) from the threshold criteria.  This is done on the Land Use Refinement
       tab.
    7.  Proceed with standard SWAT model generation ("Write Input Tables") using met data provided
       on the project FTP site (processed  weather series and station locations files). Precipitation and
       temperature use observed series; other weather data are simulated with the weather generator. It
       is advisable to screen the precipitation and air temperature files for any gross errors during the
       simulation time period. While errors are uncommon in the BASINS dataset from which these
       were derived, they do occur. Outliers and periods of flatlined values have been discovered in the
       pilot phase of the project, and the met data should be corrected and/or stations  removed if gross
       errors are found.
    8.  Specify PET option as 1 (Penman/Monteith) in General Watershed Parameters.

-------
      9.  Turn on elevation bands if necessary to account for orographic effects in areas with a sparse
         precipitation network and significant elevation changes.  This will generally be appropriate where
         elevations within subbasins span a range of 250 m or more (see Watershed Topographic Report).

      10. Assign management operations. The simulation option IURBAN in the mgt files associated with
         urban land use classes should be left at the default value of 1 (use USGS regression equations).
      11. Set instream water quality options; IWQ = 1, and start with program defaults.
      12. Use (daily) Curve Number hydrology with observed precipitation and air temperature.
         Remaining meteorological variables are simulated using the weather generator.
      13. Specify atmospheric N wet deposition concentrations.
      14. The time period for simulation should be 31 water years. The first year will be dropped from
         analysis to account for model spinup.  The remaining 30 years span a period for which the
         supplied weather data are complete and include the year 2000 (with the exception of the
         Nebraska (Loup and Elkhorn River) basins, where the weather data are complete only through
         1999). Note that the start of simulation for some of the non-pilot study areas may be one year
         prior to the complete weather data period. For this spinup year, some weather stations may be
         absent, but SWAT will fill in the missing records using the weather generator. Save and
         backup the model at this point.

      15. Run the model for the full 31 year period. Due to spin up effects and interaction with the  weather
         generator random number processing,  all model runs (calibration, validation, and scenario
         application) should use the entire model network and the entire simulation time period. (Initial
         testing can be done on a subset of the model or a reduced time period; however, it is necessary to
         run the full model extent and time period to obtain valid final results.)

      16. Undertake calibration for target HUC8. After calibration, repeat for remainder of study area.
         Calibration should first be performed for hydrology, then sediment, then nutrients. Calibration
         spreadsheet templates are distributed for Hydrology and Water Quality, as described in the next
         section.
         NOTE: When pursuing calibration through the Arc SWAT interface be sure to use the option to
         REwrite SWAT input files (and not "Write Input Files").  The latter option will cause default
         parameters to be reloaded from the geodatabase.


  2.2  HSPF MODEL SETUP

BASINS4/HSPF uses primarily two applications - Map Window GIS and WinHSPF to create, modify and
run HSPF UCI files.  The following steps should be implemented first for the Calibration HUC8  subbasins,
then repeated for the entire model watershed.

    1.   Prepare a starter.uci file defining default values for PERLND/IMPLND base numbers (see  below).
        Where previous modeling is available, the initial parameter values will be  based on that earlier
        modeling.  For areas without previous modeling, hydrologic parameters will be based on
        recommended ranges in BASINS Technical Note 6 and related to soil and meteorological
        characteristics where appropriate.

    2.   Load HSPF land cover/soils grid (as discussed in the GIS Processing Memorandum), DEM,
        subbasin, and reach file into BASINS4 Map Window interface.

    3.   Use Manual Delineation to calculate subbasin and reach parameters.

    4.   Assign subbasins to model segment groups using the Model Segmentation Specifier Tool.  (A
        segment is a group of subbasins with a unique set of PERLNDs and IMPLNDs. Segments  are used
        primarily for weather station assignment, but may also be used for other factors such as differences
        in soils, geology, etc.) Assignment will be based on proximity to weather stations, elevation bands,
^^^^^^^^^^^^^^^^^^^——^^^^^^^^^^^^^^^^^^—

-------
   and any other unique watershed characteristics identified previously.  The Tool will generate unique
   model segments within the HSPF model, using the PERLND/IMPLND numbering scheme shown
   below.
5.  Run the BASINS4 HSPF tool.
       a.   General Tab
               i. Land Use Type, assign to LU+HSG grid
              ii. Subbasins and Streams, use subbasin and reach shapefiles developed previously.
             iii. Point sources, using input shapefile developed for watershed.
             iv. Met stations, using input shapefile developed for watershed.
       b.   Land Use Tab: assign imperviousness to the developed land use classes using values
           developed from the impervious area analysis.
       c.   Streams Tab: uses default fields
       d.   Subbasins Tab: uses default fields.
       e.   Point Sources Tab: uses default fields.
       f   Met Stations Tab: uses default fields.
6.  Use separate automated processing tool to do the following:
       a.   Lump area from the four developed classes into one IMPLND and four PERLND categories.
       b.   Assign PERLND/IMPLND model segmentation using a set numbering scheme for the land
           use classes. Model segments will be implemented in groups of 25 (i.e., PERLND 1, 26, 51,
           76, and 101 are  all WATER).  Ensure that all combinations of base number and segment are
           represented throughout the model. If necessary, define a nominal area for missing
           PERLND/IMPLND values in the SCHEMATIC block. Assignments are shown below, first
           for PERLND, and then IMPLND.
LC_HSG Class
WATER
BARREN_D
WETL_D
FOREST_A
FOREST_B
FOREST_C
FOREST_D
SHRUB_A
SHRUB_B
SHRUB_C
SHRUB_D
GRASS_A
GRASS_B
GRASS_C
GRASS_D
AGRI_A
AGRI_B
LC_HSG
Value
101
4
14
21
22
23
24
31
32
33
34
41
42
43
44
51
52
HSPF PERLND
Name
WATER
BARREN_D
WETL_D
FOREST_A
FOREST_B
FOREST_C
FOREST_D
SHRUB_A
SHRUB_B
SHRUB_C
SHRUB_D
GRASS_A
GRASS_B
GRASS_C
GRASS_D
AGRI_A
AGRI_B
HSPF PERLND
Base Number
1
2
3
5
6
7
8
9
10
11
12
13
14
15
16
17
18
                                          A-6

-------
LC_HSG Class
AGRI_C
AGRI_D
DEVO_A
DEVO_B
DEVO_C
DEVO_D
DEVL_A
DEVL_B
DEVL_C
DEVL_D
DEVM_A
DEVM_B
DEVM_C
DEVM_D
DEVH_A
DEVH_B
DEVH_C
DEVH_D
LC_HSG
Value
53
54
61
62
63
64
71
72
73
74
81
82
83
84
91
92
93
94
HSPF PERLND
Name
AGRI_C
AGRI_D
DEVPERV_A
DEVPERV_B
DEVPERV_C
DEVPERV_D
DEVPERV_A
DEVPERV_B
DEVPERV_C
DEVPERV_D
DEVPERV_A
DEVPERV_B
DEVPERV_C
DEVPERV_D
DEVPERV_A
DEVPERV_B
DEVPERV_C
DEVPERV_D
(all imperv)
HSPF PERLND
Base Number
19
20
21
22
23
24
21
22
23
24
21
22
23
24
21
22
23
24
25
  7.  In WinHSPF, define hydraulic characteristics for major reservoirs and flow/load characteristics for
     major point sources. This step can be done in common with the corresponding step for SWAT.

  8.  FTABLES will be generated automatically during model creation. FTABLES can be easily adjusted
     in WinHSPF if specific information is available to the modeler. The WinHSPF FT ABLE tool also
     includes a way to recalculate FTABLES using relationships developed for three regions in the
     Eastern United States.
  9.  Adjust lapse rates as needed to account for elevation bands associated with model segments.  Lapse
     rates are needed for precipitation and air temperature only. For snowmelt, the simplified degree-day
     method for snowmelt should be employed.
  10. Add additional UCI tables as needed, using the WinHSPF interface.
  11. Undertake calibration for target HUC8. After calibration, repeat for remainder of study area.
     Calibration should first be performed for hydrology, then sediment, then nutrients.

  12. Nutrients will be modeled as inorganic N,  inorganic P, and organic matter. The latter will be
     transformed to organic N and organic P in the MASS-LINK to the stream. The buildup-washoff
     approach should be  used to simulate land surface processes; it is easy to implement and tools are
     available to translate storm Event Mean Concentrations (EMCs) to model inputs. The instream
     simulation should use GQUALS with exponential decay.  The project does not have sufficient
     resources to develop models with a full algal simulation.



3  Calibration  and  Reporting  Procedures	

Calibration will be pursued for flow, TSS, TN, and TP sequentially. The water quality calibration focuses
on replicating inferred monthly loads, although some attention must be paid to the calibration of
concentration as well.
                                           A-7

-------
As noted above, calibration starts with a single HUC8.  Parameters derived for this site are then extended
to the remainder of the watershed.  However, experience is that this procedure alone is not adequate to
provide reasonable results for the entire study area.  Therefore, secondary calibration adjustments should
be made at 2-3 additional stations in the watershed, at least one of which should be at a larger spatial scale
than HUC8 (if available).
The time period for calibration should generally be the last 10 water years of the simulation. Validation
will be applied to the preceding 10 water years (for which fit may be less good due to changes in land use
and management relative to the 2001 NLCD). Adjustments can be made based on data availability if
necessary.

The hydrologic calibration uses the Hydrocal spreadsheet. A version of this  spreadsheet will be provided
for the assigned non-pilot study areas. The spreadsheet is already loaded with a list of USGS gages with
sufficiently long periods of record. The spreadsheet will automatically download USGS flows from
NWIS and will load SWAT simulated flows from the program output.  Note  that this spreadsheet should
be opened in Excel 2003 to ensure  proper operation.  Most of the functions can be run in Excel 2007 if
needed, but, if so,  it should be saved in compatibility mode for Excel 2003. The program uses macros,
and macro security will need to be  set to "Medium" for the program to operate.
User controls for the Hydrocal spreadsheet are located on the Data Management tab. Key results are
provided on the Analysis tab. In general, the user should strive to meet the "Recommended Criteria" on
the Analysis tab.
Water quality calibration uses the WQUAL-GCRP spreadsheet. This program also uses macros and is
now designed for  Excel 2007.  The water quality calibration takes place at stations for which there are
both flow and concentration data and comes pre-loaded with a list of USGS gaging stations. Note that
one spreadsheet covers both the calibration and validation periods.
The modeler is responsible for preparing and loading observed water quality data for TSS, TN, and TP on
the "Obs" tab. At a minimum use the data available on the USGS NWIS system. (In some cases,
previous  studies have resulted in creation of more extensive, QA'd data sets that may be used instead.)
Pre-processing is required. Duplicate observations on the same day should be pre-averaged. The 'TSS"
column can be used to store both SSED and TSS data; when both are present on a given day use the
SSED results preferentially.  A special note is required regarding TN. TN is a calculated value, and
USGS reports it directly only in some years.  In many other cases the user can construct TN observations
by adding TKN plus NO2+NO3 observations. Note that the "Obs" tab contains a column for denoting
samples below detection limit (using symbol "<") for each parameter.  For TN, observations should be
flagged as non-detect only when both TKN and NO2+NO3  are non-detect. If only one of these is non-
detect use one-half the detection limit to create the sum.
After entering the  observed data, the main user controls are  on the "Interface" tab. After running the
macros, key results are shown on the GCRP tab at BH22+.  The key interest  of EPA is in the prediction of
loads.  Loads are,  of course, not observed, but are estimated using a stratified regression (TSS, TP) or an
averaging estimator (TN). The  first objective of calibration is to reduce the relative percent deviation
between simulated and estimated loads - to below 25% if possible.  (This will sometimes not be
possible.) However, the user should also examine the diagnostic plots on the TSS, TP, and TN
worksheets.
                                              A-8

-------
4  Implementing  Future Scenarios	

Both climate and land use scenarios will be implemented by working directly with the model input files,
rather than returning to the GIS interface.


4.1  CLIMATE SCENARIOS
In both HSPF and SWAT, climate scenarios are readily implemented simply by substituting new
meteorological series. We will not assume any feedback between climate changes and HSPF parameters.


4.2  LAND USE SCENARIOS
ICLUS will be used to estimate a change table by subwatershed. In SWAT, the HRU fractions in each
HRU file are changed using automation scripts.
                                     A-9

-------
(This page intentionally left blank.)
                                              A-10

-------
Appendix  B
Project Quality  Assurance Project  Plan
Section  8:  Model Calibration
This appendix reproduces Section 8.0 of the QAPP which describes model calibration
requirements; taken from:

Tetra Tech. (2008) Quality assurance project plan for watershed modeling to evaluate potential
impacts of climate and land use change on the hydrology and water quality of major U.S.
drainage basins.  Prepared for the Office of Research and Development, Global Change Research
Program, U.S. Environmental Protection Agency, Washington, DC.
                      8.0 MODEL SETUP/CALIBRATION


8.1    PROJECT QUALITY OBJECTIVES
EPA emphasizes (USEPA 2000, 2002) a systematic planning process to determine the type and
quality of output needed from modeling projects. This begins with a Modeling Needs and
Requirements Analysis, which includes the following components:

   •  Assess the need(s) of the modeling project

   •  Define the purpose and objectives of the model and the model output specifications

   •  Define the quality objectives to be associated with model outputs

The first item (needs assessment) is covered in EPA's task order. In essence, simulation models
are needed to predict future responses to  changes in climate and land use. The existing
simulation models HSPF and SWAT are believed to be sufficient to this purpose, and creation of
new models is not required.

The second item (define purpose and objectives) is the subject of EPA's Draft Analysis Plan.
This proposes both the purpose of the modeling and the specific endpoints to be evaluated as a
result of the modeling. At a general level, the objective of this modeling project is to assess the
potential effects of climate and land use change on the hydrology and water quality of major U.S.
drainage basins; however, this general objective will need to be made more specific to guide
development of the modeling effort. The Tt team is tasked with reviewing and commenting on
the Analysis Plan as part of this work—and revisions to the existing Analysis Plan could arise as
a result of these recommendations. At the end of this review, the Tt team and the EPA COR must
agree on the principal study questions to  be addressed through the  modeling.
                                     B-l

-------
The quality objectives for the model(s) follow directly from the purposes and objectives—and
can be refined in conjunction with the review of the Analysis Plan. In general, the modeling
effort needs to be designed to achieve an appropriate level of accuracy and certainty in
answering the principal study questions. This process takes into account the following elements:

    •   The accuracy and precision needed for the models to predict a given quantity at the
       application site of interest to satisfy study questions

    •   The appropriate criteria for making a determination of whether the models are accurate
       and precise enough on the basis of past general experience combined with site-specific
       knowledge and completeness of the conceptual models

    •   How the appropriate criteria would be used to determine whether model outputs achieve
       the needed quality

EPA's Draft Analysis Plan suggests that the principal study questions to be addressed by the
models are changes in (defined on the basis of modeling at a daily time step): (1) the 100-year
flood, (2) 7Q10 low flow, (3) runoff center of mass, (4) monthly sediment loads,  (5) monthly
total nitrogen loads, and (6) monthly total phosphorus loads. This list could be expanded or
modified on the basis of the review of the Draft Analysis Plan.

The models will be calibrated and validated to existing (1970-2000) data to establish their
credibility for use in forecasting responses to future change. Specific calibration and validation
targets for model acceptability (see Sections 8.2 and 8.3) will be selected in light of the intended
uses of the model, as identified in the final revisions to the Analysis Plan.

8.2    MODEL CALIBRATION AND VALIDATION
Model calibration is the process of adjusting model inputs in acceptable limits until the resulting
predictions give good correlation with observed data.  Commonly, calibration begins with the
best estimates for model input on the basis of measurements and subsequent  data analysis.
Results from initial simulations are then used to improve the concepts of the  system or to modify
the values of the model input parameters.  The use of calibrated models, the scientific veracity of
which is well defined, is of paramount importance to this project. Because the goal  is to be able
to assess the potential effects of climate and land  use change on the hydrology and water quality
of major U.S. drainage basins, model calibration and validation should strive to minimize errors
(deviations between model predictions and observed measurement data.).

The Tt Co-TOLs or lead modeler will direct the model calibration efforts. Models are often
calibrated through a subjective trial-and-error adjustment of model input data because a large
number of interrelated factors influence model output. However, the experience and judgment of
the modeler are a major factor in calibrating a model accurately and efficiently. Further, the
model should meet pre-specified quantitative measures of accuracy to establish its acceptability
in answering the principal study questions.

The model calibration process proceeds through both  qualitative and quantitative analyses.
Qualitative measures of calibration progress are commonly based on the following:

    •   Graphical time-series plots  of observed and predicted data
______^^

-------
    •   Graphical transect plots of observed and predicted data at a given time interval

    •   Scatter plots of observed versus predicted values in which the deviation of points from a
       45-degree straight line gives a sense of fit

    •   Tabulation of measured and predicted values and their deviations

After initially configuring the modeling systems, the Tt team will perform model calibration and
validation. The watershed models will be calibrated to the best available data, including literature
values, and interpolated or extrapolated values using existing field data. If multiple data sets are
available, an appropriate time period and corresponding data set will be chosen on the basis of
factors characterizing the data set, such as corresponding weather conditions, amount of data,
and temporal and spatial variability of data.

A model is considered calibrated when it reproduces data within an acceptable level of accuracy,
as described in Section 8.3 and itemized in Table 4 (quantitative measures). A set of parameters
used in a calibrated model might not accurately represent field values, and the calibrated
parameters might not represent the system under a different set of boundary conditions or
hydrologic stresses. Therefore, a model validation period helps establish greater confidence in
the  calibration and the predictive capabilities of the model. A site-specific model is considered
validated if its accuracy and predictive capability have been  proven to be within acceptable limits
of error independently of the calibration data.

       Table 4.  General percent error calibration/validation targets for watershed
       models (applicable to monthly, annual,  and cumulative values)

Hydrology/Flow
Sediment
Water Quality/Nutrients
Relative percent error
Very good
<10
<20
<15
Good
10-15
20-30
15-25
Fair
15-25
30-45
25-35
In general, model validation is performed using a data set separate from the calibration data. If
only a single time series is available, the series could be split into two subseries, one for
calibration and another for validation. If the model parameters are changed during the validation,
this exercise becomes a second calibration, and the first calibration needs to be repeated to
account for any changes. Representative stations will be used to guide parameter adjustment to
get an accurate representation of the conditions of the individual  subwatersheds and streams. The
calibration and validation process will be documented for inclusion in the technical reports.

8.3    SPECIFIED PERFORMANCE AND ACCEPTANCE CRITERIA
Model Testing
Model testing includes calibration, verification, and validation. The previous section described
model calibration and validation. Model verification is the process of testing the model code,
including program debugging, to ensure that the model implementation has been done correctly.
Testing usually begins with the best estimates for model input on the basis of measurements and
subsequent data analyses. Results from  initial simulations are then used to improve the concepts
of the system or to modify the values of the model input parameters.
                                           B-3

-------
For this project, existing tested model code will be used (HSPF and SWAT). Therefore, model
verification is required only for new bridge code, such as that required to translate climate
scenarios into model input.

The Tt team will calibrate the project models using the best available data, including literature
values and interpolated or extrapolated existing field data. The model will be considered
successfully tested when it reproduces data at an acceptable level of accuracy.

The work proposed for this project, as defined in the Draft Analysis Plan,  differs from other,
more common applications of watershed models (e.g., for TMDLs) in several ways that affect
the calibration strategy:

    •   Models will be developed at a very large spatial scale (i.e., HUC4  scale) and will be
       calibrated at a limited number of points, most of which will likely be at the HUC8 scale.

    •   Models will be developed for multiple watersheds, and calibration will be done by
       multiple teams of modelers. The different teams should all apply the same calibration
       metrics.

    •   Two separate models  (HSPF and SWAT) will be developed for some or all the
       watersheds. A common set of calibration criteria should be applied to both models to
       facilitate comparison.

    •   Models are proposed to be developed using a daily time step (based on the scale of the
       analysis), which will limit the ability to resolve extreme flows.

    •   Model application is not for regulatory purposes but to inform possible long-term effects
       of different change scenarios. While calibration to establish model credibility is essential,
       the ability to correctly simulate relative changes is most important.

    •   Comparison of observed and predicted values on a frequency-duration plot.

Quantitative acceptance criteria for the models will be selected to reflect the final set of principal
study questions in the revised Analysis Plan and incorporated into the QAPP. Given the
considerations listed above, quantitative acceptance criteria will be expressed in relative, rather
than absolute form. That is, relevant calibration outputs will be ranked on a  scale ranging from
poor to very good. Calibration will strive to obtain the best fit possible; however,  specific values
of quantitative measures will  not be proposed to define whether results should be accepted or
rejected. Rather, the level  of uncertainty determined in calibration and validation will be
documented to decision makers to aid in interpretation of results.

The current Draft Analysis Plan references  only three measures related to hydrology (100-year
flood, 7Q10 low flow, and runoff center of mass); however, accurate representation of the
general water balance is required to demonstrate that the model provides a reasonable
representation of reality that can serve as a foundation for water quality simulation. Therefore,
commonly accepted measures of model hydrologic fit will be applied.

Model simulation of water quality is, in general, more difficult than simulating hydrology, in part
because any uncertainty in the hydrologic simulation will propagate into the water quality

______^^

-------
simulation. In addition, the principal study questions related to water quality contained in the
Draft Analysis Plan address loads. Loads are not directly observed but are inferred from point-in-
time concentration data and continuous flow data. As a result, observed load estimates are
subject to considerable uncertainty.

Quantitative measures, sometimes referred to as calibration criteria, include the relative percent
error between model predictions and observations as defined generally below:
                                         Z  O-P
                                              Q
                                                    lOO,
where Erei = relative error in percent. The relative error is the ratio of the absolute mean error to
the mean of the observations and is expressed as a percent. A relative error of zero is ideal.
Additional statistics that will be applied include the correlation coefficient (R) and its squared
value, the coefficient of determination (R2), where
                                                          O - P
where the overbar indicates the sample mean.

For hydrology and the water balance, percent error tests will be applied to the following
components:

    •  Total flow volume

    •  10 percent high flows

    •  50 percent low flows

    •  Seasonal flow volumes

For water quality, the outcomes of interest defined in the current Draft Analysis Plan are monthly
loads. Therefore, similar calculations of relative percent error will be applied to the series of
predicted and observed monthly loads (where the observed monthly loads will need to be
estimated from observed flow and concentration data using an appropriate estimation technique,
such as those described in Preston et al. 1989).

These tests are relevant to monthly and annual values. General calibration/validation targets for
percent error consistent with current best modeling practices (Donigian 2000) are shown in Table
4.

For hydrology, there is also an interest in extreme high and low flows. Answering this study
question requires calibration to daily flows, rather than just monthly and annual values. Figure 3
(also from Donigian 2000) summarizes R and R2 ranges for the evaluation of daily and monthly
flows:

_____^^

-------
                              I	
              GL95
Q8            0.9
     Very Good
Good          VeryGood
       Figure 3. R and R value ranges for model performance

In addition, the Nash-Sutcliffe coefficient of model fit efficiency (COE) will be reported for all
calibration and validation runs — although no specific criteria are proposed. This is calculated as
A COE value of one indicates a perfect fit between measured and predicted values for all events.
A value of zero indicates that the model fit is not better than using the average value of all the
measured data.

Following model calibration, model validation will be conducted using separate, independent
portions of the available time series at the calibration stations. Because the Analysis Plan calls
for simulating the period 1970-2000, while land use will be based on 2001 NLCD information,
the 10-year period from 1991 through 2000 will generally be proposed for calibration, while an
earlier period (dependent on data availability) will be used for validation tests. Because the land
use distribution during the 1970-1991 period could be different in some regions than during the
1991-2000 period, it is important to note that validation results might not achieve the same
quantitative acceptance levels as for calibration.

The Tt team will document model performance over both the calibration and validation period in
the technical reports, using the quantitative measures of accuracy documented above (or any
additional measures that could be identified in modifications to this QAPP). In addition to
measures of accuracy, additional  acceptance criteria will include modeling result precision and
representativeness:

   •   Precision of model results: Precision of generated data produced by the model will be
       examined by performing replicate runs. By confirming that an identical data set is
       generated when a replicate of the previous model run will rule out numerical instability
       issues and verify the precision of the model.

   •   Representativeness of model results: The Tt team technical staff will compare the
       loadings data and measured environmental concentrations to examine sources and sinks
       of materials.

An overall assessment of the success  of the calibration can be expressed using calibration levels.
                                           B-6

-------
    Level 1: Quantitative performance measures fall within the very good range (highest degree
            of calibration).
    Level 2: Quantitative performance measures fall within the good range.
    Level 3: Quantitative performance measures fall within the fair range.
    Level 4: Quantitative performance measures fall within the poor range (lowest degree of
            calibration).

Model Sensitivity Analysis
The sensitivity to variations or uncertainty in input parameters is an important characteristic of a
model. Sensitivity analysis is used to identify the most influential parameters in determining the
accuracy and precision of model predictions. This information is of importance to the user who
must establish required accuracy and precision in model application as a function of data
quantity and quality. Sensitivity analysis quantitatively or semi-quantitatively defines the
dependence of the model's performance assessment measure on a specific parameter or set of
parameters. Sensitivity analysis can also be used to decide how to simplify the model simulation
and to improve the efficiency of the calibration process.

Model sensitivity can be expressed as the relative rate of change of selected output caused by a
unit change in the input. If the change in the input causes a large change in the output, the model
is considered to be sensitive to that input parameter. Sensitivity analysis methods are mostly
nonstatistical or even intuitive by nature. Sensitivity analysis is typically performed by changing
one input parameter at a time and evaluating the effects on the distribution of the dependent
variable. Nominal, minimum, and maximum values are specified for the selected input
parameter.

Sensitivity analysis is performed at the beginning of the calibration process to design  a
calibration strategy. After calibration is completed, a more elaborate sensitivity analysis is
performed to quantify the uncertainty in the calibrated  model caused by uncertainty in the
estimates of the model input parameters.

Informal sensitivity analyses (iterative parameter adjustments) are generally performed during
model calibration to ensure that reasonable values for model parameters will be obtained,
resulting in acceptable model results. The degree of allowable  adjustment of any parameter is
usually directly proportional to the uncertainty of its value and is limited to its expected range of
values.

8.4    ASSESSMENT AND RESPONSE ACTIONS
The ability of computer code to represent model theory accurately will be ensured by following
rigorous programming protocols, including documentation within the source code. Specific tests
will be required of all model revisions to ensure that fundamental operations are verified to the
extent possible, including testing numerical stability and convergence properties of the model
code algorithms, if appropriate. Model results will generally be checked by comparing results to
those obtained by other models or by comparing them to hand  calculations. Visualization of
model results will help determine whether model simulations are realistic. Model calculations

______^^

-------
will be compared to field data. If adjustments to model parameters are made to obtain a fit to the
data, the modelers will provide an explanation and justification that must agree with scientific
knowledge and fit within reasonable ranges of process rates as found in the literature.

As described in Section 5.1, non-project-generated data will be used for model development and
calibration. The model calibration procedure is discussed in Section 8.2. The DQOs were
discussed in Section 7.0 and 8.0 of this document. Modelers will cross-check data for bias,
outliers, normality, completeness, precision, accuracy, and other potential problems.

Data generated outside the project will be obtained primarily from quality assured databases
maintained by EPA, USGS, and other entities. Additional data may be obtained from either
published or nonpublished sources. The published data will have some degree or form of peer
review. Typically, modelers examine these data as part of a data quality assessment.
Unpublished databases are also examined in light of a data quality assessment. Data provided by
EPA or other sources will be assumed to meet precision objectives established by those entities.

The QA program under which this task order will operate includes surveillance, with
independent checks of the data obtained from sampling, analysis, and data gathering activities.
This process is illustrated in Figure 4.
                                           B-8

-------
                                             Problem Identified
                  QC Officer
                    TOL
             Other Tetra Tech Staff
                          QA Officer
                 (System Audit [Audit Report Form
                         Completed])
                 EPATOM
        Other non-Tetra Tech Personnel
                                              Tt TOL Informed

                                                    I
                                              Problem Defined
                                         Tt TOL Confirms Problem and
                                        Reviews Nature of the Problem
                                  Yes
                                        Significant Corrective Action
                                                  Needed
                                                                      No
                          QAO Informed
                          Documented on
                          Corrective Action
                       Request and Response
                          Verification Form
                                         Tt TOL Works with Task
                                          Leader and Technical
                                            Staff to Correct the
                                                Problem
                  Yes
Stop Work Decision or
  Recommendation
    Made by QAO
     Stop Work Order Issued
     Tt PIC and PM Informed by
   QAO; Tt TOL and technical staff
     informed by Tt PM (client
      notified as appropriate)
Problem and Corrective
  Action Recorded in
 Project Documents as
     Appropriate
            No
                                           Tt TOL Ensures That
                                          Corrective Action Has
                                            Been Implemented
                       No
                                     Problem Investigated
                                TtTOL Determines Cause, Assigns
                            ->• Responsibility, Develops Procedure
                                to Correct the Problem, Implements
                                        the Procedure
                                      Corrective Action
                                    Implementation Verified
                                           By QAO
                                          No
                                         •Yes I
     Start Work Order Issued
   By QAO; PIC and PM Informed
   by QAO; Tt TOL and Technical
   Staff Informed by Tt PM (client
     notified as appropropriate)
                                       Revisions Made
                                 to QA Documents, if Necessary
                                                                           Yes
              Revisions Distributed
           to Client and All Staff Involved
                                                                   Work Proceeds
                                    Corrective Action Form
                                          Completed
                                  Filed by QAO, PM, and Tt TOL
                                       Work Proceeds
           Tt = Tetra Tech
           PIC = Principal-in-Charge
Figure 4. Problem assessment and correction operations
                                                B-9

-------
The essential steps in the QA program are as follows:
    •   Identify and define the problem
    •   Assign responsibility for investigating the problem
    •   Investigate and determine the cause of the problem
    •   Assign and accept responsibility for implementing appropriate corrective action
    •   Establish the effectiveness of and implement the corrective action
    •   Verify that the corrective action has eliminated the problem

Many of the possible technical problems can be solved  on the spot by staff, for example, by
modifying the Initial Technical Approach memorandum or correcting errors or deficiencies in
implementation of the approach. Immediate corrective actions are considered SOPs, and they are
noted in records for the project. Problems that cannot be solved in this way require more
formalized, long-term corrective action.

If quality problems that require attention are identified,  Tt will determine whether attaining
acceptable quality requires either short- or long-term actions. If a failure in an analytical system
occurs (e.g., performance requirements are not met), the Tt team modeling QC officers will be
responsible for corrective action and will immediately inform the Tt Co-TOLs or the QAO, as
appropriate. Subsequent steps taken will depend on the  nature and significance of the problem,  as
illustrated in Figure 4.

              Crat™ct(™nie)	                   The Tt Co-TOLs have primary
                                                         responsibility for monitoring the
                                                         activities of this project and
                                               ~         identifying or confirming any quality
                                                         problems. The Co-TOLs will also
                                                         bring these problems to the attention
                                                         of the Tt QAO, who will initiate the
                                                         corrective action system described
                                                         above, document the nature of the
                                                         problem (using a form such as that
                                                         shown in Figure 5), and ensure that the
                                                         recommended corrective action is
                                                         carried out. The Tt QAO has the
                                                         authority to stop work on the project if
                                                         problems affecting data quality that
                                                         will require extensive effort to resolve
                                                         are identified.
Date of Assessment _
Title (of project or other) _

Project Leader	
Other Responsible Personnel	

Auditor or Initiator of This Corrective Action Request _

Problem Description:
Recommended Action:
Quality Assurance Officer


Principal-in-Charge or Program Manager

Action Taken:
Verification of Completion of Corrective Action:
                              Date to Be Completed:
  Quality Assurance Officer
  Principal-in-Charge or Program Manager             Date

Original form to be filed in QAO File; one copy to be filed in Project File and one copy in Contract File (if corrective
 action pertains to a project), or one copy to be filed in Contract File (if corrective action pertains to a contract).
           Figure 5. Example corrective
           action request and response
           verification form
                                             B-10

-------
The EPA COR, Tt PGM and Tt Co-TOLs will be notified of major corrective actions and stop
work orders. Corrective actions can include the following:

    •   Reemphasizing to staff the project objectives, the limitations in scope, the need to adhere
       to the agreed-upon schedule and procedures, and the need to  document QC and QA
       activities

    •   Securing additional commitment of staff time to devote to the project

    •   Retaining outside consultants to review problems in specialized technical areas

    •   Changing procedures

The Tt Co-TOLs may replace a staff member, as appropriate, if it is  in the best interest of the
project to do so.

Performance audits are quantitative checks on different segments of project activities; they are
most  appropriate for sampling, analysis, and data-processing activities. The Tt modeling QC
officer is responsible for overseeing work as it is performed and periodically conducting internal
assessments during the data entry and analysis phases of the project. As data entries, model
codes, calculations, or other activities are checked, the Tt modeling QC  officer will sign and date
a hard copy of the material or complete Tt's standard Technical/Editorial Review Form, as
appropriate, and provide it to the Tt Co-TOLs for inclusion in the administrative record.
Performance audits will consist of comparisons of model results with observed historical data.
Performing control calculations and post-simulation validation of predictions are major
components of the QA framework.

The Tt Co-TOLs will  perform or oversee the following qualitative and quantitative assessments
of model performance periodically to ensure that the model is performing the required task while
meeting the quality objectives:

    •   Data acquisition assessments

    •   Model calibration studies

    •   Sensitivity analyses

    •   Uncertainty analyses

    •   Data quality assessments

    •   Model evaluations

    •   Internal peer reviews

Internal peer reviews will be documented in the project and QAPP files. Documentation will
include the names, titles, and positions of the peer reviewers; their report findings;  and the
project management's documented responses to their findings.

The Tt Co-TOLs will  perform surveillance activities throughout the  duration of the project to
ensure that management and technical aspects are being properly implemented according to the

__^^

-------
schedule and quality requirements specified in this QAPP. These surveillance activities will
include assessing how project milestones are achieved and documented; corrective actions
implemented; budgets adhered to; peer reviews performed; data managed; and whether
computers, software, and data are acquired in a timely manner.

System audits are qualitative reviews of project activity to check that the overall quality program
is functioning and that the appropriate QC measures identified in the QAPP are being
implemented. If requested by the EPA COR, and EPA provides additional funding, the Tt QAO
or designee will conduct an internal system audit of the project and report the results to the EPA
COR and the Tt Co-TOLs.

8.5     DOCUMENTATION AND RECORDS
Thorough documentation of all modeling activities is necessary for interpreting study results. Tt
will prepare monthly progress reports that will address task and subtask milestones, deliverables,
adherence to schedule, and financial progression at the end of each full month while the task
order for this project is still open. Data needs and deadlines for Tt's receipt of information
needed to meet the project schedule will also be included in the progress reports and Gantt chart.
The progress in meeting modeling QA targets (QA reports) will also be included in the progress
reports. Other deliverables will be distributed to project participants as indicated by the EPA
COR. Data tables, assumptions and analyses used to develop the models will be recorded and
provided to EPA as a separate deliverable. The format of the  raw data to be used for model
parameters, model input, model calibration, and model output will be converted to the
appropriate units, as necessary.

The Tt team will save on an external hard drive all modeling  output data from all 20 watersheds
as digital computer files in a  file directory using a file-naming convention specified by the EPA
COR. In  addition, the Tt team will save on an external hard drive all scripts, project files,
calibration data, and other information used to conduct watershed modeling at each of the 20
study watersheds. Tt will deliver these external hard drives to EPA within 2 weeks of the EPA
COR's approval of the final report presenting and discussing the goals, methods, results and
conclusions of watershed modeling in all 20 study watersheds (see the  schedule in Table 2). Tt
will maintain a copy of the project files at the Cincinnati, Ohio and/or Fairfax, Virginia, office
for at least 3 years (unless otherwise directed by the EPA COR). The EPA COR and Tt Co-TOLs
will maintain files, as appropriate, as repositories for information and data used in models and
for preparing any reports and documents during the project. Electronic project files are
maintained on network computers and are backed up periodically.  The Tt Co-TOLs will
supervise the use of project materials. The following information will be included in the
electronic project files within Tt and on the external hard drives:

   •   Any reports  and documents prepared
   •   Contract and task order information

   •   Electronic copies of model input/output (for model calibration and allocation scenarios)

   •   Results of technical reviews, model tests, data quality assessments of output data, and
       audits
                                          B-12

-------
    •  Documentation of response actions during the project to correct model development or
       implementation problems
    •  Assessment reports for acquired data
    •  Statistical goodness-of-fit methods and other rationale used to decide which statistical
       distributions should be used to characterize the uncertainty or variability of model input
       parameters
    •  Communications (memoranda; internal notes; telephone conversation records; letters;
       meeting minutes; and all written correspondence among the project team personnel,
       subcontractors, suppliers, or others)
    •  Maps, photographs, and drawings
    •  Studies, reports, documents, and newspaper articles pertaining to the project
    •  Spreadsheet data files including physical measurements, analytical chemistry data, and
       microbiological data (hard copy  and on diskette)
The model application will include complete record keeping of each  step of the modeling
process. The documentation will consist of reports and files addressing the following items:
    •  Selection of study watersheds and  model calibration points
    •  Assumptions
    •  Adjustments
    •  Parameter values and sources
    •  Nature of grid, network design, or  subwatershed delineation
    •  Changes and verification of changes made in code
    •  Actual input used
    •  Output of model runs and interpretation
    •  Sensitivity analyses results
    •  Calibration and validation of the models
Formal reports submitted to EPA that are generated from the data will be maintained in the
central file (diskette and hard copy) at Tt's Cincinnati, Ohio, and Fairfax, Virginia, offices. The
data reports will include a summary of the types of data collected,  sampling dates, and any
problems or anomalies observed during sample collection.
8.6    OUTPUT ASSESSMENT AND MODEL USABILITY
Tt team technical staff will review model predictions for reasonableness, relevance, and
consistency with the requirements of the model development process through model calibration
as described in Section 8.0 of this QAPP. Tt team modeling experts will also determine
consistency with the acceptance criteria described in Sections 7.0 and 8.0 of this QAPP. The Tt
modeling QC officer will ensure that all steps of the modeling process are performed correctly.
_____^^

-------
Electronic copies of model input/output for model calibration, data quality assessments of output
data, and QA reports will be maintained as part of the project files.

                                 LITERATURE CITED
Donigian, A.S., Jr. 2000. HSPF Training Workshop Handbook and CD. Lecture #19. Calibration
   and Verification Issues, Slide #L19-22. EPA Headquarters, Washington Information Center,
   January 10-14, 2000. Presented and prepared for U.S. Environmental Protection Agency,
   Office of Water, Office of Science and Technology, Washington, DC.

Preston, S.D., V.J. Bierman, Jr., and S.E. Silliman. 1989. An evaluation of methods for the
   estimation of tributary  mass loads. Water Resources Research, 25(6): 1379-1389.

USEPA (U.S. Environmental Protection Agency). 2000. Guidance for the Data Quality
   Objectives Process (G-4). EPA 600-R-96-055. U.S. Environmental Protection Agency,
   Office of Environmental Information, Washington, DC.

USEPA (U.S. Environmental Protection Agency). 2002. Guidance for Quality Assurance Project
   Plans for Modeling (G-5M). EPA 240-R-02-007. U.S. Environmental Protection Agency,
   Office of Environmental Information, Washington, DC.
                                        B-14

-------
Appendix C
Climate  Change and  the  Frequency  and
Intensity  of Precipitation  Events
 Technical Note
 By: Charles Rodgers, Stratus Consulting Inc.


 This review has been prepared to address issues raised in the context of the preparation of
 meteorological data used as input to the SWAT and HSPF watershed models. The data
 preparation process is summarized as follows: The approach selected for this project is to use the
 U.S. Environmental Protection Agency's (EPA's) BASINS  Climate Assessment Tool (CAT;
 U.S. EPA, 2009) to modify historical meteorological records to reflect the projected impacts of
 climate change on important meteorological variables. Temperature and precipitation records
 from the National Oceanic and Atmospheric Administration (NOAA) National Climatic Data
 Center (NCDC) stations in and surrounding each pilot watershed have been identified, and
 hourly data covering the period 1970-2000 is being used in  the calibration of models and the
 simulation of historical patterns of discharge. The projected regional impacts of climate change
 will be obtained from the North American Regional Climate Change Assessment Program
 (NARCCAP) dynamically downscaled 50 kilometer Regional Climate Model (RCM) output,
 primarily at monthly resolution. The original proposal was to use 15 General Circulation Model
 (GCM)-RCM combinations to simulate a range of future projections, although due to restrictions
 on the likely availability of NARCCAP downscaled data, a combination of NARCCAP,
 statistically downscaled CMIP3 projections and direct GCM outputs will likely be used. In each
 instance, the model-projected  changes in temperature and precipitation patterns will be used to
 modify the historical climate records using CAT. The advantages to this approach include the
 preservation of short-timescale variability and other aspects of time series behavior, and the
 preservation of inter-site variability and correlation patterns, none of which are feasible using
 downscaled GCM outputs directly as model inputs.

 CAT permits the sequential modification of weather records to introduce a number of alterations,
 each reflecting various assumptions concerning the regional manifestations of climate change.
 Precipitation records can be modified by (1) multiplying all  records by  an empirical constant
 reflecting projected climate change to simulate a shift in total precipitation, applied uniformly to
 all periods and intensity classes, (2) selective application of such a multiplier to  specific seasons
 or months, (3) selective application of the multiplier to a range of months or years within the
 record, (4) selective application of the multiplier to storm events of a specific size or intensity
 class; and (5) addition or removal of storm events to simulate changes in the frequency of
                                       C-l

-------
precipitation events (U.S. EPA, 2009). Modification (4) can be iteratively applied to more than
one event size class. In summary, changes in frequency and intensity as well as changes in overall
precipitation accumulation can be represented using CAT and historical records.

Relative changes in the frequency and intensity of precipitation events associated with climate
change may prove to be more influential in determining future patterns of discharge than
projected changes in overall (annual, seasonal) precipitation. In particular, the partitioning of
precipitation into re-evaporation, runoff and percolation to  groundwater is understood to be
sensitive to the intensity and timing of precipitation events. Thus, to ensure that model
simulations embody the most important dimensions of projected climate change, particular
attention should be paid to precipitation intensity-frequency-duration (IFD) relationships.  As a
general pattern, the warming of the lower atmosphere is projected to lead to a more vigorous
hydrologic cycle, characterized by increases in global precipitation, and proportionally larger
increases in high-intensity events (Trenberth et al., 2007). This memorandum is intended to
provide a summary review of recent literature to address the following questions: (1) How
should precipitation change as a consequence of lower atmosphere warming? (2) What is the
historical evidence for increases in  precipitation intensity over the United States? (3) What do
climate models project with respect to precipitation frequency and intensity? (4) What are the
important limitations in these projections? and (5) What are the implications for the development
of meteorological time series used in the modeling study?

1.     How should precipitation  change as a consequence of lower atmosphere warming?

Physical arguments predicting increases in precipitation intensity as a  consequence of the
warming of the lower atmosphere are presented by Trenberth et al. (2003). The basic argument
can be summarized as follows: (1) The primary conditions  for precipitation to occur include
(a) availability of precipitable moisture in the atmosphere and (b) a mechanism for lifting  and
cooling parcels of air, leading to condensation and precipitation.  (2) Progressive warming of the
land surface and lower atmosphere  (i.e., climate change) will lead to increases in atmospheric
(precipitable) moisture through the  positive relationship between air temperature and saturation
vapor pressure (moisture-holding capacity). The Clausius-Clapeyron equation quantifies this
relationship, and can be used to predict an increase in atmospheric water holding capacity of
around 7% per °C at current global mean temperatures.l (3) Precipitation, when it occurs, often
exceeds the extractable fraction (typically below 30%) of available moisture in the immediate
zone of precipitation. This reflects the role  of low-level convergence in drawing moist air  into
convective zones from surrounding areas. Trenberth et al. (2003) calculate that as  an
approximate global average, a zone of precipitation is supported  by a larger region - roughly three
to five times the radius of the precipitation  zone - from which it  draws moisture. Assuming no
significant change in the efficiency of precipitation generation (i.e., maximum rate of extraction of
water from the atmosphere), the intensity of such events should therefore increase as a function of
mean temperature at roughly the same rate  as atmospheric water-holding capacity (i.e., around 7%
per °C). Finally (4), the increased atmospheric moisture supply also provides additional latent heat
                                           C-2

-------
to drive the convective process, further enhancing low-level convergence. Thus, increases in
atmospheric temperature lead to proportional increases in precipitation intensity when other
conditions required for initiating convective precipitation are present. Trenberth et al. (2003) note
that the GCMs supporting the Intergovernmental Panel on Climate Change (TPCC) assessment
simulate increases in precipitation of around 1% to 2% per °C.2 This suggests (assuming these
GCMs accurately simulate the atmospheric water balance) that the increase in precipitation
intensity predicted  by Clausius-Clapeyron (7% per °C) for intense convective events must be
compensated for by reductions in the frequency and/or intensity of light to moderate intensity
events. The latter hypothesis assumes that the durations of intense precipitation events do not also
change significantly as a function of temperature.

2.     Historical evidence of increases in precipitation intensity over the United States

If the proposed relationship between increasing air temperatures and precipitation intensity is
theoretically sound, then the predicted changes should already be evident due to observed
increases in global  and U.S. air temperatures. Analysis of instrumental records from 1850-2005
indicates that globally-averaged temperatures have increased by 0.76°C (+/- 0.19°C) over this
period, with the most rapid warming occurring in the last 50 years and the steepest increase in
global temperatures, equivalent to changes of+0.177°C per decade, occurring over the last
25 years (Trenberth et al., 2007). Within the United States, temperature increases have also been
observed at rates exceeding the global average. Present (1993-2008) U.S. temperatures are on
average over 1.1°C warmer than during the 1961-1979 period (Karl et al., 2009). Corresponding,
increasing trends in evaporation, atmospheric moisture and precipitation, particularly high-
intensity precipitation, should thus be in evidence. However,  it is not necessarily the case that  all
of these increases (if observed) should be of the magnitude predicted by Clausius-Clapeyron (7%
per °C)  since globally, evaporation is controlled by the availability of surface moisture (over
land) and by the availability of energy at the earth's surface to drive evaporation and
transpiration (Allen and Ingram, 2002).

Evaporation: Among the predicted impacts of a warming lower atmosphere, increases in actual
evaporation and transpiration (evapo-transpiration, or ET) have been the most difficult to
demonstrate, largely due to the relative absence of long-term records of direct ET measurements
(Lettenmaier et al., 2008). Physical theory (the Clausius-Clapeyron equation) predicts an
increase in potential ET., since a supply  of moisture available for ET cannot be assumed. A
relatively small number of recent land-based studies in the United States, India, China and
Australia that make use of long-term evaporation pan data conclude that actual evaporation rates
2. Trenberth et al. (2003) refer to the GCMs supporting the Third IPCC Assessment (2001).
                                           C-3

-------
have decreased. One proposed explanation for this paradox is a reduction in incoming solar
radiation due to increases in aerosols associated with air pollution (Trenberth et al., 2007).
Alternatively, Brutsaert and Parlange (1998) conjecture that as humidity supplied by the
surrounding landscape increases, pan evaporation will decrease (a reverse of the "oasis effect").
A synthesis of water balance studies of several major North American watersheds (Walter et al.,
2004), in which ET was estimated as the residual of precipitation and discharge, concludes that
actual ET has increased over the last 50 years. More direct evidence of a temperature-induced
increase in actual evaporation is provided by Yu and Weller (2007). These authors utilized
satellite remote sensing and atmospheric model re-analysis to estimate trends in evaporation over
the ocean surface, where moisture supply is not limited. They estimate that globally averaged
ocean latent heat flux (evaporation) has increased by approximately 10% over the 25-year period
1981-2005. This reflects increases in both atmospheric moisture capacity (Clausius-Clapeyron)
and sea surface temperature (SST).

Atmospheric moisture: Clausius-Clapeyron predicts an increase in absolute or specific humidity
(q) with increasing temperature, as distinct from relative humidity (RH). Climate model
simulations tend to indicate that temperature-related changes in RH are small (Trenberth et al.,
2003). Balloon-borne radiosonde has been used to estimate altitude-integrated RH, although time
series analysis based on radiosonde is subject to a number of constraints. Specifically, the density
of radiosonde observations is low, observations are unavailable over the open ocean and
radiosonde sensors have changed over time, confounding  efforts to measure decadal-scale
changes in  atmospheric water content (Dai, 2006b), effectively limiting analysis to the mid-
1970s and onward. Nevertheless, Ross and Elliott (1996; 2001) used radiosonde time series
records to estimate changes in RH and precipitable water  (up to 500 mb) over North America in
recent decades. They found that precipitable water has increased by 3% to 7% per decade
between  1973 and 1995 over the area ranging from the Caribbean to 45°N, with greater increases
in the south and smaller increases in the north. Above 45°N, changes were either uncertain or
negative  over this period.  Ross and Elliott (2001) note also that these changes appear greater and
more uniform over North America than over Eurasia.

Dai (2006b) evaluated changes in surface specific q and RH using a much wider range of
sensors, located both over land and over ocean. Near-surface measurements do not provide
altitude-integrated estimates of q and RH, although the spatial sampling is greatly improved
relative to Ross and Elliott (1996; 2001) due to the large number of records (over 15,000 surface
and ocean weather stations), and the sensor technology is  more consistent over the period of
record. Dai (2006b) found that globally averaged specific humidity (q)  increased by around
0.06 g kg"1  per decade over the 1976-2004 period. This corresponds to roughly 4.9% per degree
(°C) of warming over that period, globally averaged. The  response of q to temperature increases
over water  (i.e., not source-limited) was found to be around 5.7% per °C of warming, reasonably
consistent with the predictions of Clausius-Clapeyron (7% per °C) at constant RH. By contrast,
the response over land is around 4.3% per °C of warming, presumably reflecting spatio-temporal
                                           C-4

-------
limitations in water available for evaporation. Finally, column-integrated estimates of
atmospheric water vapor have been available since 1988 from the special sensor microwave
imager (SSM/I). On the basis of SSM/I, Trenberth et al. (2005) estimate that over the period
1988-2003,  altitude-integrated atmospheric precipitable water over oceans has increased by
around 0.40  mm per decade (1.3%). Variability over the period of analysis was found to reflect
variations in SST. Assuming relatively constant RH, the observed trend is reasonably close to the
7% per °C predicted by Clausius-Clapeyron.

Precipitation: Increases in the frequency and intensity of heavy precipitation events over the last
several decades are among the most clearly documented changes in recent U.S. climate (Kunkel,
2008). The following studies are representative of several recent studies examining trends in
precipitation in the United States and globally. Karl and Knight (1998) found that precipitation
over the U.S. increased by around  10% between the 1910s and the 1980s. These authors
examined the respective contributions of changes in both frequency and intensity of precipitation
to changes in total precipitation. Precipitation events were disaggregated into 20 intensity
classes, each encompassing 5% of observed events; and extreme intensity events,  defined as
precipitation exceeding 2 inches (50.4) mm per day, were also examined. Among  their
conclusions, Karl and Knight (1998) found that observed increases reflect both increased
frequency and intensity of rainfall  events. While the frequency of events increased for all
intensity (percentile) classes, intensity increased for heavy and extreme precipitation days only,
and the proportion of total annual precipitation attributable to these heavy and extreme events
has increased relative to more moderate events. Specifically, over half (53%) of the observed
increase was due to increases in the upper 10% of events. Karl and Knight (1998) also found that
the percentage of total area within the U.S. experiencing extreme precipitation events
(> 50.4 mm/day) had increased by roughly 20% between 1910 and the mid-1990s. Kunkel et al.
(1999) found statistically significant increasing trends in 1-year and 5-year return  period 7-day
precipitation events in the United States. However, subsequent work (Kunkel et al., 2003)
extended the period of record back to 1895, and the frequency of extreme events in the late 19th-
early 20th century was found to be similar to the late 20th Century, suggesting that natural
variability cannot be ruled out as an additional factor contributing to the observed late
20th century increases in  intensity.

Groisman et al. (2004) examined trends in several climatologic and hydrologic variables for the
conterminous U.S. potentially influenced by climate change, including total precipitation,
precipitation intensity, temperature and  streamflow. Heavy precipitation events, defined as the
upper 5% of daily events, increased by 14% over the period 1908-2002. Very heavy events
(upper 1 %)  increased by 20% over this period, and extreme events (upper 0.1 %) by 21%. The
most significant increases occurred in the upper and lower Midwest for annual events, the upper
Midwest and Great Lakes areas for summer events, and in New England for winter events.
Similar results are presented in the global analysis of Groisman et al. (2005) for the period 1910-
1999, who found that while total annual precipitation volumes over the United States increased
                                           C-5

-------
by 1.2% per decade over the period 1970-1999, the share of annual precipitation associated with
extreme events (defined as above) increased by 14% per decade over this period. These authors
note that " ...practically the entire nationwide increase in heavy and very heavy precipitation
occurred during the last three decades" (p. 1328). Alexander et al. (2006), examining global
precipitation statistics for the period 1951-2003, reached similar conclusions, specifically, that
the contribution to total annual precipitation from very wet days, defined as the upper 5% of
daily precipitation events, has increased over this period, even in many areas where total
precipitation has decreased.

3.     Model projections of trends in precipitation intensity

Evaporation and transpiration are in many circumstances controlled by factors other than the
moisture-holding capacity of the atmosphere, including availability of moisture supply over land
areas and energy available to drive the ET process (e.g., Allen and Ingram, 2002). Thus, GCMs
generally predict increases in the global hydrologic cycle that are more modest than the 7% per
°C predicted by Clausius-Clapeyron (Trenberth et al., 2003). Sun et al. (2007) have summarized
changes in total global precipitation, precipitation frequency, intensity, fraction of precipitation
from convective events and other related variables as projected by 17 of the most recent
generation of GCMs from the Program for Climate Model Diagnosis and Intercomparison
(PCMDI), used in the IPCC AR4 (2007), for emissions [Special Report on Emissions Scenarios
SRES] scenarios Bl (low), A1B (medium) and A2 (high).  Ensemble results,  averaged over
models and scenarios, indicate that global mean precipitation is projected to increase by around
1.2% per °C, and latent heat flux (evaporation) by a comparable amount, although global
precipitable  water is projected to increase by around 9.1% per °C.  These results indicate that the
atmospheric state variable (atmospheric precipitable water) responds approximately as predicted
by Clausius-Clapeyron (consistent with relatively small increases in average RH), while
atmospheric water fluxes (ET, precipitation) are constrained by other factors. Sun et al. (2007)
report that overall, the frequency of (daily) precipitation events is projected to decrease, and the
intensity of events to increase on average, consistent with Trenberth et al. (2003). However, the
frequency of heavy precipitation events is projected to increase, indicating a more dramatic
reduction in the frequency of light precipitation events. Thus, heavy (20-50 mm day"1) and very
heavy (> 50  mm day"1) precipitation events are projected to contribute a disproportionately
increasing share of total  precipitation, through the combined effects of increased frequency and
increased magnitude, with frequency effects more influential than intensity effects. These
projected impacts are most pronounced under SRES scenario A2 (high emissions). These authors
acknowledge that the (simulated) increases in intensity may not be fully captured in an analysis
based on daily precipitation totals, since high-intensity events are often of shorter duration.

Tebaldi et al. (2006) have also examined model-simulated changes in extreme events,
encompassing both temperature and precipitation events, on the basis of nine GCMs included in
the IPCC AR4 (2007). These authors use a set of indicators of precipitation intensity proposed by
                                           C-6

-------
Frich et al. (2002) including the following: (1) frequency of days with precipitation exceeding
10 mm, (2) maximum 5-day precipitation total, (3) mean precipitation intensity (total
precipitation divided by number of days with precipitation exceeding 1 mm) and (4) fraction of
total precipitation due to events exceeding the 95th percentile. Significant increases in each of
these four indices were projected by the GCMs evaluated, although not all trends were
statistically significant. Significant (increasing) trends were associated with mid- to high
latitudes in the Northern Hemisphere as well as some tropical areas within South America and
Africa. Tebaldi et al. (2006) conclude that "Models (also) agree with observations over the
historical period that there is a trend towards a world characterized by intensified precipitation,
with a greater frequency of heavy-precipitation and high-quantile events, although with
substantial geographic variability" (p. 206).

4.     How well do GCMs capture the frequency-intensity relationships observed in actual
       precipitation?

In evaluating the model-generated evidence that precipitation intensity is likely to increase as a
consequence of increasing tropospheric temperatures, it is important to recognize that many
GCMs display well-documented biases with regard to precipitation frequency and intensity.
Specifically, there is a tendency for GCMs to generate too many low-intensity events, and to
under-simulate the intensity of heavy events. There are several possible reasons for this (since
there are several convective precipitation parameterization schemes in use) although problems
associated  with the simulation of the diurnal cycle appear to play an important role (Dai, 1999;
Trenberth et al.,  2003). If conditions for the onset of moist convection in models are biased or
poorly specified, convection occurs too early in the diurnal cycle, weaker convection results in
less vigorous precipitation, and the removal of atmospheric moisture reduces the likelihood of
more intense convective events subsequently (Trenberth et al., 2003).

Sun et al. (2006) compared the performance of 18 coupled Atmosphere-Ocean General
Circulation Models (AOGCMs) used in the IPCC AR4 (2007) in simulating precipitation with
historic observational data. Most of the models were found to greatly overestimate the frequency
of summer (June-August) light precipitation events in the Northern Hemisphere, although the
frequency of heavy precipitation events was simulated with greater skill, each subject to regional
variations.  Sun et al. (2006) summarize their observations as  follows: "For light precipitation,
most of the models greatly overestimate the frequency but reproduce the observed patterns of
intensity relatively well. For heavy precipitation, most of the  models roughly reproduce the
observed frequency but underestimate the intensity" (p.  928,  emphasis in original). Light
precipitation is defined  as 1-10 mm day"1 and heavy precipitation as >10 mm day"1. These
authors emphasize the importance of getting precipitation "right for the right reasons" since
surface runoff, evaporation and soil moisture are all highly sensitive to precipitation IFD
relationships. Dai (2006a) also examines the performance of  18 models from the PCMDI (AR4)
ensemble with respect to the characterization of precipitation. Dai's (2006a) study emphasizes
                                           C-7

-------
model skill in simulating precipitation via the distinct convective and stratiform mechanisms,
noting that (in GCMs) stratiform precipitation is a grid-scale process while convective
precipitation is a subgrid-scale process. Model performance is compared to the Tropical Rainfall
Measuring Mission (TRMM) observational data. Among other results, Dai (2006a) found that
the models examined derived too much of total precipitation (28%-38%) from light precipitation
(1-5 mm day"1) relative to TRMM data (19%), and far too little (0-2%) from very heavy
precipitation (> 50 mm day"1) (7% for TRMM). Model replication of TRMM results were best
for moderate events, defined as 10-20 mm day"1. Dai (2006a) concludes that".. .the newest
generation of Coupled Ocean-Atmosphere General Circulation Models (CGCMs) still rains too
frequently, as in previous generations (...), mostly within the 1-10 mm  day"1 categories, while
heavy precipitation (> 20 mm day"1) occurs too rarely" (p. 4622).

5.     Implications for the ORD modeling study

On the basis of literature reviewed here, several observations can be made. First, the importance
of getting IFD relationships right cannot be over-emphasized. Analysis of historical data
indicates that changes in the distribution of precipitation between light-  and heavy-intensity
events are quantitatively greater than changes in overall precipitation at annual or seasonal levels
in many regions, and projected runoff estimates are likely to be quite sensitive to these IFD
relationships. Second, model-generated projections of precipitation are characterized by
documented biases with  respect to precipitation intensity and frequency. This suggests that the
relative changes in precipitation IFD relationships should be used as the basis for adjusting
historical precipitation records in CAT rather than their absolute levels.

NARCCAP has to date provided summaries of three GCM-RCM downscaled products intended
for use in modifying the  historical gauge records. These  datasets include changes in monthly
precipitation accumulation,  and changes in the contributions of precipitation by intensity class
for 25%, 50%, 70%, and 90% percentile classes. While this data is extremely useful, it is
recommended that we obtain a number of representative output series at finer time resolution -
down to 3-hourly as output by NARCCAP RCMs. Ideally, we would obtain series that sample
from at least three of the climatic zones associated with the pilot watersheds, and including the
upper Midwest (Minnesota) in particular, since many of the greatest observed changes in
precipitation intensity have occurred in this region. These time series would be processed to
obtain estimates of the change in frequency of events, by event size class,  to support the
appropriate use of CAT.  As emphasized above, the "deltas" (changes in projection period
relative to base period) would be the basis for CAT transformations rather than the projections
themselves, which potentially contain biases, as discussed.

References

Alexander, L.V., X. Zhang,  T.C. Peterson, J. Caeser,  B. Gleason, A.M.G.  Klein Tank, M.
Haylock, D. Collins, B. Trewin,  F. Rahimzadeh, A. Tagipour, K. Rupa Kumar, J. Revadekar, G.
                                           C-8

-------
Griffiths, L. Vincent, D.B. Stephenson, J. Burn, E. Aguilar, M. Brunet, M. Taylor, M. New, P.
Zhai, M. Rusticucci, and J.L. Vazquez-Aguirre. 2006. Global observed changes in daily climate
extremes of temperature and precipitation. Journal of Geophysical Research 111(D05109). doi:
10:1029/2005JD006290, 22 pp.

Allen, M.R. and WJ. Ingram. 2002. Constraints on future changes in climate and the hydrologic
cycle. Nature 419:224-232.

Brutsaert, W. and M.B. Parlange. 1998. Hydrologic cycle explains the evaporation paradox.
Nature 396:29-30.

Dai, A.  1999. Recent changes in the diurnal cycle of precipitation over the United States.
Geophysical Research Letters 26:341-344.

Dai, A. 2006a. Precipitation characteristics in eighteen coupled climate models. Journal of
Climate 19:4605-4630.

Dai, A. 2006b. Recent climatology, variability, and trends in global surface humidity. Journal of
Climate 19:3589-3606.

Frich, P., L.V. Alexander, P. Della-Marta, B. Gleason, M. Haycock, A.M.G.K. Tank, and T.
Peterson. 2002. Observed coherent changes in climatic extremes during the second half of the
twentieth century. Climatic Research 19:193-212.

Groisman, P.Y., R.W.  Knight, D. Easterling, T.R. Karl, G.C. Hegerle, and V.N. Razuvaev. 2005.
Trends in intense precipitation in the climate record. Journal of Climate 18:1326-1350.

Groisman, P.Y., R.W.  Knight, T.R. Karl, D. Easterling, B. Sun, and J.H. Lawrimore. 2004.
Contemporary changes of the hydrologic cycle over the contiguous United States: Trends
derived from in situ observations. Journal of Hydrometeorology 5:64-85.

Karl, T.R. and R.W. Knight. 1998.  Secular trends of precipitation amount, frequency, and
intensity in the United States. Bulletin of the American Meteorological Society 79(2): 231-241.

Karl, T.R., J.M. Melillo, and T.C. Peterson (eds.). 2009. Global Climate Change Impacts in the
United States. Cambridge University Press, New York.

Kunkel, K.E. (Convening Lead  Author). 2008. Observed changes in weather and  climate
extremes. Chapter 2 in Weather and Climate Extremes in a Changing Climate. Regions of Focus:
North America, Hawaii, Caribbean, and U. S.  Pacific Islands. U.S. Climate Change Science
Program Synthesis and Assessment Product 3.3.
                                          C-9

-------
Kunkel, K.E., K. Andsager, and D.R. Easterling. 1999. Long-term trends in extreme precipitation
events over the conterminous United States and Canada. Journal of Climate 12:2515-2527.

Kunkel, K.E., D.R. Easterling, K. Redmond, and K. Hubbard. 2003. Temporal variations of
extreme precipitation events in the United States: 1895-2000. Geophysical Research Letters
30(17):1900. doi:10.1029/2003GL018052.

Lettenmaier, D., D. Major, L. Poff, and S. Running. 2008. Water resources. In The effects of
climate change on agriculture, land resources, water resources, and biodiversity in the United
States. (Synthesis and Assessment Product 4.3.) A Report by the U.S. Climate Change Science
Program and the Subcommittee on Global Change Research. Washington, DC. 362 pp.

Ross, RJ. and W.P. Elliott. 1996. Tropospheric water vapor climatology and trends over North
America: 1973-93. Journal of Climate 9:3561-3574.

Ross, RJ. and W.P. Elliott. 2001. Radiosonde-based northern hemisphere tropospheric water
vapor trends. Journal of Climate 14:1602-1611.

Sun, Y., S. Solomon, A. Dai, and R.W. Portmann. 2006. How often does it rain? Journal of
Climate 19:916-934.

Sun, Y., S. Solomon, A. Dai, and R.W. Portmann. 2007. How often will it rain? Journal of
Climate 20:4801-4818.

Tebaldi, C., K. Hayhoe, J.M. Arblaster, and G.A. Meehl. 2006. Going to the extremes: An
intercomparison of model-simulated historical and future changes in extreme events. Climatic
Change 79:185-211.

Trenberth, K.E., J. Fasullo, and L. Smith. 2005. Trends and variability in column-integrated
atmospheric water vapor. Climate Dynamics 24:741-758.

Trenberth, K.E., A. Dai, R.M. Rasmussen, and D.B. Parsons. 2003. The changing character of
precipitation. Bulletin of the American Meteorological Society (September 2003): 1205-1217.

Trenberth, K.E., P.O. Jones, P. Ambenje, R. Bojariu, D. Easterling, A. Klein Tank, D. Parker, F.
Rahimzadeh, J.A. Renwick, M. Rusticucci, B. Soden, and P. Zhai. 2007. Observations: Surface
and atmospheric climate change. In Climate Change 2007: The Physical Science Basis,
Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change, S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt,
M. Tignor, and H.L.  Miller (eds.). Cambridge University Press,  Cambridge, UK and New York.
                                         C-10

-------
U.S. EPA. 2009. BASINS 4.0 Climate Assessment Tool (CAT): Supporting Documentation and
User's Manual. U.S. Environmental Protection Agency, Washington, DC. National Center for
Environmental Assessment, Office of Research and Development. 191 pp.

Walter, M.T., D.S. Wilks, J-Y. Parlange, andR.L. Schneider. 2004. Increasing
evapotranspiration from the coterminous United States. Journal of Hydrometeorology 5:405-
408.

Yu, L. and R.A. Weller. 2007. Objectively analyzed air-sea heat fluxes for the global ice-free
oceans (1981-2005). Bulletin of the American Meteorological Association doi: 10.1175/BAMS-
88-4-527. 527-539.
                                         C-ll

-------
Appendix D
Model Configuration, Calibration and
Validation

Basin: Apalachicola-Chattahoochee-Flint (ACF)
                    D-l

-------
Contents
Watershed Background	D-8
  Water Body Characteristics	D-8
  Soil Characteristics	D-11
  Land Use Representation	D-11
  Point Sources	D-16
  Meteorological Data	D-19
  Watershed Segmentation	D-22
  Calibration Data and Locations	D-24
HSPF Modeling	D-26
  Changes Made to Base Data Provided	D-29
  Assumptions	D-29
  Hydrology Calibration	D-31
  Hydrology Validation	D-36
  Hydrology Results for Larger Watershed	D-41
  Water Quality Calibration and Validation	D-53
  Water Quality Results for Larger Watershed	D-63
SWAT Modeling	D-66
  Changes Made to Base Data Provided	D-66
  Assumptions	D-66
  Hydrology Calibration	D-66
  Hydrology Validation	D-72
  Hydrology Results for Larger Watershed	D-77
  Water Quality Calibration and Validation	D-84
  Water Quality Results for Larger Watershed	D-93
References	D-96
                                               D-2

-------
Tables
Table 1.    Aggregation of NLCD land cover classes	D-14
Table 2.    Land use distribution forthe ACF River basin (2001 NLCD) (mi2)	D-15
Table 3.    Major point source discharges in the ACF River basin	D-16
Table 4.    Precipitation stations forthe ACF River basin model	D-19
Table 5.    Calibration and validation locations in the ACF River basin	D-25
Table 6.    Reservoirs represented in the ACF basin model	D-29
Table 7.    Seasonal summary at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
           period (HSPF)	D-34
Table 8.    Summary statistics at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
           period (HSPF)	D-36
Table 9.    Seasonal summary at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
           period (HSPF)	D-39
Table 10.   Summary statistics at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
           period (HSPF)	D-41
Table 11.   Seasonal summary: Model DSN 9001 vs. USGS 02358000 Apalachicola River at
           Chattahoochee, FL- calibration period (HSPF)	D-44
Table 12.   Summary statistics: Model DSN 9001 vs. USGS 02358000 Apalachicola River at
           Chattahoochee, FL- calibration period (HSPF)	D-46
Table 13.   Summary statistics (percent error) for all stations - calibration period 1993-2002 (HSPF)	D-47
Table 14.   Summary statistics (percent error) for all stations - validation period 1983-1992 (HSPF)	D-48
Table 15.   Model fit statistics (observed minus predicted) for monthly TSS loads using stratified
           regression	D-55
Table 16.   Relative errors (observed minus predicted), TSS concentration at USGS 02349605 Flint River
           at Ga 26, near Montezuma, GA (HSPF)	D-57
Table 17.   Model fit statistics (observed minus predicted) for monthly total phosphorus loads using
           stratified regression	D-58
Table 18.   Relative errors (observed minus predicted), total phosphorus concentration at USGS 02349605
           Flint River at Ga 26, near Montezuma, GA  (HSPF)	D-60
Table 19.   Model fit statistics (observed minus predicted) for monthly total nitrogen loads using a
           veraging estimator (HSPF)	D-61
Table 20.   Relative errors (observed minus predicted), total nitrogen concentration at USGS 02349605
           Flint River at Ga 26, near Montezuma, GA  (HSPF)	D-63
Table 21.   Summary statistics for water quality (observed minus predicted) for all  stations - calibration
           period 1999-2002 (HSPF)	D-64
Table 22.   Summary statistics for water quality (observed minus predicted) for all  stations - validation
           period 1986-1998 (HSPF)	D-65
Table 23.   Seasonal summary at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
           period (SWAT)	D-70
Table 24.   Summary statistics at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
           period (SWAT)	D-72
Table 25.   Seasonal summary at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
           period (SWAT)	D-75
Table 26.   Summary statistics at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
           period (SWAT)	D-77
Table 27.   Seasonal summary at USGS 02358000 Apalachicola River at Chattahoochee, FL- calibration
           period (SWAT)	D-80
Table 28.   Summary statistics at USGS 02358000 Apalachicola River at Chattahoochee, FL- calibration
           period (SWAT)	D-82
Table 29.   Summary statistics (percent error) for all stations - calibration period 1993-2002 (SWAT)	D-83
Table 30.   Summary statistics (percent error) for all stations - validation period 1983-1992 (SWAT)	D-84
                                                D-3

-------
Table 31.   Model fit statistics (observed minus predicted) for monthly TSS loads using stratified
           regression at USGS 02349605 Flint River at Ga 26, near Montezuma, GA (SWAT)	D-85
Table 32.   Relative errors (observed minus predicted), TSS concentration at USGS 02349605 Flint River
           at Ga 26, near Montezuma, GA (SWAT)	D-87
Table 33.   Model fit statistics (observed minus predicted) for monthly total phosphorus loads using
           stratified regression at USGS 02349605 Flint River at Ga 26, near Montezuma, GA (SWAT).... D-88
Table 34.   Relative errors (observed minus predicted), total phosphorus concentration, USGS 02349605
           Flint River at Ga 26, near Montezuma, GA (SWAT)	D-90
Table 35.   Model fit statistics (observed minus predicted) for monthly total nitrogen loads using
           averaging estimator at USGS 02349605 Flint River at Ga 26, near Montezuma, GA (SWAT).... D-91
Table 36.   Relative errors (observed minus predicted), total nitrogen concentration, USGS 02349605
           Flint River at Ga 26, near Montezuma, GA (SWAT)	D-93
Table 37.   Summary statistics (observed minus predicted) for water quality for all stations - calibration
           period 1999-2002 (SWAT)	D-94
Table 38.   Summary statistics (observed minus predicted) for water quality for all stations - validation
           period 1986-1998 (SWAT)	D-95

Figures
Figure 1. Location of the Apalachicola-Chattahoochee-Flint (ACF) River basin	D-10
Figure 2. Land use in the ACF River basin	D-13
Figure 3. Major point sources in the ACF River basin	D-18
Figure 4. Weather stations for the ACF River basin model	D-21
Figure 5. Model segmentation and USGS stations utilized for the ACF River basin	D-23
Figure 6. Parameter mapping utilized in the HSPF ACF River basin model	D-28
Figure 7. Mean daily flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
         period (HSPF)	D-32
Figure 8. Mean monthly flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
         period (HSPF)	D-32
Figure 9. Monthly flow regression and temporal variation at USGS 02349605 Flint River at Ga 26, near
         Montezuma, GA - calibration period (HSPF)	D-33
Figure 10. Seasonal regression and temporal aggregate at USGS 02349605  Flint River at Ga 26, near
         Montezuma, GA - calibration period (HSPF)	D-33
Figure 11. Seasonal medians and ranges at USGS 02349605 Flint River at Ga 26, near Montezuma, GA -
         calibration period (HSPF)	D-34
Figure 12. Flow exceedence at USGS 02349605 Flint River at Ga 26, near Montezuma,  GA - calibration period
         (HSPF)	D-35
Figure 13. Flow accumulation at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
         period (HSPF)	D-35
Figure 14. Mean daily flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation period
         (HSPF)	D-37
Figure 15. Mean monthly flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
         period (HSPF)	D-37
Figure 16. Monthly flow regression and temporal variation at USGS 02349605 Flint River at Ga 26, near
         Montezuma, GA - validation period (HSPF)	D-38
Figure 17. Seasonal regression and temporal aggregate at USGS 02349605  Flint River at Ga 26, near
         Montezuma, GA - validation period (HSPF)	D-38
Figure 18. Seasonal medians and ranges at USGS 02349605 Flint River at Ga 26, near Montezuma, GA -
         validation period (HSPF)	D-39
Figure 19. Flow exceedence at USGS 02349605 Flint River at Ga 26, near Montezuma,  GA - validation period
         (HSPF)	D-40
Figure 20. Flow accumulation at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation period
         (HSPF)	D-40
                                               D-4

-------
Figure 21. Mean daily flow: Model DSN 9001 vs. USGS 02358000 Apalachicola River At Chattahoochee, FL-
         calibration period (HSPF)	D-42
Figure 22. Mean monthly flow: Model DSN 9001 vs. USGS 02358000 Apalachicola River At Chattahoochee,
         FL-calibration period (HSPF)	D-42
Figure 23. Monthly flow regression and temporal variation: Model DSN 9001 vs. USGS 02358000 Apalachicola
         River At Chattahoochee, FL- calibration period (HSPF)	D-43
Figure 24. Seasonal regression and temporal aggregate: Model DSN 9001 vs. USGS 02358000 Apalachicola
         River At Chattahoochee, FL- calibration period (HSPF)	D-43
Figure 25. Seasonal medians and ranges: Model DSN 9001 vs. USGS 02358000 Apalachicola River At
         Chattahoochee, FL- calibration period (HSPF)	D-44
Figure 26. Flow exceedence: Model DSN 9001 vs. USGS 02358000 Apalachicola River At Chattahoochee,
         FL-calibration period (HSPF)	D-45
Figure 27. Flow accumulation: Model DSN 9001 vs. USGS 02358000 Apalachicola River At Chattahoochee,
         FL-calibration period (HSPF)	D-45
Figure 28. Histogram of simulated and measured elevation for Lake Lanier from 1/1/1993 to 12/31/2002	D-50
Figure 29. Histogram of simulated and measured elevation for West Point Lake from 1/1/1993 to 12/31/
         2002	D-51
Figure 30. Histogram of simulated and measured elevation for Lake Walter F. George from 1/1/1993 to
         12/31/2002	D-52
Figure 31. Histogram of simulated and measured elevation for Lake Seminole from 1/1/1993 to 12/31/2002.. D-53
Figure 32. Fit for monthly load  of TSS USGS 02349605 Flint River at Ga 26, near Montezuma, GA
         (HSPF)	D-54
Figure 33. Power plot for observed and simulated TSS at USGS 02349605 Flint River at Ga 26, near
         Montezuma, GA - calibration period (HSPF)	D-55
Figure 34. Power plot for observed and simulated TSS at USGS 02349605 Flint River at Ga 26, near
         Montezuma, GA- validation period (HSPF)	D-56
Figure 35. Time series plot of TSS concentration at USGS 02349605 Flint River at Ga 26, near Montezuma,
         GA - calibration period (HSPF)	D-57
Figure 36. Fit for monthly load  of total phosphorus at USGS 02349605 Flint River at Ga 26, near Montezuma,
         GA(HSPF)	D-58
Figure 37. Power plot for observed and simulated total phosphorus at USGS 02349605 Flint River at Ga 26,
         near Montezuma, GA  - calibration period (HSPF)	D-59
Figure 38. Power plot for observed and simulated total phosphorus at USGS 02349605 Flint River at Ga 26,
         near Montezuma, GA  - validation period (HSPF)	D-59
Figure 39. Time series plot of total phosphorus concentration at USGS 02349605 Flint River at Ga 26, near
         Montezuma, GA - calibration period (HSPF)	D-60
Figure 40. Fit for monthly load  of total nitrogen at USGS 02349605 Flint River at Ga 26, near Montezuma,
         GA(HSPF)	D-61
Figure 41. Power plot for observed and simulated total nitrogen at USGS 02349605 Flint River at Ga 26,
         near Montezuma, GA  - calibration period (HSPF)	D-62
Figure 42. Power plot for observed and simulated total nitrogen at USGS 02349605 Flint River at Ga 26,
         near Montezuma, GA  - validation period (HSPF)	D-62
Figure 43. Time series plot of total nitrogen concentration at USGS 02349605 Flint River at Ga 26, near
         Montezuma, GA - calibration period (HSPF)	D-63
Figure 44. Mean daily flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
         period (SWAT)	D-68
Figure 45. Mean monthly flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
         period (SWAT)	D-68
Figure 46. Monthly flow regression and temporal variation at USGS 02349605 Flint River at Ga 26, near
         Montezuma, GA- calibration period (SWAT)	D-69
Figure 47. Seasonal regression and temporal aggregate at USGS 02349605 Flint River at Ga 26, near
         Montezuma, GA- calibration period (SWAT)	D-69


                                               D-5

-------
Figure 48.  Seasonal medians and ranges at USGS 02349605 Flint River at Ga 26, near Montezuma, GA -
        calibration period (SWAT)	D-70
Figure 49.  Flow exceedence at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
        period (SWAT)	D-71
Figure 50.  Flow accumulation at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
        period (SWAT)	D-71
Figure 51.  Mean daily flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
        period (SWAT)	D-73
Figure 52.  Mean monthly flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
        period (SWAT)	D-73
Figure 53.  Monthly flow regression and temporal variation at USGS 02349605 Flint River at Ga 26, near
        Montezuma, GA- validation period (SWAT)	D-74
Figure 54.  Seasonal regression and temporal aggregate at USGS 02349605 Flint River at Ga 26, near
        Montezuma, GA - validation period (SWAT)	D-74
Figure 55.  Seasonal medians and ranges at USGS 02349605 Flint River at Ga 26, near Montezuma, GA -
        validation period (SWAT)	D-75
Figure 56.  Flow exceedence at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
        period (SWAT)	D-76
Figure 57.  Flow accumulation at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
        period (SWAT)	D-76
Figure 58.  Mean daily flow at USGS 02358000 Apalachicola River at Chattahoochee, FL- calibration period
        (SWAT)	D-78
Figure 59.  Mean monthly flow at USGS 02358000 Apalachicola River at Chattahoochee, FL- calibration
        period (SWAT)	D-78
Figure 60.  Monthly flow regression and temporal variation at USGS 02358000 Apalachicola River at
        Chattahoochee, FL- calibration period (SWAT)	D-79
Figure 61.  Seasonal regression and temporal aggregate at USGS 02358000 Apalachicola River at
        Chattahoochee, FL- calibration period (SWAT)	D-79
Figure 62.  Seasonal medians and ranges at USGS 02358000 Apalachicola River at Chattahoochee, FL-
        calibration period (SWAT)	D-80
Figure 63.  Flow exceedence at USGS 02358000 Apalachicola River at Chattahoochee, FL- calibration
        period (SWAT)	D-81
Figure 64.  Flow accumulation at USGS 02358000 Apalachicola River at Chattahoochee, FL- calibration
        period (SWAT)	D-81
Figure 65.  Fit for monthly load of TSS at USGS 02349605 Flint River at Ga 26, near Montezuma, GA -
        calibration period (SWAT)	D-85
Figure 66.  Power plot for observed and simulated TSS at USGS  02349605 Flint River at Ga 26, near
        Montezuma, GA- calibration period (SWAT)	D-86
Figure 67.  Power plot for observed and simulated TSS at USGS 02349605 Flint River at Ga 26, near
        Montezuma, GA - validation period (SWAT)	D-86
Figure 68.  Time series plot of TSS concentration at USGS 02349605 Flint River at Ga 26, near Montezuma,
        GA(SWAT)	D-87
Figure 69.  Fit for monthly load of total phosphorus at USGS 02349605 Flint River at Ga 26, near Montezuma,
        GA(SWAT)	D-88
Figure 70.  Power plot for observed and simulated total phosphorus at USGS 02349605 Flint River at Ga 26,
        near Montezuma, GA- calibration period (SWAT)	D-89
Figure 71.  Power plot for observed and simulated total phosphorus at USGS 02349605 Flint River at Ga 26,
        near Montezuma, GA - validation period (SWAT)	D-89
Figure 72.  Time series plot of total phosphorus concentration at USGS  02349605 Flint River at Ga 26, near
        Montezuma, GA (SWAT)	D-90
Figure 73.  Fit for monthly load of total nitrogen  at USGS 02349605 Flint River at Ga 26, near Montezuma,
        GA(SWAT)	D-91

                                              D-6

-------
Figure 74. Power plot for observed and simulated total nitrogen at USGS 02349605 Flint River at Ga 26,
         near Montezuma, GA - calibration period (SWAT)	D-92
Figure 75. Power plot for observed and simulated total nitrogen at USGS 02349605 Flint River at Ga 26,
         near Montezuma, GA - validation period (SWAT)	D-92
Figure 76. Time series plot of total nitrogen concentration at USGS 02349605 Flint River at Ga 26, near
         Montezuma, GA (SWAT)	D-93
                                               D-7

-------

The Apalachicola-Chattahoochee-Flint (ACF) River basin lies in Georgia, Alabama, and Florida, empties into the
Gulf of Mexico at Apalachicola Bay (Figure 1). It is comprised of 12 HUC8 cataloging units, and stretches across
parts of three geological physiographic provinces. The ACF basin along with the Alabama-Coosa-Tallapoosa
(ACT) River basin, are the central focus of water war that has been ongoing for over 20 years. The states of
Georgia, Alabama, and Florida have been involved in a legal controversy over the fair management of the waters
that these states share (Alabama River Alliance 2007).

Approximately 64 percent of the basin is forested. Approximately 25 percent of these forests are timberlands used
for manufacturing wood products. Agricultural land represents a mix of cropland, pasture, orchards, and areas of
confined feeding for poultry and livestock production. The dominant agricultural land use in the Piedmont
Province is pasture and confined feeding for dairy or livestock production. Most of the poultry operations in the
ACF River basin are concentrated in the upper part of the Chattahoochee River basin. Row-crop agriculture,
orchards, and silviculture  are most common in the Coastal Plain areas. Common crops in the watershed include
peanuts, corn, soybeans, wheat, and cotton. The largest concentration of urban land in the basin is in the Atlanta
area. Nearly 90 percent of the total population in the basin lives  in Georgia, and nearly 75 percent live in the
Atlanta metropolitan area.

The ACF River basin is characterized by a warm and humid, temperate climate. Precipitation is greatest in the
mountains and near the Gulf of Mexico, lowest in the center of the basin. Average annual precipitation in the
basin is about 55 inches, but ranges from a low of 45 inches in the east-central part of the basin to a high of 60
inches in the Florida panhandle. Throughout the ACF River basin, low flows usually occur from September to
November and peak flows usually occur from January to April when rainfall is high and evapotranspiration is
low.

The watershed is underlain by five major aquifer systems: crystalline rock aquifers in the Blue Ridge and
Piedmont physiographic provinces, and four aquifer systems in the Coastal Plain physiographic province.
Watershed hydrology is influenced by 16  reservoirs, 13 of which are on the Chattahoochee River. These
reservoirs play a major role in controlling flow and influencing the quality of water in the watershed (Couch
1993).


Water Body Characteristics

Chattahoochee River
The Chattahoochee River is 430 miles long, drains an area of 8,770 mi2, and has an average discharge of 11,500
cubic feet per second (cfs). The river begins in the Blue Ridge Province in the mountainous region of northeast
Georgia, which is characterized by steep topography and relatively high precipitation and runoff. Annual
precipitation ranges from  53 to 70 inches and annual runoff from 27 to 37 inches. The part of the Chattahoochee
River watershed in the Blue Ridge Province is underlain by crystalline rock, and surface water in the area is
siliceous and low in natural mineral content (Couch 1993).

Thirteen of 16  dams on mainstem locations in the ACF River basin are on the Chattahoochee River. Dam
construction in the watershed began in the early 1800s on the Chattahoochee River above the Fall Line at
Columbus,  Georgia, to take advantage of natural gradients for power production. Pronounced decreases in the
frequency of high and low flows have occurred since the start of operation of Buford Dam, which forms Lake
Sidney Lanier. Lake Sidney Lanier, West Point Lake, and Lake Walter F. George provide most  water storage
available to regulate flows in the watershed. Lake Sidney Lanier alone provides 65 percent of conservation


                                                 D-8

-------
storage, although it drains only 5 percent of the ACF River basin. In addition, West Point Lake and Lake Walter
F. George provide 18 and 14 percent, respectively, of the watershed's conservation storage (USGS 2008).

Throughout most of its length, the Chattahoochee River is controlled by hydropeaking hydroelectric plants, which
contribute to power supply during peak periods of electric demand. From Cornelia, Georgia all the way down to
Lake Seminole, the hydrograph shows the influence of hydropeaking operations and these operations can result in
daily stage fluctuations of 4 feet or more (USGS 2008).

In contrast to the mainstem Chattahoochee River, many tributaries remain free flowing. Flows of tributaries in
forested watersheds are represented by Snake Creek and flows typical of urban watersheds are represented by
Peachtree Creek. Similar to most Piedmont streams, both streams have higher sustained flows during winter
months and show response to storm events throughout the year. However, sharper peaks in the hydrography of
Peachtree Creek reflect greater influence of impervious land cover in the urban watershed (USGS  2008).

Flint River
The Flint River, which is 340 miles long and drains an area of 8,460 mi2, has an average discharge of 9,800 cfs
and begins in the Piedmont Province near Atlanta's Hartsfield International Airport. In the upper part of the Flint
River watershed annual precipitation ranges from 44 to 59 inches, and annual runoff ranges from 10 to 39 inches.
The upper part of the Flint River watershed is characterized by both broad and narrow ridges separated by narrow
valleys. Natural surface water quality in the part of the watershed in the Piedmont Province is similar to that in the
Blue Ridge Province, but the water generally has higher concentrations of dissolved minerals and higher turbidity
(Couch 1993, Cherry 1961).

Most of the larger tributaries in the ACF River basin are located in the Coastal Plain Province part of the Flint
River watershed. These tributaries include Ichawaynochaway Creek, Chickasawhatchee Creek, Kinchafoonee
Creek , and Muckalee Creek.

Apalachicola River
The Chattahoochee and Flint Rivers flow through the Piedmont and Coastal Plain Provinces to their confluence at
Lake Seminole where they form the Apalachicola River. The parts of these river watersheds that lie in the Coastal
Plain Province are underlain by unconsolidated sediments consisting of sand, gravel, and clay. Surface water
tends to be siliceous in the upper part of the Coastal Plain Province but is predominantly carbonate in
southwestern Georgia where it is in contact with limestone.  Rainfall in the lower Chattahoochee and Flint River
watersheds  ranges from 43 to 55 inches, annually. Rainfall in this area is rapidly absorbed by the permeable soils,
and annual runoff ranges from 12 to 28 inches (Couch 1993).

The Apalachicola River is 106 miles long and drains an area of about 2,400 mi2 in the lower Coastal Plain
Province. Because of the low gradient of the lower Coastal Plain Province, the channel of the Apalachicola River
meanders through a wide, swampy floodplain. The floodplain ranges in width from 0.6 miles below Lake
Seminole to 5 miles near its mouth, where the Apalachicola River flows through a system of distributaries to the
Apalachicola Bay. The Apalachicola River has an average discharge of 26,000 cfs (Couch 1993).
                                                 D-9

-------
                                                                                   Hydrography

                                                                                   Water (Nat Atlas Dataset)

                                                                                   US Census Populated Places

                                                                                   Municipalities (pop 2 50.000)

                                                                                   County Boundaries

                                                                              1   I  Watershed with HUCSs
                                                             Athenslarke
                                                             County
                                  iddle
                             Chattahoochee-
                              take Harding
 Upper
  Flint
03130005 \
                                             Ichawaynocha
                                                  3130009
                               Chattahoochee
                                      004
                                                           Lower Flint
                                                            03130008
      GCRP Model Areas - ACF River Basin
                 Base Map
Figure 1. Location of the Apalachicola-Chattahoochee-Flint (ACF) River basin.
                                                   D-10

-------
Soil Characteristics

The ACF River basin contains parts of the Blue Ridge, Piedmont, and Coastal Plain physiographic provinces that
extend throughout the southeastern United States. Similar to much of the Southeast, the watershed's physiography
reflects a geologic history of mountain building in the Appalachian Mountains, and long periods of repeated land
submergence in the Coastal Plain Province. Physiography within the major provinces is not homogeneous and has
been subdivided by the states of Alabama, Florida, and Georgia. Although similar physiography may extend
across state boundaries, districts may be assigned different names by state geologists in each state (USGS 2008).

Three major soil orders, ultisols, entisols, and spodosols, and more than 50  soil series are present in the ACF
River basin. Ultisols are characterized by sandy or loamy surface horizons and loamy or clayey subsurface
horizons. These deeply weathered soils are derived from underlying acid crystalline and metamorphic rocks.
Entisols are young soils with little or no change from parent material and with poorly developed subhorizons.
These soils are frequently infertile and droughty because they are deep, sandy, well-drained, and subject to active
erosion. Spodosols are characterized by a thin sandy subhorizon underlaying the A horizon. This sandy
subhorizon is cemented by organic matter and aluminum. The ACF River basin is similar to much of the
southeastern coastal plain in the dominance of ultisols. Entisols are found at and below the Fall Line and in the
Dougherty Plain; and spodosols are found in the Gulf Coast Lowlands (USGS 2008).

The 20 Watershed study utilized STATSGO soil survey hydrologic soil group (HSG) information during model
set-up. The descriptions of each hydrologic soil group are provided below.

       Group A Soils Have low runoff potential and high infiltration rates even when thoroughly wetted. They
                     consist chiefly of deep, well to excessively drained sands or gravels and have a high rate
                     of water transmission.

       Group B  Soils Have moderate infiltration rates when wet and consist chiefly of soils that are moderately
                     deep to deep, moderately well to well drained, and moderately fine to moderately course
                     textures.

       Group C  Soils Have low infiltration rates when thoroughly wetted and consist chiefly of soils having a
                     layer that impedes downward movement of water with moderately fine to fine structure.

       Group D Soils Have high runoff potential, very low infiltration rates and consist chiefly  of clay soils
                     with high swelling potential, soils with a permanent water table, soils with a claypan or
                     clay layer at or near the surface and shallow soils over nearly impervious material.


The ACF basin has all four HSGs in the watershed. The Upper and Middle  Chattahoochee and most of the Upper
Flint watersheds are dominated by hydrologic type B soils.  As both rivers reach and cross over the Fall Line, the
boundary between the Piedmont and Coastal Plain physiographic provinces, they flow through an area dominated
by HSG A soils. As the two rivers come together near Lake Seminole the soil distribution is equally split between
HSG A and B soils. The southernmost extents of the Apalachicola River are dominated by hydrologic soil type D.


Land Use Representation

Land use/cover in the watershed is based on the 2001 NLCD coverage (Figure 2). The 2001 NLCD land cover
was used in order to generate consistency amongst all models for the 20 Watershed project.

Chattahoochee River Watershed
The Chattahoochee River watershed above Lake Sidney Lanier is dominated by forested land with a majority of
the remaining land being pasture. As the Chattahoochee River flows out of Buford Dam to the southwest the
dominant land use starts shifting to urban. As the Chattahoochee River nears and flows through the Atlanta metro

                                                D-ll

-------
area, land use is almost entirely urban. After leaving the Atlanta Metro area, land use shifts back to forest
dominance, but with a greater amount of pasture than the area above Lake Sidney Lanier. Continuing down the
Chattahoochee and across the Fall Line, the dominant land use is still forest, but wetland areas begin to increase
while pasture areas begin to decrease. After the Chattahoochee leaves Lake Walter F. George, agriculture and
forest become equally dominant with wetlands still being prevalent. This land use/land cover pattern continues
until the Chattahoochee empties into Lake Seminole.

Flint River Watershed
The most northern portions of the Flint River watershed are almost entirely urban. As the Flint River flows south
the land use  shifts to predominately pasture and then to forest. Once the Flint River crosses the Fall Line,
agriculture become increasingly prevalent but there  is still a good portion of forest and an increase in wetlands.
The land uses of major tributaries to the Flint River are also chiefly comprised of agriculture, forest, and wetlands.
The Flint River immediately above Lake Seminole is mostly agriculture with a small portion of pasture and forest.

Apalachicola River Watershed
The Apalachicola River immediately below Lake Seminole is comprised of mostly forest, scrub land and pasture.
As the Apalachicola River flows south, the land use is almost entirely dominated by wetlands until the river
empties into the Gulf of Mexico. The Chipola, a major tributary to the Apalachicola River, is chiefly comprised of
pasture and barren lands to the north and forest and  wetlands to the south.
                                                 D-12

-------
                                                                                                                  Interstate
                                                                                                            ^^— Hydrography
                                                                                                            ^H Water (Nat. Atlas Dataset)
                                                                                                            I     | County Boundaries
                                                                                                                t State Boundaries
                                                                                                                  Watershed
                                                                                                            2001 NLCD Land Use
                                                                                                               I Open water
                                                                                                               ^ Developed, open space
                                                                                                               | Developed, low intensity
                                                                                                               I Developed, medium intensity
                                                                                                               I Developed, high intensity
                                                                                                               ^ Barren land
                                                                                                               I Deciduous forest
                                                                                                               Jj Evergreen forest
                                                                                                            |    | Mixed forest
                                                                                                               ^ Scrub/shrub
                                                                                                               ^ Grassland/hertaceous
                                                                                                               ~j PastureAiay
                                                                                                               | Cultivated crops
                                                                                                               ^ Woody wetlands
                                                                                                               ~j Emergent hertiaceous wetlands
   GCRP Model Areas - ACF River Basin
                Land Use Map
Figure 2. Land  use  in the ACF River basin.
                                                                   D-13

-------
NLCD land cover classes were aggregated according to the scheme shown in Table 1 for representation in the 20
Watershed model, and then overlain with the soils HSG grid. Pervious and impervious lands are specified
separately for HSPF, so only one developed pervious class is used, along with an impervious class. HSPF
simulates impervious land areas separately from pervious land. Impervious area distributions were also
determined from the NLCD Urban Impervious data coverage. Specifically, percent impervious area was
calculated over the entire watershed for each of the four developed land use classes. These percentages were then
used to separate out impervious land. NLCD impervious area data products are known to underestimate total
imperviousness in rural areas. However, the model requires properly connected impervious area, not total
impervious area, and the NLCD tabulation is assumed to provide a reasonable approximation of connected
impervious area. Different developed land classes are specified separately in SWAT. The WATER, BARREN,
DEVPERV, and WETLAND classes are not subdivided by HSG in HSPF; SWAT uses the built-in HRU overlay
mechanism in the ArcSWAT interface. The  distribution of land use in the watershed is summarized in Table 2.
Table 1.     Aggregation of NLCD land cover classes
NLCD Class
1 1 Water
12 Perennial ice/snow
21 Developed open space
22 Dev. Low Intensity
23 Dev. Med. Intensity
24 Dev. High Intensity
31 Barren Land
41 Forest
42 Forest
43 Forest
51-52 Shrubland
71-74 Herbaceous Upland
81 Pasture/Hay
82 Cultivated
91 -97 Wetland
98-99 Wetland
Comments
Water surface area
usually accounted for as
reach area






Deciduous
Evergreen
Mixed




Emergent & woody
wetlands
Aquatic bed wetlands (not
emergent)
SWAT class
WATR
WATR
URLD
URMD
URHD
UIDU
SWRN
FRSD
FRSE
FRST
RNGB
RNGE
HAY
AGRR
WETF, WETL,
WETN
WATR
HSPF (after processing)
WATER
BARREN, Assume HSG D
DEVPERV;
IMPERV
BARREN (D)
FOREST (A,B,C,D)
SHRUB (A,B,C,D)
GRASS (A,B,C,D), BARREN (D)
GRASS(A,B,C,D)
AGRI (A,B,C,D)
WETLAND, Assume HSG D
WATER
                                               D-14

-------
Table 2.     Land use distribution for the ACF River basin (2001 NLCD) (mi2)
HUC8
watershed
Upper
Chattahoochee
03130001
Middle
Chattahoochee
Lake Harding
03130002
Middle
Chattahoochee
Walter F. George
Reservoir
03130003
Lower
Chattahoochee
03130004
Upper Flint
03130005
Middle Flint
03130006
Kinchafoonee
Muckalee
03130007
Lower Flint
03130008
Ichawaynochaway
03130009
Spring"
03130010
Apalachicola
03130011
Chipola
03130012
Total
Open
water
66.6
82.8
75.9
8.7
28.3
17.3
4.0
6.1
4.6
36.6
14.8
10.4
356.1
Developed land9
Open
Space
226.5
218.6
116.5
41.4
143.7
62.6
40.0
62.5
31.4
33.5
25.9
68.4
1,071.0
Low
density
150.0
127.1
56.7
10.2
79.6
20.7
13.0
27.4
5.6
8.9
2.3
7.6
509.0
Medium
density
53.8
29.9
14.8
2.3
21.9
3.9
3.1
6.5
1.1
1.4
0.6
2.2
141.4
High
density
30.5
15.6
6.0
0.8
15.0
1.6
1.5
3.3
0.2
0.4
0.1
0.6
75.6
Barren
land
10.8
23.0
9.7
0.9
16.3
1.6
0.8
1.7
0.8
0.8
0.6
1.0
68.1
Forest
804.8
1,809.3
1,707.5
470.5
1,454.6
618.3
506.4
455.7
411.6
257.6
335.7
412.7
9,244.8
Shru bland
56.8
246.5
385.0
138.1
227.5
126.6
97.1
125.9
90.6
73.3
61.4
231.4
1,860.1
Pasture/Hay
176.0
388.1
157.9
94.8
376.2
120.4
83.9
82.2
82.0
70.1
25.7
100.7
1,758.2
Cultivated
0.1
1.3
142.8
294.0
79.6
416.2
232.7
348.5
303.9
347.5
35.8
194.2
2,396.6
Wetland
9.5
98.7
164.2
70.3
192.3
164.9
117.7
95.1
172.3
127.5
339.5
250.4
1,802.5
Total
1,585.5
3,040.9
2,837.0
1,132.1
2,635.0
1,554.1
1,100.1
1,214.9
1,104.0
957.7
842.2
1,279.6
19,283.2
 The percent imperviousness applied to
density(89.9%).
                                each of the developed land uses is as follows: open space (8.04%), low density (30.16%), medium density (60.71%), and high

 Delineation for Lake Seminole crossed HUC8 boundaries so whole watershed is represented in Spring HUC8.
                                                                  D-15

-------
The HSPF model is set up on a hydrologic response unit (HRU) basis. For HSPF, HRUs were formed from an
intersection of land use and hydrologic soil group, and then further subdivided by precipitation gage. Average
slopes (which tend to correlate with soils) were calculated for each HRU. Slopes in most of the watershed are
relatively mild (1-5 percent), therefore HSPF HRU's were not further subdivided by slope. The three HRUs above
Lake Lanier have average slopes of 15-24 percent, but since there were already three HRU's for four delineated
subwatersheds it was not further divided. The water land use area was adjusted to prevent double counting with
area described in HSPF reaches. SWAT HRUs are formed from an intersection of land use and SSURGO major
soils.




Facilities permitted under the National Pollutant Discharge Elimination System (NPDES) are, by definition,
considered point sources. For all models in the 20 Watershed application, it was assumed that minor dischargers
(below 1.0 MOD) were insignificant and, therefore, not included in the model setup and simulation. Data were
sought from the PCS database for the major dischargers in the ACF River basin and reflect the time period from
1991-2006. Facilities that were missing total nitrogen, total phosphorus, ortotal suspended solids (TSS)
concentrations were filled with a typical pollutant concentration value from literature based on Standard Industrial
Classification (SIC) classification. For the 20 Watershed application, the assumption was to use constant point
source flows and concentrations, for the entire simulation period, for each major discharge facility in the
watershed. Figure 3 presents the locations of the major point sources included in the models.

During the water quality calibration, it was noticed that assumptions used for total phosphorus at some facilities
were too high. An investigation into the point sources that had assumed values for total phosphorus was
conducted. It was found that point sources with assumed values for total phosphorus, that were too high, were
water pollution control plants (WPCP) and the assumed total phosphorus  concentration for those facilities was 7
mg/L. A new assumed value was needed for these facilities. The new assumed value was 1.5 mg/L, which is an
average of the total phosphorus concentration for WPCP's that do monitor for total phosphorus. It is assumed that
1.5 mg/L is a much better estimate of the true total phosphorus concentration coming out of WPCP's in the ACF
basin. The new assumed value was also applied to the SWAT simulation. Both the HSPF and SWAT models used
the exact same flows and concentrations for each of the major point sources included in the simulations for the
ACF basin.
Table 3.     Major point source discharges in the ACF River basin
NPDES ID
AL0000817
AL0022209
AL0022764
AL0023159
AL0024619
AL0024724
AL0059218
AL0061671
AL0072737
FL0002283
FL0026867
FL0031402
GA0000973
GA0001112
GA0001198
GA0001201
Name
MEADWESTVACO COATED BOARD INC
PHENIXCITYWWTP
DOTHAN CITY OF OMUSSEE WWTP
LANETT CITY OF WWTP
SOUTHERN NUCLEAR OPERATNG CO
EAST AL WATER LOWER VALLEY WTP
OPELIKA CITY OF EASTSIDE WWTP
EUFAULA CITY OF
DOTHAN CITY CYPRESS WWTP NEW
GULF PWR SCHOLZ STEAM
BLOUNTSTOWN-STP
FL STATE HOSPITAL
COLUMBUS WATER WKS-FT.BENNING
SCOVILL FASTENERS, INC.
USAF PLT #6 - LOCKHEED MARTIN
GA. PACIFIC CORP (GREAT S.P)
Design flow
(MGD)*
40.00
7.75
7.12
5.00
0.16
4.00
1.00
2.70
3.00
129.60
1.50
1.30
4.60



Observed flow
(MGD)
(1991-2006 average)
22.34
3.32
3.97
1.90
0.52
2.83
0.70
1.75
1.32
3.24
0.57
0.71
11.71
0.26
1.63
32.60
                                                D-16

-------
NPDES ID
GA0001619
GA0020052
GA0020079
GA0020168
GA0020486
GA0020516
GA0021156
GA0021326
GA0021458
GA0021482
GA0021504
GA0023167
GA0024040
GA0024333
GA0024503
GA0024678
GA0025381
GA0025585
GA0026077
GA0026140
GA0026158
GA0026433
GA0030121
GA0030341
GA0030686
GA0030791
GA0031721
GA0033511
GA0035777
GA0035807
GA0036951
GA0037222
GA0038369
GA0046019
GA0046655
GA0047201
GA0047767
Name
MERCK & CO -FLINT RVR PLNT
WEST POINT WPCP
THOMASTON-BELL CREEK WPCP
GAINESVILLE (LINWOOD DRIVE)
MONTEZUMAWPCP#2
COLUMBUS (SOUTH WPCP)
GAINESVILLE FLAT CR WPCP
DAWSON WPCP
ATLANTA (UTOY CREEK WRC)
ATLANTA (R.M. CLAYTON WPCP)
CORNELIA WPCP
BUFORD SOUTHSIDE WPCP
ATLANTA (SOUTH RIVER WRC)
FULTON CO-BIG CREEK WPCP
CORDELEWPCP
BAINBRIDGEWPCP
FULTON CO-CAMP CREEK WPCP
BLAKELYWPCP
DAHLONEGAWPCP
COBB CO-SUTTON WPCP
COBB CO.-SO. COBB WPCP
GWINNETT CO (CROOKED CRK WPCP)
THOMASTON-TOWN BRANCH WPCP
DOUGLASVILLE SOUTHSIDE WPCP
FULTON CO-JOHNS CREEK WPCP
GRIFFIN POTATO CR WPCP
NEWNAN WAHOO WPCP
DECATUR CO-IND. AIRPARK WPCP
PEACHTREE CTY (LINE CRK WPCP)
FAYETTEVILLE-WHITEWTER CR WPCP
LAGRANGE WPCP (LONG CANE CRK)
ALBANY (WPCP NO 2)
CLAYTON COUNTY (SHOAL CRK)
GUMMING WPCP
PEACHTREE CTY (ROCKAWAY WPCP)
DOUGLASVILLE (SWEETWATER CRK.)
AMERICUS MILL CRK, WPCP
Design flow
(MGD)*

1.00
2.00
2.70
1.95
42.00
10.20
2.50
40.00
100.00
3.00
2.00
48.00
24.00
5.00
2.50
13.00
1.32
1.44
40.00
40.00
36.00
2.00
3.25
7.00
2.00
3.00
1.50
2.00
3.75
12.50
32.00
4.40
2.00
4.00
3.00
4.40
Observed flow
(MGD)
(1991 -2006 average)
1.13
0.64
1.01
2.03
0.30
30.35
6.45
1.29
29.50
80.00
2.33
1.11
35.31
21.09
2.66
1.32
12.47
1.01
0.51
30.62
24.50
23.52
0.97
2.48
5.73
1.41
1.66
0.38
1.44
1.88
5.91
18.05
1.73
1.30
1.66
1.39
2.73
*Note: Facilities that
the state of Georgia
do not list a design flow are large industrial facilities.
and these permits do not report a design flow.
These industrial facilities have different permitting in
                                                         D-17

-------
                                                               SCQMLL-'FASTENERS
                                  DAHLONEGA
                                                            CORNE.lilA
                                                          GAINESVILLE
                f  '\  COBBCO.
                 DOUG~LATSVILLE
                FAYETTEVILLE
                                   FULTON GO.
                                       CLAYTON
                                       V     1
                                      PEACHTREE
                                          GRIFFIN'POTATO CR.
              NEWNAN WAHOO
               LAGRANGE
            WEST-POINT
                                         THOMASTON
             EAST AL WATER
             ^  PHENIX
                 COLUMBUS
                                                MONTEZUMA
                                                  AMERICUSMILLCR-.
            MEADWESTVACO
                                                               DAWSON
                                                             "\ 1       r*
                                                             ALBANY

                                                             MERCK
SOUTHERN NUCLEAR

     D'OTHAN
                . PACIFIC C.QRB
                    GULF PWR
                                              FL STATE HOSPITAL
                                                                      Point Sources
                                                                      Hydrography
                                                                      Interstate
                                                                      Water (Nat. Atlas Dataset)
                                                                      US Census Populated Places
                                                                      Municipalities (pop > 50,000)
                                                                      County Boundaries
                                                                      Watershed with HUCSs
                                BLOUNTSTOWN STP
                                    .	v—
                                    Florida
  GCRP Model Areas - ACF River Basin
        Major Point Sources
Figure 3. Major point sources in the ACF River basin.
                                           D-18

-------
Meteorological Data
The required meteorological data series for the 20 Watershed study are precipitation, air temperature, and
potential evapotranspiration. The 20 Watershed model does not include water temperature or algal simulation and
uses a degree-day method for snowmelt. These are drawn from the BASINS4 Meteorological Database (USEPA
2008), which provides a consistent, quality-assured set of nationwide data with gaps filled and records
disaggregated. Scenario application will require simulation over 30 years, so the available stations are those with
a common 30-year period of record (or one that can be filled from an approximately co-located station). A total of
37 precipitation stations were identified for use in the ACF basin model with a common period of record of
10/1/1972-9/30/2002 (Table 4 and Figure 4). Temperature records were sparse; where these were absent,
temperature was taken from nearby stations with an elevation correction. For each weather station, Penman-
Monteith reference evapo-transpiration was calculated for use in HSPF using observed precipitation and
temperature coupled with SWAT weather generator estimates of solar radiation, wind movement, cloud cover,
and relative humidity.

For the 20 Watershed model applications, SWAT uses daily meteorological data, while HSPF requires hourly
data. It is important to note that a majority of the meteorological stations available for the ACF basin are
Cooperative Summary of the Day stations that do not report sub-daily data. The BASINS4 dataset already has
versions of the daily data that have been disaggregated to an hourly time step using template stations. For each
daily station, this disaggregation was undertaken in reference to a single disaggregation template. Occasionally,
this automated procedure provides undesirable results, particularly when the total rainfall for the day is very
different between the subject station and the disaggregation template. This yields a small number of hourly
precipitation intensity estimates that are unrealistically high (e.g., much greater than the 100-year 1-hour event for
the region). This has only a small impact on the watershed-scale hydrologic calibration as gages are  influenced by
rainfall from multiple weather stations, but can introduce significant problems for the prediction of erosion and
sediment loads.
Table 4.     Precipitation stations for the ACF River basin model
COOP ID
GA094230
GA092283
GA091998
GA092408
GA096407
GA090444
GA092791
GA090451
GA096335
GA093570
GA094949
GA099506
GA099291
GA098661
GA098535
GA091425
GA092166
GA091372
GA095979
AL015397
GA090253
GA095394
AL012730
Name
HELEN
CORNELIA
CLERMONT 4 WSW
GUMMING 1 ENE
NORCROSS
ATLANTA BOLTON
DOUGLASVILLE 4 S
ATLANTA HARTSFIELD
INTERNATIONAL
NEWNAN 4NE
FRANKLIN
LA GRANGE
WOODBURY
WEST POINT
THOMASTON 2 S
TALBOTTON
BUTLER
COLUMBUS METRO AP
BUENA VISTA
MONTEZUMA
MIDWAY
AMERICUS 3 SW
LUMPKIN2SE
EUFAULA WILDLIFE REF
Latitude
34.6997
34.5181
34.4503
34.2214
33.9483
33.8236
33.7006
33.63
33.4428
33.2758
33.065
32.9839
32.8694
32.8664
32.6856
32.6525
32.5161
32.3178
32.2903
32.0597
32.0503
32.0306
32.0086
Longitude
-83.7261
-83.5286
-83.855
-84.1222
-84.2219
-84.4983
-84.7303
-84.4417
-84.7886
-85.0992
-85.0294
-84.5889
-85.1892
-84.3175
-84.5192
-84.1858
-84.9422
-84.5203
-84.0314
-85.4953
-84.2753
-84.7753
-85.0919
Temperature
Yes
Yes
No
No
No
No
No
Yes
Yes
No
Yes
No
Yes
Yes
Yes
No
Yes
No
No
No
Yes
Yes
Yes
Elevation
(ft)
1440
1470
1281
1306
1030
885
1002
1010
920
790
715
790
575
672
686
446
392
646
327
556
490
485
215
                                                 D-19

-------
COOP ID
GA092266
GA092570
GA092450
AL010008
GA093028
GA090140
GA090979
AL012377
GA091500
GA092153
GA090586
FL081544
FL089795
FL089566
Name
CORDELE
DAWSON
CUTHBERT
ABBEVILLE
EDISON
ALBANY 3 SE
BLAKELY
DOTHAN
CAMILLA 3 SE
COLQUITT2W
BAINBRIDGE INT PAPER
CHIPLEY
WOODRUFF DAM
WEWAHITCHKA
Latitude
31.9847
31.7819
31.7672
31.5703
31.5664
31.5339
31.3811
31.1942
31.1903
31.1681
30.8228
30.7836
30.7219
30.1192
Longitude
-83.7758
-84.4497
-84.7931
-85.2483
-84.7339
-84.1489
-84.9508
-85.3708
-84.2036
-84.7664
-84.6175
-85.4847
-84.8742
-85.2042
Temperature
Yes
No
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
No
Yes
No
Yes
Elevation
(ft)
308
355
461
456
294
180
268
275
175
153
190
130
107
42
D-20

-------
                                                                                           Legend
                                                                                                Weather Stations
                                                                                                Interstate
                                                                                                Hydrography
                                                                                                Water (Nat. Atlas Dataset)
                                                                                                US Census Populated Places
                                                                                           ^H  Municipalities (pop > 50,000)
                                                                                             ^]  County Boundaries
                                                                                                State Boundaries
                                                                                                Watershed with HUCSs
   GCRP Model Areas - ACF River Basin
            Weather Stations
             NAD_1983_Albers_meters
          Map produced 2-10-2010 - B. Tucker
 140
^•Kilometers
      100
     • Miles
                         TETRATECH
Figure 4. Weatherstations for the ACF River basin model.
                                                          D-21

-------
Watershed Segmentation
The ACF River basin was divided into 101 sub-watersheds for the purposes of modeling (Figure 5). The initial
calibration watershed (Upper Flint HUC) is highlighted. Each of the subwatershed delineations represents roughly
a HUC 10 scale watershed. Each of the major reservoirs in the ACF basin was delineated so that the each dam
outlet represents an individual watershed outlet. The delineations were done this way to ensure that any individual
lake was contained in one watershed and that the watershed was only represented by one outlet. The ACF 20
Watershed model encompasses the complete watershed and does not require specification of any boundary
conditions for application.
                                                D-22

-------
                                                                         Legend
                                                                          A  USGS Gages
                                                                         	 Hydrography
                                                                             Interstate
                                                                             Water (Nat. Atlas Dataset)
                                                                             US Census Populated Places
                                                                             Municipalities (pop > 50,000)
                                                                           ] County Boundaries
                                                                         I   I Initial Calibration Watershed
                                                                           ] Model Subbasins
                                       87  // n 53 VJ51
                                                            28 JL—02353000
     GCRP Model Areas - ACF River Basin
            Model Segmentation
    Figure 5. Model segmentation and USGS stations utilized for the ACF River basin.
Note: SWAT subwatershed numbering is shown; the HSPF model for this watershed uses the same
subwatershed boundaries with an alternative internal numbering scheme.
                                                    D-23

-------
Calibration Data and Locations

The ACF basin was selected as an early pilot site application because of previous modeling experience in parts of
the watershed and the state of Georgia. The specific site chosen for initial calibration was the Flint River at GA
26, near Montezuma, Georgia (USGS 02349605) (Table 5). This is a flow and water quality monitoring location
that approximately coincides with the pour point of the Upper Flint (03130005) 8-digit HUC (Figure 5). This
location was selected for several reasons: 1) there is a good set of flow and water quality data available, 2)
previous modeling efforts in nearby HUCS's were successful, and 3) investigations of land use, drainage area, and
percentage of drainage area controlled by flow control structures, compared with other USGS gage locations in
the ACF basin, identified this gage as the best possible choice.

There were an additional eight sites chosen for the whole ACF basin to check the performance  of the model.
These sites were chosen based on subwatershed delineation boundaries, land use, drainage area, flow control
structures, data completeness and location. The eight additional sites are in Table 5 and shown  spatially in Figure
5. The idea was to have some locations that were un-impacted by upstream flow control structures and
additionally also have some locations downstream of major reservoirs to check the model performance of the
reservoir simulation.

Three of the chosen sites were located at the outfall of two delineated subwatersheds. In reality, these gages are
slightly downstream of a tributary joining the mainstem. These sites were Flint River at Montezuma, Georgia,
Chattahoochee River near Cornelia, Georgia, and Flint River at Newton, Georgia. It is easy to add two flows
together to get the theoretical flow at the sampling location but two water quality concentrations cannot be
summed to get the theoretical concentration. In order to generate a theoretical concentration, constituent masses
must be added together and then divide by the summed volumes to determine what the water quality
concentration would be. Accordingly, in the SWAT application, constituent masses from two reaches were added
together, and then divided by the summed volumes to determine constituent concentration. The HSPF application,
dealt with this by combining the watersheds internally and generating one time series that represented the
hydrology and water quality where these subwatersheds merge. This method makes the assumption that the main
stem and tributary waters have fully mixed at the sampling location.

The gage for Peachtree Creek is not at the outlet of the subwatershed for Peachtree Creek. Both HSPF and SWAT
applications utilized an area weighting approach for this gage. The USGS published drainage area was 66 percent
of the drainage area of the subwatershed delineated for Peachtree Creek, so a multiplier of 0.66 was applied to the
time series at the output of Peachtree Creek. Although this is not exact, theoretically it should be close to reality at
the sampling location, because land use differences are insignificant for this watershed (90 percent urban).

The water quality data found in the NAWQA database for the chosen calibration and validation locations were
limited in certain situations. Therefore, additional data were utilized from Georgia Environmental Protection
Division (EPD).  Due  to earlier modeling work done in the state, these  data were readily available, vetted and
included in the water quality data for calibration and validation. While combining the NAWQA and EPD datasets
there was some overlap on a few dates. The data from both datasets were compared and they always agreed on
constituent values.  It was decided to keep the NAWQA data and remove the  EPD data. Georgia EPD submits
their monitoring data to NAWQA and in general the additional EPD data contributed to lengthening the period of
record available for calibration and validation.

Many of the locations chosen for calibration and validation did not specifically monitor for all constituents
making up total nitrogen. Many times reported values were only for ammonia and nitrate+nitrite. The sum of
those two constituents does not represent total nitrogen because it is missing  the component for organic nitrogen.
Because of this, the data available for total nitrogen were very limited. An approach was developed to bolster the
amount of total nitrogen data available for calibration and validation. The NAWQA database was investigated for
sampling dates that reported total nitrogen, ammonia, and nitrate+nitrite. These sampling dates and data were
extracted and a regression analysis of total nitrogen vs. ammonia+nitrate+nitrite was performed. The regression

                                                 D-24

-------
had an R2 vale of 0.80. Because the fit was high, the regression was applied to the ammonia+nitrate+nitrite value
and the result was an estimated value of total nitrogen based on two of the three components making up total
nitrogen. For the 20 Watershed application, it was assumed that an estimated value for total nitrogen was better
than having no value at all.

A database containing NAWQA, EPD, and calculated total nitrogen values was compiled and used in both the
SWAT and HSPF modeling applications. This ensured the data that both models were calibrated to were
consistent.

Table 5.     Calibration and validation locations in the ACF River basin
Station Name
FLINT RIVER AT MONTEZUMA, GA
CHATTAHOOCHEE RIVER NEAR
CORNELIA, GA
PEACHTREE CREEK AT ATLANTA, GA
CHATTAHOOCHEE RIVER AT ATLANTA,
GA
ICHAWAYNOCHAWAY CREEK AT
MILFORD, GA
CHATTAHOOCHEE RIVER AT WEST
POINT, GA
CHATTAHOOCHEE RIVER NEAR
COLUMBIA, AL
FLINT RIVER AT NEWTON, GA
APALACHICOLA RIVER AT
CHATTAHOOCHEE FLA
USGS ID
USGS02349605
USGS02331600
USGS02336300
USGS02336000
USGS02353500
USGS02339500
USGS02343801
USGS02353000
USGS02358000
Drainage area
(mi2)
2,900.00
315
86.8
1,450
620
3,550
8,210
5,740
17,200.00
Hydrology
calibration
X
X
X
X
X
X
X
X
X
Water quality
calibration
X
X
X
X
X
X
X
X

For hydrology, the model calibration period was set to calendar years 1993-2002 (from within the 30-year period
of record for modeling). The end date is constrained by the common period of the set of 20 Watershed
meteorological stations available for the watershed, and a ten year calibration period was desired. Hydrologic
validation was then performed on Calendar Years 1983-1992. Water quality calibration used calendar years 1999-
2002, because all gages had a decent set of data during that time period. Water quality validation was limited to
1986-1998, as very sparse data were available prior to 1986.  Some of the stations didn't have observed water
quality data prior to 1991. In these situations, the validation period represents all data available prior to January 1,
1999.
                                                 D-25

-------

Initial hydrologic parameterization for the Upper Flint calibration focus area came from a Loading Simulation
Program - C++ (LSPC) model created for the Upper Oconee watershed in north central Georgia (HUC
03070101). LSPC is a comprehensive data management and modeling system that is capable of representing both
flow and water quality loading from nonpoint and point sources and simulating in-stream processes. It is capable
of simulating flow, sediment, metals, nutrients, pesticides, and other conventional pollutants, as well as
temperature and pH for pervious and impervious lands  and waterbodies. LSPC and HSPF use the  same
parameterization, therefore, an LSPC model was chosen for initial parameterization assignments.

The LSPC model used for initial parameterization and the HSPF model set up for the 20 Watershed  study differed
in land use representation as well as soils/HRU representation. The LSPC models utilized a much more detailed
land use that was develop by the state of Georgia called Georgia Land Use Trends (GLUT).  Additionally, in the
LSPC model, subwatersheds were assigned to hydrologic soil groups by utilizing the hydrologic soil group that
had the greatest area within the subwatershed. Using this method the Upper Oconee watershed did not have any
hydrologic soil groups A or D. Therefore, the LSPC model was investigated to ensure the most representative
land use and soil type parameterization was transferred to the HSPF 20 Watershed model. Technical Note 6
(USEPA 2000) was utilized to establish initial parameterization for infiltration rates in areas that had hydrologic
soil groups A and D in the 20 Watershed application that.

Upon initial hydrologic parameterization of the focus area, both a model for the Upper Oconee watershed and a
model for the Lake Lanier watershed were  investigated as potential starting points.  The Upper Oconee watershed
parameterization did a much better job of representing measured flow than did those for the  Lake  Lanier
watershed parameters. At this point it was realized that parameterization assigned to the calibration focus area
probably will not work for all areas in the ACF basin.

The calibration focus area represents 7 HRUs. After calibrating the 7 HRUs, the calibrated parameterization was
transferred to the remaining 30 HRUs. Three locations, un-affected by upstream impoundments, therefore only
affected by parameterization, were selected to check the results with the focus area parameterization. These three
locations were Chattahoochee River near Cornelia, Georgia (USGS 02331600), Peachtree Creek at Atlanta,
Georgia (USGS 02336300), and Ichawaynochaway Creek at Milford, Georgia (USGS 02353500). All three of
these locations had a poor simulation of the observed hydrology. Because of the poor hydrologic simulation,
additional calibration was completed at each  of the three locations.

The area contributing to Peachtree  Creek at Atlanta, Georgia (USGS 02336300) was utilized as an urban area
calibration. Land use at this location is roughly 79 percent urban. Since calibration  at this location, entirely
revolved around the urban land use, the calibrated results were transferred to all other urban  areas throughout the
ACF basin.

The area contributing to Chattahoochee River near Cornelia, Georgia (USGS 02331600) was parameterized with
the Lake Lanier TMDL LSPC model, and is represented by two HRUs in the HSPF 20 Watershed model. The
initial parameterization was adjusted slightly, to account for the indirect transfer of land use  associated
parameters, from the more detailed LSPC model to the  20 Watershed model. The calibrated  parameters at this
location were transferred to one more HRU, immediately downstream. The area represented by these parameters
is closely associated with the area of the ACF basin that is in the Blue Ridge Geographic Province (Figure 6.)

The area contributing to Ichawaynochaway Creek at Milford, Georgia (USGS 02353500) was used to represent
hydrologic conditions for the HRUs in the Coastal Plain Province. To calibrate the area contributing to this gage,
the calibration focus area parameterization was adjusted until the simulated hydrology closely resembled the
                                                 D-26

-------
observed hydrology. Since adjustment was needed for multiple parameters on multiple land uses it was decided
that this gage would represent the hydrology of the Coastal Plain Province (Figure 6.)

The initial calibration focus area parameters were supplied for all other HRUs and this represents the Piedmont
Province (Figure 6.)

In summary, after realizing that parameterization assigned to the calibration focus area will not work for all areas
in the ACF basin an approach needed to be developed for assigning parameters for each HRU. After each area
that wasn't influenced by major impoundments was calibrated separate of the others it was decided that each of
the calibration areas would represent either the geologic province that each was contained in or in the case of the
Peachtree Creek gage, the dominant land use. Essentially, there  are three parameter groups assigned by geologic
province and the urban land use is parameterized the same throughout the model.

After the parameter mapping was complete for all three geologic provinces the calibration turned to reservoir
representation and operation. There is a more detailed discussion about the challenges faced during modeling
reservoirs in the HSPF Assumptions section of this report.

Once the hydrology calibration was complete for the whole ACF basin, the focus turned to sediment and water
quality representation. Initial parameterization for sediment and water quality simulation was taken from a LSPC
model developed for the Lake Allatoona watershed. The Lake Allatoona TMDL model was utilized rather than
the Lake Lanier TMDL or  Upper Oconee watershed LSPC models because the Lake Allatoona model utilized the
same general water quality approach that is utilized for the 20 Watershed application.  The water quality
simulation also generally reflected the need to assign parameters by geologic province, therefore, water quality
was calibrated at the same  locations as hydrology and the parameterization was transferred to the same HRUs as
the hydrology parameters.
                                                 D-27

-------
                                                                           Legend
                                                                             — Hydrography
                                                                               Interstate
                                                                           ^H Water (Nat. Atlas Dataset)
                                                                           ^B US Census Populated Places
                                                                              | Municipalities (pop > 50,000)
                                                                           I   I County Boundaries
                                                                               Model Subbasins
                                                                           Parameter Mapping Region
  GCRP Model Areas - ACF River Basin
     Parameter Mapping Regions
Figure 6. Parameter mapping utilized in the HSPF ACF River basin model.
                                                  D-28

-------
                   to

No changes were made to the meteorological or land use base data. The impoundments of Lake Blackshear and
Lake Seminole created an odd subwatershed connectivity when developed by BASINS. These two lakes both
have large tributaries contributing to them and BASINS delineated the watersheds by having all of the individual
tributaries pouring into the next downstream watershed. This made it difficult to represent the dam operation. The
connectivity was modified so that there was only one watershed representing the outflow of each impoundment.
The upstream subwatersheds, for all of the tributaries entering the lakes, were adjusted to pour into the
subwatershed containing the lake. This change was also made in the SWAT model. Before this change was added,
the HSPF model was in operation with the original connectivity, and BASINS generated f-tables. The simulation
results below both of the lakes, for both original and updated connectivity, showed very similar results. This
suggests that the update to the connectivity should not pose any problems.

As discussed earlier in the Point Sources section, an amendment was made to WPCP's that had default values
assigned for total phosphorus. Additionally, as discussed earlier, GaEPD data were used to supplement the
observed water quality data found in the NAWQA database.
Reservoirs
The Chattahoochee and the Flint Rivers represent two very different types if rivers. The Chattahoochee River has
many impoundments while the Flint River has one of the longest unimpeded stretches of flow in the United
States. The base data supplied point coverage for nine dams in the ACF basin and each was at the outlet of a
delineated subwatershed. Three of those given dams were assumed to operate as run of the river (Oliver, Bartlett's
Ferry, and George W. Andrews) and they were not included in the simulation. Table 6 identifies the dams and
corresponding reservoirs represented in the ACF basin 20 Watershed model.


Table 6.     Reservoirs represented in the ACF basin model
Dam Name
Buford
West Point
Walter F George Lock,
Dam, Powerhouse
Crisp County (Warwick)
Muckafoonee Creek
Dam
Jim Woodruff Dam
Other Name
Lake Sidney Lanier
West Point Lake
Eufala
Lake Blackshear
Lake Worth
Lake Seminole
River
Chattahoochee
Chattahoochee
Chattahoochee
Flint
Flint
Flint
Owner
USAGE
USAGE
USAGE
Crisp County Power
Commission
Georgia Power
Company
USAGE
All of the Army Corps of Engineers' (USAGE) lakes simulated in the model have data published including
elevation, inflow, discharge, and power generation, since the facility became operational (USAGE 2010). The
USAGE has also made available a graph of the area capacity curve for each of the lakes. For the 20 Watershed
application, it was assumed that the best representation of the reservoirs was to try to simulate them without
supplying time series operations or boundary conditions. If time series operations were supplied, it would be
difficult to predict what the boundary condition would be in the future. Therefore, the area-capacity curves were
developed into an f-table, consistent with Technical Note 1 (USEPA 2007) and supplied for the subwatershed

                                                D-29

-------
containing the lake. By having the controlling feature on the lake be an f-table future climatic conditions will not
be affected by boundary conditions of the past.

Elevation-Storage relationships were not available for Lake Blackshear and Lake Worth. Research was done to
look for the average depth and average surface area for each of the lakes. This information was utilized to come
up with a reservoir of similar size and storage by using the stage-storage-discharge relationships tool in the HSPF
BMP Toolkit on the EPA website.

One of the biggest challenges was trying to represent reservoir operations with only an f-table. When the model
was first set up, the focus was on calibrating to areas un-impacted by flow control devices, and the reservoirs
simply had the f-table created by BASINS. The hydrology results at the gages used to check reservoir operations
actually looked pretty decent considering lake storage and dam releases were not accounted for. F-tables were
developed and inserted into the model for all reservoirs included in the simulation. The physical relationship
between surface area and storage was left unchanged and reservoir calibration focused on assigning a discharge to
a particular depth of water. When there were elevation and discharge data available, they were used as a guideline
for assignment of the outflow for a particular depth. The results on the mainstem Flint River changed very little
but the results on the mainstem Chattahoochee changed drastically and the simulation became very poor. Much
work was done  on the Chattahoochee River simulation to try to represent the reservoirs properly. Since the
reservoirs are highly controlled by peaking hydro electric operations and targeted elevations based on the  season,
the approach used for the 20 Watershed application did not do a very good job at representing observed flows
below the dams on the Chattahoochee River. Much of the error on the Chattahoochee can be attributed to the
improper operation of Buford Dam. The discharges at Buford Dam impact the discharges at all other dams on the
Chattahoochee and this is also the case in the 20 Watershed model.

Withdrawals
It is not known what water withdrawals by municipal and industrial facilities will look like in the future,
therefore, they were not included in the 20 Watershed model application. Recent court rulings suggest that current
withdrawals below Buford dam may change in as little as three years.

Irrigation
Irrigation in the Lower Flint, its tributaries, and the Lower Chattahoochee is used quite extensively when needed.
The model, for the 20 Watershed application, is not using the irrigation module.  HSPF requires that a land use be
associated with irrigation and applying irrigation to all agricultural land may greatly over estimate the amount  of
irrigation that is actually taking place. Additionally, no one knows what agricultural irrigation may look like in the
future. There have been numerous studies commissioned in the past decade to look into the amount of irrigation in
the state of Georgia. A majority of the irrigation is from groundwater sources and this would represent new water
to the HSPF model. It was assumed that an irrigation component would not benefit the model for the 20
Watershed application.

Snow Simulation
Previous modeling experience in Georgia did not utilize snow simulation. The model for the 20 Watershed
application is to include snow simulation using the degree-day method for snowmelt. With no previous models to
obtain initial parameters for snow simulation, the initial parameters needed to be developed.  Technical Note 6
(USEPA 2000)  was used as a guideline for parameterization. The  parameters for the physical properties of each
HRU are assigned by HRU but  all other snow simulation parameters are the same for each HRU. These values are
assumed to be appropriate and the initial parameterization was not adjusted.
                                                 D-30

-------
Hydrology Calibration
As explained above, the starting parameters for ACF 20 Watershed application came from an LSPC model for the
Upper Oconee watershed. Differences amongst the model set up between the LSPC application and HSPF 20
Watershed application meant that not all parameters in the LSPC application were directly transferable to the
HSPF 20 Watershed application. When it did not make sense to utilize parameters from the LSPC model,
Technical Note 6 (USEPA 2000) was utilized to determine a good starting value. The parameters from the Upper
Oconee watershed simulated flows in the range of the observed flows but minor adjustments needed to be made to
better fit the simulated flows to the observed flows. Calibration adjustments focused on the following parameters:

    •   INFILT (index to mean soil infiltration rate): The LSPC model did not represent hydrologic soil groups A
       or D. A representative value was obtained from Technical Note 6 (USEPA 2000). Very minor
       modifications were made to these estimated values because the calibration focus area is dominated by B
       and C soils. Additionally, INFILT was adjusted slightly by land use to account for the land use
       differences between the LSPC model and the HSPF 20 Watershed model. INFILT values for the
       hydrologic soil groups are in the range of those stated in Technical note 6.
    •   AGWRC (Groundwater recession rate): Adjusted slightly in order to replicate groundwater recession in
       the  observed data.
    •   LZSN (lower zone nominal soil moisture storage): This was increased slightly upward due to baseflow
       contributions  severely tapering off during extreme dry weather. The changes to INFILT and AGWRC
       resulted in small modifications in this parameter.
    •   BASETP (ET by riparian vegetation): Even with the modifications mentioned above, simulation of low
       flows was not that good. It was assumed that the Flint River watershed has a greater amount of riparian
       vegetation than that of the Upper Oconee watershed. Slightly increasing the BASETP value made the
       simulation of low flows much better.

Initial calibrations were performed for the upper Flint River, comparing model results to data from USGS
02349605, and are summarized in Figures 7 through 13 and Tables 7 and 8. The model fit is of high quality but
always simulates a little bit high. This could be because municipal and industrial withdrawals were not included
in the in the simulation. None of the metrics fall out of those set for the 20 Watershed study. The model
calibration period was set to calendar years 1993-2002.
                                                D-31

-------
         Avg Monthly Rainfall (in)
        -Avg Observed Flow (1/1/1993 to 12/31/2002)
        •Avg Modeled Flow (Same Period)
i9nnnn
mnnnn
onnnn
finnnn
/nnnn
onnnn -

Hlf 	





J
	 II

pi





, A

T






Jan-93 Jul-94







in

r







i


1
Jan-96











ill












1
JL
Jul-97






J^
















f'l





JL
Jan-99 Jul-00









•*^_
Jan-02






1
9
4 ^
R £*
b ^
'ro
Q ry
^s
m 'ra
1U ro
19


                                              Date
  Figure 7. Mean daily flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
               period (HSPF).
i
o
    30000
   20000
    10000 -
        J-93
                                                        Avg Monthly Rainfall (in)
                                                       -Avg Observed Flow (1/1/1993 to 12/31/2002
                                                       •Avg Modeled Flow (Same Period)
J-94
J-02
  Figure 8. Mean monthly flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
               period (HSPF).
                                                D-32

-------
             Avg Flow (1/1/1993 to 12/31/2002 )
             •Line of Equal Value
             Best-Fit Line
    30000
 LJ-  20000
    10000
 O)
 TO
0        10000       20000      30000
    Average Observed Flow (cfs)
                                               O
                                               0)
                                               o
                                               ro
                                               m
                                                _
                                               0)
                                               TO
                                          100%
                                          90% -
                                          80%
                                          70% -
                                          60%
                                          50%
                                                 Avg Observed Flow (1/1/1993 to 12/31/2002 )
                                                 Avg Modeled Flow (1/1/1993 to 12/31/2002 )
                                                -Line of Equal Value
                                          40% -
                                          30% -
                                          20% -
                                          10% -
                                           0%

                                                      J-93    J-94   J-96
                                                                         J-97   J-99
                                                                          Month
                                                                                     J-00   J-02
   Figure 9. Monthly flow regression and temporal variation at USGS 02349605 Flint River at Ga 26, near
                 Montezuma, GA - calibration period (HSPF).
  8000
t
36000
o
           Avg Flow (1/1/1993 to 12/31/2002)
           •Line of Equal Value
           Best-Fit Line
  y = 1.0261x-
      R- = 0.9946
              2000    4000     6000
             Average Observed Flow (cfs)
                               8000
                                               Avg Monthly Rainfall (in)
                                              -Avg Observed Flow (1/1/1993 to 12/31/2002)
                                               Avg Modeled Flow (Same Period)
                                                  8000
                                          6000
                                                  4000
                                                  2000
                                                        1  2  3   4   5   6   7  8  9  10  11  12
   Figure 10. Seasonal regression and temporal aggregate at USGS 02349605 Flint River at Ga 26, near
                 Montezuma, GA - calibration period (HSPF).
                                                  D-33

-------
     Average Monthly Rainfall (in)
     -Median Observed Flow (1/1/1993 to 12/31/2002)
    10000 n
I Observed (25th, 75th)
Modeled (Median, 25th, 75th)
                                                                  10     11
                                                                              12
   Figure 11. Seasonal medians and ranges at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
                - calibration period (HSPF).
Table 7.    Seasonal summary at USGS 02349605 Flint River At Ga 26, near Montezuma, GA
            calibration period (HSPF)
MONTH

Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
3955.13
5767.27
6976.42
3970.93
2277.06
1783.49
3617.54
1440.71
1260.28
1816.83
2297.06
3137.85
3665.00
3915.00
5150.00
3065.00
1810.00
1670.00
1290.00
1035.00
964.00
1105.00
1660.00
2240.00
2110.00
2572.50
2932.50
2210.00
1400.00
1180.00
977.25
752.00
770.50
775.75
1220.00
1490.00
5155.00
6887.50
8377.50
4297.50
2470.00
2170.00
1765.00
1420.00
1252.50
1855.00
2685.00
3497.50
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
4080.88
5840.28
7509.10
4138.24
2259.06
1976.13
3734.42
1646.45
1584.54
1920.32
2506.66
3202.84
3477.76
4036.36
5064.71
3192.51
2007.62
1497.68
1309.40
1108.31
917.93
1202.58
1807.51
2167.47
2037.87
2764.25
2921.96
2099.52
1330.76
1168.42
880.28
803.03
712.46
785.75
878.28
1344.21
5135.68
6514.59
8921.61
4715.87
2653.59
2453.92
2131.87
1567.80
1731.82
2195.35
3533.72
3347.46
                                             D-34

-------
            •Observed Flow Duration (1/1/1993 to 12/31/2002 )
             Modeled Flow Duration (1/1/1993 to 12/31/2002 )
o
D)
ro
Q
    1000000
     100000
      10000
       1000
        100
                  10%
                          20%     30%      40%     50%     60%     70%     80%
                               Percent of Time that Flow is Equaled or Exceeded
                                                                                   90%
                                                                                           100%
  Figure 12. Flow exceedence at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
                period (HSPF).
  o
  o
ro
T3
I
.0


E
"o

^3
LL
T3
.N
(0
                •Observed Flow Volume (1/1/1993 to 12/31/2002 )
                 Modeled Flow Volume (1/1/1993 to 12/31/2002 )
       120%
       100%
        80%
        60%
        40%
        20%
          Jan-93
                      Jul-94
                                  Jan-96
                                              Jul-97
                                                          Jan-99
                                                                       Jul-00
                                                                                  Jan-02
  Figure 13. Flow accumulation at USGS 02349605 Flint River at Ga 26, near Montezuma, GA •
                calibration period (HSPF).
                                                 D-35

-------
Table 8.     Summary statistics at USGS 02349605 Flint River At Ga 26, near Montezuma, GA
             calibration period (HSPF)
HSPF Simulated Flow
REACH OUTFLOW FROM DSN 1001
10-Year Analysis F^riod: 1/1/1993 - 12/31/2002
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9)
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3):
Simulated Srjring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
NasjT^SuteliffeJDpjsfficj^^
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
15.61

6.10
2.71

2.73
2.98
6.67
3.23

4.86
1.00
Error Statistics
5.50
-1.49
6.32
10.14

USGS 02349605 FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA
Hydrologic Unit Code: 3130006
Latitude: 32.29305556
Longitude: -84.04361 1 1
Drainage Area (sq-rri): 2920
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow Volume (1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
5.17 » | 30
4.48
4.21
1.25
-2.14
0.707
0.568
0.934
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0





14.80

5.74
2.75

2.48
2.83
6.39
3.10

4.80
1.02





Clear [













Hydrology Validation

Validation for the Upper Flint calibration focus area was performed at the same location but for calendar years
1983-1992. Results are presented in Figures 14 through 20 and Tables 9 and 10. Similarly to the calibration years,
the validation years' model fit is of high quality but always simulates a little bit high. None of the metrics fall out
of the range set for the 20 Watershed study. The model validation period was set to calendar years 1983-1992.
                                              D-36

-------
         Avg Monthly Rainfall (in)
        -Avg Observed Flow (1/1/1983 to 12/31/1992)
        •Avg Modeled Flow (Same Period)
    70000

    60000

_  5000°
^,  40000

1   30000
LJ_
    20000

    10000
        Jan-83       Jul-84       Jan-86       Jul-87       Jan-89

                                               Date
                                             Jul-90
Jan-92
   Figure 14. Mean daily flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
                period (HSPF).
 I
 o
    15000
    10000
     5000
         J-83
                                                         Avg Monthly Rainfall (in)
                                                        -Avg Observed Flow (1/1/1983 to 12/31/1992 )
                                                        •Avg Modeled Flow (Same Period)
J-84
                                                                         c
                                                                         ro
  J-92
   Figure 15. Mean monthly flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA -
                validation period (HSPF).
                                                 D-37

-------
 •  Avg Flow (1/1/1983 to 12/31/1992 )
• - - - • Line of Equal Value
	Best-Fit Line
   15000
   10000
    5000
O)
ro
        0        5000      10000     15000
            Average Observed Flow (cfs)
                                              O
                                              0)
                                              o
                                      ro
                                      m
                                              0)
                                              to
                                         100%
                                          90% -
                                          80%
                                          70% -
                                          60% -
                                                         Avg Observed Flow (1/1/1983 to 12/31/1992 )
                                                         Avg Modeled Flow (1/1/1983 to 12/31/1992 )
                                                        -Line of Equal Value
                                          50%
                                                   0%
                                             J-83   J-84   J-86
                                                                J-87   J-89
                                                                 Month
                                                                            J-90   J-92
  Figure 16. Monthly flow regression and temporal variation at USGS 02349605 Flint River at Ga 26, near
                Montezuma, GA - validation period (HSPF).
 8000

          Avg Flow (1/1/1983 to 12/31/1992)
          •Line of Equal Value
          Best-Fit Line
          y = 0.981 6x + 228.77
              R2 = 0.9888
             2000     4000     6000
            Average Observed Flow (cfs)
                               8000
                                               Avg Monthly Rainfall (in)
                                              -Avg Observed Flow (1/1/1983 to 12/31/1992)
                                               Avg Modeled Flow (Same Period)
                                                 8000
                                                 6000
                                                 4000
                                                 2000
                                                          2  3
                                                        4  5  6  7  8  9  10  11  12
                                                             Month
  Figure 17. Seasonal regression and temporal aggregate at USGS 02349605 Flint River at Ga 26, near
                Montezuma, GA - validation period (HSPF).
                                                 D-38

-------
     Average Monthly Rainfall (in)
     •Median Observed Flow (1/1/1983 to 12/31/1992)
• Observed (25th, 75th)
Modeled (Median, 25th, 75th)
    8000
           1      23456789     10    11
                                                                              12
   Figure 18. Seasonal medians and ranges at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
                - validation period (HSPF).
Table 9.     Seasonal summary at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
            validation period (HSPF)
MONTH

Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
4168.00
5242.97
5976.65
4328.17
2451 .47
1753.24
1908.70
1632.54
1283.93
1425.06
2275.14
3667.95
3115.00
4150.00
4310.00
3275.00
2135.00
1300.00
1330.00
1150.00
1120.00
1020.00
1330.00
2275.00
2152.50
2850.00
2830.00
2277.50
1420.00
986.75
927.50
816.50
881 .75
869.50
1120.00
1602.50
5350.00
6430.00
6542.50
4880.00
2937.50
2072.50
2257.50
1830.00
1500.00
1390.00
2277.50
4277.50
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
4169.05
5353.14
6371 .53
4468.43
2557.55
2100.76
2323.54
1765.46
1577.80
1695.02
2252.76
3559.80
3287.48
5008.00
4667.85
3626.93
2256.24
1331.44
1314.50
1414.88
1249.76
1001.13
1293.90
2145.93
1922.32
3062.19
2944.79
2094.33
1533.87
985.52
866.44
867.56
876.51
717.97
680.19
1079.64
5410.41
6675.95
6916.63
5365.85
3083.22
2134.09
2921.29
2433.92
1816.29
1808.37
2259.65
5244.57
                                             D-39

-------
            •Observed Flow Duration (1/1/1983 to 12/31/1992 )

             Modeled Flow Duration (1/1/1983 to 12/31/1992 )
    100000
o

g
o
D)
ro
Q
     1 0000
      1000  =
       100
                 10%
                         20%      30%     40%     50%     60%      70%     80%


                               Percent of Time that Flow is Equaled or Exceeded
                                                                                   90%
                                                                                           100%
  Figure 19. Flow exceedence at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
                period (HSPF).
  O
  o
  ro

  T3


  I


  .0





  E


  "o




  ^3

  LL

  T3

  .N

  (0
                •Observed Flow Volume (1/1/1983 to 12/31/1992 )

                 Modeled Flow Volume (1 /1/1983 to 12/31/1992 )
       120%
       100%
        80%
        60%
        40%
        20%
          Jan-83
                      Jul-84
                                  Jan-8
                                              Jul-87
                                                          Jan-89
                                                                      Jul-90
                                                                                  Jan-92
  Figure 20. Flow accumulation at USGS 02349605 Flint River at Ga 26, near Montezuma, GA •

                validation period (HSPF).
                                                D-40

-------
Table 10.    Summary statistics at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
             validation period (HSPF)
HSPF Simulated Flow
REACH OUTFLOW FROM DSN 1001
10-Year Analysis F^riod: 1/1/1983 - 12/31/1992
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9)
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3):
Simulated Srjring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
NasjT^SuteliffeJDpjsfficj^^
Baseline adjusted coefficient (Garrick), E':
Monthly NSE ~
14.76

5.05
2.70

2.22
2.94
6.09
3.52

4.30
0.55
Error Statistics
5.79
-3.16
3.69
17.40

USGS 02349605 FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA
Hydrologic Unit Code: 3130006
Latitude: 32.29305556
Longitude: -84.04361 1 1
Drainage Area (sq-rri): 2920
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow Volume (1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
1 .92 » | 30
3.33
6.93
-2.74
9.14
0.651
0.551
0.899
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0





13.95

4.87
2.78

1.89
2.88
5.89
3.29

4.42
0.51





Clear [













Hydrology Results for Larger Watershed

As discussed above, the parameters from the calibration focus were not fully transferrable to other gages in the
ACF basin. Therefore, at each of the gages un-impacted by flow control devices, an additional level of calibration
was performed. Please refer to the discussion and Figure 6 for details on how the additional calibration areas'
parameterization was assigned to the other areas in the watershed.

As stated above, the Upper Chattahoochee borrowed parameterization from a model done for the Lake Lanier
TMDL. Due to careful transferring of parameters there wasn't any adjustment made to achieve calibration. The
model fit was of high quality in the TMDL model and also has a very high goodness of fit in the 20 Watershed
model. The statistics for Upper Chattahoochee gage are within the range defined for the 20 Watershed application.
The calibration and validation results for this region are shown in Tables 11 and 12 (station 02331600).

The Ichawaynochaway Creek subwatershed did not have a model to borrow parameterization from. The calibrated
results of the calibration focus area were assigned and then adjusted until the simulated flows closely matched the
observed flows. All of the parameters adjusted were for the baseflow component of the simulation. Baseflow was
being simulated too high so the goal was to lower the amount of baseflow reaching the stream. This was achieved

                                               D-41

-------
through increasing the amount of ET satisfied by riparian vegetation and direct evaporation from groundwater.
The model fit is fairly good except for the extreme low flows. The high simulation of the extreme low flows may
be explained by not having simulated irrigation. The calibration and validation results for this region are shown in
Tables 13 and 14 (station 02353500).
As discussed in the Assumptions section for Reservoirs, the simulation at all gages on the mainstem
Chattahoochee, below Lake Lanier, is very poor. The approach taken to simulate the reservoirs in the ACF did not
simulate Lake Lanier very well, but simulation of the other reservoirs was acceptable. This could be because the
reservoirs other than Lake Lanier are mostly managed as inflow equals outflow and the discharges at  Lake Lanier
usually control most of the inflow. Once the Chattahoochee and Flint converge and leave Lake Seminole,  the
simulation and model fit is once again of high quality and shown for the calibration period in Figures 21 through
27 and Tables 11 and 12.
          Avg Monthly Rainfall (in)
          -Avg Observed Flow (1/1/1993 to 12/31/2002)
          •Avg Modeled Flow (Same Period)
      350000

      300000

      250000

      200000

      150000

      100000

       50000
                                                                            Jan-02
                                                                                       14
    Figure 21. Mean daily flow: Model DSN 9001 vs. USGS 02358000 Apalachicola River At
                 Chattahoochee, FL- calibration period (HSPF).
     100000

      80000
                                                         Avg Monthly Rainfall (in)
                                                        -Avg Observed Flow (1/1/1993 to 12/31/2002 )
                                                        •Avg Modeled Flow (Same Period)
           J-:
J-02
                                                                                            c
                                                                                            'ro
                                                                                            OL
    Figure 22. Mean monthly flow: Model DSN 9001 vs. USGS 02358000 Apalachicola River At
                 Chattahoochee, FL- calibration period (HSPF).
                                                D-42

-------
JD
LL
0)
E
0)
0)
   100000
    80000
    60000
    40000
    20000
            Avg Flow (1/1/1993 to 12/31/2002 )
            •Line of Equal Value
            Best-Fit Line
          0.9871X
          R2 - Q
                 + 1737
                9334
=2
         0   20000 40000  60000  80000 100000
            Average Observed Flow (cfs)
                                              O
                                              £=
                                              _
                                              ro
                                              m
                                              y
                                              "co
                                                        Avg Observed Flow (1/1/1993 to 12/31/2002 )
                                                        Avg Modeled Flow (1/1/1993 to 12/31/2002 )
                                                        -Line of Equal Value
                                                     J-93   J-94
                                                                 J-96
                                                                       J-97   J-99
                                                                        Month
                                                                                   J-OO   J-02
  Figure 23. Monthly flow regression and temporal variation: Model DSN 9001 vs. USGS 02358000
                Apalachicola River At Chattahoochee, FL- calibration period (HSPF).
           Avg Flow (1/1/1993 to 12/31/2002)
Best-Fit Line
cnnnn
4nnnn -
onnnn
onnnn
10000
n
y = 1.
R



.'
OOOSx +
2 = 0 99'

*


1470
>o
s
£fr*


>
^f'



.•'




            10000  20000  30000  40000 50000
            Average Observed Flow (cfs)
                                                       Avg Monthly Rainfall (in)
                                                      -Avg Observed Flow (1/1/1993 to 12/31/2002)
                                                      •Avg Modeled Flow (Same Period)
                                                 50000
                                                           2  3
                                                                4  5  6  7  8  9  10  11  12
                                                                      Month
  Figure 24. Seasonal regression and temporal aggregate: Model DSN 9001 vs. USGS 02358000
                Apalachicola River At Chattahoochee, FL- calibration period  (HSPF).
                                                 D-43

-------
     Average Monthly Rainfall (in)
     •Median Observed Flow (1/1/1993 to 12/31/2002)
• Observed (25th, 75th)
Modeled (Median, 25th, 75th)
                                                                       11
                                                                             12
                                                                                     c
                                                                                     'co

                                                                                     >,
                                                                                     o
   Figure 25. Seasonal medians and ranges: Model DSN 9001 vs. USGS 02358000 Apalachicola River At
                Chattahoochee, FL- calibration period (HSPF).
Table 11.   Seasonal summary: Model DSN 9001 vs. USGS 02358000 Apalachicola River At
            Chattahoochee, FL- calibration period (HSPF).
MONTH
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
24730.16
34453.44
41056.97
25249.20
15972.58
12939.23
18873.90
12411.87
10850.80
12458.10
13896.10
19332.77
19950.00
29050.00
36850.00
19800.00
15250.00
12600.00
11950.00
11050.00
8235.00
10750.00
13500.00
14550.00
13500.00
18275.00
19300.00
16000.00
9570.00
8237.50
7665.00
7102.50
6530.00
6112.50
6557.50
9150.00
30475.00
43900.00
53300.00
30250.00
19675.00
17600.00
15575.00
14300.00
12900.00
13900.00
18250.00
23800.00
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
26069.98
35291 .43
43511.83
26617.77
16032.37
14124.72
20000.73
13379.10
12790.03
15607.91
16250.34
20267.76
20647.35
27706.10
32896.58
21242.91
14369.52
12931.98
11560.84
11102.23
9367.20
11368.60
12148.80
15105.64
14491.39
17766.08
19444.48
15995.22
12295.58
9530.52
9687.78
8740.66
7476.25
8238.77
9171.31
10019.73
32838.93
43841 .35
51756.70
30764.85
17570.34
17294.53
16487.60
14248.45
13854.01
16874.80
19639.35
22663.99
                                             D-44

-------
o
g
o
en
ro
Q
            •Observed Flow Duration (1/1/1993 to 12/31/2002 )
             Modeled Flow Duration (1/1/1993 to 12/31/2002 )
    1000000
     1 00000
      10000
       1000
                  10%
                          20%     30%     40%     50%     60%     70%     80%
                               Percent of Time that Flow is Equaled or Exceeded
                                                                                  90%
                                                                                          100%
  Figure 26. Flow exceedence: Model DSN 9001 vs. USGS 02358000 Apalachicola River At
               Chattahoochee, FL- calibration period (HSPF).
  o
  o
  ro
  T3
  I
  .0


  E
  "o

  ^3
  LL
  T3
  .N
  (0
                •Observed Flow Volume (1/1/1993 to 12/31/2002 )
                 Modeled Flow Volume (1/1/1993 to 12/31/2002 )
       120%
       100%
        80%
        60%
        40%
        20%
          Jan-93
                      Jul-94
                                  Jan-96
                                              Jul-97
                                                         Jan-99
                                                                      Jul-00
                                                                                 Jan-02
  Figure 27. Flow accumulation: Model DSN 9001 vs. USGS 02358000 Apalachicola River At
               Chattahoochee, FL- calibration period (HSPF).
                                                D-45

-------
Table 12.   Summary statistics: Model DSN 9001 vs. USGS 02358000 Apalachicola River At
           Chattahoochee, FL- calibration period (HSPF).
HSPF Simulated Flow
REACH OUTFLOW FROM DSN 9001
10-Year Analysis F^riod: 1/1/1993 - 12/31/2002
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9)
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3):
Simulated Srjring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
NasjT^SuteliffeJDpjsfficj^^
Baseline adjusted coefficient (Garrick), E':
Monthly NSE ~
17.06

5.55
4.09

3.07
3.46
6.82
3.72

4.27
0.72
Error Statistics
7.35
12.13
8.50
9.50

USGS 02358000 APALACHICOLA RIVER AT CHATTAHOOCHEE FLA
Hydrologic Unit Code: 3130011
Latitude: 30.701 0251
Longitude: -84.8590871
Drainage Area (sq-rri): 17200
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow Volume (1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
14.07 » | 30
4.69
4.78
15.09
12.02
0.769
0.575
0.922
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0





15.89

5.11
3.65

2.80
3.03
6.51
3.55

3.71
0.64





Clear [













                                        D-46

-------
The calibration and validation statistical measurements, at all USGS gages used in the ACF basin for the 20
Watershed project, are shown in Tables  13 and 14 respectively.
Table 13.    Summary statistics (percent error) for all stations - calibration period 1993-2002
             (HSPF)
Station
Error in
total
volume:
Error in
50%
lowest
flows:
Error in
10%
highest
flows:
Seasonal
volume
error -
Summer:
Seasonal
volume
error- Fall:
Seasonal
volume
error -
Winter:
Seasonal
volume
error -
Spring:
Error in
storm
volumes:
Error in
summer
storm
volumes:
Daily
Nash-
Sutcliffe
Coefficient
of
Efficiency,
E:
Monthly
Nash-
Sutcliffe
Coefficient:
of
02349605

5.50


-1.49


6.32

10.14


5.17


4.48


4.21


1.25


-2.14



0.707


0.934


02331600

0.14


3.09


4.62

2.28


6.04


4.19


-10.90


-0.25


-28.59



0.640


0.862


02336300

17.16


2.56


6.29

1.30


19.26


27.56


13.02


8.76


3.05



0.536


0.477


02353500

3.93


17.11


9.69

1.33


-8.97


9.57


9.42


6.61


-34.98



0.339


0.652


02336000

24.16


59.65


-7.60

8.20


42.13


16.81


35.91


-56.00


-62.85



0.539


0.683


02339500

15.40


74.07


-0.36

9.33


19.50


9.17


27.67


-70.75


-87.05



0.591


0.821


02343801

16.79


49.13


8.63

11.54


26.84


10.78


24.10


-37.96


-60.30



0.717


0.858


02353000

8.33


4.70


14.97

16.06


7.59


7.54


4.65


4.85


12.41



0.607


0.928


02358000

7.35


12.13


8.50

9.50


14.07


4.69


4.78


15.09


12.02



0.769


0.922


                                              D-47

-------
Station
Efficiency
02349605

02331600

02336300

02353500

02336000

02339500

02343801

02353000

02358000

Table 14.   Summary statistics (percent error) for all stations - validation period 1983-1992
           (HSPF)
Station
Error in
total
volume:
Error in
50%
lowest
flows:
Error in
10%
highest
flows:
Seasonal
volume
error -
Summer:
Seasonal
volume
error- Fall:
Seasonal
volume
error -
Winter:
Seasonal
volume
error -
Spring:
Error in
storm
volumes:
Error in
summer
storm
volumes:
Daily
Nash-
Sutcliffe
Coefficient
of
Efficiency,
E:
Monthly
Nash-
Sutcliffe
Coefficient:
of
02349605

5.79



-3.16



3.69


17.40



1.92


3.33



6.93



-2.74


9.14




0.651





0.899


02331600

-8.32



-4.98



-6.88


-9.94



-4.13


-2.73



-17.08



-22.55


-50.35




0.696





0.865


02336300

13.73



4.87



3.05


5.82



19.78


19.85



7.49



6.12


7.78




0.553





0.654


02353500

1.19



-1.92



6.48


23.48



-14.67


-3.02



7.15



-0.18


-22.89




0.385





0.652


02336000

12.01



47.20



-14.34


-2.55



8.67


22.43



19.52



-61.77


-71.06




0.479





0.845


02339500

10.86



80.89



-5.02


6.67



-1.60


16.08



21.34



-74.01


-87.14




0.566





0.797


02343801

13.28



37.90



7.57


13.98



8.04


14.33



15.92



-56.18


-77.18




0.698





0.858


02353000

6.91



-1.14



11.27


17.92



7.94


4.62



3.64



-10.57


-23.32




0.682





0.890


02358000

2.21



-9.05



8.49


3.55



1.33


4.57



-1.81



20.87


-5.49




0.707





0.914


                                          D-48

-------
Station
Efficiency
02349605

02331600

02336300

02353500

02336000

02339500

02343801

02353000

02358000

D-49

-------
An additional check was done on the reservoirs of the Chattahoochee River mainstem. Observed and modeled
reservoir elevations were compared by using a histogram approach. This check ensured that the storage contained
within a reservoir was accounted for even though the flow calibration downstream of the reservoir didn't simulate
well. These comparisons were performed for the time period from January 1993 to December 2002. All lake
elevation simulations closely compared to the observed elevations except for Lake Lanier (Figures 28 through
31).
                        Lake Lanier-Simulated and Measured Elevation
      1000
    I  500
                                         Stage (ft)
    Figure 28. Histogram of simulated and measured elevation for Lake Lanier from 1/1/1993 to
                12/31/2002.
                                               D-50

-------
                West Point - Simulated and Observed Elevation
    600
    500
    400
 Q
    300
    200
    100
                                                              PI  n  n
        
-------
            Walter F. George - Simulated and Observed Elevation
   1800
   1600
   1400
   1200
   1000
ra
Q
   800
   600
   400
   200
                                                                    CD
                                                               05    05
                                   Stage (ft)
Figure 30. Histogram of simulated and measured elevation for Lake Walter F. George from 1/1/1993 to

            12/31/2002.
                                         D-52

-------
Lake Seminole - Simulated and Measured Elevation
1 Ann
1 onn

onn -
1



0






n

PI

— .












—











O IO O LO O


n
m o o o o o o

D Simulated
• Measured

Stage (ft)
    Figure 31. Histogram of simulated and measured elevation for Lake Seminole from 1/1/1993 to
                 12/31/2002.
Water Quality Calibration and Validation
The 20 Watershed models are designed to provide water quality simulation for total suspended solids (TSS), total
nitrogen, and total phosphorus. TSS is simulated with the standard HSPF approach (USEPA 2006). In contrast to
TSS, total nitrogen and total phosphorus are simulated in this application in a simplistic fashion, as HSPF general
quality constituents (GQUALs) subject to an exponential decay rate during transport.

The water quality calibration focuses on the replication of monthly loads, as specified in the project QAPP. Given
the approach to water quality simulation in the 20 Watershed model, a close match to individual concentration
observations cannot be expected. Comparison to monthly loads presents challenges, as monthly loads are not
observed. Instead, monthly loads must be estimated from scattered concentration grab samples and continuous
flow records. As a result, the monthly load calibration is inevitably based on the comparison of two uncertain
numbers. Nonetheless, calibration is able to achieve a reasonable agreement. Further, the load comparisons were
supported by detailed examinations of the relationships of flows to loads and concentrations and the distribution
of concentration prediction errors versus flow, time, and season, as well as standard time series plots.

For application on a nationwide basis, the 20 Watershed protocols assume that TSS and total phosphorus loads
will likely exhibit a strong positive correlation to flow (and associated erosive processes), while total nitrogen
loads, which often have a dominant groundwater component, will not. Accordingly, TSS and total phosphorus

                                                D-53

-------
loads were estimated from observations using a flow-stratified log-log regression approach, while total nitrogen
loads were estimated using a flow-stratified averaging estimator, consistent with the findings of Preston et al.
(1989).

Similarly to hydrology, initial calibration and validation of water quality was done on the Upper Flint River,
comparing model results to data from USGS 02349605. The initial calibration used calendar years 1999-2002 for
calibration and calendar years 1991-1998 for validation as there were no data available prior to 1992 for this gage.
As stated above, initial water quality parameters were obtained from an LSPC model for a TMDL done for the
Lake Allatoona watershed in North Georgia. The Lake Allatoona watershed LSPC model was parameterized with
values from literature, information collected in the field, and from previous modeling work done in the state of
Georgia. With the exception of shrub lands, both the TMDL model and 20 Watershed model had similar land
uses. Shrub lands had parameters assigned from forest lands since these two land uses should behave similarly to
each other.

Time series of simulated and estimated TSS loads at the Upper Flint gage for both periods are shown in Figure 32
and statistics for the two periods are provided separately in Table 15. Results of the TSS calibration are generally
acceptable. Visually, the model is roughly simulating the trends contained in the observed data but the loading
estimates are on the high side. The statistics performed on the comparison between the simulated results and
observed data also indicate that TSS  loading is slightly high. The key statistic in the table (consistent with the
QAPP) is the relative percent error, which shows the error in the prediction of monthly load normalized to the
estimated load. The table also shows the relative average absolute error, which is the average of the relative
magnitude of errors in individual monthly load predictions. This number is inflated by outlier months in which the
simulated and estimated loads differ  by large amounts (which may be as easily due to uncertainty in the estimated
load due to limited data) and the third statistic, the relative median absolute error, is likely more relevant and
shows acceptable agreement.
                                                 TSS
      1,000,000
   o
   «
   I
-Regression Loads
-Simulated Loads
    Figure 32.  Fit for monthly load of TSS USGS 02349605 Flint River at Ga 26, near Montezuma, GA
                 (HSPF).
                                                 D-54

-------
Table 15.    Model fit statistics (observed minus predicted) for monthly TSS loads using
             stratified regression
Statistic
Relative Percent Error
Relative Average Absolute Error
Relative Median Absolute Error
Calibration period
(1999-2002)
-117%
129%
38.1%
Validation period
(1991-1998)
-78%
110%
44.5%
A variety of other diagnostics were also pursued to ensure agreement between the model and observations. These
are available in full in the calibration spreadsheets, but a few examples are provided below. First, load-flow power
plots were compared for individual days (Figures 33 and 34). These confirm that the relationship between flow
and load is consistent across the entire range of observed flows, for both the calibration and validation periods.
TSS Load, tons/day
FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1999-2002
1 nnnnn
I UUUUU
10000 -
1000 -
100 -
10
1
0.1 -
0.01 -
Onni
* »
" : j^
l^K^^^
JiSsir *'
AJpf*
i) :
: *
4
*

1 10 100 1000 10000 100000
Flow, cfs
« Simulated A Observed ^^"Fbwer (Simulated) ^^~ Power (Observed)
    Figure 33. Power plot for observed and simulated TSS at USGS 02349605 Flint River at Ga 26, near
                Montezuma, GA - calibration period (HSPF).
                                               D-55

-------
               FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1986-1998
   -1
   "5>
    8

1000000

 100000

  10000

   1000

     100

      10

       1

     0.1
                         10       100       1000     10000    100000   1000000
                                          Flow, cfs
              »  Simulated  A  Observed
                                 Power (Simulated)
Power (Observed)
    Figure 34.  Power plot for observed and simulated TSS at USGS 02349605 Flint River at Ga 26, near
                Montezuma, GA - validation period (HSPF).


Standard time series plots (Figure 35) show that observed and simulated concentrations achieve good agreement,
although individual observations may deviate. Plots of concentration error versus flow and versus month (not
shown) were used to guard against hydrologic and temporal bias. Finally, statistics on concentration (Table 16)
show that acceptable median errors are achieved for the calibration period.
                                              D-56

-------
           FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA


_. I .

A
Observed
       1000
        100
   CO
          1999
2000
2001
Year
2002
    Figure 35. Time series plot of TSS concentration at USGS 02349605 Flint River at Ga 26, near
                Montezuma, GA - calibration period (HSPF).
Table 16.    Relative errors  (observed minus predicted), TSS concentration at USGS 02349605
             Flint River at Ga 26, near Montezuma, GA (HSPF).
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1999-2002)
49
3.30%
19.78%
Validation period
(1991-1998)
24
-97.85%
-33.27%
The general quality constituent parameters in the Lake Allatoona TMDL model for total phosphorus were
essentially directly transferred to the 20 Watershed model. As stated earlier, shrub land was parameterized
similarly to forested lands. The model simulates total phosphorus from the uplands as having both sediment-
associated and buildup-washoff components. The sediment-associated component of the surface load reflects
mineral phosphorus, while the buildup-washoff component addresses  organic phosphorus.

Initial parameter assignments for phosphorus for the Upper Flint calibration focus area performed well. Minor
adjustments were made to the interflow and groundwater component.  The same percent adjustment was made to
all land uses in order to keep the land use associated loading rates, developed in the TMDL model, intact. All
streams in the calibration focus area were supplied with the same first order decay rate. This decay rate was
obtained from the TMDL model and is consistent with other modeling work conducted throughout the state of
Georgia. Adjustment was not made to this parameter while calibration was performed on the calibration focus
area.

Monthly loading series for total phosphorus are shown in Figure 36 and load statistics are summarized in Table
17. In general, the observed and simulated total phosphorus loads attain an acceptable match for both the
calibration and validation periods. There are a few locations where the simulation is not trending and the
                                               D-57

-------
simulated loads are higher than the observed loads. These errors are most likely attributed to the error in the TSS
simulation during the same time period.
                                              Total P
     1000
      100 --
   o
   E
   In

   I
-Regression Loads

-Simulated Loads
       10 -----
    Figure 36. Fit for monthly load of total phosphorus at USGS 02349605 Flint River at Ga 26, near
                Montezuma, GA (HSPF).

Table 17.    Model fit statistics (observed minus predicted) for monthly total phosphorus loads
             using stratified regression
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1999-2002)
-59%
69%
35.3%
Validation period
(1986-1998)
-23%
35%
18.5%
As with TSS, additional diagnostics for total phosphorus included flow-load power plots (Figures 37 and 38),
time series plots (Figure 39) and analysis of concentration errors (Table 18). All show acceptable agreement.
                                               D-58

-------
FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1999-2002
•\ nn
I UU
10 -
>
ra
•o
1 1
3
§ 0.1
Q.
0.01 -
Onni


»
««»
•>••••**•*
«**i» * *£**

-------
                   FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA
                          A A A AAA  A AAA
          1999
2000
2001
Year
2002
    Figure 39. Time series plot of total phosphorus concentration at USGS 02349605 Flint River at Ga 26,
                near Montezuma, GA- calibration period (HSPF).
Table 18.    Relative errors (observed minus predicted), total  phosphorus concentration at
             USGS 02349605 Flint River at Ga 26, near Montezuma, GA (HSPF)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1999-2002)
48
-31 .2%
-26.5%
Validation period
(1986-1998)
80
-2.2%
-12.5%
As discussed above in the Calibration Data and Locations section of this report, the number of measured total
nitrogen observations was very limited for the ACF. The approach used to estimate the observed total nitrogen
should give reasonable values for total nitrogen since two of the three components making up total nitrogen were
measured. Similarly to total phosphorus, total nitrogen parameters in the Lake Allatoona TMDL model were
easily transferred to the 20 Watershed model. Also, similarly to total phosphorus, initial interflow and
groundwater total nitrogen concentrations were adjusted for all land uses together in order to keep the land use
associated loading rates, developed in the  TMDL model, intact.

Results for total nitrogen are summarized  in Figures 40 through 43 and Tables  19 and 20, following the same
format as total phosphorus. The results are acceptable, and generally better than those for total phosphorus. This is
because nitrogen is not sediment-associated, therefore, problems with sediment are not reflected in the calibration
for total nitrogen.
                                               D-60

-------
     3,000
                                           Total N
                                                                              -Averaging Loads
                                                                              -Simulated Loads
   Figure 40. Fit for monthly load of total nitrogen at USGS 02349605 Flint River at Ga 26, near
               Montezuma, GA (HSPF).
Table 19.    Model fit statistics (observed minus predicted) for monthly total nitrogen loads
            using averaging estimator (HSPF)
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1999-2002)
-30%
54%
26.1%
Validation period
(1986-1998)
-22%
42%
18.2
                                            D-61

-------
FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1999-2002
•\ nnn
I UUU
> 100-
ra
•o
«
J. 10-
•o
ra
o
* 1
01


•
„_._
» » »
.. .. ».«.•«
* ,».>»V>
* »*s
°t*tf ¥>
ji^1
-#r-

10 100 1000 10000 100000
Flow, cfs
• Simulated A Observed ^^"Fbwer (Simulated) ^^~ Power (Observed)
Figure 41. Power plot for observed and simulated total nitrogen at USGS 02349605 Flint River at Ga

            26, near Montezuma, GA - calibration period (HSPF).
            FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1986-1998
   1000
    100
ra
•o

«
c
o
•o
ra
o
                  10       100       1000     10000     100000   1000000


                                   Flow, cfs
        •  Simulated  A  Observed ^^™Power (Simulated) ^^™Power (Observed)
Figure 42. Power plot for observed and simulated total nitrogen at USGS 02349605 Flint River at Ga

            26, near Montezuma, GA - validation period (HSPF).
                                         D-62

-------
                        FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA
         1999
2000
2001
Year
2002
    Figure 43. Time series plot of total nitrogen concentration at USGS 02349605 Flint River at Ga 26,
                near Montezuma, GA- calibration period (HSPF).
Table 20.    Relative errors (observed minus predicted), total nitrogen concentration at USGS
             02349605 Flint River at Ga 26, near Montezuma, GA (HSPF)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1999-2002)
48
25.5%
19.3%
Validation period
(1986-1998)
76
27.2%
14.7%
Water Quality Results for Larger Watershed
Similar to hydrology, the Upper Flint water quality parameterization was not directly transferable to other areas
of the watershed. The Upper Flint parameters were utilized as starting parameters at the other calibration
locations. Once those locations reasonably agreed with the observed data, they were transferred to other parts of
the watershed, as with the hydrology calibration. The decay rates assigned to the streams in the calibration focus
area were also assigned to all streams in the ACF basin.

Upon initial water quality parameterization mapping, the water quality simulation below the reservoirs was
checked. At this point the simulation suggested that each of the reservoirs consumed all of the nitrogen and
phosphorus entering the reservoir and the water leaving the reservoir was free of all total nitrogen and total
phosphorus. This was taken to mean that due to longer residence time within reservoir reaches, the first order
decay rate applied to the reservoirs was too high. Total nitrogen and total phosphorus first order decay rates were
lowered until the simulation below each of the reservoirs better matched the observed water quality. This resulted
                                               D-63

-------
in decay rates that were an order of magnitude lower than those applied to the other stream reaches. HSPF
assumes that the water within a reach is both vertically and horizontally mixed; therefore it does not take into
account nutrient transformations and cycling occurring within a reservoir. The lowered decay rate was utilized as
a parameter to capture nutrient dynamics within the reservoirs, therefore they decay rate was the only parameter
adjusted in order to get a reasonable representation of water quality below the reservoirs.

Summary statistics for the water quality calibration and validation at all stations in the watershed are provided in
Tables 21 and 22.  The results of the water quality calibration and validation are much better at some gages than
others. The gages that are simulating poorly are probably doing so because of a lack of data to reasonably
construct observed monthly loadings. Another source of error that can balloon error statistics is poor hydrology
simulation at some of the monitoring locations. These two errors coupled together can severely impact the error
statistics presented.
Table 21.    Summary statistics for water quality (observed minus predicted) for all stations -
             calibration period 1999-2002 (HSPF)
Station
Relative Percent
Error TSS Load
TSS
Concentration
Median Percent
Error
Relative Percent
Error TP Load
TP
Concentration
Median Percent
Error
Relative Percent
Error TN Load
TN
Concentration
Median Percent
Error
02349605
-117%
19.78%
-59%
-26.5%
-30%
19.3%
02331600
-4%
16.73%
-1%
12.8%
-25%
4.2%
02336300
74%
10.16%
24%
-13.8%
-38%
16.8%
02353500
-438%
-38.90%
-205%
-224.5%
-38%
46.7%
02336000
-215%
1.72%
-77%
-40.1%
-35%
35.8%
02339500
-141%
-120.07%
-272%
-759.0%
2%
46.4%
02343801
34%
28.57%
-202%
-1107.9%
3%
54.8%
02353000
-63%
-64.66%
-82%
-89.2%
32%
53.1%
                                                 D-64

-------
Table 22.   Summary statistics for water quality (observed minus predicted) for all stations -
           validation period 1986-1998 (HSPF)
Station
Relative Percent
Error TSS Load
TSS
Concentration
Median Percent
Error
Relative Percent
Error TP Load
TP
Concentration
Median Percent
Error
Relative Percent
Error TN Load
TN
Concentration
Median Percent
Error
02349605
-78%
-33.3%
-23%
-12.5%
-22%
14.7%
02331600
90%
7.2%
53%
43.6%
-8%
1.3%
02336300
89%
3.5%
47%
5.9%
-21%
20.7%
02353500
-570%
-103.8%
-94%
-17.0%
-20%
32.7%
02336000
18%
6.7%
-16%
-9.1%
10%
20.0%
02339500
-7%
-36.9%
-64%
-85.2%
25%
-28.2%
02343801
84%
8.0%
54%
-499.6%
7%
31.2%
02353000
-77%
-1.0%
-22%
-26.1%
16%
-39.9%
                                          D-65

-------
 a a a a a a a a  a a a a a a  a a a a a a a a a a a a a  a a a a a a  a a a a a a a a a a a a a  a a a a a a  a a a a a a a a a a a a a  a a a a a a  a a a a a a a a a a a a a  a a
                   to
As mentioned in the above section for the HSPF model, no changes were made to the meteorological or land use
base data for the SWAT model. However, an error in the connectivity of reaches was found in the predefined
reach file, which was rectified. The impoundments of Lake Blackshear and Lake Seminole have large tributaries
contributing to them. The pre-defined reach file had reaches of the upstream tributaries draining into the next
downstream reach rather than into the reach within the subwatershed containing the lakes. This made the
contributing drainage areas for these lakes incorrect. The connectivity was modified so that the tributaries pour
into the reaches within the subwatersheds containing the lakes. Incorrect connectivity of reaches would pose a
problem because of the incorrect drainage area contributing flow to the impoundments. In the SWAT model,
reservoirs are modeled to be simulated at the  outlet of the subwatershed in which they are located and receive
flow from all the upstream drainage area.
Jim Woodruff Dam, Muckafoonee Creek Dam, West Point, and Buford (Table 6) reservoirs were represented in
the ACF basin 20 Watershed SWAT model. Pertinent reservoir information including surface area and storage at
principal (normal) and emergency spillway levels for the reservoirs modeled were obtained from the National
Inventory of Dams (NID) database (USAGE 1982). The SWAT model provides four options to simulate reservoir
outflow: 1) measured daily outflow, 2) measured monthly outflow, 3) average annual release rate for uncontrolled
reservoir, and 4) controlled outflow with target release. Keeping the goals of the 20 Watershed climate change
impact evaluation application, it was assumed that the best representation of the reservoirs was to simulate them
without supplying time series of outflow records. Therefore, a target release approach was used in the GCRP-
SWAT model. The average release rate was estimated using the outflow data available at the (USAGE 2010).
The number of days to reach target storage was assumed to be 90 days for all lakes except Lake Worth, which was
assumed to be 10 days.
Croplands occupy about eight percent of the total watershed area. It was found that irrigation occurred on about
5.5 percent of the total watershed area with 4.18 percent being irrigated by groundwater and 1.35 percent being
irrigated by surface water (Hook 2009). To simulate irrigation in the SWAT model, the auto-irrigation feature was
used in the management set-up on those HRUs that represented cotton and peanut crops.

The SWAT model setup for the ACF basin was set up fresh, with no prior-existing SWAT model for the
watershed. The model calibration period was set to calendar years 1993-2002.

Consistent with the HSPF modeling efforts, the specific site chosen for initial calibration was the Flint River at
GA 26, near Montezuma, GA (USGS 02349605) (Table 5). Most of the calibration efforts were geared toward
getting a closer match between simulated and observed flows at the outlet of the calibration focus area. Initially,
the parameters set for this area were applied across the watershed and the model performance was verified at other

                                                D-66

-------
stations. Model performance was not the same as it was for the calibration focus area, mostly due to the
dominance of different land uses in different parts of the watershed. In response to the variations in spatial
characteristics of the subwatersheds, a systematic adjustment of parameters individually, by land use type has
been adopted and the same adjustment is applied throughout the watershed. Observed data at other gaging stations
with the dominance of a different land use type was used to adjust the corresponding parameters. For example, at
the gaging station that drains predominantly urban land, the area was used to set the parameters for the urban land
areas.

It is acknowledged that a hydrologic/water quality model can be precisely calibrated, given the degree of freedom,
resources, time, and data. Keeping in view the interests of this project, which are to study the land use change and
climate change impacts on flow and water quality, a site specific calibration was deliberately not attempted. To
some extent, the limitation of this approach is that the local differences in soil, weather, management, and
hydrology is not thoroughly accounted for. This approach will provide an idea of the model performance when it
is not spatially-tightly calibrated and what to expect when transferring the parameters to other ungaged
watersheds or to watersheds where detailed modeling  is not practical due to limited resources.


Land  Use/Soil/Slope Definition
A 5/10/5 percent threshold was used for land use/soil/slope in the SWAT model while defining the HRUs. The
cropland HRUs were split into cotton, peanuts, and corn in proportions of 48, 30, and 22 percent, respectively.
Further these classes and the urban (including current and future urban class types) classes were exempt from
applying the thresholds.

The calibration focus area represents 21 subwatersheds that, together, consist of 1,342 HRUs. The parameters
were adjusted within the practical range to obtain a reasonable fit between the simulated and measured flows in
terms of Nash-Sutcliffe modeling efficiency and the high flow and low flow components as well as the seasonal
flows.  The general  land use characteristics of the watershed were represented well in the calibration focus area.
Two other locations: one predominantly forested (Chartahoochee River near Cornelia, GA; USGS 02331600) and
the other, predominantly urban (Peachtree Creek at Atlanta, GA; USGS 02336300), were chosen to set the
parameters for forest and urban areas, respectively. These parameters were then applied across the  entire
watershed. There is essentially one set of parameters for a land use type for the entire watershed.

During calibration, parameters were carefully adjusted such that different components of streamflow contribution
were adequately simulated. For instance, the observed and simulated baseflow and surface runoff contributions to
streamflow, as well as seasonal flows, matched well. Reasonable estimations of actual ET and crop yields were
also given consideration.

Calibration adjustments focused  on the following parameters:
    •   Curve numbers (varied systematically by land use)
    •   ESCO (soil evaporation compensation factor)
    •   SURLAG (surface runoff lag coefficient)
    •   Groundwater "revap" rates
    •   Baseflow factor
    •   GW_DELAY (groundwater delay time)
    •   GWQMN (threshold depth of water in the shallow aquifer required for return flow to occur)
    •   RevapMN  (threshold depth of water in the shallow aquifer required for "revap" or percolation to the deep
       aquifer to occur
    •   CANMAX (maximum canopy storage)
    •   Manning's "n" value for overland flow, main  channels, and tributary channels
    •   Sol_AWC  (available water capacity of the soil layer, mm water/mm of soil)


                                                 D-67

-------
Initial calibrations were performed for the Flint River at GA 26, near Montezuma, GA and are summarized in
Figures 44 through 50 and Tables 23 and 24. As evidenced through the time series plots and the Nash-Sutcliffe
modeling efficiency, the model performed well in simulating the timing and magnitude of streamflow for various
seasons.
         I Avg Monthly Rainfall (in)
         -Avg Observed Flow (1/1/1993 to 12/31/2002 )
         Avg Modeled Flow (Same Period)
i?nnnn
mnnnn
onnnn
finnnn
4nnnn
onnnn







Ljjt
"|







Jt
Hf'F





i
J
Jan-93 Jul-94
	 r





i,
m






,!
i
Jan-96


Ifll If





ill I
PT|
'









I
iJ
Jt ,
Jul-97
Date






Jj^
W






']


P|||"l '









_i









IP






_ o
4


m
19

Jan-99 Jul-00 Jan-02
    Figure 44.  Mean daily flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration
                 period (SWAT).
     30000
     20000
     10000
         J-93
                                                     Avg Monthly Rainfall (in)
                                                     -Avg Observed Flow (1/1/1993 to 12/31/2002 )
                                                     Avg Modeled Flow (Same Period)
                                                                         J-02
                                                                                      •E
                                                                                      o
    Figure 45.  Mean monthly flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
                 calibration period (SWAT).
                                                 D-68

-------
i
I
   30000
T3
JD
 CD
T3
 O
         •   Avg Flow (1/1/1993 to 12/31/2002)

        	Line of Equal Value

        	Best-Fit Line
   20000
 CD
 D)

 5
 CD

 <
   10000
                 10000      20000     30000

            Average Observed Flow (cfs)
                                             T3
                                             O
                                             CD
                                             O
                                             C
                                            JS
                                             ro
                                            m

                                             (D
                                            "5
                                                       Avg Observed Flow (1/1/1993 to 12/31/2002)
                                                       Avg Modeled Flow (1/1/1993 to 12/31/2002 )
                                                    	Line of Equal Value
                                                   J-93   J-94   J-96
                                                                     J-97   J-99

                                                                      Month
                                                                                 J-00   J-02
   Figure 46.  Monthly flow regression and temporal variation at USGS 02349605 Flint River at Ga 26,
                 near Montezuma, GA- calibration period (SWAT).
           Avg Flow (1/1/1993 to 12/31/2002)
          • Line of Equal Value

          -Best-Fit Line
  8000
£
| 6000
                                                     Avg Monthly Rainfall (in)
                                                    -Avg Observed Flow (1/1/1993 to 12/31/2002)
                                                    -Avg Modeled Flow (Same Period)
                                                8000
                                             I
                                             I
                                                6000
                                                4000
                                                2000
             2000    4000     6000

            Average Observed Flow (cfs)
                                      8000
                                                        2  3  4  5  6  7  8   9  10 11  12

                                                                   Month
   Figure 47.  Seasonal regression and temporal aggregate at USGS 02349605 Flint River at Ga 26, near
                 Montezuma, GA - calibration period (SWAT).
                                                    D-69

-------
     Average Monthly Rainfall (in)
    -Median Observed Flow (1/1/1993 to 12/31/2002)

    12000 -
 I
 I
• Observed (25th, 75th)
Modeled (Median, 25th, 75th)
                                                                   11
                                                                        12
   Figure 48. Seasonal medians and ranges at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
                - calibration period (SWAT).
Table 23.   Seasonal summary at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
            calibration period (SWAT)
MONTH
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
3955.13
5767.27
6976.42
3970.93
2277.06
1783.49
3617.54
1440.71
1260.28
1816.83
2297.06
3137.85
3665.00
3915.00
5150.00
3065.00
1810.00
1670.00
1290.00
1035.00
964.00
1105.00
1660.00
2240.00
2110.00
2572.50
2932.50
2210.00
1400.00
1180.00
977.25
752.00
770.50
775.75
1220.00
1490.00
5155.00
6887.50
8377.50
4297.50
2470.00
2170.00
1765.00
1420.00
1252.50
1855.00
2685.00
3497.50
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
4206.77
6093.10
7590.84
4303.82
2405.19
1937.85
3856.12
1715.12
1538.55
1874.05
2484.96
3082.68
3393.41
3996.73
5958.04
3329.58
2322.73
1659.04
1342.07
1139.52
957.90
1162.76
1619.56
1810.72
1700.42
2395.72
2700.58
2078.74
1358.75
909.57
644.18
673.31
710.14
673.08
854.25
1116.26
5504.55
7738.14
10476.33
5450.56
3121.85
2654.73
2287.50
1722.66
1670.33
2586.26
3456.35
361 1 .32
                                              D-70

-------
            •Observed Flow Duration (1/1/1993 to 12/31/2002 )


            •Modeled Flow Duration (1/1/1993 to 12/31/2002 )
    1000000
i

o
0
D)
ro
Q
     100000
      10000
       1000 =
       100
          0%     10%     20%     30%    40%     50%     60%     70%     80%     90%     100%



                              Percent of Time that Flow is Equaled or Exceeded
  Figure 49.  Flow exceedence at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - calibration

                period (SWAT).
  o
  o
  T3


  I
  0
  0


  _g
  o



  o
  T3
  0
  N
                •Observed Flow Volume (1/1/1993 to 12/31/2002 )


                Modeled Flow Volume (1/1/1993 to 12/31/2002 )
       120%
       100%
        80% -
60%
40%
20%
          Jan-93
                     Jul-94
                                Jan-96
                                            Jul-97
                                                       Jan-99
                                                                   Jul-00
                                                                              Jan-02
  Figure 50.  Flow accumulation at USGS 02349605 Flint River at Ga 26, near Montezuma, GA

                calibration period (SWAT).
                                                  D-71

-------
Table 24.   Summary statistics at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
            calibration period (SWAT)
SWAT Simulated Flow
REACH OUTFLOW FROM OUTLET 38
10-Year Analysis F^riod: 1/1/1993 - 12/31/2002
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9):
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3):
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe CoefficjejTtjjfJEfficjejTC^JE^
Baseline adjusted coefficient (Garrick), E':
Monthly NSE 1
15.88

6.39
2.49

2.79
2.91
6.85
3.33

4.22
0.90
Error Statistics
7.28
-9.39
11.34
12.46
Observed Flow Gage
USGS 02349605 FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA
Hydrologic Unit Code: 3130006
Latitude: 32.29305556
Longitude: -84.04361 11
Drainage Area (sq-rri): 2920
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow Volume (1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
2.56 » | 30
7.19
7.64
-12.03
-12.28
0.624
0.442
0.876 1
30
30
20
50
Model accuracy increases
as E or E' approaches 1.0





14.80

5.74
2.75

2.48
2.83
6.39
3.10

4.80
1.02





Clear [













Hydrology Validation
Consistent with HSPF modeling efforts, validation for the Upper Flint calibration focus area was performed at the
same location but for calendar years 1983-1992. Results are presented in Figures 51 through 57 and Tables 25 and
26. Although, the Nash-Sutcliffe modeling efficiency is not as good as it was for the calibration period, the model
performance was adequate for the validation period. None of the metrics fall out of the range set for the 20
Watershed study.
                                              D-72

-------
         Avg Monthly Rainfall (in)
        -Avg Observed Flow (1/1/1983to 12/31/1992)
         Avg Modeled Flow (Same Period)
    70000
                    Jul-84
           Jan-86
Jul-87       Jan-89
     Date
Jul-90
Jan-92
                                                                                                c
                                                                                                'ro
                                                                                                ro
                                                                                                Q
 Figure 51. Mean daily flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
               period (SWAT).
I
    15000
    10000
    5000
         J-83
J-84
                                                           Avg Monthly Rainfall (in)
                                                           -Avg Observed Flow (1/1/1983 to 12/31/1992 )
                                                           Avg Modeled Flow (Same Period)
                                      J-92
 Figure 52. Mean monthly flow at USGS 02349605 Flint River at Ga 26, near Montezuma, GA -
               validation period (SWAT).
                                               D-73

-------
             Avg Flow (1/1/1983 to 12/31/1992)

            • Line of Equal Value

             Best-Fit Line
t


_o

U_

•a
_CD

CD
•a
o
CD
O)
   15000
10000
    5000
                  5000
                          10000
15000
                                                 100% i

                                             f   90% ^
                                                         Avg Observed Flow (1/1/1983 to 12/31/1992 )

                                                         Avg Modeled Flow (1/1/1983 to 12/31/1992 )

                                                        - Line of Equal Value
            Average Observed Flow (cfs)
                                                     J-83   J-84    J-86   J-87   J-89   J-90   J-92


                                                                         Month
  Figure 53.  Monthly flow regression and temporal variation at USGS 02349605 Flint River at Ga 26,

               near Montezuma, GA - validation period (SWAT).
  8000
        •  Avg Flow (1 /1 /1983 to 12/31 /1992)

       	Line of Equal Value

           Best-Fit Line
              2000     4000     6000


             Average Observed Flow (cfs)
                                     8000
                                                I
                                                    8000
                                                    6000
                                                    4000
                                                    2000
                                                       Avg Monthly Rainfall (in)

                                                      -Avg Observed Flow (1/1/1983 to 12/31/1992)

                                                      •Avg Modeled Flow (Same Period)
                                                            Illllllllll
                    2345678


                                Month
                                          9  10 11  12
  Figure 54.  Seasonal regression and temporal aggregate at USGS 02349605 Flint River at Ga 26, near

               Montezuma, GA - validation period (SWAT).
                                                D-74

-------
      Average Monthly Rainfall (in)
    -Median Observed Flow (1/1/1983 to 12/31/1992)

    8000
• Observed (25th, 75th)
 Modeled (Median, 25th, 75th)
 I

 I
                                                                            11
                                   12
                                            to
   Figure 55. Seasonal medians and ranges at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
               - validation period (SWAT).
Table 25.   Seasonal summary at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
            validation period (SWAT)
MONTH

Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
4168.00
5242.97
5976.65
4328.17
2451 .47
1753.24
1908.70
1632.54
1283.93
1425.06
2275.14
3667.95
3115.00
4150.00
4310.00
3275.00
2135.00
1 300.00
1 330.00
1150.00
1120.00
1 020.00
1 330.00
2275.00
2152.50
2850.00
2830.00
2277.50
1420.00
986.75
927.50
816.50
881 .75
869.50
1120.00
1602.50
5350.00
6430.00
6542.50
4880.00
2937.50
2072.50
2257.50
1830.00
1500.00
1390.00
2277.50
4277.50
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
4143.78
5285.84
6060.01
4531.12
2658.75
2095.76
2160.97
1732.26
1534.60
1509.13
2284.09
3322.66
2886.66
4532.61
4515.09
3509.51
2418.76
1629.70
1141.19
1482.31
1359.25
945.55
1154.11
1848.91
1525.30
2618.46
2739.77
2517.97
1 824.58
1 060.96
757.51
655.52
812.62
654.00
608.47
833.94
5836.80
7460.70
7304.52
5265.76
3102.27
2490.67
2715.43
2563.83
2005.97
1823.54
2252.94
4660.44
                                             D-75

-------
             •Observed Flow Duration (1/1/1983 to 12/31/1992)
              Modeled Flow Duration (1/1/1983 to 12/31/1992)
t
o
O)
CO
Q
    100000
     10000
    1000
       100
                  10%
                        20%     30%      40%     50%      60%     70%      80%

                              Percent of Time that Flow is Equaled or Exceeded
                                                   90%     100%
  Figure 56.  Flow exceedence at USGS 02349605 Flint River at Ga 26, near Montezuma, GA - validation
               period (SWAT).
  o
  o
  CO
  T3
  .Q
  O
o


_o
LJ_
T3
N
"CD
                 •Observed Flow Volume (1/1/1983 to 12/31/1992)

                  Modeled Flow Volume (1/1/1983 to 12/31/1992)
       120%
       100%
        80% -
        60% -
        40% -
        20%
          Jan-83
                     Jul-84
Jan-86
Jul-87
Jan-89
Jul-90
Jan-92
  Figure 57.  Flow accumulation at USGS 02349605 Flint River at Ga 26, near Montezuma, GA -
               validation period (SWAT)
                                               D-76

-------
Table 26.    Summary statistics at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
             validation period (SWAT)
SWAT Simulated Flow
REACH OUTFLOW FROM OUTLET 37
10-Year Analysis F^riod: 1/1/1983 - 12/31/1992
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9):
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3):
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient _of_EfficiencyJ_Ej___
Baseline adjusted coefficient (Garrick), E':
Monthly NSE 1
14.42

5.01
2.57

2.12
2.78
5.93
3.58

3.67
0.46
Error Statistics
3.33
-7.57
2.89
12.43
Observed Flow Gage
USGS 02349605 FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA
Hydrologic Unit Code: 3130006
Latitude: 32.29305556
Longitude: -84.04361 1 1
Drainage Area (sq-rri): 2920
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Obsei-ved Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow Volume (1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
-3.46 » | 30
0.66
8.82
-16.93
-10.04
0.559
0.444
0.833
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0





13.95

4.87
2.78

1.89
2.88
5.89
3.29

4.42
0.51





Clear [













Hydrology Results for Larger Watershed
Calibration and validation results at all gages are summarized in Tables 29 and 30. As discussed above, a detailed
spatial calibration was not conducted for the GCRP-SWAT model of the ACF basin. The parameterization is
identical across the entire watershed, although, measured flow and water quality data at other stations (other than
the calibration focus area) where land use dominance occurred were used to set the corresponding parameters. A
better model fit could perhaps be achieved if the model was more tightly calibrated but this was not attempted
deliberately keeping in view the intended bigger scope of the project.

In general, the model performance was good, as noticed from the Nash-Sutcliffe modeling efficiency statistics,
except for forest and urban dominated subwatersheds. The model over-predicted flow in winter months and
under-predicted flow in spring months. Also, the Ichawaynochaway Creek subwatershed (station 02353500)
simulation statistics are relatively lower than those at the other stations. The model over-predicted high flows and
under-predicted low flows. The simulated peak flows correlated well with the observed high rainfall events;
however, watershed response, as noticed from the measured streamflows, didn't result in such high flows.
                                               D-77

-------
All the reservoir parameters were set such that inflow equals outflow and this approach worked well for all
reservoirs except Lake Lanier. Similar to the HSPF results, after the confluence of Chattahoochee and Flint rivers
and downstream of Lake Seminole, the simulation and the model fit greatly improved as shown in Figures 58
through 64 and Tables 27 and 28.
       ^H Avg Monthly Rainfall (in)
       	Avg Observed Flow (1/1/1993 to 12/31/2002)
          Avg Modeled Flow (Same Period)
      250000

      200000

 %   150000

 £   100000

       50000
          Jan-93
Jul-94
Jan-96
Jul-97
Jan-99
Jul-00
Jan-02
                                                Date
    Figure 58.  Mean daily flow at USGS 02358000 Apalachicola River at Chattahoochee, FL- calibration
                 period (SWAT)
      100000

      80000
           J-93
                                                          Avg Monthly Rainfall (in)
                                                         -Avg Observed Flow (1/1/1993 to 12/31/2002 )
                                                         •Avg Modeled Flow (Same Period)
 J-94
                                               J-02
                                                                                             f
                                                                                             ro
                                                                                             01
                                                                                             o
    Figure 59.  Mean monthly flow at USGS 02358000 Apalachicola River at Chattahoochee, FL-
                 calibration period (SWAT)
                                                 D-78

-------
         •  Avg Flow (1/1/1993 to 12/31/2002 )
        • - - - • Line of Equal Value
        	Best-Fit Line
   100000
O)
TO
   20000
                                             O
                                             0)
                                             o
                                              ro
                                             m
                                              _
                                              0)
                                             to
100%
 90% -
 80%
 70% -
 60%
 50%
       Avg Observed Flow (1/1/1993 to 12/31/2002 )
      I Avg Modeled Flow (1/1/1993 to 12/31/2002 )
      -Line of Equal Value
 40% -
 30% -
 20% -
 10% -
  0%
         0   20000  40000 60000 80000 100000
            Average Observed Flow (cfs)
    J-93   J-94   J-96
                      J-97   J-99
                       Month
                                  J-00   J-02
  Figure 60.  Monthly flow regression and temporal variation at USGS 02358000 Apalachicola River at
               Chattahoochee, FL- calibration period (SWAT)
                                                      Avg Monthly Rainfall (in)
                                                     -Avg Observed Flow (1/1/1993 to 12/31/2002)
                                                      Avg Modeled Flow (Same Period)
                                                 50000
                                                 40000
Best-Fit Line
c;nnnn -,
&
-ii/innnn
O
LL
-n ^nnnn
"oi
T3
° onnnn
0
D)
2 1 nnnn
^j 1UUUU
n -
y = 0.8975X +
p2 = n 9fl




»
itf*
.,--
3900.1
R2

,x



X



f '




           10000  20000 30000  40000  50000
            Average Observed Flow (cfs)
                                                                                  10  11  12
  Figure 61.  Seasonal regression and temporal aggregate at USGS 02358000 Apalachicola River at
               Chattahoochee, FL- calibration period (SWAT)
                                                 D-79

-------
      Average Monthly Rainfall (in)
    -Median Observed Flow (1/1/1993 to 12/31/2002)

    60000
    50000
    40000
    30000
    20000 -
    10000
• Observed (25th, 75th)
Modeled (Median, 25th, 75th)
                                                                         11
                                                                               12
                                          c
                                          'co

                                          >,
                                                                                        o
    Figure 62.  Seasonal medians and ranges at USGS 02358000 Apalachicola River at Chattahoochee,
                FL- calibration period (SWAT)
Table 27.    Seasonal summary at USGS 02358000 Apalachicola River at Chattahoochee, FL-
             calibration period (SWAT)
MONTH
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
24730.16
34453.44
41056.97
25249.20
15972.58
12939.23
18873.90
12411.87
10850.80
12458.10
13896.10
19332.77
19950.00
29050.00
36850.00
19800.00
15250.00
12600.00
11950.00
11050.00
8235.00
10750.00
13500.00
14550.00
13500.00
18275.00
19300.00
16000.00
9570.00
8237.50
7665.00
7102.50
6530.00
6112.50
6557.50
9150.00
30475.00
43900.00
53300.00
30250.00
19675.00
17600.00
15575.00
14300.00
12900.00
13900.00
18250.00
23800.00
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
26487.83
33462.26
41547.96
27986.54
18337.31
15481.95
21090.83
15037.00
14122.01
15677.74
16163.41
18812.57
20770.32
29041 .02
36338.79
24610.79
18704.41
16359.52
15028.16
12370.73
10885.75
11788.04
12783.91
13546.71
12429.88
15105.85
17536.38
15952.52
11760.67
9413.12
10404.58
9075.87
7701 .25
8509.07
9989.64
9380.46
33656.64
44319.91
56838.96
37760.21
22816.81
19775.33
18080.23
15658.52
17196.48
17775.64
20016.35
22589.03
                                              D-80

-------
o
D)
ro
Q
            •Observed Flow Duration (1/1/1993 to 12/31/2002 )
             Modeled Flow Duration (1/1/1993 to 12/31/2002 )
  1000000


   100000


    10000


     1000


      100


       10
          1
        0.1
          0%     10%     20%      30%     40%     50%     60%     70%      80%

                               Percent of Time that Flow is Equaled or Exceeded
                                                                                    90%
                                                                                            100%
  Figure 63. Flow exceedence at USGS 02358000 Apalachicola River at Chattahoochee, FL- calibration
                period (SWAT)
  o
  o
ro
T3

I

.a



E

"o


^3
LL
T3

.N
(0
                •Observed Flow Volume (1/1/1993 to 12/31/2002 )

                 Modeled Flow Volume (1/1/1993 to 12/31/2002 )
       120%
       100%
        80%
        60%
        40%
        20%
          Jan-93
                      Jul-94
                                  Jan-96
                                               Jul-97
                                                           Jan-99
                                                                       Jul-00
                                                                                   Jan-02
  Figure 64. Flow accumulation at USGS 02358000 Apalachicola River at Chattahoochee, FL-
                calibration period (SWAT)
                                                 D-81

-------
Table 28.   Summary statistics at USGS 02358000 Apalachicola River at Chattahoochee, FL-
           calibration period (SWAT)

REACH OUTFLOW FROM OUTLET 13
10-Year Analysis F^riod: 1/1/1993 - 12/31/2002
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9):
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3):
Simulated Srjring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
_Nas^3ut£liffeJDpj5^
Baseline adjusted coefficient (Garrick), E':
Monthly NSE H
17.35

5.19
4.21

3.34
3.36
6.60
4.05

3.13
0.62
Error Statistics
9.16
15.49
1.57
19.16
Observed Flow Gage
USGS 02358000 APALACHICOL/
Hydrologic Unit Code: 3130011
Latitude: 30.701 0251
Longitude: -84.8590871
Drainage Area (sq-rri): 17200

i RIVER AT CHATTAHOOCHEE FLA
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow Volume (1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
10.82 » | 30
1.39
14.12
-15.66
-3.56
0.793
0.543
0.919 \
30
30
20
50
Model accuracy increases
as E or E' approaches 1.0





15.89

5.11
3.65

2.80
3.03
6.51
3.55

3.71
0.64





Clear [













                                        D-82

-------
Table 29.    Summary statistics (percent error) for all stations - calibration period 1993-2002
            (SWAT)
Station
Error in total
volume:
Error in 50%
lowest flows:
Error in 10%
highest flows:
Seasonal
volume error
- Summer:
Seasonal
volume error
- Fall:
Seasonal
volume error
- Winter:
Seasonal
volume error
- Spring:
Error in storm
volumes:
Error in
summer
storm
volumes:
Daily Nash-
Sutcliffe
Coefficient of
Efficiency, E:
Monthly
Nash-
Sutcliffe
Coefficient of
Efficiency, E:
02349605
7.28
-9.39
11.34
12.46

2.56


7.19


7.64

-12.03

-12.28

0.624


0.876


02331600
1.74
-44.46
28.59
9.11

22.61


16.05


-27.05

71.60

107.67

0.358


0.560


02336300
-10.17
-22.69
-27.08
26.93

-15.44


10.95


-24.11

-38.66

-49.29

0.334


0.336


02336000
17.52
34.07
-10.28
-4.30

25.99


13.79


37.03

-69.57

-77.02

0.489


0.631


02353500
5.89
-20.35
37.75
0.78

-2.69


25.04


-21.27

25.40

-10.26

0.335


0.532


02339500
6.26
33.88
-7.29
-8.34

0.51


5.34


25.48

-75.17

-82.70

0.542


0.837


02343801
16.53
62.17
-10.00
28.70

19.75


-2.93


41.14

-52.52

-58.88

0.642


0.837


02353000
10.45
-2.48
16.02
17.53

1.02


9.50


15.31

-48.51

-42.31

0.697


0.830


02358000
9.16
15.49
1.57
19.16

10.82


1.39


14.12

-15.66

-3.56

0.793


0.919


                                           D-83

-------
Table 30.    Summary statistics (percent error) for all stations - validation period 1983-1992
             (SWAT)
Station
Error in total
volume:
Error in 50%
lowest flows:
Error in 10%
highest flows:
Seasonal
volume error
- Summer:
Seasonal
volume error
- Fall:
Seasonal
volume error
- Winter:
Seasonal
volume error
- Spring:
Error in storm
volumes:
Error in
summer
storm
volumes:
Daily Nash-
Sutcliffe
Coefficient of
Efficiency, E:
Monthly
Nash-
Sutcliffe
Coefficient of
Efficiency, E:
02349605
3.33
-7.57
2.89
12.43
-3.46
0.66
8.82
-16.93
-10.04
0.559
0.833
02331600
-4.33
-49.69
26.93
-13.83
9.43
11.99
-27.97
56.71
66.81
0.347
0.578
02336300
-12.32
-23.46
-29.96
-18.32
-12.56
-3.15
-18.51
-40.06
-43.98
0.420
0.587
02336000
5.62
18.44
-16.50
-15.29
-3.92
23.35
18.15
-73.11
-81.91
0.369
0.484
02353500
1.03
-38.73
35.73
14.60
-15.20
9.61
-11.43
26.32
14.15
0.222
0.543
02339500
0.92
36.21
-12.47
-11.16
-15.65
12.01
15.39
-77.02
-83.37
0.481
0.758
02343801
11.47
50.10
-10.38
25.32
1.66
1.04
27.43
-65.86
-76.19
0.599
0.829
02353000
6.48
-7.51
5.33
21.21
-0.29
3.04
8.68
-52.34
-52.71
0.685
0.785
02358000
2.17
-7.83
-2.31
9.39
-3.96
1.04
3.84
-8.64
-6.41
0.770
0.901
Water Quality Calibration and Validation
Initial calibration and validation of water quality was done on the Flint River near Montezuma (USGS02349605),
using calendar years 1999-2002 for calibration and calendar years 1991-1998 for validation. As with hydrology,
calibration was performed on the later period as this better reflects the land use included in the model. The start of
the validation period is constrained by data availability.

Calibration adjustments for sediment focused on the following parameters:
    •  PRF (Peak rate adjustment factor for sediment routing in the main channel)
    •  SPCON and SPEXP (Linear and Exponent parameters for estimating maximum amount of sediment that
       can be reentrained during channel sediment routing)
    •  RSDCO (Residue decomposition coefficient)
    •  USLE-P (USLE equation support practice factor
                                              D-84

-------
Simulated and estimated sediment loads at the Montezuma station for both periods are shown in Figures 65
through 68 and statistics for the two periods are provided separately in Tables 31 and 32. The key statistic in
Table 31 is the relative percent error, which shows the error in the prediction of monthly load normalized to the
estimated load. Table 31 also shows the relative average absolute error, which is the average of the relative
magnitude of errors in individual monthly load predictions. This number is inflated by outlier months in which the
simulated and estimated loads differ by large amounts (that may be as easily due to uncertainty in the estimated
load due to limited data as to problems with the model) and the third statistic, the relative median absolute error,
is likely more relevant and shows good agreement.
     1,000,000
                                                TSS
                                                                                   -Regression Loads
                                                                                   -Simulated Loads
    Figure 65. Fit for monthly load of TSS at USGS 02349605 Flint River at Ga 26, near Montezuma, GA -
                calibration period (SWAT).
Table 31.    Model fit statistics (observed minus predicted) for monthly TSS loads using
             stratified regression at USGS 02349605 Flint River at Ga 26, near Montezuma, GA
             (SWAT)
Statistic
Relative Percent Error
Relative Average Absolute Error
Relative Median Absolute Error
Calibration period
(1999-2002)
-9%
66%
39.1%
Validation period
(1991-1998)
17%
42%
25.8%
                                               D-85

-------
           FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1999-2002
o
•o
ra
o
CO

V)
                      10
100        1000


    Flow, cfs
    10000
100000
              Simulated  A  Observed
   Fbw er (Simulated)
 Power (Observed)
Figure 66.  Power plot for observed and simulated TSS at USGS 02349605 Flint River at Ga 26, near

            Montezuma, GA - calibration period (SWAT).
           FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1986-1998
^

"5>
•o
ra
o
CO

V)
                       10
 100        1000


    Flow, cfs
    10000
100000
           •  Simulated  A  Observed
  Row er (Simulated)
Power (Observed)
Figure 67. Power plot for observed and simulated TSS at USGS 02349605 Flint River at Ga 26, near

            Montezuma, GA - validation period (SWAT).
                                         D-86

-------
A low baseflow recession factor used in the calibrated model setup allowed streamflow to reach very low values,
which in turn simulated numerous days with extremely low sediment values as seen in Figure 68.
           FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA
                                                        • Simulated    A Observed
      1000
        100
    O)

   eo
   CO
          1999
2000
2001
Year
2002
    Figure 68.  Time series plot of TSS concentration at USGS 02349605 Flint River at Ga 26, near
                Montezuma, GA (SWAT).
Table 32.    Relative errors (observed minus predicted), TSS concentration at USGS 02349605
             Flint River at Ga 26, near Montezuma, GA (SWAT)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1999-2002)
49
63.01%
64.14%
Validation period
(1991-1998)
24
60.61%
52.79%
Calibration adjustments for total phosphorus and total nitrogen focused on the following parameters:
    •   PPERCO (phosphorus percolation coefficient)
    •   NPERCO (nitrogen percolation coefficient)
    •   PHOSKD (phosphorus soil partitioning coefficient)
    •   HLIFE_NGW (half life of nitrate in the shallow aquifer)
    •   SOL_CBN1 (organic carbon in the first soil layer)
    •   QUAL2E parameters such as algal, organic nitrogen, and organic phosphorus settling rate in the reach,
       benthic source arte for dissolved phosphorus and NH4-N in the reach, fraction of algal biomass that is
       nitrogen and phosphorus, Michaelis-Menton half-saturation constant for nitrogen and phosphorus
                                              D-87

-------
In general, the match between observed and measured total phosphorus and total nitrogen was acceptable. Total
phosphorus and total nitrogen calibration results are presented in Figures 69 through 76 and Tables 33 through 36.
                                            Total P
     1000
      100 --
   o
   «
   I
-Regression Loads
-Simulated Loads
       10 ---
    Figure 69. Fit for monthly load of total phosphorus at USGS 02349605 Flint River at Ga 26, near
                Montezuma, GA (SWAT).
Table 33.   Model fit statistics (observed minus predicted) for monthly total phosphorus loads
            using stratified regression at USGS 02349605 Flint River at Ga 26, near Montezuma,
            GA (SWAT)
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1999-2002)
-50%
72%
31.8%
Validation period
(1991-1998)
-30%
49%
18.9%
                                             D-88

-------
5
"5>
•o
ra
o
    0.01
   0.001
            FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1999-2002
                     10
100         1000


    Flow, cfs
10000       100000
             Simulated  A  Observed     Power (Simulated) ^^™Power (Observed)
Figure 70. Power plot for observed and simulated total phosphorus at USGS 02349605 Flint River at

            Ga 26, near Montezuma, GA - calibration period (SWAT).
            FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1986-1998
   1000
   0.01
                    10
100         1000


    Flow, cfs
10000
100000
           Simulated  A  Observed ^^™Power (Simulated) ^^™Power (Observed)
Figure 71. Power plot for observed and simulated total phosphorus at USGS 02349605 Flint River at

            Ga 26, near Montezuma, GA - validation period (SWAT).
                                         D-89

-------
                 FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA
       1999
2000
2001

Year
2002
   Figure 72. Time series plot of total phosphorus concentration at USGS 02349605 Flint River at Ga 26,
              near Montezuma, GA (SWAT).
Table 34.   Relative errors (observed minus predicted), total phosphorus concentration, USGS
           02349605 Flint River at Ga 26, near Montezuma, GA (SWAT)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1999-2002)
48
-32.67%
-41.28%
Validation period
(1991-1998)
80
-8.3%
-11.99%
                                         D-90

-------
                                          Total N
                                                                             -Averaging Loads
                                                                             -Simulated Loads
                                                       O    T-    CN
                                                       999
                                                       c    c    c
                                                       03    03    03
   Figure 73. Fit for monthly load of total nitrogen at USGS 02349605 Flint River at Ga 26, near
               Montezuma, GA (SWAT).
Table 35.    Model fit statistics (observed minus predicted) for monthly total nitrogen loads
            using averaging estimator at USGS 02349605 Flint River at Ga 26, near Montezuma,
            GA (SWAT)
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1999-2002)
-18%
31%
15.7%
Validation period
(1991-1998)
9%
30%
19.9%
                                           D-91

-------
            FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1999-2002
   100
                   10
100         1000


    Flow, cfs
10000
100000
         •  Simulated  A  Observed ^^~Power (Simulated) ^^"Fbwer (Observed)
Figure 74. Power plot for observed and simulated total nitrogen at USGS 02349605 Flint River at Ga

            26, near Montezuma, GA - calibration period (SWAT).
            FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA 1986-1998
   1000
    100
5
"5>
•o
ra
o
                    10
 100        1000


    Flow, cfs
10000
100000
           Simulated  A  Observed ^^™Power (Simulated) ^^"Fbwer (Observed)
Figure 75. Power plot for observed and simulated total nitrogen at USGS 02349605 Flint River at Ga

            26, near Montezuma, GA - validation period (SWAT).
                                         D-92

-------
       1999
                       FLINT RIVER AT GA 26, NEAR MONTEZUMA, GA
2000
2001
Year
2002
    Figure 76. Time series plot of total nitrogen concentration at USGS 02349605 Flint River at Ga 26,
                near Montezuma, GA (SWAT).
Table 36.   Relative errors (observed minus predicted), total nitrogen concentration, USGS
            02349605 Flint River at Ga 26, near Montezuma, GA (SWAT)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1999-2002)
48
-14.52%
-10.44%
Validation period
(1991-1998)
76
2.26%
8.07%
Water Quality Results for Larger Watershed
As with hydrology, the Flint River watershed parameters for water quality were directly transferred to other
portions of the watershed. In general, simulated sediment was low and simulated total phosphorus and total
nitrogen was high at most of the stations. Ortho phosphorus and mineral nitrogen made up most of the total
phosphorus and total nitrogen as organic components corresponded to the sediment fraction. Summary statistics
for the water quality calibration and validation at other stations in the watershed are provided in Tables 37 and 38,
respectively.
                                             D-93

-------
Table 37.   Summary statistics (observed minus predicted) for water quality for all stations -
           calibration period 1999-2002 (SWAT)
Station
Relative
Percent Error
TSS Load
TSS
Concentration
Median
Percent Error
Relative
Percent Error
TP Load
TP
Concentration
Median
Percent Error
Relative
Percent Error
TN Load
TN
Concentration
Median
Percent Error
02349605
-9


64.1



-50


-41.3



-18


-10.4



02331600
37


-26.0



25


0.0



13


54.3



02336300
91


-20.4



40


-21.6



-59


15.8



02336000
-3


3.2



-29


-56.4



-117


-46.1



02353500
33


41.9



-305


-124.5



0


63.9



02339500
71


16.9



-317


-1136.7



-606


-461.8



02343801
83


32.6



-85


-814.9



-310


-335.7



02353000
66


33.7



-45


-24.8



26


25.8



                                          D-94

-------
Table 38.   Summary statistics (observed minus predicted) for water quality for all stations -
           validation period 1986-1998 (SWAT)
Station
Relative
Percent
Error TSS
Load
TSS
Concentrati
on Median
Percent
Error
Relative
Percent
Error TP
Load
TP
Concentrati
on Median
Percent
Error
Relative
Percent
Error TN
Load
TN
Concentrati
on Median
Percent
Error
02349605
17



53.8




-30



-12.0




9



8.1




02331600
91



0.9




62



46.6




44



59.7




02336300
93



1.7




18



11.2




-75



37.4




02336000
33



14.4




14



2.8




-52



-50.3




02353500
42



89.9




-177



-15.1




14



70.6




02339500
87



54.9




-59



-134.0




-280



-467.1




02343801
95



31.1




72



-291.1




-144



-229.4




02353000
65



33.8




-10



-12.1




37



20.2




                                          D-95

-------
References

Alabama River Alliance. 2007. Water Wars Background: http://www.alabamarivers.org/current_work/water-wars
(Accessed January 2, 2010).

Cherry, R.N. 1961. Chemical Quality of Water of Georgia Stream. 1957-58, U.S. Geological Survey Bulletin No.
69.

Couch, C.A. 1993. Environmanetal Seetting of the Apalachicola-Chattahoochee_Flint River Basin. Proceedings
of the 1993 Georgia Water Resources Conference. April 20-21, 1993, University of Georgia.

Hook, J.E., 2009. Agricultural Irrigation Water Demand (1 June 2009):
http: //www .ne spal.org/SIRP/waterinfo/State/common/AgWaterDemand .htm.

Preston, S.D., V.J. Bierman, Jr., and S.E. Silliman.  1989. An evaluation of methods for the estimation of tributary
mass loads. Water Resources Research. 25(6): 1379-1389.

USAGE (Unites States Army Corps of Engineers). 2010. Apahalachicola-Chattahoochee-Flint River (1/23/2009):
http://water.sam.usace.army.mil/acfframe.htm (Accessed August 10, 2010).

U. S. Army Corps of Engineers (USAGE). 1982. National inventory of dams database in card format, available
from National Technical Information Service,  Springfield, VA 22162, #ADA 118670.

USEPA (United States Environmental Protection Agency). 2008. Using the BASINS  Meteorological Database
(Version 2006). BASINS Technical Note 10.  Office of Water, U.S. Environmental Protection Agency.
Washington, DC. http://water.epa.gov/scitech/datait/models/basins/upload/2009_04_13_BASINSs_tecnotel0.pdf
(Accessed June, 2009).

USEPA (United States Environmental Protection Agency). 2007. Creating Hydraulic  Function Tables
(FTABLES) for Reservoirs in BASINS. BASINS Technical Note 1. Office of Water,  U.S. Environmental
Protection Agency. Washington, DC.
http://water.epa.gov/scitech/datait/models/basins/upload/2009_04_13_BASINSs_tecnotel.pdf (Accessed June,
2009).

USEPA (United States Environmental Protection Agency). 2006. Sediment Parameter and Calibration Guidance
for HSPF.  BASINS Technical Note 8. Office of Water, U.S. Environmental Protection Agency. Washington, DC.
http://water.epa.gov/scitech/datait/models/basins/upload/2006_02_02_BASINS_tecnote8.pdf (Accessed June,
2009).

USEPA (United States Environmental Protection Agency). 2000. Estimating Hydrology and Hydraulic
Parameters for HSPF. BASINS Technical Note 6. EPA-823-ROO-012. Office of Water, U.S. Environmental
Protection Agency. Washington, DC.
http://water.epa.gov/scitech/datait/models/basins/upload/2000_08_14_BASINS_tecnote6.pdf (Accessed June,
2009).

USGS (United States Geological Survey). 2008.  Description of the ACF River Basin  Study Area (July 17, 2008):
http://ga.water.usgs.gov/nawqa/basinall.html, Jan. 2, 2010.
                                               D-96

-------
Appendix E
Model Configuration, Calibration and
Validation

Basin: Arizona: Salt, Verde, and San
Pedro Rivers (Ariz)
                   E-l

-------
Contents
Watershed Background	E-8
  Water Body Characteristics	E-8
  Soil Characteristics	E-13
  Land Use Representation	E-13
  Topography	E-18
  Point Sources	E-20
  Meteorological Data	E-22
  Watershed Segmentation	E-26
  Calibration Data and Locations	E-29
  Other Relevant Features	E-29
HSPF Modeling	E-31
  Changes Made to Base Data Provided	E-31
  Assumptions	E-31
  Hydrology Calibration	E-31
  Hydrology Validation	E-37
  Hydrology Results for Larger Watershed	E-42
  Water Quality Calibration and Validation	E-49
  Water Quality Results for Larger Watershed	E-59
SWAT Modeling	E-61
  Changes Made to Base Data Provided	E-61
  Assumptions	E-61
  Hydrology Calibration	E-62
  Hydrology Validation	E-67
  Hydrology Results for Larger Watershed	E-72
  Water Quality Calibration and Validation	E-79
  Water Quality Results for Larger Watershed	E-89
References	E-91
                                               E-2

-------
Tables
Table 1.     Aggregation of NLCD land cover classes	E-16
Table 2.     Land use distribution for the Arizona basins (2001 NLCD) (mi2)	E-17
Table 3.     Major point source discharges in the Arizona basins	E-20
Table 4.     Precipitation stations for the Arizona models	E-22
Table 5.     Calibration and validation locations in the Arizona basins	E-29
Table 6.     Seasonal summary at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
            (HSPF)	E-35
Table 7.     Summary statistics at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
            (HSPF)	E-37
Table 8.     Seasonal summary at USGS 09504000 Verde River near Clarkdale, AZ - validation period
            (HSPF)	E-40
Table 9.     Summary statistics at USGS 09504000 Verde River near Clarkdale, AZ - validation period
            (HSPF)	E-42
Table 10.    Summary statistics (percent error) for all stations - calibration period WY 1992-2002 (HSPF) ...E-43
Table 11.    Seasonal summary at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
            (HSPF)	E-46
Table 12.    Summary statistics at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
            (HSPF)	E-48
Table 13.    Summary statistics for all stations - validation period WY 1982-1992 (HSPF)	E-49
Table 14.    Model fit statistics (observed minus predicted) for monthly sediment loads using stratified
            regression  at USGS 095 04000 Verde River near Clarkdale, AZ (HSPF)	E-51
Table 15.    Relative errors (observed minus predicted) for TSS concentration at USGS 09504000 Verde
            River near Clarkdale, AZ (HSPF)	E-53
Table 16.    Model fit statistics (observed minus predicted) for monthly total phosphorus loads using
            stratified regression	E-54
Table 17.    Relative errors (observed minus predicted) for total phosphorus concentration at USGS
            09504000 Verde River near Clarkdale, AZ (HSPF)	E-56
Table 18.    Model fit statistics (observed minus predicted) for monthly total nitrogen loads using
            averaging estimator (HSPF)	E-57
Table 19.    Relative errors (observed minus predicted) for total nitrogen concentration at USGS
            09504000 Verde River near Clarkdale, AZ (HSPF)	E-59
Table 20.    Summary statistics (observed minus predicted) for water quality for all stations - calibration
            period 1993-2002 (HSPF)	E-60
Table 21.    Summary statistics (observed minus predicted) for water quality for all stations - validation
            period 1986-1992 (HSPF)	E-60
Table 22.    Seasonal summary at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
            (SWAT)	E-65
Table 23.    Summary statistics at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
            (SWAT)	E-66
Table 24.    Seasonal summary at USGS 09504000 Verde River near Clarkdale, AZ - validation period
            (SWAT)	E-70
Table 25.    Summary statistics at USGS 09504000 Verde River near Clarkdale, AZ - validation period
            (SWAT)	E-72
Table 26.    Summary statistics (percent error) at all stations - calibration period WY 1992-2002 (SWAT) ...E-73
Table 27.    Seasonal summary at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
            (SWAT)	E-76
Table 28.    Summary statistics at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
            (SWAT)	E-78
Table 29.    Summary statistics at all stations - validation period WY 1982-1992 (SWAT)	E-79
Table 30.    Model fit statistics (observed minus predicted) for monthly sediment loads using stratified
            regression  at USGS 09504000 Verde River near Clarkdale, AZ (SWAT)	E-81

                                                E-3

-------
Table 31.    Relative errors (observed minus predicted) for TSS concentration at USGS 09504000 Verde
            River near Clarkdale, AZ (SWAT)	E-83
Table 32.    Model fit statistics (observed minus predicted) for monthly total phosphorus loads using
            tratified regression (SWAT)	E-84
Table 33.    Relative errors (observed minus predicted) for total phosphorus concentration at USGS
            09504000 Verde River near Clarkdale, AZ (SWAT)	E-86
Table 34.    Model fit statistics (observed minus predicted) for monthly total nitrogen loads using
            averaging estimator (SWAT)	E-87
Table 35.    Relative errors (observed minus predicted) for total nitrogen concentration at USGS
            09504000 Verde River near Clarkdale, AZ (SWAT)	E-89
Table 36.    Summary statistics (observed minus predicted) for water quality at all stations - calibration
            period 1993-2002 (SWAT)	E-90
Table 37.    Summary statistics (observed minus predicted) for water quality at all stations - validation
            period 1986-1992 (SWAT)	E-90


Figures
Figure 1.   The Arizona basins -Verde and Salt River sections	E-ll
Figure 2.   The Arizona basins - San Pedro River section	E-12
Figure 3.   Land use in the Arizona basin - Verde and Salt River section	E-14
Figure 4.   Land use in the Arizona basin - San Pedro River section	E-15
Figure 5.   Topography of the Arizona Basins	E-19
Figure 6.   Major point sources in the Arizona basins	E-21
Figure 7.   Weather stations for the Arizona basins model -Verde and Salt River section	E-24
Figure 8.   Weather stations for the Arizona basins model - San Pedro River section	E-25
Figure 9.   Model segmentation USGS stations utilized for the Arizona basins - Verde and Salt River
           section	E-27
Figure 10.  Model segmentation USGS stations utilized for the Arizona basins - San Pedro River section	E-28
Figure 11.  Mean daily flow at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
           (HSPF)	E-32
Figure 12.  Mean monthly flow at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
           (HSPF)	E-33
Figure 13.  Monthly flow regression and temporal variation at USGS 09504000 Verde River near Clarkdale,
           AZ - calibration period (HSPF)	E-33
Figure 14.  Seasonal regression and temporal  aggregate at USGS 09504000 Verde River near Clarkdale,
           AZ - calibration period (HSPF)	E-34
Figure 15.  Seasonal medians and ranges at USGS 09504000 Verde River near Clarkdale, AZ - calibration
           period (HSPF)	E-34
Figure 16.  Flow exceedance at USGS 09504000 Verde River near Clarkdale, AZ - caalibration period
           (HSPF)	E-35
Figure 17.  Flow accumulation at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
           (HSPF)	E-36
Figure 18.  Mean daily flow at USGS 09504000 Verde River near Clarkdale, AZ - validation period
           (HSPF)	E-38
Figure 19.  Mean monthly flow at USGS 09504000 Verde River near Clarkdale, AZ - validation period
           (HSPF)	E-38
Figure 20.  Monthly flow regression and temporal variation at USGS 09504000 Verde River near Clarkdale,
           AZ - validation period (HSPF)	E-39
Figure 21.  Seasonal regression and temporal  aggregate at USGS 09504000 Verde River near Clarkdale,
           AZ - validation period (HSPF)	E-39
                                                E-4

-------
Figure 22.  Seasonal medians and ranges at USGS 09504000 Verde River near Clarkdale, AZ - validation
           period (HSPF)	E-40
Figure 23. Flow exceedance at USGS 09504000 Verde River near Clarkdale, AZ - validation period
           (HSPF)	E-41
Figure 24. Flow accumulation at USGS 09504000 Verde River near Clarkdale, AZ - validation period
           (HSPF)	E-41
Figure 25.  Mean daily flow at USGS 09498500 Salt River near Roosevelt, AZ - calibration period (HSPF)..E-44
Figure 26.  Mean monthly flow at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
           (HSPF)	E-44
Figure 27.  Monthly flow regression and temporal variation at USGS 09498500 Salt River near Roosevelt,
           AZ - calibration period (HSPF)	E-45
Figure 28.  Seasonal regression and temporal aggregate at USGS 09498500 Salt River near Roosevelt, AZ -
           calibration period (HSPF)	E-45
Figure 29.  Seasonal medians and ranges at USGS 09498500 Salt River near Roosevelt, AZ - calibration
           period (HSPF)	E-46
Figure 30.  Flow duration at USGS 09498500 Salt River near Roosevelt, AZ - calibration period (HSPF)	E-47
Figure 31.  Flow accumulation at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
           (HSPF)	E-47
Figure 32.  Fit for monthly load of TSS at USGS 09504000 Verde River near Clarkdale, AZ (HSPF)	E-50
Figure 33.  Power plot for observed and simulated TSS at USGS 09504000 Verde River near Clarkdale, AZ -
           calibration period (HSPF)	E-51
Figure 34.  Power plot for observed and simulated TSS at USGS 09504000 Verde River near Clarkdale, AZ -
           validation period (HSPF)	E-52
Figure 35.  Time series plot of TSS concentration at USGS 09504000 Verde River near Clarkdale, AZ
           (HSPF)	E-53
Figure 36.  Fit for monthly load of total phosphorus at USGS 09504000 Verde River near Clarkdale, AZ
           (HSPF)	E-54
Figure 37.  Power plot for observed and simulated total phosphorus at USGS 09504000 Verde River near
           Clarkdale, AZ - calibration period (HSPF)	E-55
Figure 38.  Power plot for observed and simulated total phosphorus at USGS 09504000 Verde River near
           Clarkdale, AZ - validation period (HSPF)	E-55
Figure 39.  Time series plot of total phosphorus concentration at USGS 09504000 Verde River near
           Clarkdale, AZ (HSPF)	E-56
Figure 40.  Fit for monthly load of total nitrogen at USGS 09504000 Verde River near Clarkdale, AZ
           (HSPF)	E-57
Figure 41.  Power plot for observed and simulated total nitrogen at USGS 09504000 Verde River near
           Clarkdale, AZ - calibration period (HSPF)	E-58
Figure 42.  Power plot for observed and simulated total nitrogen at USGS 09504000 Verde River near
           Clarkdale, AZ - validation period (HSPF)	E-58
Figure 43.  Time series plot of total nitrogen concentration at USGS 09504000 Verde River near Clarkdale,
           AZ(HSPF)	E-59
Figure 44.  Mean daily flow at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
           (SWAT)	E-62
Figure 45.  Mean monthly flow at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
           (SWAT)	E-63
Figure 46.  Monthly flow regression and temporal variation at USGS 09504000 Verde River near Clarkdale,
           AZ - calibration period (SWAT)	E-63
Figure 47.  Seasonal regression and temporal aggregate at USGS 09504000 Verde River near Clarkdale,
           AZ - calibration period (SWAT)	E-64
Figure 48.  Seasonal medians and ranges at Verde River near Clarkdale, AZ - calibration period (SWAT)	E-64
Figure 49.  Flow exceedance at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
           (SWAT)	E-65


                                               E-5

-------
Figure 50.   Flow accumulation at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
           (SWAT)	E-66
Figure 51.  Mean daily flow at USGS 09504000 Verde River near Clarkdale, AZ - validation period
           (SWAT)	E-68
Figure 52. Mean monthly flow at USGS 09504000 Verde River near Clarkdale, AZ - validation period
           (SWAT)	E-68
Figure 53.  Monthly flow regression and temporal variation at USGS 09504000 Verde River near Clarkdale,
           AZ - validation period (SWAT)	E-69
Figure 54.  Seasonal regression and temporal aggregate at USGS 09504000 Verde River near Clarkdale, AZ -
           validation period (SWAT)	E-69
Figure 55.  Seasonal medians and ranges at USGS 09504000 Verde River near Clarkdale, AZ - validation
           period (SWAT)	E-70
Figure 56.  Flow exceedance at USGS 09504000 Verde River near Clarkdale, AZ - validation period
           (SWAT)	E-71
Figure 57.  Flow accumulation at USGS 09504000 Verde River near Clarkdale, AZ - validation period
           (SWAT)	E-71
Figure 58.  Mean daily flow at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
           (SWAT)	E-74
Figure 59.  Mean monthly flow at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
           (SWAT)	E-74
Figure 60.  Monthly flow regression and temporal variation at USGS 09498500 Salt River near Roosevelt,
           AZ - calibration period (SWAT)	E-75
Figure 61.  Seasonal regression and temporal aggregate at USGS 09498500 Salt River near Roosevelt, AZ -
           calibration period (SWAT)	E-75
Figure 62.  Seasonal medians and ranges at USGS 09498500 Salt River near Roosevelt, AZ - calibration
           period (SWAT)	E-76
Figure 63.  Flow exceedence at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
           (SWAT)	E-77
Figure 64.  Flow accumulation at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
           (SWAT)	E-77
Figure 65.  Fit for monthly load of TSS at USGS 09504000 Verde River near Clarkdale, AZ (SWAT)	E-81
Figure 66.  Power plot for observed and simulated TSS at USGS 09504000 Verde River near Clarkdale,
           AZ - calibration period (SWAT)	E-82
Figure 67.  Power plot for observed and simulated TSS at USGS 09504000 Verde River near Clarkdale,
           Z - validation period (SWAT)	E-82
Figure 68.  Time series plot of TSS concentration at USGS 09504000 Verde River near Clarkdale, AZ
           (SWAT)	E-83
Figure 69.  Fit for monthly load of total phosphorus  at USGS 09504000 Verde River near Clarkdale, AZ
           (SWAT)	E-84
Figure 70.  Power Plot for Observed and Simulated total phosphorus at USGS 09504000 Verde River near
           Clarkdale, AZ - calibration period (SWAT)	E-85
Figure 71.  Power plot for observed and simulated total phosphorus at USGS 09504000 Verde River near
           Clarkdale, AZ - validation period (SWAT)	E-85
Figure 72.  Time series plot of total phosphorus concentration at USGS 09504000 Verde River near
           Clarkdale, AZ (SWAT)	E-86
Figure 73.  Fit for monthly load of total nitrogen at USGS 09504000 Verde River near Clarkdale, AZ
           (SWAT)	E-87
Figure 74.  Power plot for observed and simulated total nitrogen at USGS 09504000 Verde River near
           Clarkdale, AZ - calibration period (SWAT)	E-88
Figure 75.  Power plot for observed and simulated total nitrogen at USGS 09504000 Verde River near
           Clarkdale, AZ - validation period (SWAT)	E-88


                                                E-6

-------
Figure 76   Time series plot of total nitrogen concentration at USGS 09504000 Verde River near Clarkdale,
           AZ(SWAT)	E-89
                                                E-7

-------

The Arizona (Salt, Verde, and San Pedro) basins are located in central and southern Arizona in EPA Region 9
(Cordy et al. 2000). The watershed includes large parts of two hydrologic provinces—the Central Highlands in
the north and the Basin and Range Lowlands in the south. Five major river systems drain the area: the Gila, Salt,
Verde, Santa Cruz and San Pedro Rivers. The selected model area includes perennial portions of the Salt and
Verde River basins (in HUC 1506) that lie upstream of major impoundments, along with the San Pedro River
(HUC 1505), for a total of 10 HUCSs with an area of 14,910 mi2 (Figure 1 and Figure 2).

Land cover is primarily desert scrub and rangeland at low elevations with sparse forest at higher elevations
(USGS, 2004; Cordy et al., 2000). The two major population centers of Arizona, Phoenix and Tucson, are  located
just downstream of the model area, while portions of Flagstaff, Prescott, and several smaller towns are within the
Verde River watershed. Population growth is resulting in increasing demands on the limited water resources of
the area. The climate is arid to semiarid and is characterized by variability from place to place as well as large
differences in precipitation from one year to the next. Precipitation can be three times greater in wet years than in
dry years.

The Verde and Salt River watersheds are in the Central Highlands hydrologic province, characterized by
mountainous terrain with shallow, narrow intermountain basins. Forests and rangeland cover most of the area.
The largest town in the  province is Prescott and other small rural towns dot the region. Most of the perennial
streams in the study area are in the Central Highlands. These streams derive their flow from precipitation in the
mountains and from rainfall and snowmelt along the northeastern border of the basins. Many of the major streams
with headwaters in the Central Highlands are perennial in their upper reaches but are captured for water supply for
metropolitan Phoenix, power generation, and flood control before they reach the Basin and Range Lowlands.

The San Pedro watershed is in the Basin and Range Lowlands hydrologic province. The Basin and Range
Lowlands are characterized by ephemeral streams, the largest water demands, and reliance on groundwater. Deep,
broad alluvial basins separated by mountain ranges of small areal extent characterize this hydrologic province.
There is very little natural streamflow because of an average annual rainfall of less than  10 to  15 inches except at
the highest elevations. With the exception of some small, higher elevation streams and sections of the San  Pedro
River, most perennial streams in the Basin and Range Lowlands are dependent on treated wastewater effluent for
their year-round flow. Water use in the Basin and Range Lowlands represents 96 percent of all water use in the
Arizona basins. Agriculture is the largest water user. Because of the general lack of surface water resources in the
Basin and Range Lowlands, groundwater is relied upon heavily to meet agricultural and municipal demands.

The lower portions of the rivers in the Arizona basins have been extensively engineered for water supply purposes
(e.g., the Salt River Project) and also contain many reaches that flow only intermittently. Larger reservoirs are
problematic for scenario simulations as future demands  and reservoir management are not fully known, while
intermittent streams are difficult to calibrate and can present problems for model performance. Therefore, the
portions of the watershed chosen for simulation are upstream of major reservoirs and focus on perennial streams.
The resulting three distinct study areas are the Verde, Salt, and San Pedro rivers (Figure 1 and Figure 2).


Water Body Characteristics

Verde River
The first area of study is the Verde River watershed upstream of Horseshoe Reservoir (Figure  1). The Verde River
watershed comprises approximately 6,577 square miles  (mi2), while the area upstream of the northern end  of
Horseshoe Reservoir contains approximately 5,563 mi2. The watershed trends south-southeast from Fraziers Well,

                                                 E-8

-------
immediately south of the Colorado River watershed and the Grand Canyon National Park to its confluence with
the Salt River on the east side of Phoenix. The study area ranges in elevation from over 11,000 feet (ft) where it
drains a portion of Humphreys Peak north of Flagstaff to around 2,100 ft at the confluence of the Verde River
with Horseshoe Reservoir.

The Verde Valley, which descends into the Central Highland province, is bounded by the Mogollon Rim to the
north and northeast and by the Black Hills to the southwest (Owen-Joyce and Bell 1983). The headwaters of the
Verde River are considered to be just below Sullivan Lake, an impoundment of Big Chino Wash. Upstream of
this point lies a large drainage area that is dominated by intermittent flow (HUC 15060201). Within the Upper
Verde watershed (HUC 15060202) from Sullivan Lake to Camp Verde, the Verde River flows through rugged
country and drains high mountains to the north and east. Perennial flow in the Verde River is usually considered
to start at the confluence with Granite Creek, just below Sullivan Lake. Granite Creek and its two tributaries
originate in the mountainous area outside of Prescott. All three of these tributaries are dammed to provide water to
the city of Prescott and the Chino Valley Irrigation District. Flow  in Granite Creek is ephemeral at the point of
confluence with the Verde River; however, about 25 percent of the baseflow in the Verde River at this point is
believed to derive from groundwater transport out of the Granite Creek drainage (ADWR 2000).

The baseflow in the upper reaches of the Verde River is supported by groundwater discharges between Granite
Creek and Paulden (Owen-Joyce and Bell 1983). From Paulden to Sycamore Creek the river gains additional
groundwater discharges, primarily at Mormon Pocket. Sycamore Creek is an important tributary of the Verde
River, draining the area west of Flagstaff, and has a spring-fed baseflow. The net result of these groundwater
sources is a nearly constant baseflow of around 75 to 80 cfs at Clarkdale.

Groundwater throughout the Big Chino subwatershed occurs under both confined and unconfmed conditions.
Groundwater levels range from above surface due to confined conditions to over 200 feet below surface, with a
depth to water in most wells of less than 80 ft (Schwab 1995). The major source of recharge for the Big Chino
subwatershed is infiltration of runoff from the mountain fronts and flow within the major washes. Only a small
percentage of the annual precipitation in the subwatershed reaches the groundwater table because the majority
occurs in high intensity summer storm events and is lost as surface runoff, evaporation and transpiration by
vegetation (Schwab 1995).

ADWR (2000) examined water budgets for 1996-97 for the Big Chino subwatershed plus the uppermost part of
the Verde River to the USGS gage at Paulden and concluded that there was no net change in the groundwater
storage. Inflows were estimated to be 26,760 acre-feet from natural recharge plus 8,010 acre-feet from incidental
anthropogenic recharge. Of the total discharges, 19,050 acre-feet (55 percent) occurred as flow in the Verde River
near Paulden and the remainder as groundwater pumpage.

An additional important factor in the hydrology of the Verde River watershed, particularly upstream of Paulden,
is the construction of numerous stock pond impoundments used to capture surface runoff to  support cattle
ranching. These impoundments may act as recharge basins, but impede the flow of runoff that would otherwise
have occurred. A survey of small impoundments upstream of Camp Verde was conducted in 1996. Approximately
2,635 impoundments ranging in size from 0.1 acres to approximately 350 acres in surface area were identified
(ADWR 2000). No estimate of recharge has been calculated for these impoundments and no determination of the
impact from restricting and/or impounding the natural runoff has ever been studied.


Salt River
The Salt River (including Tonto Creek) lies immediately to the east of the Verde River watershed and shares
many similar characteristics. The model simulates these streams down to Roosevelt Reservoir (Figure 1). Like the
Verde River watershed, the Salt River and Tonto Creek watersheds have high relief and are bounded by the
Mogollon Rim. However, unlike the Verde watershed, these watersheds have less in the way of teleconnections to
deep groundwater. Perennial springs are important in the upper reaches of the Salt; however, most of the water

                                                 E-9

-------
discharged by these springs appears to derive from local sources. The Salt River watershed also has much less
human influence than the Verde, with only 1.5 percent of the land area in private ownership. The bulk of the
watershed is under tribal or US Forest Service ownership.


San Pedro River
The San Pedro River, a tributary of the Gila River, flows northward from the Arizona-Mexico border (Figure 2).
The watershed consists of a large alluvial valley flanked by mountain ranges. The river is perennial in the
southern (upstream) reaches, but only intermittent in the northern (downstream) reaches. As with the Verde River
and Salt River watersheds, precipitation and temperature vary strongly with elevation, with most of the
precipitation occurring at the higher elevations. The perennial portions of the river support important desert
riparian forest habitat, and most of this section is contained within the San Pedro Riparian National Conservation
Area.
                                                 E-10

-------
    Hydrography

    Water (Nat. Atlas Dataset)

    US Census Populated Places

    Municipalities (pop 2 50,000)

    County Boundaries

  ^ Watershed witfi HUCSs
      Salt and Verde
      River Basins
              Big
       Chino-William
            Valley
          (15060201
                       Upper Verde
                       (15060202)
                                  Lower Verd
                                           Tonto
                                         (15060105)
                                                                         Carrizo
                                                                       (15060104)
                                                     Upper Salt
                                                     (15060103)
  White
(15060102)
sevelt
ervoir
     V
                                                                                       ack
                                                                                  (15060101)
GCRP Model Areas - Salt and Verde River Basins
                Base Map
Figure 1.  The Arizona basins - Verde and Salt River sections.
                                                    E-ll

-------
                                  Lower San Pedro
                                      (15050203)
                                          Upper San Pedro
                                             (15050202)
                                                                                         Hydrography
                                                                                    	 Interstate
                                                                                    ^B Water (Nat. Atlas Dataseti
                                                                                         US Census Populated Places
                                                                                    ^H Municipalities (pop 2 50.000)
                                                                                         County Boundaries
                                                                                    1    I Watershed with HUCBs
 GCRP Model Areas - San Pedro River Basin
                Base Map
             NAD 1983 Albers_metets
20
 Kilometers
Figure 2.  The Arizona basins - San Pedro River section.
                                                       E-12

-------
Soil  Characteristics

The hydrology of the Arizona Basin is strongly influenced by the soils and underlying geology of the watershed.
These in turn reflect the complex geologic history of Arizona, which includes periods of marine inundation,
volcanism, and uplift.

One of the most important characteristics of soils for watershed modeling is their hydrologic soil group (HSG).
The 20 Watershed study utilized STATSGO soil survey HSG information during model set-up. Soils are
classified into four hydrologic groups (SCS 1986), separated by runoff potential, as follows:

       A     Low runoff potential and high infiltration rates even when thoroughly wetted. Chiefly deep, well
              to excessively drained sands or gravels. High rate of water transmission (> 0.75 cm/hr).
       B     Moderate infiltration rates when thoroughly wetted. Chiefly moderately deep to deep, moderately
              well to well drained soils with moderately fine to moderately coarse textures. Moderate rate of
              water transmission (0.40—0.75 cm/hr).
       C     Low infiltration rates when thoroughly wetted. Chiefly soils with a layer that impedes downward
              movement of water, or soils with moderately fine to fine texture. Low rate of water transmission
              (0.15—0.40 cm/hr).
       D     High runoff potential. Very low infiltration rates when thoroughly wetted. Chiefly clay soils with
              a high swelling potential, soils with a permanent high water table, soils with a claypan or clay
              layer at or near the surface, or shallow soils over nearly impervious material. Very low rate of
              water transmission (0—0.15 cm/hr).

The soils in the Verde River watershed are predominantly hydrologic group B soils while soils in the San Pedro
River watershed are predominantly hydrologic group C soils. The Salt River watershed contains almost equal
amounts of B, C, and D soils with a slight dominance of B soils.


Land Use Representation

Land use in the watershed is based on the 2001 National Land Cover Database (NLCD) coverage (Figure 3 and
Figure 4) and is predominantly scrub/shrub chaparral blending into Sonoran paloverde at lower elevations and
pinyon-juniper evergreen forest at higher elevations. Only a few small municipalities are located in the study
watersheds and much of the land is in federal ownership.
                                                E-13

-------
                                                                                                Hydrography
                                                                                                Interstate
                                                                                                Water (Nat. Atlas Dataset)
                                                                                           I    I County Boundaries
                                                                                                Watershed
                                                                                           2001 NLCD Land Use
                                                                                           ^B °Pen water
                                                                                              ^ Developed, open space
                                                                                              | Developed, low intensity
                                                                                              | Developed, medium intensity
                                                                                             | Developed, high intensity
                                                                                              ^ Barren land
                                                                                              | Deciduous forest
                                                                                              | Evergreen forest
                                                                                              ^ Mixed forest
                                                                                           I    | Scrub/shrub
                                                                                              ] Grassland/herbaceous
                                                                                                Pasture/hay
                                                                                             ^| Cultivated crops
                                                                                              ^ Woody wetlands
                                                                                              n Emergent herbaceous wetlands
Verc/e
R/Ver
GCRP Model Areas - Salt and Verde River Basins
                Land Use Map
Figure 3.  Land use  in the Arizona basin - Verde and Salt River section.
                                                            E-14

-------
                                                                                          Legend
                                                                                          	 Hydrography
                                                                                          ^^= Interstate
                                                                                          ^B Water (Nat. Atlas Dataset)
                                                                                            3 County Boundaries
                                                                                          n Watershed
                                                                                          2001 NLCD Land Use
                                                                                          ^^| Open water
                                                                                            ^ Developed, open space
                                                                                            | Developed, low intensity
                                                                                            | Developed, medium intensity
                                                                                          ^H Developed, high intensity
                                                                                            ^ Barren land
                                                                                            | Deciduous forest
                                                                                            | Evergreen forest
                                                                                            ^ Mixed forest
                                                                                          I    | Scrub/shrub
                                                                                          I    | Grassland/herbaceous
                                                                                            ^ Pasture/hay
                                                                                            | Cultivated crops
                                                                                            ^ Woody wetlands
                                                                                            ^ Emergent herbaceous wetlands
                                                                                     San Pedro
                                                                                     River
          MEXICO
  GCRP Model Areas - San Pedro River Basin
                Land Use Map
               NAD_1983_Albers_meters
             Map produced 2-8-2010- B. Tucker
40
• Miles
                                                                                                        TETRATECH
Figure 4. Land  use in the Arizona basin - San Pedro River section.
                                                           E-15

-------
NLCD land cover classes were aggregated according to the scheme shown in Table 1 then overlain with the soils
HSG grid. Minor land uses with less than 5 percent coverage within a subwatershed were reassigned to more
dominant classes. Pervious and impervious lands are specified separately for HSPF, so only one developed
pervious class is used, along with an impervious class. HSPF simulates impervious land areas separately from
pervious land.  Impervious area distributions were  also determined from the NLCD Urban Impervious data
coverage.  Specifically, percent impervious area was calculated over the whole basin for each of the four
developed land use classes. These percentages were then used to separate out impervious land. NLCD
impervious area data products are known to underestimate total imperviousness in rural areas. However, the
model properly requires connected impervious area, not total impervious area, and the NLCD tabulation is
assumed to provide a reasonable approximation of connected impervious area. Different developed land classes
are specified separately in SWAT. In HSPF the WATER, BARREN, DEVPERV, and WETLAND classes are not
subdivided by HSG;  SWAT uses the built-in HRU overlay mechanism in the ArcSWAT interface.
Table 1.  Aggregation of NLCD land cover classes
NLCD Class
1 1 Water
12 Perennial ice/snow
21 Developed open space
22 Dev. Low Intensity
23 Dev. Med. Intensity
24 Dev. High Intensity
31 Barren Land
41 Forest
42 Forest
43 Forest
51-52 Shrubland
71-74 Herbaceous Upland
81 Pasture/Hay
82 Cultivated
91 -97 Wetland
98-99 Wetland
Comments
Water surface area
usually accounted for as
reach area






Deciduous
Evergreen
Mixed




Emergent & woody
wetlands
Aquatic bed wetlands (not
emergent)
SWAT class
WATR
WATR
URLD
URMD
URHD
UIDU
SWRN
FRSD
FRSE
FRST
RNGB
RNGE
HAY or GRASS
AGRR
WETF, WETL,
WETN
WATR
HSPF (after processing)
WATER
BARREN, Assume HSG D
DEVPERV;
IMPERV
BARREN (D)
FOREST (A,B,C,D)
SHRUB (A,B,C,D)
GRASS (A,B,C,D), BARREN (D)
GRASS (A,B,C,D)
AGRI (A,B,C,D)
WETLAND, Assume HSG D
WATER
The distribution of land use in the watershed is summarized in Table 2. Note that the small areas in crop and hay
production along the Verde mainstem and elsewhere do not meet the 5 percent threshold requirement in SWAT
and are thus not explicitly included in the model; instead, the developed pervious land use implicitly includes
those areas in crop production for SWAT.
                                                E-16

-------
Table 2. Land use distribution for the Arizona basins (2001 NLCD) (mi2)
HUC8
watershed
Upper San
Pedro
15050202
Lower San
Pedro
15050203
Black
15060101
White
15060102
Upper Salt
15060103
Carrizo
15060104
Tonto
15060105
Big Chino-
Williamson
Valley
15060201
Upper
Verde
15060202
Lower
Verde
15060203
Total
Open
Water
0.1
1.4
1.8
1.6
16.7
0.0
0.0
0.1
1.2
0.3
23.2
Developed9
Open
space
19.2
8.3
1.8
4.5
10.4
2.7
4.6
11.8
48.0
9.8
121.2
Low
Density
4.9
2.2
0.2
1.2
5.1
0.2
1.2
3.5
25.1
4.0
47.7
Medium
Density
1.2
0.3
0.0
0.2
1.4
0.0
0.1
0.2
5.3
0.4
9.0
High
Density
0.2
0.1
0.0
0.0
0.3
0.0
0.0
0.0
0.6
0.1
1.3
Barren
Land
0.4
6.6
0.6
0.5
14.0
1.4
0.6
3.8
15.3
1.0
44.3
Forest
64.2
179.1
1,026.3
537.2
950.1
586.4
443.7
615.8
1,256.8
582.0
6,241.7
Shru bland
1,129.2
1,791.7
217.9
89.9
1,251.9
118.5
494.9
1,514.8
1,143.2
601.3
8,353.2
Pasture/Hay
1.5
1.5
0.0
0.1
0.7
0.0
0.3
1.4
2.5
1.9
10.0
Cultivated
9.9
7.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
17.4
Wetland
4.5
12.5
2.0
3.2
4.7
0.4
1.9
0.9
6.1
4.6
40.7
Total
1,235.2
2,011.2
1,250.7
638.4
2,255.2
709.8
947.1
2,152.4
2,504.1
1,205.4
14,909.6
 The percent imperviousness applied to each of the developed land uses is as follows: open space (7.37%), low density (29.66%), medium density (53.71%), and high
density (73.85%).
                                                                 E-17

-------
The HSPF model is set up on a hydrologic response unit (HRU) basis. For HSPF, HRUs were formed from an
intersection of land use and hydrologic soil group, then further subdivided by precipitation gage and slope. SWAT
HRUs were formed from an intersection of land use and SSURGO major soils.


Topography

The Salt, Verde, and San Pedro River watersheds are characterized by high relief (Figure 5) and precipitation and
temperature vary greatly with elevation. The largest precipitation amounts and lowest temperatures occur at the
high elevations along the Mogollon Rim on the north and east sides of the Salt and Verde River watersheds.
                                                E-18

-------
 GCRP Model Areas - Verde and Salt Rivers
        Elevation Map
                             o    25  so
 100
• Kilometers
                             0  12.5  25
                                         50
                                        • Miles
                                                             TETRATECH
GCRP Model Areas - San Pedro Basin
    Elevation Map
                                          Legend
                                                Interstate
                                          ^H Water (Nat. Atlas Dataset)
                                             | Municipalities (pop > 50,000)
                                                County Boundaries
                                          I     | Watershed
                                          Elevation
                                          Meters
                                                High : 3848
     Low: 585
 Figure 5. Topography of the Arizona Basins.
                                        E-19

-------
Only the two major dischargers with a design flow greater than 1 MOD are included in the simulation (Table 3
and Figure 6). These dischargers are Page Springs Fish Hatchery and the Final Creek Wastewater Treatment Plant
(WWTP) in Globe, Arizona. Because of the arid climate and low population in the study watersheds, much of the
wastewater that is generated is either used for irrigating golf courses or discharged to ephemeral washes that lack
a direct surface connection to the river system.

Table 3.  Major point source discharges  in the Arizona basins
NPDES ID
AZ0021245
AZ0020249
Name
Page Springs Fish Hatchery (AZ Game and
Fish Department)
Final Creek WWTP (City of Globe, AZ)
Design Flow (MGD)
20.35
1.20
Observed Flow (MGD)
(1991-2006 average)
21.92
12.54
The discharges from Page Springs Fish Hatchery to Oak Creek are largely composed of natural groundwater.
Some of this groundwater arises within the local subwatershed, and is thus already accounted for in the model. To
prevent double-counting of this water, the reported discharges were reduced significantly to provide an
approximate match to observed base flows in Oak Creek.

Several other smaller discharges reported in the study area were determined to be used primarily for irrigation or
discharge to dry washes, do not cause live stream discharges and so are not explicitly included in the model. The
San Jose WWTP major discharge at Bisbee, Arizona is in part used for irrigation, but also discharges to
Greenbush Draw, tributary to the San Pedro. However, it enters the San Pedro upstream of the modeled area.
                                                E-20

-------
                                                                           Legend
                                                                                Point Sources
                                                                                Hydrography
                                                                                Interstate
                                                                                Water (Nat. Atlas Dataset)
                                                                                US Census Populated Places
                                                                                Municipalities (pop > 50,000)
                                                                              ] County Boundaries
                                                                                Watershed with HUCSs
                               Verde
                               River
    PAGE SPRINGS /    //
    FISH HATCHERY
                            Rgpseyelt
                            Reservoir
                                                          PINAL CR. WWTP
 GCRP Model Areas - Salt and Verde River Basins
            Major Point Sources
Figure 6.  Major point sources in the Arizona basins.
                                                 E-21

-------
Meteorological Data
The required meteorological data series for the 20 Watershed study are precipitation, air temperature, and
potential evapotranspiration. The 20 Watershed model does not include water temperature or algal simulation and
uses a degree-day method for snowmelt. These meteorological data are drawn from the BASINS4 Meteorological
Database (USEPA 2008), which provides a consistent, quality-assured set of nationwide data with gaps filled and
records disaggregated. Scenario application will require simulation over 30 years, so the available stations are
those  with a common 30-year period of record (or one that can be filled from an approximately co-located station)
that covers the year 2001. A total of 29 precipitation stations were identified for use in the Arizona basins model
with a common period of record of 10/1/1972-9/30/2002 (Table 4 and Figure 7 and Figure 8). Temperature
records are sparser; where these are absent, temperature is taken from nearby stations with an elevation correction.
For each weather station, Penman-Monteith reference evapotranspiration was calculated for use in HSPF using
observed precipitation and temperature coupled with SWAT weather generator estimates of solar radiation, wind
movement, cloud cover, and relative humidity.

For the 20 Watershed model applications, SWAT uses daily meteorological data, while HSPF requires hourly
data. It is important to note that a majority of the meteorological stations available for the Arizona basins are
Cooperative Summary of the Day stations that do not report sub-daily data. The BASINS4 dataset already has
versions of the daily data that have been disaggregated to an hourly time step using template stations. For each
daily station, this disaggregation was undertaken in reference to a single disaggregation template. Occasionally,
this automated procedure provides undesirable results, particularly when the total rainfall for the day is very
different between the subject station and the disaggregation template.
Table 4.  Precipitation stations for the Arizona models
COOP ID
AZ020159
AZ020487
AZ020670
AZ020683
AZ020808
AZ021231
AZ021330
AZ021614
AZ021654
AZ021870
AZ022140
AZ023010
AZ023828
AZ024453
AZ025512
AZ026323
AZ026601
AZ026653
AZ026796
AZ026840
Name
Alpine
Ash Fork 3
Beaver Creek
Benson 6 SE
Black River Pumps
Canelo 1 NW
Cascabel
Childs
Chino Valley
Cochise 4 SSE
Coronado NM Hdqtrs
Flagstaff AP
Happy Jack RS
Jerome
Miami
Payson
Pinetop 2E
Pleasant Valley RS
Prescott
Punkin Center
Latitude
33.8493
35.199
34.6418
31.8803
33.4783
31.559
32.3208
34.3495
34.757
32.059
31.3457
35.1442
34.7433
34.7523
33.4045
34.2315
34.1243
34.099
34.5706
33.8557
Longitude
-109.146
-112.488
-111.783
-110.24
-109.751
-110.529
-110.413
-111.698
-112.456
-109.89
-110.254
-111.666
-111.413
-112.111
-110.87
-111.339
-109.921
-110.944
-112.432
-111.306
Temperature
X

X
X
X

X
X
X
X
X
X
X
X
X
X

X
X
X
Elevation
(ft)
8049
5074
3523
1125
6065
1527
959
2650
4749
1274
1598
7003
7478
4950
3559
4907
7200
5048
5202
2326
                                                 E-22

-------
COOP ID
AZ027281
AZ027530
AZ027708
AZ027716
AZ028619
AZ028650
AZ029158
AZ029271
AZ029359
Name
Roosevelt 1 WNW
San Manuel
Sedona
Seligman
Tombstone
Tonto Creek Fish Hatchery 2
Walnut Creek
Whiteriver 1 SW
Williams
Latitude
33.6731
32.6014
34.8957
35.3323
31.7057
34.3839
34.9282
33.8169
35.2407
Longitude
-111.15
-110.633
-111.764
-112.879
-110.056
-111.097
-112.809
-109.983
-112.19
Temperature
X
X
X
X
X

X
X
X
Elevation
(ft)
2204
1055
4218
5248
1405
6389
5087
5120
6747
Orographic effects on precipitation and temperature are important throughout the region. This is addressed
through use of the elevation bands option and the imposition of precipitation and temperature lapse rates in the
SWAT  model. All SWAT model subwatersheds are assigned at least one elevation band, and multiple elevation
bands are used when the interquartile range of elevations within a subwatershed exceeds 375 m. For HSPF,
whenever the precipitation station was located outside or near the edge of a model segment, a multiplier was
applied to the data based on the ratio of the estimated median annual rainfall from isohyetal information and the
long term annual average for the station. The evaporation data appeared to be estimates of pan evaporation, and
ranged from 70 to 100 inches per year. They were therefore adjusted by a factor of 0.7 to reduce them to potential
evapotranspiration. Some of the multipliers were adjusted slightly during the hydrology calibration.
                                                 E-23

-------
                                                                                A  Weather Stations
                                                                                 — Hydrography
                                                                                   Interstate
                                                                                   Water (Nat. Atlas Dataset)
                                                                                   US Census Populated Places
                                                                                   Municipalities (pop > 50,000)
                                                                                 ^] County Boundaries
                                                                                   Watershed with HUCSs
               Verde
               River
                          AZ020487
                             AZ029359
 AZ029-158
       ^
AZ021654

 AZ026796

 AZ024453
AZ020670
     7
AZ028650
         AZ026653
                    /   AZ026323
 GCRP Model Areas - Salt and Verde River Basin
             V\feather Stations
              NADJ 983_Albere_metere
           Map produced 02-11-2010- B. Tucker
Figure 7. Weather stations for the Arizona basins model - Verde and Salt River section.
                                                   E-24

-------
                                                                                      A  Weather Stations
                                                                                       — Hydrography
                                                                                         Interstate
                                                                                         Water (Nat. Atlas Dataset)
                                                                                         US Census Populated Places
                                                                                         Municipalities (pop > 50,000)
                                                                                       ^] County Boundaries
                                                                                         Watershed with HUCSs
                                                                            San Pedro
                                                                            River
  GCRP Model Areas - San Pedro River Basin
              Weather Stations
               NADJ 983_Albere_metere
            Map produced 02-11-2010- B. Tucker
Figure 8. Weather stations for the Arizona basins model - San Pedro River section.
                                                      E-25

-------
Watershed Segmentation
The Arizona basins were divided into 81 subwatersheds for the purposes of modeling (Figure 9 and Figure 10) -
30 in the Verde, 28 in the Salt, and 23 in the San Pedro river models. Initial calibration was conducted on the
Verde River at Clarkdale. However, the parameters derived at this station were not fully transferable to other
portions of the watershed, and additional calibration was conducted at multiple gage locations.

The Verde and Salt River models encompass entire watersheds upstream of major reservoirs - thus upstream
boundary conditions are not required. However, for the Verde River watershed, boundary conditions are needed to
account for the large influx of deep groundwater (much of it ultimately derived from infiltration many miles away
in the Chino watershed) that enters the river in the reach near Paulden, Arizona.

The San Pedro River watershed extends into Mexico; however, the geospatial and meteorological data used to
build the 20 Watershed models do not cover Mexico. Therefore, the San Pedro is simulated with an upstream
boundary condition at the USGS gage on the San Pedro River at Charleston, Arizona (09471000). This is the most
upstream gage with near complete records for the simulation period; the gage  at Palominas (09470500), although
closer to the Mexican border, has long periods of missing records.

Major reservoirs are generally avoided in the model setup; however, it is also  necessary to account for storage in
smaller reservoirs and stock ponds. For SWAT, these are  specified using the Ponds option, based on information
in ADWR (2009) and, for the Verde watershed, Tetra Tech (2001). Significant pond storage is considered in
subwatersheds 11, 13, 15, 16, 17, 18, 19, 23, 25, 26, and 27 for the Verde River watershed. In the Salt River
watershed there are reservoirs with nominal storage capacity near 25,000 acre feet in subwatersheds 20 and 21,
although the normal capacity is only a fraction of this total. As these are headwater subwatersheds, these
reservoirs are also treated as ponds. No reservoirs or ponds are simulated in the San Pedro watershed.

It should be noted that Sullivan Lake, at the head of the perennial portion of the Verde River watershed
(subwatershed 14), intercepts flows out of the Chino watershed and has a significant impact on the progression of
flood waves downstream. This lake is not directly represented in the model due to lack of information on storage
characteristics.
                                                 E-26

-------
                                                                                Legend
                                                                                    USGS Gages
                                                                                    Hydrography
                                                                                    Interstate
                                                                                    Water (Nat. Atlas Dataset)
                                                                                    US Census Populated Places
                                                                                    Municipalities (pop > 50,000)
                                                                                  ^| County Boundaries
                                                                                    Model Subbasins
                                                                                  ] Initial Calibration Watershed
                                                                                    Minnesota River Basin
                                    USGS 09503700
                                                    USGS 09504420
                                                 USGS 09504500
                                                     USGS 09505800
       USGS 09504000 '
USGS 09507980
         USGS 09506000
                 USGS 09508500
                             USGS 09498500
   GCRP Model Areas - Salt and Verde River Basins
               Model Segmentation
                 NAD_1983_Albers_meters
              Map produced 2-11-2010- B. Tucker
    Figure 9.  Model segmentation USGS stations utilized for the Arizona basins - Verde and Salt River
               section.
Note: SWAT subwatersheds numbering is shown; the HSPF model for this watershed uses the same subwatershed boundaries with an
alternative internal numbering scheme.
                                                       E-27

-------
                                                                                  Legend
                                                                                       USGS Gages
                                                                                       Hydrography
                                                                                       Interstate
                                                                                       Water (Nat. Atlas Dataset)
                                                                                       US Census Populated Places
                                                                                       Municipalities (pop > 50,000)
USGS 09473000
                                                                                     ^| County Boundaries
                                                                                       Model Subbasins
                                                                                     ] Initial Calibration Watershed
                                                                                       Minnesota River Basin
                                                              USGS 09471000
      GCRP Model Areas - San Pedro River Basin
               Model Segmentation
                 NADJ 983_Albers_meters
               Map produced 2-11-2010 - B. Tucker
    Figure 10.  Model segmentation USGS stations utilized for the Arizona basins - San Pedro River
               section.
Note: SWAT subwatersheds numbering is shown; the HSPF model for this watershed uses the same subwatershed boundaries with an
alternative internal numbering scheme.
                                                         E-28

-------
Calibration Data and Locations
The site selected for initial calibration was the Verde River near Clarkdale, AZ (USGS gage 09504000); however,
calibration and validation were pursued at multiple locations (Table 5, Figure 9 and Figure 10).

Table 5.  Calibration and validation locations in the Arizona basins
Station name
San Pedro River near Redington, AZ
Aravaipa Creek near Mammoth, AZ
Salt River near Roosevelt, AZ
Verde River near Paulden, AZ
Verde River near Clarkdale, AZ
Oak Creek near Cornville, AZ
West Clear Creek near Camp Verde, AZ
Verde River near Camp Verde, AZ
East Verde River near Childs, AZ
Verde River below Tangle Creek
USGS ID
09472000
09473000
09498500
09503700
09504000
09504500
09505800
09506000
09507980
09508500
Drainage area
(mi2)
2,927
537
4,306
2,507
3,503
355
241
5,009
331
5,858
Hydrology
calibration
X
X
X
X
X
X
X
X
X
X
Water quality
calibration


X

X


X

X
The model hydrology calibration period was set to Water Years 1993-2002, with some variation according to
gage variability. The end date was constrained by the common period of the set of 20 Watershed meteorological
stations available for the watershed, and a 10 year calibration period was desired. Calibration was done on the
later data, because of concerns that there may have been changes in land use and management over time.
Hydrologic validation was then performed on Water Years 1983-1992. Water quality calibration used calendar
years 1993-2002, while validation used 1986-1992, as limited data were available prior to 1986.


Other Relevant Features

Along the mainstem of the Verde River between Clarkdale and Camp Verde and on several tributaries there are
substantial water diversions to support riparian agricultural production, primarily hay. ADWR (2000) identifies 24
diversion structures on the Verde River proper from near Clarkdale to below Camp Verde, 32 diversions along
Oak  Creek, and 12 diversions along Wet Beaver Creek, as well as several in other locations, and estimates that the
total agricultural diversion amount between Perkinsville and Horseshoe Reservoir (most of it occurring in the
Verde Valley, Oak Creek, and Wet Beaver Creek) amounts to 31,668 acre-feet per year. Very little of the water
diverted for irrigation returns as surface flow (Owen-Joyce and Bell 1983); however, a substantial portion may
return as subsurface flow. The water applied from these diversions is represented as irrigation applications in the
model. During development of the previous SWAT model for the Verde it was found that a direct linkage of
irrigation applications to river withdrawals did not provide satisfactory results and indeed tended to cause model
instability. Therefore, the withdrawals and irrigation are uncoupled in the model: irrigation is represented as
nominally occurring from an external source, while withdrawals from the river are specified separately as a
consumptive use that occurs during the April-September growing season. Consumptive use withdrawals are
applied to Verde model subwatersheds 6, 8, 9, 10, and 30. The status of agricultural diversions in the Salt and San
Pedro watersheds is not fully known; however, growing season diversions from the river are assigned to improve
flow closure, being assigned to subwatershed 8 in the San Pedro watershed and subwatershed 5 in the Salt River
watershed.
                                                E-29

-------
A separate representation is used for the Prescott Valley area of the Verde River watershed model (subwatershed
16). Here, agriculture is supported by water stored in two small reservoirs (Granite and Willow). For this
subwatershed, irrigation is represented as linked to and derived from water stored in these reservoirs (represented
as ponds in the SWAT model).

Special notes are required regarding the East Verde River (Verde model subwatershed 4). The town of Payson,
Arizona obtains its municipal supply from groundwater, which is pumped from the alluvium of the East Verde.
The groundwater supply appears to be directly connected to surface water, and causes the East Verde to go dry at
times. However, there is also a source of imported water in the East Verde, as water is brought from across the
Mogollon Rim divide and discharged into the East Verde to augment Payson supplies. Detailed documentation
was not obtained. For the purposes of the 20 Watershed model it is assumed that the imported water is essentially
all consumed by Payson. Therefore, the stream is  simulated as a losing reach, but the imported water is not
explicitly simulated. Payson's wastewater discharges leave the watershed, and are primarily used for golf course
irrigation in the Tonto Creek watershed (subwatershed 26 in the Salt River model).
                                                 E-30

-------
 a a a a a a a a  a a a a a a  a a a a a a a a a a a a a  a a a a a a a a a a a a a  a a a a a a  a a a a a a a a a a a a a  a a a a a a  a a a a a a a a a a a a a  a a a a a a  a a si
No changes were made to the meteorological or land use base data. Similar to the SWAT modeling, the Globe,
Arizona point source total nitrogen and total phosphorus concentrations were reduced, since the recommended
concentrations did not permit a reasonable calibration of total nitrogen and total phosphorus at the Roosevelt
station.
An important feature of the Arizona basins is the complex interaction of surface and groundwater. As noted
earlier, neither SWAT nor HSPF is capable of providing a detailed, process-based simulation of groundwater
flow. It is therefore assumed that interactions with groundwater can be handled with the following simplifying
assumptions:

    •   The local (within subwatershed) accumulation and discharge of shallow groundwater is adequately
       addressed by HSPF's active groundwater formulation.
    •   Discharges to streams from deep groundwater are represented as constant point sources, with discharge
       rates set based on flow information on major springs identified in the Arizona Water Atlas (ADWR
       2009). This means that the model cannot account for seasonal variability in deep groundwater discharge,
       nor can it evaluate how such discharges may evolve in response to climate change.
    •   Losses to groundwater from stream reaches in alluvial basins are simulated based on channel
       conductivity. This makes such losses a function of flow and depth in the affected reaches. The HSPF
       model formulation, which is incorporated in the stream reach FTABLEs as a volume-based loss term does
       not take into account changes in local groundwater head; instead, the loss occurs continuously.

In the Verde River watershed, irrigation withdrawals and applications were modeled  (similarly to the SWAT
model) in the Verde Valley between Clarkdale and Camp Verde based on information provided in ADWR (2009).
A total of 31,668 ac-ft/yr is withdrawn during April-September from selected reaches based on the relative
amounts of grass and developed area in the reach watersheds. The HSPF irrigation module was used to apply this
water to the grass and developed PERLND's using a constant application of 0.11 inches/day during the April-
September period.
The starting parameters for the Arizona HSPF model were developed from an HSPF model of the San Francisco
Bay area watersheds, particularly watersheds in eastern Alameda County. After the starting parameters were
inserted into the model input files, average annual potential evapotranspiration values were computed and
compared to published values. Through this process it was determined the input potential evapotranspiration time
series should be reduced by multipliers, since the computation of these time series produced more PET on an
average annual basis than the published values indicate. The default multipliers used for PET were 0.70; however,
some of the multipliers were adjusted slightly during the hydrology calibration. Calibration adjustments focused
on the following parameters:

   •    LZSN (lower zone nominal storage): LZSN was generally reduced from the initial values to shift flows to
        the wet period and reduce them in the summer.  It was also used to increase total runoff.

                                                E-31

-------
    •   INFILT (index to mean soil infiltration rate): Infiltration was generally decreased from the high initial
       values to increase storm peaks, reduce low flows, and increase surface runoff.
    •   DEEPFR (fraction of groundwater inflow that will enter deep groundwater): small values of DEEPFR
       were used to attempt to reduce low flows and to reduce total flow volume. In the Salt River, the initial
       low values were not adjusted. In the Verde River at Paulden, DEEPFR was increased to a high value to
       represent the recharge losses in the Chino Basin; some of this groundwater returns to the river below the
       Paulden gage.
    •   BASETP (ET by riparian vegetation): Generally BASETP was increased over the initial values in order to
       provide some ET by riparian vegetation and improved the  simulation of low flows.
    •   LZETP (lower zone E-T parameter): LZETP was generally increased to reduce flow, particularly the low
       flows, and to reduce total volumes.
    •   AGWRC (Groundwater recession rate): AGWRC was typically reduced from the initial values to help
       reproduce the brief, sudden storms that are experienced in  the Arizona basin.

Obtaining a high quality fit to hydrology in the  Arizona basin is difficult with HSPF due to the importance of
groundwater, which is  simplistically represented in the model. As in the SWAT model, the specification of
groundwater discharges as constant values and the simulation of reach losses by channel conductivity without
feedback from local groundwater elevations both introduce uncertainty.

Initial calibrations were performed for the two Verde River gages at Paulden and Clarkdale. The calibration
period was  set to the 10 water years from 10/01/1992 to 09/30/2002. The results at Clarkdale are summarized in
Figures 11 through 17 and Tables 6 and 7.  The  fit at Clarkdale is fairly good, although the summer storm volumes
are over-simulated. Predictions at Clarkdale are largely determined by model fit upstream at Paulden, where flows
about 95 percent of the time consist of approximately constant base flow. Spring peaks occasionally push through
from the Chino subwatershed.  Accuracy in simulating these peaks  is primarily affected by lack of an accurate
representation of the hydraulic behavior of Sullivan Lake, and somewhat caused by the necessity of specifying
constant values for channel conductivity to account for transmission losses, when in fact these loss rates are likely
much reduced during the spring wet period. Parameter modifications to improve the peak spring flows out of the
Chino subwatershed result in significant over-prediction of summer storm events.
          Avg Monthly Rainfall (in)
         -Avg Observed Flow (10/1/1992 to 9/30/2002 )
         •Avg Modeled Flow (Same Period)
      35000


                    Apr-94
Oct-95
Apr-97

     Date
Oct-98
Apr-00
Oct-01
                                                                                       6

                                                                                       8

                                                                                       10

                                                                                       12

                                                                                       14

    Figure 11. Mean daily flow at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
             (HSPF).
                                                 E-32

-------
•6
    6000
    4000
    2000
                                                           Avg Monthly Rainfall (in)
                                                          -Avg Observed Flow (10/1/1992 to 9/30/2002 )
                                                          •Avg Modeled Flow (Same Period)
        0-92
A-94
0-01
  Figure 12. Mean monthly flow at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
            (HSPF).
_ 6000
I
I
LJ- 4000
   2000
O)
TO
         •   Avg Flow (10/1 /1992 to 9/30/2002 )
        • - - - • Line of Equal Value
        	Best-Fit Line
                 2000       4000
            Average Observed Flow (cfs)
                                        6000
                                              O
                                              £=
                                              _
                                              ro
                                              m
                              100%
                              90%
                              80%
                              70% -
                              60%
                              50%
                              40% -
                              30%
                              20%
                              10% -
                               0%
                                     Avg Observed Flow (10/1/1992 to 9/30/2002 )
                                     Avg Modeled Flow (10/1/1992 to 9/30/2002 )
                                    -Line of Equal Value
                                                                              .JL.
    I  Ij,
                                 O-92   A-94   O-95
                                                    A-97   O-98
                                                     Month
                                                                A-00   O-01
  Figure 13. Monthly flow regression and temporal variation at USGS 09504000 Verde River near
            Clarkdale, AZ - calibration period (HSPF).
                                                  E-33

-------
           Avg Flow (10/1/1992 to 9/30/2002)
          • Line of Equal Value
           Best-Fit Line
  800

¥
I600
T3
0)
^400
o
0)
  200
y=1.|)498x-3.8J436
     ° = 0.9747
                                            Avg Monthly Rainfall (in)
                                           -Avg Observed Flow (10/1/1992 to 9/30/2002)
                                            Avg Modeled Flow (Same Period)
                                                 800
                                      600 4
                                      400 ]
   200      400      600
  Average Observed Flow (cfs)
                                       800
                                      200 -4—               V
                                                     10 11 12  1  2  3  4  5  6  7
                                                                    Month
   Figure 14.  Seasonal regression and temporal aggregate at USGS 09504000 Verde River near
             Clarkdale, AZ - calibration period (HSPF).
        • Observed (25th, 75th)
        -Median Observed Flow (10/1/1992 to 9/30/2002)
                                                  Average Monthly Rainfall (in)
                                                  Modeled (Median, 25th, 75th)
o
    600
    400
    200
           10      11      12      1
   Figure 15. Seasonal medians and ranges at USGS 09504000 Verde River near Clarkdale, AZ -
             calibration period (HSPF).
                                                  E-34

-------
Table 6.  Seasonal summary at USGS 09504000 Verde River near Clarkdale, AZ - calibration
         period (HSPF)
MONTH
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
87.85
84.09
101.70
370.90
529.79
290.55
129.11
79.98
76.29
78.20
83.85
100.71
80.00
84.00
83.50
87.00
87.00
92.00
83.00
79.00
75.00
77.00
78.00
81.00
77.00
78.00
80.00
81.00
80.00
80.00
77.00
72.00
69.00
71.00
75.00
73.00
84.00
89.00
90.00
93.00
101.50
232.00
93.00
87.00
81.00
82.00
84.00
85.00
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
96.91
99.66
108.76
324.23
600.69
282.20
126.31
87.10
70.29
77.73
88.43
104.90
82.51
85.28
81.17
90.51
101.20
120.53
100.56
73.67
68.05
69.49
79.82
83.32
71.73
72.52
75.07
74.56
78.14
74.05
72.53
62.50
59.97
64.23
69.25
70.17
99.47
120.36
108.70
122.50
136.39
227.47
157.49
108.75
76.39
78.65
97.29
112.11
             •Observed Flow Duration (10/1/1992 to 9/30/2002 )

             Modeled Flow Duration (10/1/1992 to 9/30/2002 )
  o
      100000
      10000
       1000
  'ro
  Q
        100  =    =3
                  10%
                         20%     30%    40%     50%    60%     70%    80%


                              Percent of Time that Flow is Equaled or Exceeded
                                                                              90%    100%
    Figure 16. Flow exceedance at USGS 09504000 Verde River near Clarkdale, AZ - caalibration period

             (HSPF).
                                               E-35

-------
              •Observed Flow Volume (10/1/1992 to 9/30/2002 )
               Modeled FlowVolume (10/1/1992 to 9/30/2002 )
     120%
o
o

-------
Table 7. Summary statistics at USGS 09504000 Verde River near Clarkdale, AZ - calibration
         period (HSPF)
HSPF Simulated Flow
REACH OUTFLOW FROM DSN 101
10-Year Analysis F^riod: 10/1/1992 - 9/30/2002
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

_Jotak3fj3irrujlatejdjT^^
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9)
Simulated Fall Flow Volume (months 10-12^
Simulated Winter Flow Volume (months 1-3^:
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
^EjTOnnJ^yTlih^SLIlP^J-^
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Sjpring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of Efficiency, E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE ~
0.66

0.35
0.14

0.09
0.10
0.38
0.09

0.25
0.02
Error Statistics
2.43
-7.64
-1.58
3.15
Observed Flow Gage
USGS 09504000 VERDE RIVER NEAR CLARKDALE, AZ
Hydrologic Unit Code: 15060202
Latitude: 34.8522416
Longitude: -112.065994
Drainage Area (sq-rri): 3503
Total Observed In-stream Flow:

Total of Observed higjiesMO%Jows:_^
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow VolumeJ10-12):
Observed Winter Flow VolumeJ1-3):
Observed Spring Flow Volume (4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
11.52 »| 30
0.83
-0.50
-15.79
44.47
0.481
0.485
0.803
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0





0.64

0.36
0.15

0.09
0.09
0.38
0.09

0.29
0.01





Clear [













Hydrology Validation

Like the SWAT modeling, validation for the Verde River near Clarkdale was performed for the period 10/1/1982
through 9/30/1992. Results are presented in Figures 18 through 24 and Tables 8 and 9. The HSPF validation
results are fair, but are generally worse than during the calibration period. In particular, the storm peak volumes
are under-predicted, likely due to the effort to reduce summer storm peaks in the calibration.
                                               E-37

-------
         Avg Monthly Rainfall (in)
        -Avg Observed Flow (10/1/1982 to 9/30/1992 )
        •Avg Modeled Flow (Same Period)
i
I
onnn
vnnn
finnn

A nnn

onnn
1 nnn

1 1



1


i
yi








I,



























11111 	







1
If'i'l





1
,1
1 1








Jn
















ll
njFTt
, pi.








• " |i[r -
























1
1








t








j

2
4
fi

8
10
12

                                                                         c
                                                                         'ro
       Oct-82       Apr-84       Oct-85      Apr-87      Oct-88       Apr-90       Oct-91
                                               Date

   Figure 18. Mean daily flow at USGS 09504000 Verde River near Clarkdale, AZ - validation period
            (HSPF).
    1500
i
 I
    1000 -
     500 - -
        O-82
A-84
                                                          Avg Monthly Rainfall (in)
                                                         -Avg Observed Flow (10/1/1982 to 9/30/1992
                                                         •Avg Modeled Flow (Same Period)
O-85
A-87
   Month
O-88
A-90
O-91
                                                           )

                                                           10
                                                           9
                                                           8
                                                           7
                                                           6
                                                           5
                                                           4
                                                           3
                                                           2
                                                           1
                                                           0
                                                                          'ro
                                                                          a:
   Figure 19. Mean monthly flow at USGS 09504000 Verde River near Clarkdale, AZ - validation period
             (HSPF).
                                                 E-38

-------
            Avg Flow (10/1/1982 to 9/30/1992 )
            • Line of Equal Value
            Best-Fit Line
   1500
I
o
LJ-  1000 -
O)
TO
   y = 0.5957X + 72.209
       R2 = 0.666
    500 - — —»	-.'-
0         500        1000        1500

    Average Observed Flow (cfs)
                                                  100%
                                                  Avg Observed Flow (10/1/1982 to 9/30/1992 )
                                                  Avg Modeled Flow (10/1/1982 to 9/30/1992 )
                                                 -Line of Equal Value
                                        en
                                       O
                                        8
                                        ro
                                       m
                                        o>
                                       to
                                                            A-84  O-85
                                                                         A-87   O-88
                                                                          Month
                                                                                     A-90  O-91
  Figure 20. Monthly flow regression and temporal variation at USGS 09504000 Verde River near
            Clarkdale, AZ - validation period (HSPF).
          Avg Flow (10/1/1982 to 9/30/1992)
         •Line of Equal Value
         •Best-Fit Line
 500 n
           100    200    300    400
           Average Observed Flow (cfs)
                                       500
                                              Avg Monthly Rainfall (in)
                                             -Avg Observed Flow (10/1/1982 to 9/30/1992)
                                              Avg Modeled Flow (Same Period)
                                                     ONDJFMAMJJAS
                                             10 11  12  1  2  3  4  5  6  7
                                                           Month
  Figure 21. Seasonal regression and temporal aggregate at USGS 09504000 Verde River near
             Clarkdale, AZ - validation period (HSPF).
                                                  E-39

-------
         • Observed (25th, 75th)
         -Median Observed Flow (10/1/1982 to 9/30/1992)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
    600
    500
    400
 t

 o
           10     11      12      1
                                       CD
                                      or
   Figure 22. Seasonal medians and ranges at USGS 09504000 Verde River near Clarkdale, AZ -
            validation period (HSPF).
Table 8. Seasonal summary at USGS 09504000 Verde River near Clarkdale, AZ - validation
         period (HSPF)
MONTH
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
101.69
129.11
160.55
97.85
232.11
422.76
206.73
86.31
80.68
89.01
104.45
146.81
85.00
87.00
89.00
89.00
99.00
175.50
88.00
83.00
79.00
81.00
83.00
83.00
79.00
84.00
85.00
86.00
85.00
87.00
83.00
80.00
76.75
77.00
79.00
79.00
88.00
89.00
98.00
94.75
251.00
510.50
113.00
88.00
85.00
86.75
93.00
89.00
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
150.26
148.72
165.82
159.89
242.85
293.60
191.47
119.35
86.81
105.71
152.50
157.50
89.46
105.69
151.17
143.90
156.86
219.66
149.35
99.69
75.84
87.23
102.41
95.00
80.07
81.49
88.08
116.52
99.85
118.37
98.93
76.30
68.37
76.53
80.53
78.80
122.85
160.11
189.05
188.96
273.47
364.94
249.64
136.06
89.21
109.19
154.85
132.27
                                              E-40

-------
          •Observed Flow Duration (10/1/1982 to 9/30/1992 )
           Modeled Flow Duration (10/1/1982 to 9/30/1992 )
  10000
               10%
                       20%     30%     40%     50%     60%     70%     80%

                            Percent of Time that Flow is Equaled or Exceeded
                                                                               90%     100%
Figure 23.  Flow exceedance at USGS 09504000 Verde River near Clarkdale, AZ - validation period
          (HSPF).
o
O
CO
JD
o
o
I
"(D
E
              •Observed Flow Volume (10/1/1982 to 9/30/1992 )

              Modeled Flow Volume (10/1/1982 to 9/30/1992 )
     120%
     100%
      80%  —
      60%  —
40%  —
      20%  —
        Oct-82
                    Apr-84
                                Oct-85
                                     Apr-87
                                                       Oct-88
Apr-90
                                                                               Oct-91
Figure 24.  Flow accumulation at USGS 09504000 Verde River near Clarkdale, AZ - validation period
          (HSPF).
                                                E-41

-------
Table 9. Summary statistics at USGS 09504000 Verde River near Clarkdale, AZ - validation
          period (HSPF)
HSPF Simulated Flow
REACH OUTFLOW FROM DSN 101
10-Year Analysis F^riod: 10/1/1982 - 9/30/1992
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9)
Simulated Fall Flow Volume (months 10-12^
Simulated Winter Flow Volume (months 1-3^:
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Sjpring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of Efficiency, E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE ~
0.64

0.21
0.16

0.14
0.15
0.22
0.13

0.17
0.05
Error Statistics
6.31
3.34
-27.25
22.38
Observed Flow Gage
USGS 09504000 VERDE RIVER NEAR CLARKDALE, AZ
Hydrologic Unit Code: 15060202
Latitude: 34.8522416
Longitude: -112.065994
Drainage Area (sq-rri): 3503
Total Observed In-stream Flow:

Total of Observed highesMO%Jows:_^
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow VolumeJ10-12):
Observed Winter Flow VolumeJ1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
18.80 » | 30
-7.82
6.64
-27.53
39.20
0.451
0.325
0.655
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0





0.60

0.29
0.16

0.11
0.13
0.24
0.12

0.23
0.03





Clear [













Hydrology Results for Larger Watershed
As described above, parameters determined through calibration to the Verde River near Paulden and Clarkdale
gages were not fully transferable to other gages in the watershed. Hydrology calibration was performed at a total
of 10 gages in the Arizona basins - 6 in the Verde River watershed, 1 in the Salt River watershed, and 2 in the
San Pedro River watershed. Only the gage at Roosevelt provides a long period of record for the Salt River
watershed. The mainstem gage for the San Pedro River does not provide a rigorous calibration test because its
flow is largely determined by the upstream boundary condition. Therefore, calibration was also performed on
perennial Aravaipa Creek. The San Pedro gages ceased operation in 1995; therefore calibration was pursued over
an earlier time period without a separate validation test.

Calibration results at all gages are summarized in Table 10 and are generally of similar quality to the fit obtained
on the Verde River near Clarkdale. The generally close match between observed and predicted flow at the Salt
River gage is shown in Figures 25 through 31 and Tables 11 and 12. Results of the validation exercise are
summarized in Table 13. In general, the quality of fit during the validation period is similar to that in the
calibration period, with some reductions in fit for some of the seasonal volume error terms, and also some
improvements. The model is judged to be useful for scenario evaluation.
                                                E-42

-------
Table lO.Summary statistics (percent error) for all stations - calibration period WY 1992-2002
         (HSPF)


Station
Error in
total
volume:
Error in
50%
lowest
flows:
Error in
10%
highest
flows:
Seasonal
volume
error -
Summer:
Seasonal
volume
error- Fall:
Seasonal
volume
error -
Winter:
Seasonal
volume
error -
Spring:
Error in
storm
volumes:
Error in
summer
storm
volumes:
Daily
Nash-
Sutcliffe
Coefficient,
E:
Monthly
Nash-
Sutcliffe
Coefficient,
E:
09472000
San
Pedro nr
Redington
(1972-95)*

8.73


NA*


-5.26

22.67


8.64


-3.27


25.98


-4.82


11.07


0.574



0.908


09473000
Aravaipa
Crk nr
Mammoth
(1972-95)

-2.49


-3.74


1.89

-0.61


-7.68


2.64


-11.01


-13.40


-3.99


0.553



0.425



09498500
Salt River
nr
Roosevelt

4.48


2.24


7.56

20.18


11.78


2.79


-1.82


30.67


21.62


0.529



0.930



09503700
Verde
River nr
Paulden

9.41


8.37


5.57

20.17


15.98


5.12


9.84


-12.00


111.20


0.624



0.835



09504000
Verde
River nr
Clarkdale

2.43


-7.64


-1.58

3.15


11.52


0.83


-0.50


-15.79


44.47


0.481



0.803



09504500
Oak
Creek nr
Cornville

2.64


-12.56


0.51

-5.58


24.57


7.67


-33.23


-18.12


-8.02


0.078



0.443


09505800
W Clear
Cr nr
Camp
Verde

7.50


-7.97


2.60

-3.67


13.45


1.00


37.02


-33.88


-44.53


0.451



0.803


09506000
Verde
River nr
Camp
Verde

-2.41


-34.63


6.79

-34.15


-5.83


6.53


-18.47


-10.81


-42.63


0.661



0.803



09507980
E Verde
River nr
Childs

1.00


64.01


-1.53

-4.49


2.62


1.47


-0.25


-11.92


-57.76


0.689



0.946


09508500
Verde R
below
Tangle
Cr

-5.03


-17.25


-2.42

-19.87


0.44


-1.97


-14.97


-16.42


-36.53


0.703



0.921


*Note that median flow for the San Pedro River nr Redington is 0.
                                             E-43

-------
         Avg Monthly Rainfall (in)
        -Avg Observed Flow (10/1/1992 to 9/30/2002 )
        •Avg Modeled Flow (Same Period)
     100000
     90000
     80000
     70000
     60000
     50000
Q:    40000
     30000
     20000
     10000
1







1
ip '






,
IL







—
,1
m
I1





—
Hi
in .





i
i








—

i'|i'"|'






	
, j







	








—
•1









' i i








T '







,







	
|
"ii






T








	


i





—
.
          o
         Oct-92
                                                                         603
                                                                        *£

                                                                   -8   rS
                                                                        ^s
                                                                     10  g

                                                                   - 12
                                                                     14
Apr-94      Oct-95      Apr-97      Oct-98       Apr-00      Oct-01

                           Date
   Figure 25. Mean daily flow at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
            (HSPF).
i
 o
    15000
    10000 - —
     5000 - -
         0-92
                                                          Avg Monthly Rainfall (in)
                                                         -Avg Observed Flow (10/1/1992 to 9/30/2002 )
                                                         •Avg Modeled Flow (Same Period)
A-94
0-01
   Figure 26. Mean monthly flow at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
            (HSPF).
                                                 E-44

-------
    15000
I


 o
    10000
    5000
 O)
 TO
             Avg Flow (10/1/1992 to 9/30/2002 )

            •Line of Equal Value

             Best-Fit Line
  y = 0.9806X + 54.209

      R2 = 0.9333
0        5000       10000       15000


    Average Observed Flow (cfs)
                                                  100%
                                                 Avg Observed Flow (10/1/1992 to 9/30/2002 )

                                                 Avg Modeled Flow (10/1/1992 to 9/30/2002 )

                                                -Line of Equal Value
                                      en
                                      JD

                                      O
                                               ro
                                               m
                                                _
                                               0)

                                               to
                                                            A-94  O-95
                                                                         A-97   O-98


                                                                          Month
                                                                                     A-00   O-01
   Figure 27. Monthly flow regression and temporal variation at USGS 09498500 Salt River near

             Roosevelt, AZ - calibration period (HSPF).
           Avg Flow (10/1/1992 to 9/30/2002)

          • Line of Equal Value

           Best-Fit Line
  2500
          y = CJ.9922X f 44.478

 -20001-   -^B^=Loapa5_
-0 1500 -
  1000 -
0
D)
ro
   500 -
    500   1000   1500   2000


    Average Observed Flow (cfs)
                                       2500
                                               Avg Monthly Rainfall (in)

                                           —•-Avg Observed Flow (10/1/1992 to 9/30/2002)

                                               Avg Modeled Flow (Same Period)

                                         2500 T—-—-—-—-—-—-—-—-—-—-—-—r 0
                                         2000 -
                                      •g   1500
                                      r?   1000
                                                   500 - —a
                                                                                            3.5
10 11  12  1  23456789


              Month
   Figure 28. Seasonal regression and temporal aggregate at USGS 09498500 Salt River near Roosevelt,

             AZ - calibration period (HSPF).
                                                   E-45

-------
         • Observed (25th, 75th)
         -Median Observed Flow (10/1/1992 to 9/30/2002)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
    3000
    2500
    2000
 t

 o
   Figure 29. Seasonal medians and ranges at USGS 09498500 Salt River near Roosevelt, AZ -
             calibration period (HSPF).
Table 11.Seasonal summary at USGS 09498500 Salt River near Roosevelt, AZ - calibration
         period (HSPF)
MONTH

Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
256.42
450.48
507.61
1831.48
1511.59
1875.51
1429.55
768.86
277.08
226.12
377.92
368.88
200.50
243.00
267.50
267.00
455.50
1080.00
953.50
535.00
187.50
185.50
267.50
249.50
168.25
195.00
210.00
204.00
206.00
211.50
224.00
150.25
121.00
129.25
205.50
174.00
264.00
352.75
356.25
413.25
1135.00
2682.50
1722.50
1047.50
335.25
269.50
389.50
402.25
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
376.63
490.17
490.28
1674.43
1731.21
1974.34
1465.01
718.83
247.91
272.76
452.19
444.35
271.58
278.03
291.03
251.55
340.99
725.57
750.52
397.70
194.37
193.92
229.13
316.73
214.51
229.85
228.54
204.76
182.31
200.23
218.88
136.42
104.06
123.47
158.00
198.86
372.78
404.40
484.15
491.75
1074.86
2429.07
1974.37
761 .58
307.05
274.39
366.45
415.12
                                              E-46

-------
(0
Q
            •Observed Flow Duration (10/1/1992 to 9/30/2002 )
             Modeled Flow Duration (10/1/1992 to 9/30/2002 )
    100000
     10000
      1000
       100 - =
                  10%
                                                                                    90%
                                                                                            100%
                          20%     30%      40%     50%     60%      70%     80%

                               Percent of Time that Flow is Equaled or Exceeded

  Figure 30. Flow duration at USGS 09498500 Salt River near Roosevelt, AZ - calibration period (HSPF).
  o
  o
  ro
  T3
  (D
  
-------
Table 12.Summary statistics at USGS 09498500 Salt River near Roosevelt, AZ - calibration
        period (HSPF)
HSPF Simulated Flow
REACH OUTFLOW FROM DSN 103
10-Year Analysis F^riod: 10/1/1992 - 9/30/2002
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

_JoJak3fj3imuljjt^
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9)
Simulated Fall Flow Volume (months 10-12^
Simulated Winter Flow Volume (months 1-3^:
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
^EjTOnnJ^yTlih^SLIlP^J-^
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Sjpring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of Efficiency, E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE 1
2.70

1.67
0.29

0.31
0.36
1.40
0.64

1.22
0.11
Error Statistics
4.48
2.24
7.56
20.18
Observed Flow Gage
USGS 09498500 SALT RIVER NEAR ROOSEVELT, AZ
Hydrologic Unit Code: 15060103
Latitude: 33.61 94949
Longitude: -110.9215037
Drainage Area (sq-rri): 4306
Total Observed In-stream Flow:

Total of Observed higjiesMO%Jows:_^
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow VolumeJ10-12):
Observed Winter Flow VolumeJ1-3):
Observed Spring Flow Volume (4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
11.78 »| 30
2.79
-1.82
30.67
21.62
0.529
0.539
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0





2.59

1.56
0.29

0.26
0.32
1.36
0.65

0.94
0.09





Clear [












0.930
                                         E-48

-------
Table 13.Summary statistics for all stations - validation period WY 1982-1992 (HSPF)
Station
Error in total
volume:
Error in 50%
lowest flows:
Error in 10%
highest flows:
Seasonal
volume error -
Summer:
Seasonal
volume error -
Fall:
Seasonal
volume error -
Winter:
Seasonal
volume error -
Spring:
Error in storm
volumes:
Error in
summer storm
volumes:
Daily Nash-
Sutcliffe
Coefficient of
Efficiency, E:
Monthly Nash-
Sutcliffe
Coefficient of
Efficiency, E
09498500 Salt
River nr
Roosevelt
-7.09
-9.03
6.64
16.38
4.66
-26.48
1.49
16.98
40.28
0.354
0.786
09503700
Verde River
nr Paulden
8.35
5.07
-11.57
-4.29
13.22
16.64
8.64
-14.00
-20.94
0.443
0.614
09504000
Verde River
nrClarkdale
6.31
3.34
-27.25
22.38
18.80
-7.82
6.64
-27.53
39.20
0.451
0.320
09504500
Oak Creek nr
Cornville
15.32
3.54
-9.47
61.67
15.76
6.84
12.03
-17.61
86.24
0.545
0.755
09505800 W
ClearCr nr
Camp Verde
-9.69
-9.99
-17.07
-26.78
-37.41
7.66
-17.93
-51.19
-74.51
0.232
0.655
09507980
E Verde
River nr
Childs
31.18
8.49
45.18
-17.15
28.39
65.55
-26.11
34.57
-43.69
-0.119
0.335
09508500
Verde R
below Tangle
Cr
-6.93
-21.11
-13.91
24.90
-10.34
-5.86
-25.48
-16.03
78.31
0.510
0.809
Water Quality Calibration and Validation
The 20 Watershed models are designed to provide water quality simulation for total suspended solids (TSS), total
nitrogen, and total phosphorus. The water quality calibration focuses on the replication of monthly loads, as
specified in the project QAPP. Given the simplified approach to water quality simulation in the 20 Watershed
model a close match to individual concentration observations cannot be expected. However, comparison to
monthly loads presents challenges, as monthly loads are not observed. Instead, monthly loads must be estimated
from scattered concentration grab samples and continuous flow records. Such estimation presents some
uncertainty because it depends on the degree and form in which concentration and flow are correlated with one
another. Further, the bulk of the load of sediment and sediment-associated phosphorus is likely to move through
the system in a limited number of high flow events, which usually have not been monitored. As a result, the
monthly load calibration is inevitably based on the comparison of two uncertain numbers. Nonetheless,
calibration is able to achieve a fair agreement. The load comparisons were supported by detailed examinations of
                                                E-49

-------
the relationships of flows to loads and concentrations and the distribution of concentration prediction errors versus
flow, time, and season, as well as standard time series plots.

For application on a nationwide basis, the 20 Watershed protocols assume that TSS and total phosphorus loads
will likely exhibit a strong positive correlation to flow (and associated erosive processes), while total nitrogen
loads, which often have a dominant groundwater component, will not. Accordingly, TSS and total phosphorus
loads were estimated from observations using a flow-stratified log-log regression approach, while total nitrogen
loads were estimated using a flow-stratified averaging estimator, consistent with the findings of Preston et al.
(1989).

Water quality calibration and validation was done on the Verde River near Clarkdale, using 1993-2002 for
calibration and 1986-1992 for validation. As with hydrology, calibration was performed on the later period as this
better reflects the land use included in the model. The start of the validation period is constrained by data
availability.

TSS calibration was performed by adjusting the coefficients in the soil detachment (KRER) and soil washoff
(KSER) equations along with changes to the seasonal vegetation COVER. Furthermore, it was necessary to model
scour of the soil matrix (i.e., gully erosion) in addition to losses of detached sediment. The washoff of detached
sediment did not provide sufficient sediment losses to calibrate the model without severely degrading the channel
bed.

Time series of simulated and estimated TSS loads at the Clarkdale station for both periods are shown in Figure 32
and statistics for the two periods are provided separately in Table 14. Visually, the model is roughly simulating
the trends contained in the observed data. The key statistic in Table  14 (consistent with the QAPP) is the relative
percent error, which shows the error in the prediction of monthly load normalized to the estimated load. Table  14
also shows the relative average absolute error, which is the average  of the relative magnitude of errors in
individual monthly load predictions. This number is inflated by outlier months in which the simulated and
estimated loads differ by large amounts (which may be  as easily due to uncertainty in the  estimated load due to
limited data as to problems with the model) and the third statistic, the relative median absolute error, is likely
more relevant and shows good agreement.
                                               TSS
     1,000,000
                                                                                   - Regression Loads
                                                                                   -Simulated Loads
                 OO
                            O->-rMCO-^-ir)(DI^
                            O)O)O)O)O)O)O)O)
g   8  5  8
    Figure 32.  Fit for monthly load of TSS at USGS 09504000 Verde River near Clarkdale, AZ (HSPF).
                                                 E-50

-------
Table 14. Model fit statistics (observed minus predicted) for monthly sediment loads using
         stratified regression at USGS 09504000 Verde River near Clarkdale, AZ (HSPF)
Statistic
Relative Percent Error
Relative Average Absolute Error
Relative Median Absolute Error
Calibration period
(1993-2002)
31%
40%
1.1%
Validation period
(1986-1992)
-41%
123%
8.5%
Several other diagnostics were also examined to evaluate agreement between the model and observations. These
are available in full in the calibration spreadsheets, but a few examples are provided below. First, load-flow power
plots were compared for individual days (Figures 33 and 34). These show that the relationship between flow and
load is reasonably consistent across the entire range of observed flows, for both the calibration and validation
periods.
Verde River -Clarkdale
1993-2002
-I nnnnnn
I UUUUUU
100000
>
| 10000
I/)
I 1000
§ 100
| 10
1
0-1
. 1 n
1
••
^^^
• • >v-.-*-t^i^^
4^^'-^>
jjKj&**_

IT

10 100 1000 10000 100000
Flow, cfs
• Simulated A Observed ^^™ Power (Simulated) Power (Observed)
    Figure 33.  Power plot for observed and simulated TSS at USGS 09504000 Verde River near Clarkdale,
            AZ - calibration period (HSPF).
                                              E-51

-------
       100000
    ra
    •o
    ra
    o


    (0
10000



 1000



  100



    10



     1  -f



   0.1
1
                             Verde River - Clarkdale

                                     1986-1992

                              10
                                     100


                                   Flow, cfs
  1000
10000
                 Simulated  A Observed
                                 Power (Simulated)
Power (Observed)
    Figure 34. Power plot for observed and simulated TSS at USGS 09504000 Verde River near Clarkdale,

            AZ - validation period (HSPF).



A standard time series plot (Figure 35) shows that observed and simulated concentrations achieve at best a fair

agreement, and the model may deviate substantially from individual observations. However, the concentration

statistics (Table 15) show that reasonably low median errors are achieved.
                                              E-52

-------
                              Verde River - Clarkdale
                                      1993-2002
• Simulated   A Observed
       10000
    O)
            1993   1994  1995  1996  1997   1998  1999  2000  2001  2002
                                             Year
    Figure 35. Time series plot of TSS concentration at USGS 09504000 Verde River near Clarkdale, AZ
             (HSPF).
Table 15.Relative errors (observed minus predicted) for TSS concentration at USGS 09504000
         Verde River near Clarkdale, AZ (HSPF).
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1993-2002)
47
-104%
-19%
Validation period
(1986-1992)
62
7.3%
-1.0%
For simulation of total phosphorus, calibration was performed primarily through adjustment of the potency factors
and the subsurface concentrations. Total nitrogen calibration was accomplished primarily by adjusting the
subsurface concentrations and secondarily by the accumulation-washoff parameters. Monthly loading time series
for total phosphorus are shown in Figure 36 and the load statistics are summarized in Table 16. The model
reproduces the general trend in monthly loads, but is significantly lower than the peak loads predicted by the
regression method, resulting in high relative percent errors for both the calibration and validation periods. It
should be noted that the available data are limited, particularly for high flow events. Thus, the estimates of
"observed" load are also subject to considerable uncertainty.
                                               E-53

-------
     10000
                                            Total P
                                                               9   9  9
                                                                                - Regression Loads
                                                                                -Simulated Loads
    Figure 36. Fit for monthly load of total phosphorus at USGS 09504000 Verde River near Clarkdale, AZ
             (HSPF).
Table 16. Model fit statistics (observed minus predicted) for monthly total phosphorus loads
         using stratified regression
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1993-2002)
87%
87%
0.6%
Validation period
(1986-1992)
66%
78%
5.1%
Additional diagnostics for total phosphorus included flow-load power plots (Figures 37 and 38), concentration
time series plots (Figure 39) and analysis of concentration errors (Table 17). While these show approximate
agreement, the model often overpredicts total phosphorus concentrations under lower flow conditions.
                                               E-54

-------
                        Verde River - Clarkdale

                               1993-2002
w
•o
•o
a
o
  100




   10




    1




  0.1




 0.01




0.001
                     10
                             100         1000


                                 Flow, cfs
     10000
100000
          •  Simulated  A  Observed
                                Row er (Simulated)
 Power (Observed)
Figure 37. Power plot for observed and simulated total phosphorus at USGS 09504000 Verde River

         near Clarkdale, AZ - calibration period (HSPF).
                        Verde River - Clarkdale

                               1986-1992
      10
w
•o

w
c

s

•a
a
o
  0.1
    0.01
   0.001
                        10
                                    100


                                 Flow, cfs
   1000
 10000
            Simulated  A  Observed
                               Fbw er (Simulated)
Fbw er (Observed)
Figure 38. Power plot for observed and simulated total phosphorus at USGS 09504000 Verde River

         near Clarkdale, AZ - validation period (HSPF).
                                          E-55

-------
                               Verde River - Clarkdale
                                      1993-2002
• Simulated  A  Observed
          1993  1994  1995  1996  1997   1998  1999  2000  2001  2002
                                             Year
    Figure 39. Time series plot of total phosphorus concentration at USGS 09504000 Verde River near
            Clarkdale, AZ (HSPF).
Table 17. Relative errors (observed minus predicted) for total phosphorus concentration at
         USGS 09504000 Verde River near Clarkdale, AZ (HSPF)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1993-2002)
57
16%
-23%
Validation period
(1986-1992)
75
54%
-8.5%
Fewer data are available for total nitrogen because many sampling events omitted one or more nitrogen species.
This increases the uncertainty of the comparison. Results for total nitrogen are summarized in Figures 40 through
43 and Tables 18 and 19 following the same format as total phosphorus. The loading results are fair, and are
generally better than those obtained for total phosphorus; however, there is significant uncertainty in the
prediction of individual total nitrogen observations. Total nitrogen concentrations at base flow are generally over-
predicted.
                                             E-56

-------
     180
        c
        ro
                                          Total N
                                                                             -Averaging Loads


                                                                             -Simulated Loads
c
ro
c
ro
9
c
ro
                                      •fincor^oocno-j-CN
                                      o>o>o>o>o>o>ooo
c
ro
c
ro
c
ro
c
ro
c
ro
c
ro
   Figure 40. Fit for monthly load of total nitrogen at USGS 09504000 Verde River near Clarkdale, AZ

            (HSPF).
Table 18. Model fit statistics (observed minus predicted) for monthly total nitrogen loads using

         averaging estimator (HSPF)
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1993-2002)
1 .6%
45%
17%
Validation period
(1986-1992)
-2.7%
37%
18%
                                             E-57

-------
                         Verde River - Clarkdale

                                1993-2002
TN Load, tons/day
    100
O
      1
    0.1
   0.01
                    10
                           100         1000


                              Flow, cfs
      10000
100000
          •  Simulated A Observed ^^™ Power (Simulated) ^^"Fbwer (Observed)
Figure 41. Power plot for observed and simulated total nitrogen at USGS 09504000 Verde River near

         Clarkdale, AZ - calibration period (HSPF).
                        Verde River - Clarkdale

                                1986-1992
     10
re
5
In
c
o
•4-i

-6
re
o
1
    0.1
   0.01
                       10
                                 100


                              Flow, cfs
    1000
 10000
           Simulated A Observed
                            Fbw er (Simulated)
Row er (Observed)
Figure 42. Power plot for observed and simulated total nitrogen at USGS 09504000 Verde River near

         Clarkdale, AZ - validation period (HSPF).
                                          E-58

-------
       2.5
    D)
         2
       1.5
                               Verde River - Clarkdale
                                       1993-2002
•Simulated  A  Observed
         1993   1994   1995   1996   1997   1998  1999  2000   2001   2002
                                              Year
    Figure 43. Time series plot of total nitrogen concentration at USGS 09504000 Verde River near
             Clarkdale, AZ (HSPF).
Table 19. Relative errors (observed minus predicted) for total nitrogen concentration at USGS
         09504000 Verde River near Clarkdale, AZ (HSPF)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1993-2002)
46
-63%
-70%
Validation period
(1986-1992)
75
22%
1.9%
Water Quality Results for Larger Watershed
Summary statistics for the water quality calibration and validation at two other stations in the watershed (i.e., Salt
River near Roosevelt and Verde River below Tangle Creek) are provided in Tables 20 and 21 along with the
Clarkdale statistics. Water quality was not calibrated at the Camp Verde station on the Verde River because of the
lack of observed data at that location. And no water quality calibration was done in the San Pedro River. In most
cases, total nitrogen loads are better predicted than total phosphorus and TSS loads. Simulated TSS and total
phosphorus loads in the Verde River are lower than those estimated from observations, but this may reflect, in
part, the uncertainty in the regression-based load estimates as water quality observations during high flows are
sparse. Water quality results in the Salt River are generally better than the Verde River.
                                              E-59

-------
Table 20. Summary statistics (observed minus predicted) for water quality for all stations
        calibration period 1993-2002 (HSPF)
Station
Relative Percent Error
TSS Load
TSS Concentration
Median Percent Error
Relative Percent Error
TP Load
TP Concentration
Median Percent Error
Relative Percent Error
TN Load
TN Concentration
Median Percent Error
09498500
Salt River nr
Roosevelt
-10
0.80
-29
-0.53
-8.2
-3.5
09504000
Verde River nr
Clarkdale
31
-19
87
-23
1.6
-70
09508500
Verde River below
Tangle Cr
81
-5.6
75
-17
-2.7
-64
Table 21. Summary statistics (observed minus predicted) for water quality for all stations
        validation period 1986-1992 (HSPF)
Station
Relative Percent Error
TSS Load
TSS Concentration
Median Percent Error
Relative Percent Error
TP Load
TP Concentration
Median Percent Error
Relative Percent Error
TN Load
TN Concentration
Median Percent Error
09498500
Salt River nr
Roosevelt
4.8
1.8
52
8.9
10
-21
09504000
Verde River nr
Clarkdale
-41
-1.0
78
-8.5
-2.7
1.9
09508500
Verde River below
Tangle Cr
2.1
-28
46
-12
10
-15
                                          E-60

-------
a a a a a a a a  a a a a a a  a a a a a a  a a a a a a a a a a a  a a a a a a  a a a a a a  a a a a a a a a a a a a a a a a a  a a a a a a  a a a a a a a a a a a a a a a a a  a a a a a
A SWAT model already exists for the Verde River portion of the watershed (Tetra Tech 2001). This model was
calibrated for hydrology and nutrients. The existence of this earlier model provides a useful basis for parameter
initialization, and is one of the reasons that the Arizona basins were selected as a pilot site. However, there are
also significant differences to the 20 Watershed model. Because of these differences in approach the two models
are substantively different, and not all model parameters are transferable. Nonetheless, the earlier model does
provide important insights and parameter starting values that are incorporated into the 20 Watershed model.

A key aspect of the Arizona basins models is the intimate linkage of surface and groundwater hydrology.
Perennial flow in various river segments is supported by groundwater discharge that, in the case of the Verde
watershed, may arise from distant teleconnections. Many of the river reaches in alluvial valleys also lose flow to
groundwater, at least on a seasonal basis. Both SWAT and HSPF models include  simplified mass-balance
accounting of groundwater at a local (subwatershed) scale, but neither model contains a detailed simulation of
surface and groundwater interactions. In SWAT, the presence of deep groundwater discharges derived from
sources outside a model subwatershed can be addressed through specification of these inflows as point sources.
This is adequate for calibration, but there is no provision in the model for direct consideration of how these
sources may change in the face of climate change. SWAT also provides for simulation of losing river reaches
through specification of a rate of bed conductivity. This approach does not account for interaction with the
seasonal water table. In many cases, alluvial river reaches may gain from groundwater during the wet season
when water tables are high and lose to the alluvial aquifer during dry seasons. SWAT is, however, constrained to
simulation of these interactions solely as a bed conductivity rate. This means that the behavior of seasonally
losing reaches can only be roughly approximated in the model.


                   to

No changes were made to the meteorological or land use base data. For the Globe, Arizona point source it is clear
that phosphorus concentrations downstream during baseflow conditions are much less than would be expected
from the estimated phosphorus load from this WWTP. This may be due to a rapid loss of phosphorus in the near
field immediately downstream of the discharge, likely due to settling of particulate matter. Discounting total
phosphorus load in the effluent to 1/8 of the nominal value provided resolved this problem during calibration.
A key feature of the Arizona basins is the complex interaction of surface and groundwater. As noted above,
neither SWAT nor HSPF is capable of providing a detailed, process-based simulation of groundwater flow. It is
therefore assumed that interactions with groundwater can be handled with the following simplifying assumptions:
    •  The local (within a subwatershed) accumulation and discharge of shallow groundwater is adequately
       addressed by SWAT's linear storage reservoir formulation.
    •  Discharges to stream from deep groundwater are represented as constant point sources, with discharge
       rates set based on flow information on major springs identified in the Arizona Water Atlas (ADWR
       2009). This means that the model cannot account for seasonal variability in deep groundwater discharge,
       nor can it evaluate how such discharges may evolve in response to climate change.
    •  Losses to groundwater from stream reaches in alluvial basins are simulated based on channel
       conductivity. This  makes such losses a function of flow, wetted perimeter, and travel time in the affected
       reaches. The SWAT model formulation does not take into account changes in local groundwater head;
       instead, the loss occurs continuously. In addition, SWAT partitions channel losses to deep groundwater
       and bank storage using a fixed ratio.


                                                 E-61

-------
Hydrology Calibration
Obtaining a high quality fit to hydrology in the Arizona basins is difficult with SWAT due to the importance of
groundwater interaction terms, which are simplistically addressed in the model. The specification of groundwater
discharges as constant values and the simulation of reach losses by channel conductivity without feedback from
local groundwater elevations both introduce uncertainty. In addition, the model formulation requires specification
of several factors at a global level, not allowing for spatial variability, including the fraction of transmission losses
assigned to deep groundwater (TRNSRCH), the intensity of direct evaporation from the channel (EVRCH), and
the parameters controlling snow melt.

Calibration adjustments focused on the following parameters:
    •    Curve numbers (varied systematically by land use)
    •    ESCO (soil evaporation compensation factor)
    •    Reach conductivity
    •    Bank storage and recession rates
    •    Groundwater "revap" rates

Initial calibrations were performed for the Verde River at Clarkdale and are summarized in Figures 44 through 50
Tables 22 and 23. The fit is fair at best, although the total volume errors are small. However, predictions at
Clarkdale are largely determined by model fit upstream at Paulden, where flows about 95 percent of the time
consist of approximately constant baseflow. Spring peaks occasionally push through from the Chino watershed.
Accuracy in simulating these peaks appears to be affected by 1) lack of an accurate representation of the hydraulic
behavior of Sullivan Lake, and 2) the necessity of specifying constant values  for channel conductivity to account
for transmission losses, when in fact these loss rates are likely much reduced  during the spring wet period.
Modifications to increase spring flows out of the Chino watershed rapidly results in severe over-prediction of
response to summer storm events. The Nash-Sutcliffe coefficient is near zero because flows tend to remain
approximately constant during dry periods.
          Avg Monthly Rainfall (in)
         -Avg Observed Flow (10/1/1992 to 9/30/2002 )
         •Avg Modeled Flow (Same Period)
      35000
    30000

_  2500°
;§,  20000

1   15000
      10000
                                          Tyyy

                            	1	1	1	1	1	]	1	
                                                                                        14
                                Oct-95
                                         Apr-97       Oct-98

                                              Date
Apr-00
Oct-01
    Figure 44. Mean daily flow at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
             (SWAT).
                                                 E-62

-------
i
o
    4000
    3000 -
2000 -
    1000 - -
        0-92
                                                          Avg Monthly Rainfall (in)
                                                         -Avg Observed Flow (10/1/1992 to 9/30/2002 )
                                                         •Avg Modeled Flow (Same Period)
                A-94
0-01
  Figure 45. Mean monthly flow at USGS 09504000 Verde River near Clarkdale, AZ - calibration period
            (SWAT).
            Avg Flow (10/1/1992 to 9/30/2002 )
            • Line of Equal Value
            Best-Fit Line
   4000
           1000     2000     3000    4000

        Average Observed Flow (cfs)
                                              £
                                              _
                                              ro
                                              m
                                                     Avg Observed Flow (10/1/1992 to 9/30/2002 )
                                                     Avg Modeled Flow (10/1/1992 to 9/30/2002 )
                                                    -Line of Equal Value
                                                     O-92  A-94  O-95
                                                                       A-97   O-98
                                                                         Month
                                                                                    A-00   O-01
  Figure 46. Monthly flow regression and temporal variation at USGS 09504000 Verde River near
            Clarkdale, AZ - calibration period (SWAT).
                                                  E-63

-------
       •  Avg Flow (10/1 /1992 to 9/30/2002)
      	Line of Equal Value
          Best-Fit Line
                                                Avg Monthly Rainfall (in)
                                               -Avg Observed Flow (10/1/1992 to 9/30/2002)
                                                Avg Modeled Flow (Same Period)
  600
                                           i
                                           I
                                              600
                                              400
                                              200
               200        400

           Average Observed Flow (cfs)
                                     600
                                               10 11  12  1   23456789

                                                             Month
   Figure 47.  Seasonal regression and temporal aggregate at USGS 09504000 Verde River near
             Clarkdale, AZ - calibration period (SWAT).
         i Observed (25th, 75th)
         •Median Observed Flow (10/1/1992 to 9/30/2002)
                                                          Average Monthly Rainfall (in)
                                                          Modeled (Median, 25th, 75th)
•e
1
    800
    600
400
    200
           10      11      12
   Figure 48. Seasonal medians and ranges at at USGS 09504000 Verde River near Clarkdale, AZ -
             calibration period (SWAT).
                                                   E-64

-------
Table 22.Seasonal summary at USGS 09504000 Verde River near Clarkdale, AZ - calibration
         period (SWAT)
MONTH
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
87.85
84.09
101.70
370.90
529.79
290.55
129.11
79.98
76.29
78.20
83.85
100.71
80.00
84.00
83.50
87.00
87.00
92.00
83.00
79.00
75.00
77.00
78.00
81.00
77.00
78.00
80.00
81.00
80.00
80.00
77.00
72.00
69.00
71.00
75.00
73.00
84.00
89.00
90.00
93.00
101.50
232.00
93.00
87.00
81.00
82.00
84.00
85.00
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
152.63
107.60
138.79
235.56
340.35
181.43
119.69
90.37
86.15
117.83
154.15
228.15
91.66
92.48
89.91
104.44
99.91
103.13
94.57
72.78
72.08
84.24
113.30
107.68
82.39
82.01
80.83
78.08
76.04
73.67
72.58
64.29
60.86
66.98
86.81
79.88
113.64
112.22
114.00
147.50
154.28
161.90
127.94
102.97
89.61
114.69
163.60
147.35
            •Observed Flow Duration (10/1/1992 to 9/30/2002 )

            •Modeled Flow Duration (10/1/1992 to 9/30/2002 )
     100000
  o
  
-------
                •Observed Flow Volume (10/1/1992 to 9/30/2002 )

                •Modeled Flow Volume (10/1/1992 to 9/30/2002 )
        120%
   o
   o
   T3

    I
    
-------
SWAT Simulated Flow
REACH OUTFLOW FROM OUTLET 10
10-Year Analysis F^riod: 10/1/1992 - 9/30/2002
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9):
Simulated Fall Flow Volume (months 10-12^
Simulated Winter Flow Volume (months 1-3^:
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error injH)%Jik|hest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error- Winter:
Seasonal volume error - Sjpring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of Efficiency, E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
0.63

0.29
0.14

0.16
0.13
0.24
0.10

0.28
0.08
Error Statistics
-2.46
-1.74
-19.34
89.89
Observed Flow Gage
USGS 09504000 VERDE RIVER NEARCLARKDALE, AZ
Hydrologic Unit Code: 15060202
Latitude: 34.852241 6
Longitude: -11 2.065994
Drainage Area (sq-rri): 3503
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow VolumeJ10-12):
Observed Winter Flow VolumeJ1-3):
Observed Spring Flow Volume (4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
46.00 » | 30
-36.45
3.88
-3.32
593.89
0.030
0.236
0.685
30
30
20
50
Model accuracy increases
as E or E' approaches 1.0





0.64

0.36
0.15

0.09
0.09
0.38
0.09

0.29
0.01





Clear [













Hydrology Validation
Validation for the Verde River near Clarkdale was performed for the period 10/1/1982 through 9/30/1992. Results
are presented in Figures 51 through 57 and Table 24 and 25. The validation results are generally similar to those
of the calibration period, and indeed are better on  some statistics, indicating that the model is not over-fit to the
specific conditions of the calibration period.

It is important to recognize that the validation uses the 2001 land use as a static representation. While the
watershed has remained largely in National Forest, Indian Reservations and unoccupied rangeland, important
temporal changes have occurred as a result of intermittent wildfires. Areas of recent burns typically have
decreased evapotranspiration and increased direct runoff. These, however, are not represented in the model. In
addition, the PET estimates for the 20 Watershed model use SWAT weather generator statistics for solar
radiation, cloud cover, wind, and relative humidity - essentially assuming that the central tendency of these
factors has not changed over time.
                                                  E-67

-------
         Avg Monthly Rainfall (in)
         Avg Observed Flow (10/1/1982 to 9/30/1992 )
         Avg Modeled Flow (Same Period)
18000
16000
_ 14000
.(/) -i onnn
7" 10000
Q: 8000
6000
4000
2000
I'll » I
	 1-




r, j" •





jttltr
	 1 	




"




tF,

	




"
±f.

	 } 	 ] 	 i 	


j j

1 ~[ 1
ii tirH
Oct-82 Apr-84 Oct-85 Apr-87 Oct-88
m "•) |1






	 j-L-




rr
if
li






" r "
,t
-2
-4 1
603
•E
-8 |
-10 f
- 12
Apr-90 Oct-91
                                               Date

  Figure 51. Mean daily flow at USGS 09504000 Verde River near Clarkdale, AZ - validation period
            (SWAT).
    1500
i
I
    1000 -
     500 - -
        O-82
A-84
                                                         Avg Monthly Rainfall (in)
                                                        -Avg Observed Flow (10/1/1982 to 9/30/1992
                                                        •Avg Modeled Flow (Same Period)
O-85
A-87
   Month
O-88
A-90
O-91
                                                           )

                                                           10
                                                           9
                                                           8
                                                           7
                                                           6
                                                           5
                                                           4
                                                           3
                                                           2
                                                           1
                                                           0
                                                                          I
                                                                          'S
                                                                          a:
  Figure 52. Mean monthly flow at USGS 09504000 Verde River near Clarkdale, AZ - validation period
            (SWAT).
                                                 E-68

-------
            Avg Flow (10/1/1982 to 9/30/1992 )
            • Line of Equal Value
            Best-Fit Line
   1500
I

JD
LJ- 1000--
O)
TO
   y = 0.4995x + 86.113
       R2 = 0.368f
    500 - -     *  -.'--
0         500        1000       1500

    Average Observed Flow (cfs)
                                                  100%
                                                  Avg Observed Flow (10/1/1982 to 9/30/1992 )
                                                  Avg Modeled Flow (10/1/1982 to 9/30/1992 )
                                                 -Line of Equal Value
                                       en
                                       O
                                       8
                                       ro
                                       m
                                       o>
                                       to
                                                            A-84  O-85
                                                                        A-87   O-88
                                                                          Month
                                                                                     A-90   O-91
  Figure 53. Monthly flow regression and temporal variation at USGS 09504000 Verde River near
            Clarkdale, AZ - validation  period (SWAT).
Best-Fit Line
&
Q
400
o
rr
-n ^nn

-------
         • Observed (25th, 75th)
         -Median Observed Flow (10/1/1982 to 9/30/1992)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
    600
    500
  o
           10     11      12      1
   Figure 55. Seasonal medians and ranges at USGS 09504000 Verde River near Clarkdale, AZ -
            validation period (SWAT).
Table 24. Seasonal summary at USGS 09504000 Verde River near Clarkdale, AZ - validation
         period (SWAT)
MONTH
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
101.69
129.11
160.55
97.85
232.11
422.76
206.73
86.31
80.68
89.01
104.45
146.81
85.00
87.00
89.00
89.00
99.00
175.50
88.00
83.00
79.00
81.00
83.00
83.00
79.00
84.00
85.00
86.00
85.00
87.00
83.00
80.00
76.75
77.00
79.00
79.00
88.00
89.00
98.00
94.75
251 .00
510.50
113.00
88.00
85.00
86.75
93.00
89.00
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
133.41
177.41
189.36
148.63
201.41
209.87
131.86
106.03
91.66
146.37
194.00
230.93
95.42
102.19
113.19
115.42
112.81
125.61
101.34
89.26
80.87
102.31
123.61
100.94
80.47
85.14
93.62
103.86
98.46
94.45
88.26
80.14
72.39
82.18
91.04
84.97
115.23
132.23
145.09
146.28
161.08
179.31
122.62
104.86
92.45
143.73
185.62
121.65
                                              E-70

-------
o
0
D)
0
ro
Q
            •Observed Flow Duration (10/1/1982 to 9/30/1992 )

            •Modeled Flow Duration (10/1/1982 to 9/30/1992 )
    100000
     10000
      1000
       100
         0%      10%     20%     30%     40%     50%     60%     70%     80%     90%     100%


                              Percent of Time that Flow is Equaled or Exceeded



  Figure 56. Flow exceedance at USGS 09504000 Verde River near Clarkdale, AZ - validation period

             (SWAT).
  o
  o

  in
  ro

  T3
  0


  0
  
-------
Table 25. Summary statistics at USGS 09504000 Verde River near Clarkdale, AZ - validation
          period (SWAT)
SWAT Simulated Flow
REACH OUTFLOW FROM OUTLET 10
10-Year Analysis F^riod: 10/1/1982 - 9/30/1992
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9):
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3):
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe CoefficjejTtjDfJEfficiency^E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
0.63

0.26
0.16

0.19
0.16
0.18
0.11

0.26
0.10
Error Statistics
5.68
5.17
-9.70
68.04
Observed Flow Gage
USGS 09504000 VERDE RIVER NEAR CLARKDALE, AZ
Hydrologic Unit Code: 15060202
Latitude: 34.852241 6
Longitude: -11 2.065994
Drainage Area (sq-rri): 3503
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow Volume (1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
27.70 » | 30
-25.96
-11.56
12.44
197.67
-0.996
0.121
0.320
30
30
20
50
Model accuracy increases
as E or E' approaches 1.0





0.60

0.29
0.16

0.11
0.13
0.24
0.12

0.23
0.03





Clear [













Hydrology Results for Larger Watershed

As described above, parameters determined through calibration to the Verde River near Clarkdale gage were not
fully transferable to other gages in the watershed. Therefore, calibration was pursued at a total of 10 gages in the
Arizona basins - six in the Verde River watershed, one in the Salt River watershed, and two in the San Pedro
River watersheds. Only the gage at Roosevelt provides a long period of record for the Salt River watershed. The
mainstem gage for the San Pedro River does not provide a rigorous calibration test because its flow is largely
determined by the upstream boundary condition. Therefore, calibration was also pursued on perennial Aravaipa
Creek. The two San Pedro gages ceased operation in 1995; therefore calibration was pursued over an earlier time
period without a separate validation test.

Calibration results at all gages are summarized in Table 26 and are generally of similar quality to the fit obtained
on the Verde River near Clarkdale. The fit for the East Verde River is believed to be relatively poor due to the
influences of Payson water withdrawals. The generally close match between observed and predicted flow at the
Salt River gage is shown in Figures 58 through 64 and Tables 27 and 28. Results of the validation exercise are
summarized in Table 29. In general, the quality of fit during the validation period is similar to that in the
calibration period. Thus, the model is judged to be useful for scenario evaluation.
                                                E-72

-------
Table 26. Summary statistics (percent error) at all stations - calibration period WY 1992-2002
         (SWAT)
Station
Error in
total
volume:
Error in
50%
lowest
flows:
Error in
10%
highest
flows:
Seasonal
volume
error -
Summer:
Seasonal
volume
error -
Fall:
Seasonal
volume
error -
Winter:
Seasonal
volume
error -
Spring:
Error in
storm
volumes:
Error in
summer
storm
volumes:
Daily
Nash-
Sutcliffe
Coefficient
of
Efficiency,
E:
Monthly
Nash-
Sutcliffe
Coefficient
of
Efficiency,
E:
09472000
San
Pedro nr
Redington
(1972-95)*
-6.74
NA*
-7.97
-4.82
8.09
-20.25
-7.04
1.49
-3.89
-0.362
0.690
09473000
Aravaipa
Crknr
Mammoth
(1972-95)
3.46
-1.25
2.62
53.40
-3.31
-10.77
7.82
-27.95
73.05
0.629
0.717
09498500
Salt River
nr
Roosevelt
9.43
-7.16
4.52
58.85
103.52
6.95
-51.64
16.40
103.32
0.222
0.783
09503700
Verde
River nr
Paulden
9.15
-1.14
8.65
110.89
56.82
-24.75
-1.39
19.48
875.65
-0.864
0.777
09504000
Verde
River nr
Clarkdale
-2.46
-1.74
-19.34
89.89
46.00
-36.45
3.88
-3.32
593.89
0.030
0.320
09504500
Oak
Creek nr
Cornville
-2.63
-17.90
-10.31
96.08
27.01
-22.05
-4.10
-10.98
205.71
0.454
0.728
09505800
W Clear
Crnr
Camp
Verde
9.45
-10.22
7.29
49.24
26.17
2.87
2.23
21.24
96.30
-0.021
0.804
09506000
Verde
River nr
Camp
Verde
7.68
8.49
-5.00
100.98
44.07
-13.37
3.38
9.94
153.62
0.225
0.880
09507980
E Verde
River nr
Childs
-6.21
56.51
-6.56
156.62
58.54
-35.02
0.08
17.92
338.25
0.311
0.736
09508500
Verde R
below
Tangle
Cr
1.68
-17.85
-7.27
96.88
47.60
-19.30
-7.60
16.89
262.34
0.250
0.860
! Note that median flow for the San Pedro River nr Redington is 0.
                                               E-73

-------
         Avg Monthly Rainfall (in)
        -Avg Observed Flow (10/1/1992 to 9/30/2002 )
        •Avg Modeled Flow (Same Period)
t
o
100000
 90000
 80000
 70000
 60000
 50000
 40000
 30000
 20000
 10000
     0
IJT' '
I


fi
|l
u
u
jfpt



	



,l
1 1 1
['1 1 I1
I1

	




ill



	

1
,
J
1



	








	 1 	
i



.1 tf J



	




1*





















	 1 	
i

• I
t
JLL -i
n"l T


	




| J

t f

—
j



illl
                                                                                        - 2
                                                                                          8
                                                                                          10   «
                                                                                          12
                                                                                          14
         Oct-92       Apr-94      Oct-95      Apr-97      Oct-98

                                                Date
                                                              Apr-00
Oct-01
   Figure 58. Mean daily flow at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
             (SWAT).
i
 o
    15000
    10000 -
     5000 - -
                                                          Avg Monthly Rainfall (in)
                                                         -Avg Observed Flow (10/1/1992 to 9/30/2002 )
                                                          Avg Modeled Flow (Same Period)
         O-92
                A-94
   O-01
   Figure 59. Mean monthly flow at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
             (SWAT).
                                                 E-74

-------
             Avg Flow (10/1/1992 to 9/30/2002 )

            •Line of Equal Value

             Best-Fit Line
    15000
I


 o
y = 0.8091 x +234.63

    R2 = 0.786
    10000
    5000
 O)
 TO
                                                   100%
                                               Avg Observed Flow (10/1/1992 to 9/30/2002 )

                                               Avg Modeled Flow (10/1/1992 to 9/30/2002 )

                                              -Line of Equal Value
                                     en
                                    JD

                                    O
                                                ro
                                               m
                                                _
                                                0)

                                               to
         0        5000       10000       15000


            Average Observed Flow (cfs)
                                           O-92   A-94   O-95
                                                              A-97  O-98

                                                               Month
                                                                           A-00   O-01
   Figure 60. Monthly flow regression and temporal variation at USGS 09498500 Salt River near

             Roosevelt, AZ - calibration period (SWAT).
           Avg Flow (10/1/1992 to 9/30/2002)

          • Line of Equal Value

           Best-Fit Line
                                             Avg Monthly Rainfall (in)

                                            -Avg Observed Flow (10/1/1992 to 9/30/2002)

                                             Avg Modeled Flow (Same Period)
  2500


 I


 '2000
-a 1500 -
  1000 -
0
D)
ro
   500 -
                                                   2500 -i
             500   1000   1500   2000


             Average Observed Flow (cfs)
                             2500
10 11  12 1  23456789


               Month
   Figure 61. Seasonal regression and temporal aggregate at USGS 09498500 Salt River near Roosevelt,

             AZ - calibration period (SWAT).
                                                   E-75

-------
         • Observed (25th, 75th)
         -Median Observed Flow (10/1/1992 to 9/30/2002)
                                     Average Monthly Rainfall (in)
                                     Modeled (Median, 25th, 75th)
    3000
    2500
    2000
    1500
    1000
     500
            10
11
           ^   J-U
12
1
            10     11
       12
             234

                     Month
0


1


2  ^
                                                                        3   5
                                                                        4   £
                                                                           c
                                                                           o

                                                                        5   S
   Figure 62. Seasonal medians and ranges at USGS 09498500 Salt River near Roosevelt, AZ -
             calibration period (SWAT).
Table 27.Seasonal summary at USGS 09498500 Salt River near Roosevelt, AZ - calibration
         period (SWAT)
MONTH

Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
256.42
450.48
507.61
1831.48
1511.59
1875.51
1429.55
768.86
277.08
226.12
377.92
368.88
200.50
243.00
267.50
267.00
455.50
1080.00
953.50
535.00
187.50
185.50
267.50
249.50
168.25
195.00
210.00
204.00
206.00
211.50
224.00
150.25
121.00
129.25
205.50
174.00
264.00
352.75
356.25
413.25
1135.00
2682.50
1722.50
1047.50
335.25
269.50
389.50
402.25
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
608.17
791 .68
1067.82
2253.82
1755.74
1584.27
683.25
324.01
191.56
275.63
628.63
643.08
344.34
218.51
403.31
489.18
642.09
794.67
331.92
236.85
199.95
180.78
216.21
293.31
198.53
204.55
196.21
188.76
204.73
261.51
207.45
184.00
141.41
113.41
140.01
144.25
657.22
804.13
1094.30
1909.69
2372.74
2058.55
998.65
380.12
225.64
207.08
844.51
699.35
                                              E-76

-------
I
ro
Q
            •Observed Flow Duration (10/1/1992 to 9/30/2002 )
             Modeled Flow Duration (10/1/1992 to 9/30/2002 )
    100000
     10000
      1000
       100 - =
                  10%
                          20%     30%      40%     50%     60%      70%     80%

                               Percent of Time that Flow is Equaled or Exceeded
                                                                                    90%
                                                                                            100%
  Figure 63. Flow exceedence at USGS 09498500 Salt River near Roosevelt, AZ - calibration period
            (SWAT).
  o
  o
  ro
  T3
  (D
  
-------
Table 28. Summary statistics at USGS 09498500 Salt River near Roosevelt, AZ - calibration
        period (SWAT)
SWAT Simulated Flow
REACH OUTFLOW FROM OUTLET 3
10-Year Analysis F^riod: 10/1/1992 - 9/30/2002
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

_Jotak3fj3irrujlatejdjT^^
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9)
Simulated Fall Flow Volume (months 10-12^
Simulated Winter Flow Volume (months 1-3^:
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring;
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of Efficiency, E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
2.83

1.63
0.27

0.41
0.65
1.46
0.31

1.09
0.18
Error Statistics
9.43
-7.16
4.52
58.85
Observed Flow Gage
USGS 09498500 SALT RIVER NEAR ROOSEVELT, AZ
Hydrologic Unit Code: 15060103
Latitude: 33.61 94949
Longitude: -110.9215037
Drainage Area (sq-rri): 4306
Total Observed In-stream Flow:

Total of Observed higjiesMO%Jows:_^
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow VolumeJ10-12):
Observed Winter Flow VolumeJ1-3):
Observed Spring Flow Volume (4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
103.52 » | 30
6.95
-51.64
16.40
103.32
0.222
0.314
0.783
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0
Run (n-1)
2.65
3.27
-3.14
73.05
2.59

1.56
0.29

0.26
0.32
1.36
0.65

0.94
0.09
Run (n-2)
5.65
13.54
-2.75
86.02
77.83 Clear [ 91.08
0.17
-57.34
5.26
110.10
0.363
0.335
0.25
-57.27
5.60
112.10
0.360
0.317

                                         E-78

-------
Table 29. Summary statistics at all stations - validation period WY 1982-1992 (SWAT)
Station
Error in total
volume:
Error in 50%
lowest flows:
Error in 10%
highest flows:
Seasonal
volume error -
Summer:
Seasonal
volume error -
Fall:
Seasonal
volume error -
Winter:
Seasonal
volume error -
Spring:
Error in storm
volumes:
Error in
summer storm
volumes:
Daily Nash-
Sutcliffe
Coefficient of
Efficiency, E:
Monthly Nash-
Sutcliffe
Coefficient of
Efficiency, E:
09498500 Salt
River nr
Roosevelt
-1.76
-18.12
-2.86
58.01
54.85
-0.60
-60.95
6.42
99.23
-0.072
0.446
09503700
Verde River
nr Paulden
38.73
-2.09
78.10
50.18
67.27
31.00
-1.16
108.63
107.87
-2.028
-0.034
09504000
Verde River
nr Clarkdale
5.68
5.17
-9.70
68.04
27.70
-25.96
-11.56
12.44
197.67
-0.996
0.320
09504500
Oak Creek nr
Cornville
15.80
13.08
-3.77
197.59
24.82
-16.44
-8.90
18.06
370.81
-0.061
0.395
09505800 W
ClearCr nr
Camp Verde
-0.57
-28.71
-0.02
18.18
-10.18
11.25
-33.19
34.57
-9.73
-1.666
-0.161
09507980 E
Verde River
nrChilds
23.55
1.12
53.33
141.05
45.31
-6.66
-18.58
91.61
310.39
-1.193
0.736
09508500
Verde R
below Tangle
Cr
7.96
11.91
-2.37
157.67
17.43
-17.53
-30.60
19.33
332.99
-0.191
0.441
Water Quality Calibration and Validation
The 20 Watershed models are designed to provide water quality simulation for total suspended solids (TSS), total
nitrogen, and total phosphorus. The water quality calibration focuses on the replication of monthly loads, as
specified in the project QAPP. Given the simplified approach to water quality simulation in the 20 Watershed
model a close match to individual concentration observations cannot be expected. However, comparison to
monthly loads presents challenges, as monthly loads are not observed. Instead, monthly loads must be estimated
from scattered concentration grab samples and continuous flow records. Such estimation is fraught with
uncertainty because it depends on the degree and form in which concentration and flow are correlated with one
another. Further, the bulk of the load of sediment and sediment-associated phosphorus is likely to move through
the system in a limited number of high flow events that typically are not monitored. As a result, the monthly load
calibration is inevitably based on the comparison of two uncertain numbers. Nonetheless, calibration is able to
achieve a fair agreement. The load comparisons were supported by detailed examinations of the relationships of
                                                E-79

-------
flows to loads and concentrations and the distribution of concentration prediction errors versus flow, time, and
season, as well as standard time series plots.

For application on a nationwide basis, the 20 Watershed protocols assume that sediment and phosphorus loads
will likely exhibit a strong positive correlation to flow (and associated erosive processes), while total nitrogen
loads, which often have a dominant groundwater component, will not. Accordingly, sediment and phosphorus
loads were estimated from observations using a flow-stratified log-log regression approach, while total nitrogen
loads were estimated using a flow-stratified averaging estimator, consistent with the findings of Preston et al.
(1989).

Initial calibration and validation of water quality was done on the Verde River near Clarkdale, using 1993-2002
for calibration and  1986-1992 for validation. As with hydrology, water quality calibration was performed on the
later period as this belter reflects the land use included in the model. The start of the validation period was
constrained by data availability.

Sediment concentrations in larger, higher-order alluvial streams are largely determined by channel scour,
deposition, and transport capacity rather than by upland sediment load. The SWAT representation of these
channel processes is rather simplistic. First, the maximum transport capacity concentration (Cmx) is determined as
C^ = SPCON • VpkSPEXP, where Vpk is the peak velocity, estimated by a simple ratio to the  average rate of flow,
and SPCON and SPEXP are user-defined parameters. When the predicted sediment concentration in the reach
exceeds Cmx the excess is assumed to  settle out. If the predicted sediment concentration in the reach is less than
C^, additional sediment may be scoured from the channel to make up the difference, depending on  the channel
erodibility factor, Kch (cm/hr/Pa). There is no provision for the different transport characteristics of different
sediment size fractions. Further, the SPCON and SPEXP parameters are specified at the global level and do not
vary by reach. As a result, the ability of SWAT to match individual TSS observations is rather limited.

By judicious adjustment of the SPCON, SPEXP, and Kch parameters combined with the SWAT default MUSLE
representation of upland sediment yield a reasonable representation of sediment load can be obtained. Time series
of simulated and estimated sediment loads at the Clarkdale station for both periods  are shown in Figure 65 and
statistics for the two periods are provided separately in Table 30. The key statistic in the table is the  relative
percent error, which shows the error in the prediction of monthly load normalized to the estimated load. The table
also shows the relative average absolute error, which is the average of the relative magnitude of errors in
individual monthly load predictions. This number is inflated by outlier months in which the simulated and
estimated loads differ by large amounts (which may be as easily due to uncertainty  in the estimated  load due to
limited data as to problems with the model) and the third statistic, the relative median absolute error, is likely
more relevant and shows good agreement.
                                                 E-80

-------
                                               TSS
     1,000,000
                                                                                   -Regression Loads
                                                                                   -Simulated Loads
    Figure 65. Fit for monthly load of TSS at USGS 09504000 Verde River near Clarkdale, AZ (SWAT).
Table 30. Model fit statistics (observed minus predicted) for monthly sediment loads using
         stratified regression at USGS 09504000 Verde River near Clarkdale, AZ (SWAT)
Statistic
Relative Percent Error
Relative Average Absolute Error
Relative Median Absolute Error
Calibration period
(1993-2002)
16.9%
64%
0.5%
Validation period
(1986-1992)
-42.6%
122%
0.8%
A variety of other diagnostics were also examined to evaluate agreement between the model and observations.
These are available in full in the calibration spreadsheets, but a few examples are provided below. First, load-flow
power plots were compared for individual days (Figures 66 and 67). These show that the relationship between
flow and load is reasonably consistent across the entire range of observed flows, for both the calibration and
validation periods.
                                               E-81

-------
                         Verde River - Clarkdale

                                 1993-2002
•o
«
c
o
•o
ra
o


(0
(0
                        10
                              100        1000


                                 Flow, cfs
     10000
100000
             •  Simulated  A Observed
                                Row er (Simulated)
   Fbwer (Observed)
Figure 66. Power plot for observed and simulated TSS at USGS 09504000 Verde River near Clarkdale,

         AZ - calibration period (SWAT).
                         Verde River - Clarkdale

                                1986-1992
1
•a
ra
o
w
   100000
    10000
1000
      1 00
                                                                    10000
             Simulated  A  Observed
                              Fbwer (Simulated)
Power (Observed)
Figure 67. Power plot for observed and simulated TSS at USGS 09504000 Verde River near Clarkdale,

         AZ - validation period (SWAT).
                                          E-82

-------
Standard time series plots (Figure 68) show that observed and simulated concentrations achieve at best a fair
agreement, and the model may deviate substantially from individual observations. However, statistics on
concentration (0) show that low median errors are achieved (Table 31).
                    Verde River - Clarkdale
                            1993-2002
• Simulated  A  Observed
       10000
        1000
         100
   (0
   (0
            1993  1994  1995   1996   1997   1998  1999  2000   2001  2002
                                            Year
    Figure 68. Time series plot of TSS concentration at USGS 09504000 Verde River near Clarkdale, AZ
             (SWAT).
Table 31. Relative errors (observed minus predicted) for TSS concentration at USGS 09504000
         Verde River near Clarkdale, AZ (SWAT)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1993-2002)
47
-22%
3.3%
Validation period
(1986-1992)
62
32%
9.3%
For simulation of total phosphorus and total nitrogen, calibration was advanced primarily through adjustment of
the PPERCO and NPERCO coefficients. Monthly loading series for total phosphorus are shown in Figure 69 and
load statistics are summarized in Table 32. The model reproduces the general trend in monthly loads, but is
significantly lower than the peak loads predicted by the regression method, resulting in high relative percent
errors for both the calibration and validation periods. It should be noted that the available data are limited,
particularly for high flow events. Thus, the estimates of "observed" load are also subject to considerable
uncertainty.
                                               E-83

-------
                                              Total P
     10000
                                                                                    Regression Loads
                                                                                    Simulated Loads
                                                                    _  8  8  _
                                                                    CM  CM  CM  CM
    Figure 69. Fit for monthly load of total phosphorus at USGS 09504000 Verde River near Clarkdale, AZ
             (SWAT).
Table 32. Model fit statistics (observed minus predicted) for monthly total phosphorus loads
         using stratified regression (SWAT)
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1993-2002)
83.5%
93%
0.3%
Validation period
(1986-1992)
31.4%
83%
13.1%
As with TSS, additional diagnostics for total phosphorus included flow-load power plots (Figures 70 and 71),
time series plots (Figure 72) and analysis of concentration errors (Table 33). While these show approximate
agreement, the model often overpredicts total phosphorus concentrations under lower flow conditions, although
not on a consistent basis.
                                               E-84

-------
                         Verde River - Clarkdale

                                1993-2002
      100
ra
•o

«
c
o
•o
ra
o
    0.001
   0.0001
                     10
100        1000


    Flow, cfs
     10000      100000
           • Simulated  A Observed ^^™Power (Simulated) ^^™Power (Observed)
Figure 70. Power Plot for Observed and Simulated total phosphorus at USGS 09504000 Verde River

         near Clarkdale, AZ - calibration period (SWAT).
                         Verde River - Clarkdale

                                1986-1992
      100
                                                                    10000
          •  Simulated A Observed
 Power (Simulated)
Power (Observed)
Figure 71. Power plot for observed and simulated total phosphorus at USGS 09504000 Verde River

         near Clarkdale, AZ - validation period (SWAT).
                                          E-85

-------
                    Verde River - Clarkdale
                           1993-2002
• Simulated  A  Observed
        1993  1994   1995   1996   1997   1998  1999   2000   2001   2002
                                         Year
    Figure 72. Time series plot of total phosphorus concentration at USGS 09504000 Verde River near
            Clarkdale, AZ (SWAT).
Table 33. Relative errors (observed minus predicted) for total phosphorus concentration at
         USGS 09504000 Verde River near Clarkdale, AZ (SWAT)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1993-2002)
57
-11%
-33%
Validation period
(1986-1992)
75
31%
-17%
For total nitrogen, fewer data are available because many sampling events omitted one or more nitrogen species.
This increases the uncertainty of the comparison. Results for total nitrogen are summarized in Figures 73 through
76 and Tables 34 and 35, following the same format as total phosphorus. The loading results are acceptable, and
generally better than those obtained for total phosphorus. However, there is significant uncertainty in the
prediction of individual nitrogen observations.
                                              E-86

-------
                                              Total N
     120
                                                                                     Averaging Loads
                                                                                     Simulated Loads
         COCOCOCOCOCO
         O5O5O5O5O5O5O5O5O5O5O5O5O5O5O5O5O5O5O5
0000
O  O  O  O
CM  CM  CM  CM
    Figure 73. Fit for monthly load of total nitrogen at USGS 09504000 Verde River near Clarkdale, AZ
             (SWAT).
Table 34. Model fit statistics (observed minus predicted) for monthly total nitrogen loads using
         averaging estimator (SWAT)
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1993-2002)
-14.4%
84%
13%
Validation period
(1986-1992)
-15.9%
67%
48%
                                               E-87

-------
                         Verde River - Clarkdale

                                1993-2002
    100
ra
•o

«
c
o
•o
ra
o
    0.1
   0.01
                    10
100         1000


    Flow, cfs
10000
100000
          •  Simulated  A Observed ^^™Power (Simulated) ^^™Power (Observed)
Figure 74. Power plot for observed and simulated total nitrogen at USGS 09504000 Verde River near

         Clarkdale, AZ - calibration period (SWAT).
                         Verde River - Clarkdale

                                1986-1992
   0.01
                                                                     10000
         •  Simulated  A  Observed ^^™Power (Simulated) ^^"Fbwer (Observed)
Figure 75. Power plot for observed and simulated total nitrogen at USGS 09504000 Verde River near

         Clarkdale, AZ - validation period (SWAT).
                                          E-88

-------
      2.5 -i
                   Verde River - Clarkdale
                           1993-2002
• Simulated  A  Observed
        1993   1994   1995  1996  1997   1998   1999  2000  2001   2002
                                         Year
   Figure 76. Time series plot of total nitrogen concentration at USGS 09504000 Verde River near
            Clarkdale, AZ (SWAT).
Table 35. Relative errors (observed minus predicted) for total nitrogen concentration at USGS
         09504000 Verde River near Clarkdale, AZ (SWAT)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1993-2002)
46
-83%
-90%
Validation period
(1986-1992)
75
1.3%
-18%
Water Quality Results for Larger Watershed
Summary statistics for the water quality calibration and validation at other stations in the watershed are provided
in Tables 36 and 37, respectively. In most cases, nitrogen loads are better predicted than phosphorus and TSS
loads. In a majority of cases simulated TSS and total phosphorus loads are lower than those estimated from
observations, but this may reflect in part the uncertainty in the regression-based load estimates as water quality
observations during high flows are sparse.
                                             E-89

-------
Table 36. Summary statistics (observed minus predicted) for water quality at all stations
        calibration period 1993-2002 (SWAT)
Station
Relative Percent Error TSS
Load
TSS Concentration Median
Percent Error
Relative Percent Error TP
Load
TP Concentration Median
Percent Error
Relative Percent Error TN
Load
TN Concentration Median
Percent Error
09498500 Salt River
nr Roosevelt
41.1
-15
61.0
-2
9.5
-4
09504000 Verde River
nrClarkdale
16.9
3.4
83.5
-33
-14.4
-90
09508500 Verde River
below Tangle Cr
17.7
2.0
33.8
-19.9
17.0
-60
Table 37. Summary statistics (observed minus predicted) for water quality at all stations
        validation period 1986-1992 (SWAT)
Station
Relative Percent Error TSS
Load
TSS Concentration Median
Percent Error
Relative Percent Error TP
Load
TP Concentration Median
Percent Error
Relative Percent Error TN
Load
TN Concentration Median
Percent Error
09498500 Salt River
nr Roosevelt
-0.6
-16
54.2
-18
18.0
-36
09504000 Verde River
nrClarkdale
-42.6
9.3
31.4
-17
-15.9
-18
09508500 Verde River
below Tangle Cr
-55.1
0.35
-41.4
-5.8
-20.3
-13
                                          E-90

-------
a a a  a a a a a a a  a a a a a a a  a a a a a a a  a a a a a a a a a a a a a a a a a a a a a a  a a a a a a a  a a a a a a a  a a a a a a a  a a a a a a a  a a a a a a a a a a a a a a
ADWR (Arizona Department of Water Resources). 2000. Verde River Watershed Study. Arizona Department of
Water Resources. Phoenix, AZ. https://portal.azoah.com/oedf/documents/08A-AWS001-
DWR/Omnia/200004%20DWR%20Verde%20River%20Watershed%20Report.pdf

ADWR (Arizona Department of Water Resources). 2009. Arizona Water Atlas. Arizona Department of Water
Resources. Phoenix, AZ. Published online at
http://www.azwater.gov/AzDWR/StatewidePlanningAVaterAtlas/default.htm.

Anderson, J. R., E.E. Hardy, J.T. Roach, and R.E. Witmer. 1976. A Land Use and Land Cover Classification
System for Use with Remote Sensor Data. Professional Paper 964, U.S. Geological Survey. Reston, VA.

Cordy, G.E., D.J. Gellenbeck, J.B. Gebler, D.W. Anning, A.L. Goes, R.J. Edmonds, J. A.H. Rees, and H.W.
Sanger. 2000. Water Quality in the Arizona Basins, Arizona, 1995-98. U.S. Geological Survey Circular 1213.
Reston, VA.

Owen-Joyce, S.J., and C. K. Bell. 1983. Appraisal of Water Resources in the Upper Verde River Area: Yapai and
Coconino Counties, Arizona. Arizona Department of Water Resources Bulletin 2. Prepared by U.S. Geological
Survey. Phoenix, Arizona.

Preston, S.D., V.J. Bierman, Jr., and S.E. Silliman. 1989. An evaluation of methods for the estimation of tributary
mass loads. Water Resources Research, 25(6): 1379-1389.

Schwab, K.J. 1995. Maps Showing Groundwater Conditions in the Big Chino Sub-Basin of the Verde River
Basin, Coconino and Yavapai Counties, Arizona—1992. Arizona Department of Water Resources. Phoenix,
Arizona.

SCS (Soil Conservation Service). 1986. Urban Hydrology for Small Watersheds. Technical Release 55 (2nd
edition). U.S. Department of Agriculture. Washington, DC.

Tetra Tech. 2001. Verde River Assimilative Capacity  Study, Modeling and Assimilative Capacity Report.
Prepared for Arizona Department of Environmental Quality. Tetra Tech, Inc. Research Triangle Park, NC.

USEPA (United States Environmental Protection Agency). 2008. Using the BASINS Meteorological Database
(Version 2006). BASINS Technical Note 10. Office of Water, U.S. Environmental Protection Agency.
Washington, DC. http://water.epa.gov/scitech/datait/models/basins/upload/2009_04_13_BASINSs_tecnotel0.pdf
(Accessed June, 2009).

Wallace, B.L. and R. L. Laney. 1976. Maps Showing Ground-water Conditions in Lower Big Chino Valley and
Williamson Valley Areas, Yavapai and Coconino Counties, Arizona, 1975 - 1976. Water Resources
Investigations 76 - 78. U.S. Geological Survey. Washington, DC.
                                                E-91

-------
Appendix F
Model Configuration, Calibration and
Validation

Basin: Susquehanna River
                  F-l

-------
Contents
Watershed Background	F-7
  Water Body Characteristics	F-7
  Soil Characteristics	F-10
  Land Use Representation	F-10
  Point Sources	F-14
  Meteorological Data	F-20
  Watershed Segmentation	F-24
  Calibration Data and Locations	F-26
HSPF Modeling	F-27
  Changes Made to Base Data Provided	F-28
  Assumptions	F-28
  Hydrology Calibration	F-28
  Hydrology Validation	F-34
  Hydrology Results for Larger Watershed	F-39
  Water Quality Calibration and Validation	F-46
  Water Quality Results for Larger Watershed	F-55
SWAT Modeling	F-57
  Changes Made to Base Data Provided	F-57
  Assumptions	F-57
  Hydrology Calibration	F-57
  Hydrology Validation	F-63
  Hydrology Results for Larger Watershed	F-68
  Water Quality Calibration and Validation	F-75
  Water Quality Results for Larger Watershed	F-84
References	F-86
                                               F-2

-------
Tables
Table 1.    Aggregation ofNLCD land cover classes	F-ll
Table 2.    Land use distribution for the Susquehanna River watershed (2001 NLCD, mi2)	F-13
Table 3.    Major point source discharges in the Susquehanna River watershed	F-14
Table 4.    Precipitation stations for the Susquehanna River watershed model	F-20
Table 5.    Calibration and validation locations in the Susquehanna River watershed	F-26
Table 6.    Seasonal summary at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
           calibration period (HSPF)	F-32
           Figure 12. Flow accumulation at USGS 01562000 Raystown Branch Juniata River at Saxton,
           PA - calibration period (HSPF)	F-33
Table 7.    Summary statistics at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
           calibration period (HSPF)	F-34
Table 8.    Seasonal summary at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
           validation period (HSPF)	F-37
Table 9.    Summary statistics at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
           validation period (HSPF)	F-39
Table 10.   Seasonal summary at USGS 01576000 Susquehanna River at Marietta, PA - calibration
           period (HSPF)	F-42
Table 11.   Summary statistics at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
           (HSPF)	F-44
Table 12.   Summary statistics (percent error) for all stations - calibration period (HSPF)	F-45
Table 13.   Summary statistics (percent error) for all stations - validation period (HSPF)	F-46
Table 14.   Model fit statistics (observed minus predicted) for monthly sediment loads using stratified
           regression - USGS 01562000 Raystown Branch Juniata River at Saxton, PA (HSPF)	F-47
Table 15.   Relative errors,(observed minus simulated) for TSS concentrations at USGS 01562000
           Raystown Branch Juniata River at Saxton, PA (HSPF)	F-49
Table 16.   Model fit statistics (observed minus predicted) for monthly total phosphorus loads using
           stratified regression - USGS 01562000 Raystown  Branch Juniata River at Saxton, PA (HSPF).. F-50
Table 17.   Relative errors (observed minus simulated) for total phosphorus concentrations at USGS
           01562000 Raystown Branch Juniata River at Saxton, PA (HSPF)	F-52
Table 18.   Model fit statistics (observed minus predicted) for monthly total nitrogen loads using averaging
           estimator - USGS 01562000 Raystown Branch Juniata River at Saxton, PA (HSPF)	F-53
Table 19.   Relative errors (observed minus simulated) for total nitrogen concentration at USGS 01562000
           Raystown Branch Juniata River at Saxton, PA (HSPF)	F-55
Table 20.   Summary statistics for water quality for all stations - calibration period (HSPF)	F-5 6
Table 21.   Summary statistics for water quality for all stations - validation period (HSPF)	F-5 6
Table 22.   Seasonal summary at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
           calibration period (SWAT)	F-61
Table 23.   Summary statistics at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
           calibration period (SWAT)	F-63
Table 24.   Seasonal summary at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
           validation period (SWAT)	F-66
Table 25.   Summary statistics at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
           validation period (SWAT)	F-68
Table 26.   Summary statistics (percent error) for all stations - calibration period (SWAT)	F-69
Table 27.   Seasonal summary at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
           (SWAT)	F-72
Table 28.   Summary statistics at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
           (SWAT)	F-74
Table 29.   Summary statistics (percent error) for all stations - validation period (SWAT)	F-75
Table 30.   Model fit statistics (observed minus predicted) for monthly sediment loads using stratified
           regression - USGS 01562000 Raystown Branch Juniata River at Saxton, PA (SWAT)	F-76

                                                ~

-------
Table 31.   Relative errors (observed minus simulated) for TS S concentrations at USGS 015 62000 Raystown
           Branch Juniata River at Saxton, PA (SWAT)	 F-78
Table 32.   Model fit statistics (observed minus predicted) for monthly phosphorus loads using stratified
           regression - USGS 01562000 Raystown Branch Juniata River at Saxton, PA (SWAT)	F-79
Table 33.   Relative errors (observed minus predicted), total phosphorus concentration, USGS 01562000
           Raystown Branch Juniata River at Saxton, PA (SWAT)	F-81
Table 34.   Model fit statistics (observed minus predicted) for monthly total nitrogen loads using averaging
           estimator - USGS 01562000 Raystown Branch Juniata River at Saxton, PA (SWAT)	F-82
Table 35.   Relative errors (observed minus predicted), total nitrogen concentration, USGS 01562000
           Raystown Branch Juniata River at Saxton, PA (SWAT)	F-84
Table 36.   Summary statistics for water quality for all stations - calibration period (SWAT)	F-85
Table 37.   Summary statistics for water quality for all stations - validation period (SWAT)	F-85


Figures
Figure 1.  Location of the  Susquehanna River watershed	F-9
Figure 2.  Land use in the  Susquehanna River watershed	F-12
Figure 3.  Major point sources in the Susquehanna River watershed	F-19
Figure 4.  Weather stations forthe Susquehanna River watershed  model	F-23
Figure 5.  Model segmentation and USGS stations utilized for the Susquehanna River watershed	F-25
Figure 6.  Mean daily flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - calibration
          period (HSPF)	F-30
Figure 7.  Mean monthly flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
          calibration period (HSPF)	F-30
Figure 8.  Monthly flow regression and temporal variation at USGS 01562000 Raystown Branch Juniata River at
          Saxton, PA - calibration period (HSPF)	F-31
Figure 9.  Seasonal regression and temporal aggregate at USGS 01562000 Raystown Branch Juniata River at
          Saxton, PA - calibration period (HSPF)	F-31
Figure 10. Seasonal medians and ranges at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
          calibration period (HSPF)	F-32
Figure 11. Flow exceedence at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - calibration
          period (HSPF)	F-33
Figure 13. Mean daily flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - validation
          period (HSPF)	F-35
Figure 14. Mean monthly flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - validation
          period (HSPF)	F-35
Figure 15. Monthly flow regression and temporal variation at USGS 01562000 Raystown Branch Juniata
          River at Saxton, PA - validation period (HSPF)	F-36
Figure 16. Seasonal regression and temporal aggregate at USGS 01562000 Raystown Branch Juniata River at
          Saxton, PA - validation period (HSPF)	F-36
Figure 17. Seasonal medians and ranges at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
          validation period (HSPF)	F-37
Figure 18. Flow exceedence at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - validation
          period (HSPF)	F-38
Figure 19. Flow accumulation at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - validation
          period (HSPF)	F-38
Figure 20. Mean daily flow at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
          (HSPF)	F-40
Figure 21. Mean monthly flow at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
          (HSPF)	F-40
                                                F-4

-------
Figure 22. Monthly flow regression and temporal variation at USGS 01576000 Susquehanna River at Marietta,
          PA - calibration period (HSPF)	F-41
Figure 23. Seasonal regression and temporal aggregate at USGS 01576000 Susquehanna River at Marietta,
          PA - calibration period (HSPF)	F-41
Figure 24. Seasonal medians and ranges at USGS 01576000 Susquehanna River at Marietta, PA - calibration
          period (HSPF)	F-42
Figure 25. Flow exceedence at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
          (HSPF)	F-43
Figure 26. Flow accumulation at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
          (HSPF)	F-43
Figure 27. Fit for monthly load of TSS at Raystown Branch Juniata River at Saxton, PA (HSPF)	F-47
Figure 28. Power plot for observed and simulated TSS at Raystown Branch Juniata River at Saxton, PA -
          calibration period (HSPF)	F-48
Figure 29. Power plot for observed and simulated TSS at Raystown Branch Juniata River at Saxton, PA -
          validation period (HSPF)	F-48
Figure 30. Time series plot of TSS concentration at Raystown Branch Juniata River at Saxton, PA - calibration
          period (HSPF)	F-49
Figure 31. Fit for monthly load of total phosphorus at Raystown Branch Juniata River at Saxton, PA
          (HSPF)	F-50
Figure 32. Power plot for observed and simulated total phosphorus at Raystown Branch Juniata River at
          Saxton, PA - calibration period (HSPF)	F-51
Figure 33. Power plot for observed and simulated total phosphorus at Raystown Branch Juniata River at
          Saxton, PA - validation period (HSPF)	F-51
Figure 34. Time series plot of total phosphorus concentration, at Raystown Branch Juniata River at Saxton,
          PA (HSPF)	F-52
Figure 35. Fit for monthly load of total nitrogen at Raystown Branch Juniata River at Saxton, PA (HSPF)	F-53
Figure 36. Power plot for observed and simulated total nitrogen at Raystown Branch Juniata River at Saxton,
          PA - calibration period (HSPF)	F-54
Figure 37. Power plot for observed and simulated total nitrogen at Raystown Branch Juniata River at Saxton,
          PA - validation period (HSPF)	F-54
Figure 38. Time series plot of total nitrogen concentration, at Raystown Branch Juniata River at Saxton, PA
          (HSPF)	F-55
Figure 39. Mean daily flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - calibration
          period (SWAT)	F-59
Figure 40. Mean monthly flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - calibration
          period (SWAT)	F-59
Figure 41. Monthly flow regression and temporal variation at USGS 01562000 Raystown Branch Juniata River
          at Saxton, PA - calibration period (SWAT)	F-60
Figure 42. Seasonal regression and temporal aggregate at USGS 01562000 Raystown Branch Juniata River at
          Saxton, PA - calibration period (SWAT)	F-60
Figure 43. Seasonal medians and ranges at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
          calibration period (SWAT)	F-61
Figure 44. Flow exceedence at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - calibration
          period (SWAT)	F-62
Figure 45. Flow accumulation at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - calibration
          period (SWAT)	F-62
Figure 46. Mean daily flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - validation
          period (SWAT)	F-64
Figure 47. Mean monthly flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - validation
          period (SWAT)	F-64
Figure 48. Monthly flow regression and temporal variation at USGS 01562000 Raystown Branch Juniata
          River at Saxton, PA - validation period (SWAT)	F-65


                                                ~

-------
Figure 49. Seasonal regression and temporal aggregate at USGS 01562000 Raystown Branch Juniata River at
          Saxton, PA - validation period (SWAT)	F-65
Figure 50. Seasonal medians and ranges at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
          validation period (SWAT)	F-66
Figure 51. Flow exceedence at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - validation
          period (SWAT)	F-67
Figure 52. Flow accumulation at USGS 01562000 Raystown Branch Juniata River at Saxton, PA - validation
          period (SWAT)	F-67
Figure 53. Mean daily flow at USGS 01576000 Susquehanna River At Marietta, PA - calibration period
          (SWAT)	F-69
Figure 54. Mean monthly flow at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
          (SWAT)	F-70
Figure 55. Monthly flow regression and temporal variation at USGS 01576000 Susquehanna River at Marietta,
          PA - calibration period (SWAT)	F-70
Figure 56. Seasonal regression and temporal aggregate at USGS 01576000 Susquehanna River at Marietta,
          PA - calibration period (SWAT)	F-71
Figure 57. Seasonal medians and ranges at USGS 01576000 Susquehanna River at Marietta, PA - calibration
          period (SWAT)	F-71
Figure 58. Flow exceedence at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
          (SWAT)	F-72
Figure 59. Flow accumulation at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
          (SWAT)	F-73
Figure 60. Fit for monthly load of TSS at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
          (SWAT)	F-76
Figure 61. Power plot for observed and simulated TSS at USGS 01562000 Raystown Branch Juniata River at
          Saxton, PA - calibration period (SWAT)	F-77
Figure 62. Power plot for observed and simulated TSS at USGS 01562000 Raystown Branch Juniata River at
          Saxton, PA - validation period (SWAT)	F-77
Figure 63. Correlation between observed and predicted TSS concentration at USGS 01562000 Raystown
          Branch Juniata River at Saxton, PA (SWAT)	F-78
Figure 64. Fit for monthly load of total phosphorous at USGS 01562000 Raystown Branch Juniata River at
          Saxton, PA (SWAT)	F-79
Figure 65. Power plot for observed and simulated total phosphorus at USGS 01562000 Raystown Branch
          Juniata River at Saxton,  PA - calibration period (SWAT)	F-80
Figure 66. Power plot for observed and simulated total phosphorus at USGS 01562000 Raystown Branch
          Juniata River at Saxton,  PA - validation period (SWAT)	F-80
Figure 67. Time series plot of total  phosphorus concentration at USGS 01562000 Raystown Branch Juniata
          River at Saxton, PA (SWAT)	F-81
Figure 68. Fit for monthly load of total nitrogen at USGS 01562000 Raystown Branch Juniata River at
          Saxton, PA (SWAT)	F-82
Figure 69. Power plot for observed and simulated total nitrogen at USGS 01562000 Raystown Branch Juniata
          River at Saxton, PA - calibration period (SWAT)	F-83
Figure 70. Power plot for observed and simulated total nitrogen at USGS 01562000 Raystown Branch Juniata
          River at Saxton, PA - validation period (SWAT)	F-83
Figure 71. Time series plot of total  nitrogen concentration at USGS 01562000 Raystown Branch  Juniata
          River at Saxton, PA (SWAT)	F-84
                                               F-6

-------

Water Body Characteristics
The Susquehanna River drains about 27,500 mi2 in the states of New York, Pennsylvania, and Maryland and
includes a total of 19 HUCSs in HUC 2050 (Figure 1). The watershed makes up 43 percent of the Chesapeake
Bay's drainage area and consists of six major subwatersheds (Chemung, Upper Susquehanna, West Branch
Susquehanna, Middle Susquehanna, Juniata, and Lower Susquehanna). The Susquehanna River flows about 444
miles from its headwaters at Otsego Lake in Cooperstown, New York to Havre de Grace, Maryland, where the
river flows into the Chesapeake Bay. The river is the largest tributary to the Chesapeake Bay, providing 50
percent of its freshwater flows (SRBC 2008).

The Susquehanna River watershed includes three physiographic provinces: the Appalachian Plateau, the Valley
and Ridge, and the Piedmont Provinces. The Appalachian Plateau Province is characterized by high, flat-topped
hills and deep valleys cut by the Susquehanna River and its tributaries. The Valley and Ridge physiographic
province contains steep mountains and ridges separated by valleys. The Piedmont physiographic province consists
of uplands and lowlands. The Piedmont physiographic province generally has terrain that is gently rolling to hilly.

Sixty-nine percent of the watershed is forested. However, the well-drained areas with rolling hills and valleys in
the southern part of the watershed contain most of the population and some of the most productive agricultural
land in the US. Groundwater maintains the base flow of perennial streams during periods of little or no
precipitation and constitutes an average of 50 percent of the flow of most streams at other times.

The U.S. Army Corps of Engineers (USAGE) operates and maintains 13 dams and reservoirs that are located in
all six major subwatersheds. USAGE also regulates the operation of a state of Pennsylvania reservoir (George B.
Stevenson) in the West Branch Susquehanna subwatershed for the purpose of flood damage reduction. These 14
reservoirs provide most of the floodwater storage in the watershed. The Natural Resources Conservation Service
and the state of Pennsylvania have also constructed reservoirs in the watershed that reduce flood damages;
however, these reservoirs are typically smaller in scale than the USAGE reservoirs.

In addition to the many flood storage dams and reservoirs, there are 20 major electric power generating plants
located in the  Susquehanna River watershed that use water resources in their operation. Many of these
hydroelectric dams are located in the lower Susquehanna watershed. Just below Harrisburg, Pennsylvania the
Susquehanna River flows through a series of gorges dammed by hydroelectric power facilities. There are also 13
approved water diversions from the Susquehanna River watershed.

The Susquehanna River basin has a continental type of climate. The average annual temperature in the basin
ranges from about 44 degrees in the northern part of the basin to about 53 degrees in the southern part. Average
annual precipitation is about 40 inches over the entire basin and ranges from 33 inches in the northern part of the
basin to 46 inches in the southern part. Virtually all the major streams experience their highest flows in March,
April, and May, when melting snows combine with spring rains. These three months account for about one-half of
the yearly runoff. Flows are lowest in these streams during the summer and early fall months, with most streams
falling to their lowest levels in September. The Susquehanna River basin is one of the country's most flood prone
areas. Generally, floods occur each year somewhere in the basin, and major floods can occur in all seasons of the
year, and a major flood occurs on average every 13 years.

Groundwater flow maintains the base flow of perennial streams  during periods of little or no precipitation and
constitutes an average of 50 percent of the flow of most streams at other times. The use of groundwater resources

                                                 ~

-------
in the basin is extensive. Groundwater plays a critical role in supplying drinking water and maintaining economic
viability. Outside of the major population centers, drinking water supplies are heavily dependent on groundwater
wells. Approximately 20 percent of the basin population is served by public water suppliers that use groundwater
as a source.
                                                  F-8

-------
                             Lake Ontario
                                                                      •Upper Susquehanna
                                                                           (2050101)
                                                    Chenango
                                                    (2050102)
                                                 o-Wappasem
                                                 (2050103)
                           Ghemung
                            2050105}
                                           Upper Susquehanna-Tunkhannoc
                                                      (2050106)
                       me Pennsylvania/^
                         (2050205)
                                         iddle West Branch I
                                           Susquehanna
                                            (2050203)*y   ';
                                                              \ Uppey
                                                       Susquehanna-L&ckawanna
                                                              (2050,ltT7)
Sinnemahomng
   (2050*202)  "    Lower West Branc
                   Susquehanna
                   " (2050206)
                          Bald Eagle
                          (2050204)
                                         Lower Susquehanna
                                              (2050301)
                                                                           Pehrisvlvania
                                ower Juniata
                                 (2050304)
                                                                              Philadelphia
                                     yrSusquehanna-Swatara
                                            '(2050305)
                                                         Lower Susquehanna
                                                             (2050306)
               Raystown
               (2050303)
                                                                                  Hydrography

                                                                                  Water (Nat. Atlas Dasaseti

                                                                                  US Census Populated Places

                                                                                  Municipalities (pop a 50,000)

                                                                                  County Boundaries

                                                                                  Watershed with HUC8s
                                                         „.
                                                     Baltimore
  GCRP Model Areas - Susquehanna River Basin
                 Base Map
Figure 1. Location of the Susquehanna River watershed.
                                                    F-9

-------
Soil Characteristics

One of the most important characteristics of soils for watershed modeling is their hydrologic soil group (HSG).
The 20 Watershed study utilized STATSGO soil survey HSG information during model set-up. Soils are
classified into four hydrologic groups (SCS 1986), separated by runoff potential, as follows:

       Group A Soils Have low runoff potential and high infiltration rates even when thoroughly wetted. They
                     consist chiefly of deep, well to excessively drained sands or gravels and have a high rate
                     of water transmission.

       Group B Soils Have moderate infiltration rates when wet and consist chiefly of soils that are moderately
                     deep to deep, moderately well to well drained, and moderately fine to moderately course
                     textures.
       Group C Soils Have low infiltration rates when thoroughly wetted and consist chiefly of soils having a
                     layer that impedes downward movement of water with moderately fine to fine structure.

       Group D Soils Have high runoff potential, very low infiltration rates and consist chiefly of clay soils
                     with high swelling potential, soils with a permanent water table, soils with a claypan or
                     clay layer at or near the surface and shallow soils over nearly impervious material.


The Susquehanna River watershed contains all four HSGs in the watershed. However, soils in the watershed, as
described in STATSGO soil surveys, fall primarily into HSGs B (moderately high infiltration capacity) and C
(low infiltration capacity). Hydrologic group C soils dominate the northern portion of the watershed while a
mixture of B and C soils dominate the southern portion of the watershed.
Land Use Representation

Land use/cover in the watershed is based on the 2001 National Land Cover Database (NLCD) coverage and is
predominantly forested with some agricultural and developed land (Figure 2). Agriculture and pasture are more
predominant in the downstream, eastern portions of the watershed. Urban development is found throughout the
watershed; however, the major concentration is in the eastern portions of the watershed.

NLCD land cover classes were aggregated according to the scheme shown in Table 1 for representation in the 20
Watershed model, then overlain with the soils HSG grid. Pervious and impervious lands are specified separately
for HSPF, so only one developed pervious class is used, along with an impervious class. HSPF simulates
impervious land areas separately from pervious land. Impervious area distributions were also determined from the
NLCD Urban Impervious data coverage. Specifically, percent impervious area was calculated over the whole
basin for each of the four developed land use classes. These percentages were then used to separate out
impervious land. NLCD impervious area data products are known to underestimate total imperviousness in rural
areas. However, the model properly requires connected impervious area, not total impervious area, and the NLCD
tabulation is assumed to provide a reasonable  approximation of connected impervious area. Different developed
land classes are specified separately in SWAT. In HSPF the WATER, BARREN, DEVPERV, and WETLAND
classes are not subdivided by HSG; SWAT uses the built-in HRU overlay mechanism in the ArcSWAT interface.

The distribution of land use in the watershed is summarized in Table 2.
                                                F-10

-------
Table 1.    Aggregation of NLCD land cover classes
NLCD Class
1 1 Water
12 Perennial ice/snow
21 Developed open space
22 Dev. Low Intensity
23 Dev. Med. Intensity
24 Dev. High Intensity
31 Barren Land
41 Forest
42 Forest
43 Forest
51-52 Shrubland
71-74 Herbaceous Upland
81 Pasture/Hay
82 Cultivated
91 -97 Wetland
98-99 Wetland
Comments
Water surface area
usually accounted for as
reach area






Deciduous
Evergreen
Mixed




Emergent & woody
wetlands
Aquatic bed wetlands (not
emergent)
SWAT class
WATR
WATR
URLD
URMD
URHD
UIDU
SWRN
FRSD
FRSE
FRST
RNGB
RNGE
HAY
AGRR
WETF, WETL,
WETN
WATR
HSPF (after processing)
WATER
BARREN, Assume HSG D
DEVPERV;
IMPERV
BARREN (D)
FOREST (A,B,C,D)
SHRUB (A,B,C,D)
GRASS (A,B,C,D), BARREN (D)
GRASS(A,B,C,D)
AGRI (A,B,C,D)
WETLAND, Assume HSG D
WATER
                                        F-ll

-------
 Legend
    —  Hydrography
 ^^=  Interstate
 ^H  Water (Nat. Atlas Dataset)
    ^]  County Boundaries
 n  Watershed
 2001 NLCD Land Use
    I  Open water
    ^  Developed, open space
    |  Developed, low intensity
    H  Developed, medium intensity
    I  Developed, high intensity
    ^  Barren land
    H  Deciduous forest
    |  Evergreen forest
 |    |  Mixed forest
    ]  Scrub/shrub
    ^  Grassland/herbaceous
    ^  PastureAiay
    ^  Cultivated crops
    |  Woody wetlands
    ^  Emergent herbaceous wetlands
GCRP Model Areas - Susquenanna River Basin
 Figure 2. Land use in the Susquehanna River watershed.
                                                                 F-12

-------
Table 2.     Land use distribution for the Susquehanna River watershed (2001 NLCD, mi2)
HUC8
watershed
02050101
02050102
02050103
02050104
02050105
02050106
02050107
02050201
02050202
02050203
02050204
02050205
02050206
02050301
02050302
02050303
02050304
02050305
02050306
Total
Open
water
29.7
13.2
7.8
7.3
9.2
24.3
32.4
10.6
1.4
5.6
4.4
2.4
14.1
32.8
3.3
17.3
12.4
33.5
51.5
313.2
Developed3
Open
space
89.0
56.8
45.3
47.4
55.7
88.2
117.5
90.3
12.1
12.8
45.9
16.9
80.6
87.1
54.4
51.5
72.7
139.6
109.8
1,273.6
Low
density
18.1
14.3
18.3
7.1
18.1
9.4
60.5
14.7
1.7
2.3
19.0
2.4
23.6
24.7
26.5
12.6
18.0
113.1
99.7
504.1
Medium
density
5.7
4.8
6.9
2.5
5.9
2.9
42.9
2.5
0.5
0.9
4.5
0.7
8.5
7.8
8.6
2.9
4.1
35.6
40.4
188.6
High
density
1.5
1.1
1.7
0.5
1.2
0.7
12.1
0.5
0.1
0.1
1.5
0.2
2.1
2.4
3.5
0.8
1.1
15.1
16.0
62.5
Barren
land
0.5
1.7
0.5
2.7
1.2
1.9
11.8
30.1
2.5
2.1
1.4
2.1
2.3
9.3
1.8
1.4
0.3
3.5
23.0
100.0
Forest
1,345.3
909.5
659.5
833.2
704.5
1,177.4
1,083.8
1,229.8
921.2
728.9
543.1
816.2
1,170.3
860.0
688.8
655.6
1,019.2
858.3
603.5
16,808.1
Shrubland
73.0
75.2
25.0
39.8
45.4
29.0
23.7
32.9
56.5
18.6
0.0
42.0
33.1
0.0
0.0
0.0
0.0
0.0
0.0
494.4
Pasture/Hay
459.3
319.1
200.8
299.3
196.1
340.7
230.2
147.0
27.9
21.1
64.8
68.0
282.9
219.9
101.9
140.8
185.0
391.9
1,015.3
4,712.1
Cultivated
163.0
146.5
56.5
137.3
152.5
314.0
126.1
36.2
2.8
7.7
70.0
25.9
184.5
201.7
102.2
78.8
137.2
273.5
489.1
2,705.6
Wetland
104.6
65.0
23.7
6.0
22.1
18.1
22.9
1.9
7.3
3.1
0.1
4.3
7.8
3.0
0.1
0.0
0.7
12.3
39.4
342.2
Total
2,289.7
1,607.3
1,046.1
1,383.2
1,211.8
2,006.7
1,764.0
1,596.7
1,033.8
803.2
754.6
981.0
1,809.9
1,448.7
991.1
961.7
1,450.7
1,876.5
2,487.5
27,504.3
aThe percent imperviousness applied to each of the developed land uses is as follows: open space (6.90%), low density (31.26%), medium density (60.90%), and high
density (85.41%).
                                                                F-13

-------
Point Sources
There are numerous point source discharges in the watershed. For the purposes of 20 Watershed modeling, only
the 147 major dischargers with a design flow greater than 1 MGD are included in the simulation (Table 3 and
Figure 3). The major dischargers account for the majority of the facilities, so the effect of the omitted sources
distributed throughout the watershed will be relatively small, except during extreme low flow conditions. The
major dischargers are represented at long-term average flows, without accounting for changes over time or
seasonal variations.
Data from 1991-2006 were compiled from the PCS database and the median total nitrogen, total phosphorus, and
TSS values were estimated. The facilities that were missing a total nitrogen, total phosphorus, and TSS
concentration value were filled with atypical pollutant concentration value from literature  (Tetra Tech 1990)
based on the SIC classification. The median concentrations for the nutrient species were estimated based on the
values reported in the Chesapeake Bay Phase 5 Model documentation (USEPA 2010).
Table 3.     Major point source discharges in the Susquehanna River watershed
NPDES ID
MD0002518
NY0003824
NY0003859
NY0003867
NY0003875
NY0004057
NY0004081
NY0004138
NY0004146
NY0004243
NY0020672
NY0021423
NY0021431
NY0022357
NY0022730
NY0023591
NY0023647
NY0023906
NY0024414
NY0025712
Name
EXELON POWER GENERATION
AEROSPACE OPERATIONS
AES HICKLING, LLC
AES-JENNISON, LLC
AES WESTOVER
SYSTEMS INTEGRATION - OWEGO
MOTOR COMPONENTS, LLC
OSG NORWICH PHARMACEUTICALS
WOODS CORNER PLANT
KERRY BIO-SCIENCE
HAMILTON (V)WPCP
NORWICH (C) WWTP
BATH (V) WWTP
ALFRED (V) WWTP
OWEGO (T) SD#1
COOPERSTOWN (V) STP
HORNELL(C)WPCP
ERWIN (T) STP
BINGHAMTON-JOHNSON (C) JNT STP
PAINTED POST (V) STP
Design flow
(MGD)
47.74
1.35
77.34
65.34
101.90
1.61
0.93
0.80
0.30
2.30
0.85
2.20
1.00
0.98
0.50
0.52
4.00
0.80
20.00
0.50
Observed flow
(MGD)
(1991 -2006 average)
0.00
0.22
1.10
0.03
0.44
0.62
3.00
0.08
0.07
0.39
0.53
2.14
0.74
0.60
0.67
0.65
2.94
0.64
22.20
0.25
                                                F-14

-------
NPDES ID
NY0025721
NY0025798
NY0027561
NY0027669
NY0029262
NY0029271
NY0031151
NY0035742
NY0036986
PA0007498
PA0007919
PA0008231
PA0008265
PA0008281
PA0008303
PA0008419
PA0008443
PA0008451
PA0008508
PA0008575
PA0008869
PA0008885
PA0008923
PA0009024
PA0009164
PA0009202
PA0009229
PA0009253
PA0009270
PA0009733
PA0009920
PAOO 10031
Name
CORNING (C)STP
OWEGO WPCP #2
LE ROY R SUMMERSON WWTF
ENDICOTT (V) WPCP
OWEGO (V) STP
SIDNEY (V) WWTP
ONEONTA (C) WWTP
CHEMUNG CO ELMIRA SD STP
CHEMUNG CO SD#1 STP
WISE FOODS INC
CASCADES TISSUE GROUP - PA INC
GOLD MILLS INC
APPLETON PAPERS INC - SPRING M
PPL BRUNNER ISLAND LLC
ISG STEELTON LLC
MERCK & CO INC
PPL MONTOUR LLC
SUNBURY GENERATION LLC
BURLE BUSINESS PARK LP
WIREROPE WORKS INC
PH GLATFELTER CO
PROCTER & GAMBLE PRODUCTS CO
CORNING ASAHI VIDEO PROD CO
OSRAM SYLVANIA PRODUCTS INC
STANDARD STEEL LLC
CERRO METAL PRODUCTS CO
NORFOLK SOUTHERN RAILWAY CO -
UNITED DEFENSE LP
DEL MONTE CORP
EXELON GENERATION CO LLC - PEA
AMERGEN ENERGY CO LLC - THREE
RELIANT ENERGY MID-ATLANTIC PO
Design flow
(MGD)
2.13
2.00
10.00
10.00
1.00
1.70
4.00
12.00
9.50
0.59
1.25
2.00
4.84
621.00
27.60
12.20
0.46
3.38
0.32
0.07
13.70
7.60
1.97
1.22
1.45
0.23
0.50
0.03
0.67
0.05
81.02
0.01
Observed flow
(MGD)
(1991 -2006 average)
1.56
1.14
6.96
7.48
0.57
0.62
2.47
7.49
7.94
0.24
1.56
17.94
4.45
6.46
28.01
8.70
8.44
6.50
0.11
0.05
28.27
7.68
1.27
0.93
15.56
2.36
0.24
0.03
0.41
0.12
20.07
0.70
F-15

-------
NPDES ID
PAOO 10430
PAOO 13862
PA0020273
PA0020320
PA0020486
PA0020567
PA0020664
PA0020826
PA0020885
PA0020893
PA0020923
PA0021067
PA0021571
PA0021687
PA0021814
PA0021890
PA0022209
PA0022535
PA0023108
PA0023248
PA0023531
PA0023558
PA0023744
PA0024040
PA0024287
PA0024325
PA0024406
PA0024431
PA0024759
PA0024902
PA0025933
PA0026077
Name
HANOVER FOODS CORP
CORIXA CORP
MILTON REGIONAL SEW AUTH
LITITZSEWAUTH
BELLEFONTE BORO
NORTHUMBERLAND SEW AUTH
MIDDLETOWN BORO AUTH
DOVER TWP
MECHANICSBURG BORO
MANHEIM BORO
NEW OXFORD MUN AUTH
MOUNT JOY BORO AUTH
MARYSVILLE BORO COUNCIL
WELLSBORO MUN AUTH
MANSFIELD BORO MUN AUTH
NEW HOLLAND BORO
BEDFORD BORO MUN AUTH
MILLERSBURG AREA AUTH
ELIZABETHTOWN BORO
BERWICK AREA JNT SEW AUTH
DANVILLE BORO
ASHLAND BORO
NORTHEASTERN YORK CO SEW AUTH
HIGHSPIREBORO
PALMYRA BORO STP
MUNCY BORO MUN AUTH
MOUNT CARMEL MUN AUTH
DILLSBURG AREA AUTH
CURWENSVILLE MUN AUTH
UPPER ALLEN TWP BRD OF COMMRS
LOCK HAVEN CITY
CARLISLE BORO
Design flow
(MGD)
0.23
0.32
3.42
3.85
3.22
1.13
2.20
8.00
2.08
1.14
1.79
1.53
1.25
2.00
1.00
1.34
1.50
1.00
3.00
3.70
3.62
1.30
1.70
2.00
1.42
1.40
1.50
1.53
0.75
1.10
3.75
4.63
Observed flow
(MGD)
(1991 -2006 average)
0.18
0.18
78.54
2.62
2.21
11.68
1.32
3.71
1.05
6.29
1.11
0.78
0.80
1.07
0.53
0.96
0.79
0.47
2.10
1.38
2.13
12.19
0.73
1.11
0.83
0.76
1.01
0.68
0.41
5.81
2.29
3.34
F-16

-------
NPDES ID
PA0026107
PA0026123
PA0026191
PA0026239
PA0026263
PA0026280
PA0026310
PA0026361
PA0026441
PA0026484
PA0026492
PA0026557
PA0026620
PA0026654
PA0026727
PA0026735
PA0026743
PA0026808
PA0026875
PA0026921
PA0027014
PA0027022
PA0027049
PA0027057
PA0027065
PA0027090
PA0027171
PA0027189
PA0027197
PA0027316
PA0027324
PA0027405
Name
WYOMING VALLEY SAN AUTH
COLUMBIA MUN AUTH
HUNTINGTON BORO
UNIVERSITY AREA JOINT AUTH -
YORK CITY SEW AUTH
LEWISTOWN BORO
CLEARFIELD MUN AUTH
LOWER LACKAWANNA VLY SAN AUTH
LEMOYNE BORO
DERRY TWP MUN AUTH - CLEARWATE
SCRANTON CITY SEW AUTH
SUNBURY CITY MUN AUTH
MILLERSVILLE BORO
NEW CUMBERLAND BORO
TYRONE BORO
SWATARA TWP AUTH
LANCASTER CITY
SPRINGETTSBURY TWP
HANOVER BOROUGH
GREATER HAZELTON JNT SEW AUTH
ALTOONA CITY AUTH - EAST
ALTOONA CITY AUTH - WEST
WILLIAMSPORT SAN AUTH-WEST
WILLIAMSPORT SAN AUTH-CENTRAL
LACKAWANNA RIVER BASIN SEW AUT
LACKAWANNA RIVER BASIN SAN AUT
BLOOMSBURG MUN AUTH
LOWER ALLEN TWP AUTH
HARRISBURG AUTHORITY
LEBANON CITY
SHAMOKIN-COAL TWP JNT SEW AUTH
EPHRATA BORO AUTH - WWTP #1
Design flow
(MGD)
32.00
2.00
5.90
6.00
26.00
2.82
4.50
6.00
2.09
5.00
20.00
4.20
1.85
1.25
9.00
6.30
29.73
15.00
5.50
8.90
8.00
9.00
3.92
10.50
6.00
7.00
4.29
6.25
37.70
8.00
7.00
3.80
Observed flow
(MGD)
(1991 -2006 average)
24.07
0.80
2.50
4.62
10.66
1.70
3.46
3.36
1.72
3.48
13.03
3.40
0.65
3.98
5.97
3.71
19.72
10.04
4.16
7.20
6.77
8.03
2.89
7.38
2.78
5.28
2.58
4.38
25.06
5.31
3.74
2.56
F-17

-------
NPDES ID
PA0027553
PA0028142
PA0028461
PA0028576
PA0028665
PA0028681
PA0028746
PA0030643
PA0032883
PA0034576
PA0037150
PA0037966
PA0038415
PA0042269
PA0043257
PA0043273
PA0043681
PA0044661
PA0045985
PA0047325
PA0062219
PA0070041
PA0070386
PA0080314
PA0083011
PA0087181
PA01 10582
PA01 10965
PA0111759
PA0208779
PA0209228
PA0228818
Name
PINE CREEK MUN AUTH
PA NATIONAL GUARD - FORT INDIA
MIFFLINBURG BORO
CLARKS SUMMIT - S ABINGTON JSA
JERSEY SHORE BORO
KELLY TWP MUN AUTH
HAMPDEN TWP
SHIPPENSBURG BORO
DUNCANSVILLE BORO
TOWANDA MUN AUTH
PENN TWP BOARD OF COMMISSIONER
MOSHANNON VALLEY JT SEW AUTH
EAST PENNSBORO TWP
LANCASTER AREA SEW AUTH
NEW FREEDOM BORO AUTH
HOLLIDAYSBURG SEW AUTH
VALLEY JOINT SEW AUTH
LEWISBURG AREA JOINT SEW AUTH
MOUNTAINTOP AREA JNT SAN AUTH
PPL SUSQUEHANNA LLC
FRACKVILLE AREA MUN AUTH
MAHANOY CITY SEW AUTH
SHENANDOAH MUN SEW AUTH
HAMPDEN TWP - ROTH LANE
NEWBERRY TWP MUN AUTH
EPHRATA BORO AUTH - WWTP #2
EASTERN SNYDER CO REG AUTH
MID-CENTRE COUNTY AUTH
CARGILL MEAT SOLUTIONS CORP
CLEARFIELD LEATHER INC DBA WIC
LYCOMING CO WATER & SEWER AUTH
FIRST QUALITY TISSUE LLC
Design flow
(MGD)
1.30
1.00
1.40
2.50
1.05
3.75
1.76
3.30
1.22
1.16
4.20
1.73
3.70
15.00
7.20
6.00
2.25
2.42
4.16
0.08
1.40
1.38
2.00
4.65
1.30
2.30
2.80
1.00
0.80
0.12
1.50
3.95
Observed flow
(MGD)
(1991 -2006 average)
39.24
1.84
0.77
2.38
0.69
14.25
1.30
1.79
0.73
0.75
1.75
1.53
3.03
7.52
2.17
3.46
1.04
1.30
2.64
0.11
1.05
0.57
1.35
2.11
0.53
1.24
1.62
10.99
0.56
0.13
0.61
1.63
F-18

-------
     Point Sources
     Hydrography
     Interstate
     Water (Nat. Atlas Dataset)
     US Census Populated Places
     Municipalities (pop > 50,000)
   ] County Boundaries
     Watershed with HUCSs
                                                                               Pennsylvania
GCRP Model Areas - Susquehanna River Basin
           Major Point Sources
Figure 3. Major point sources in the Susquehanna River watershed.
                                                    F-19

-------
Meteorological Data
The required meteorological data series for the 20 Watershed model simulations are precipitation, air temperature,
and potential evapotranspiration (PET). The 20 Watershed model does not include water temperature or algal
simulation and uses a degree-day method for snowmelt, so additional meteorological variables such as solar
radiation are needed only for the calculation of PET. These meteorological data are drawn from the BASINS4
Meteorological Database (USEPA 2008), which provides a consistent, quality-assured set of nationwide data with
gaps filled and records disaggregated. Scenario application will require simulation over 30 years, so the available
stations are those with a common 30-year period of record (or one that can be filled from an approximately co-
located station) that covers the year 2001. A total of 62 precipitation stations were identified for use in the
Susquehanna River model with a common period of record of 2/1/1972-3/31/2004 for the entire watershed (Table
4 and Figure 4). The majority of the stations are available through at least 9/30/2005. Temperature records are
sparser; where these are absent temperature is taken from nearby stations with an elevation correction. For each
weather station, Penman-Monteith reference evapotranspiration was calculated for use in HSPF using observed
precipitation and temperature coupled with SWAT weather generator estimates of solar radiation, wind
movement, cloud cover, and relative humidity.

SWAT uses daily meteorological data for the 20 Watershed model applications, while HSPF requires hourly data.
It is important to note that a majority of the meteorological stations available for the Susquehanna River
watershed are Cooperative Summary of the Day stations that do not report sub-daily data. The BASINS4 dataset
already has versions of the daily data that have been disaggregated to an hourly time step using template stations.
Table 4.     Precipitation stations for the Susquehanna River watershed model
COOP ID
NY300085
NY300270
NY300448
NY300687
NY301168
NY301173
NY301424
NY301752
NY302454
NY302610
NY303979
NY303983
NY304772
NY305512
NY306085
NY307195
NY308498
NY309442
Name
ALFRED
ARNOT FOREST
BATH
BINGHAMTON WSO AP
CANDOR 2 SE
CANISTEO 1 SW
CHEPACHET
COOPERSTOWN
EAST SIDNEY
ELMIRA
HORNBY
HORNELL ALMOND DAM
LINDLEY2N
MORRISVILLE 6 SW
NORWICH
ROCKDALE
THURSTON
WHITNEY POINT DAM
Latitude
42.2614
42.2670
42.3500
42.2078
42.1947
42.2667
42.9097
42.7150
42.3333
42.1000
42.2330
42.3500
42.0500
42.8333
42.5011
42.3833
42.2000
42.3500
Longitude
-77.7850
-76.6330
-77.3500
-75.9814
-76.3133
-77.6167
-75.1108
-74.9283
-75.2333
-76.8000
-77.0500
-77.7000
-77.1333
-75.7333
-75.5194
-75.4000
-77.3330
-75.9670
Temperature
Yes
No
Yes
Yes
No
No
No
Yes
No
Yes
No
No
No
Yes
Yes
No
No
No
Elevation (ft)
1,770
1,200
1,120
1,600
920
1,155
1,320
1,200
1,155
844
1,795
1,325
1,040
1,300
1,020
1,030
1,620
1,040
                                                 F-20

-------
COOP ID
PA360140
PA360147
PA360457
PA360482
PA360656
PA360725
PA360763
PA361087
PA361480
PA361833
PA361961
PA362013
PA362245
PA362629
PA362721
PA363130
PA364047
PA364763
PA364778
PA364853
PA364896
PA364992
PA365344
PA365915
PA366289
PA366916
PA367409
PA367727
PA367730
PA367931
PA368057
PA368073
PA368449
Name
ALTOONA 3 W
ALVIN R BUSH DAM
BEAR GAP
BEAVERTOWN 1 NE
BIGLERVILLE
BLAIN 5SW
BLOSERVILLE 1 N
BUFFALO MILLS
CLARENCE
COVINGTON 2 WSW
CURWENSVILLE LAKE
DANVILLE
DRIFTWOOD
EMPORIUM
EVERETT
GALETON
HONEY BROOK 2 SSE
LANCASTER 2 NE FILT PLANT
LANDISVILLE 2 NW
LAURELTON CENTER
LEBANON 2 W
LEWISTOWN
MAHANOY CITY 2 N
MONTROSE
NEW PARK
PHILIPSBURG8E
RENOVO
RUSHVILLE
SABINSVILLE 3 SE
SELINSGROVE2S
SHICKSHINNY3N
SHIPPENSBURG
STATE COLLEGE
Latitude
40.4950
41.3670
40.8236
40.7667
39.9356
40.3000
40.2636
39.9461
41.0456
41.7331
40.9500
40.9483
41.3419
41.5067
40.0136
41.7356
40.0789
40.0500
40.1167
40.9017
40.3333
40.5869
40.8344
41.8667
39.7350
40.9167
41.3297
41.7833
41.8422
40.7831
41.2000
40.0500
40.7933
Longitude
-78.4667
-77.9330
-76.4983
-77.1500
-77.2578
-77.5833
-77.3639
-78.6458
-77.9453
-77.1167
-78.5330
-76.6036
-78.1403
-78.2275
-78.3653
-77.6519
-75.8975
-76.2742
-76.4333
-77.2139
-76.4667
-77.5697
-76.1353
-75.8500
-76.5061
-78.0667
-77.7381
-76.1167
-77.4747
-76.8611
-76.1500
-77.5167
-77.8672
Temperature
Yes
Yes
No
No
Yes
No
Yes
No
No
No
No
No
No
Yes
Yes
No
No
No
Yes
Yes
Yes
Yes
No
Yes
No
No
Yes
No
No
Yes
No
Yes
Yes
Elevation (ft)
1,320
930
900
540
720
820
700
1,310
1,390
1,745
1,165
475
820
1,040
1,000
1,345
665
270
360
800
450
460
1,710
1,420
800
1,945
660
870
1,999
420
780
680
1,170
F-21

-------
COOP ID
PA368491
PA368692
PA368905
PA368959
PA369408
PA369705
PA369714
PA369728
PA369823
PA369933
PA369950
Name
STILLWATER
SUSQUEHANNA
TOWANDA 1 S
TROY 1 NE
WELLSBORO 4 SW
WILKES BARRE SCRANTON
WSOAP
WILLIAMSBURG
WILLIAMSPORT RGNL AP
WOLFSBURG
YORK 3 SSW PUMP STN
YORK HAVEN
Latitude
41.6830
41.9483
41.7506
41.7833
41.7003
41.3389
40.4667
41.2433
40.0417
39.9167
40.1167
Longitude
-75.4830
-75.6050
-76.4428
-76.7833
-77.3894
-75.7267
-78.2000
-76.9217
-78.5278
-76.7500
-76.7167
Temperature
No
No
Yes
No
Yes
Yes
No
Yes
No
Yes
No
Elevation (ft)
1,650
910
750
1,110
1,818
930
840
520
1,185
390
310
F-22

-------
Legend
^ Weather Stations
= Interstate
	 Hydrography
•1 Water (Nat. Atlas Dataset)
   US Census Populated Places
HI Municipalities (pop > 50,000)
  ^] County Boundaries
l~~l Watershed with HUCSs
                                            'tnajiiio'    •/*
\ t    f & y^Y5P2*54
NY309442' ~k^~^
                                              ~~~f\f^J-f\
                                              NY300687 \   New York
          N"°±i^^ip^     >4;iyiA368491
            ^.i.«%^  x^   ^   y>fA368959A^  /T   y  &?^i
          r^^^4^U\PA367'730\  %J^>^PA368905^/^nV/J*
          W363130V1>^-V io.il.,,X^  "f^^V/  IS [    ;
Figure 4. Weatherstations for the Susquehanna River watershed model.
                                 F-23

-------
Watershed Segmentation

At about 27,000 square miles, the Susquehanna River basin is one of the largest modeling areas considered for the
20 Watershed project. It encompasses a complete drainage area, with no need for upstream boundary conditions.
There is also an existing detailed HSPF model of the basin (the Chesapeake Bay Model or CBM; USEPA 2010).
Watershed segmentation for the Susquehanna River basin is based on the segmentation used in the CBM,
resulting in 278 subbasins for modeling (Figure 5). The initial calibration watershed (Raystown Branch Juniata
River) is highlighted.

The model subbasins approximate the HUC-10 scale, but are  subdivided as needed to account for the connection
of tributaries and location of flow gages. The subbasins range in size from 1.04 to 516 mi2.
                                               F-24

-------
 Legend
  A  USGS Gages
 	 Hydrography
      Interstate
      Water (Nat. Atlas Dataset)
      US Census Populated Places
      Municipalities (pop > 50,000)
   ] County Boundaries
 I    | Initial Calibration Watershed
      Model Subbasins
 Pennsylvania
   USGS 1562000
GCRP Model Areas - Susquenanna River Basin
           Model Segmentation
 Figure 5. Model segmentation and USGS stations utilized for the Susquehanna River watershed.
                                                  F-25

-------
Calibration Data and Locations
The Susquehanna River was selected as a pilot site because of extensive previous experience with the CBM.
Given the existence of this calibrated model, the approach to the Susquehanna pilot study was to start from
parameters derived from the existing model, evaluate the parameterization through detailed application to an
initial calibration focus area, then evaluate the quality of the fit through comparison to data at two monitoring
points representing the larger watershed. Spatial calibration at multiple locations was not pursued. It should be
noted, however, that the 20 Watershed approach is based on a different land use coverage and uses different
weather data and, in  particular, a different estimator of PET (Penman-Monteith PET using local temperature and
weather generator insolation and auxiliary variables for 20 Watershed versus Hamon PET in the Phase 5 CBM
and Penman pan evaporation at first-order weather stations in early versions of the CBM). These differences may
result in systematic differences in model parameters.

Three sites with known high quality flow gaging and water quality data were selected for both hydrology and
water quality calibration (Table 5). The first of these (Raystown Branch Juniata River) was selected for initial
calibration, with subsequent adjustments based on comparison to data at two stations on the mainstem
Susquehanna.
Table 5.     Calibration and validation locations in  the Susquehanna River watershed
Station Name
Raystown Branch Juniata River at Saxton, PA
Susquehanna River at Danville, PA
Susquehanna River at Marietta, PA
USGS ID
01562000
01540500
01576000
Drainage Area
(mi2)
756
11,220
25,990
Hydrology
Calibration
X
X
X
Water Quality
Calibration
X
X
X
The model hydrology calibration period for Raystown Branch Juniata River was set to Water Years 1995-2005,
while the mainstem stations use Water Years 1993-2003 to fall within the common period of record of the
weather data. Calibration was done on the later data due to concerns that there have been significant changes in
land use and agricultural management practices since the 1980s. Hydrologic validation was then performed on the
10 water years priorto the calibration period. The bulk of available water quality data are from the early 1990s.
Therefore, water quality calibration used calendar year periods beginning in 1991, with earlier data reserved for
validation.
                                                F-26

-------

Initial hydrologic parameterization for the Susquehanna River calibration focus area came from the Chesapeake
Bay Program's (CBP) model for the Susquehanna River watershed (CBM). The CBM has undergone a series of
revisions over many years, with the most current version of the model being known as Phase 5 (USEPA 2010).
The Phase 5 parameters were obtained from the CBP and reviewed. Through this process it was identified that the
CBM set for Phase 5 was much more complex than that used for a typical HSPF model, with parameters and land
use categories provided in a multi-file database, with values changing over the time span of the simulation. In the
CBP Phase 5 model, HSPF input sequences are developed through an elaborate scripting scheme, with multiple
HSPF input sequences developed for more than 20 land use categories (including various agricultural
management practices) and multiple time spans to determine flows and loads on a unit area basis. These unit area
flows and loads are then combined through scripts to create input for an in-stream simulation.

The complexity of the Phase 5 model far exceeded the constraints of this study, so rather than using the Phase 5
parameters, it was decided to parameterize Susquehanna River watershed model using the previous version of the
CBM, Phase 4. The Phase 4 parameters are readily available, as they are incorporated into the BASINS
companion program HSPFParm, and these parameters are available in a format directly akin to that needed by
HSPF. The six pervious (forest, high till cropland, low till cropland, pasture, urban, and hay) and two impervious
(animal/feedlot and urban) land use categories of the Phase 4 model are more analogous to the categories of the
Susquehanna model for this study.

Even using the simpler Phase 4 model from the CBP, there was no one-to-one correlation between land use
categories of the CBM and this study. The Phase 4 model of the Susquehanna River watershed consisted of 12
land segments in 4 UCI files, and these parameters needed to be applied to the 21 land uses in this project.
Moreover, for this project the land use categories explicitly represented hydrologic soil groups, while the CBM
parameters did not. A method was developed for creating the Susquehanna model parameters for this study from
these Phase 4 parameters, where approximate  average values were applied from the CBP Upper Susquehanna
simulations for model section 020501, from the CBP Western Branch Susquehanna simulation for model section
020502, and from the Lower Susquehanna simulation for model section 020503. Since there was no explicit
representation of hydrologic soil groups in the CBM, parameter values were assigned the  same values across each
of the Susquehanna model soil  classifications. For example, the CBM forest parameters for the Lower
Susquehanna were applied to the land use categories Forest_A through Forest_D in the model parameters for
020503.

The USGS gage on the Raystown Branch of the Juniata River at Saxton, PA (USGS 02050303) was used as the
primary calibration location, while the gages on the Susquehanna River at Danville, PA (USGS 01540500), on the
West Branch Susquehanna River at Lewisburg, PA (USGS 01553500), and on the Susquehanna River at Marietta,
PA (USGS 01576000) were used as additional calibration checks. Calibrated parameters from the Raystown
Branch gage were applied to the Lower Susquehanna portion of the study area (020503), while calibration
adjustments for the Danville and Lewisburg gages were applied to the Upper and Western sections of the study
area respectively (020501 and 020502). The Susquehanna River at Marietta, PA (USGS 01576000), the most
downstream of the gages, was used to verify that the calibration parameters applied at the two upstream gages
were applicable to the entire watershed.

Once the hydrology calibration was complete  for the entire Susquehanna River watershed, the focus turned to
sediment and water quality representation. Extracting parameters  from the CBP Susquehanna model for sediment
and water quality was even less straightforward than that for hydrology parameters, since the CBM used the more
complicated NITR and PHOS modules of the HSPF pervious land simulation operations. Initial parameterization
for sediment and water quality  simulation was taken from the loadings used to set up the Willamette model in this
20 Watershed study, adjusted based on the parameters from the CBP Phase 4 model where available.

-------
Changes Made to Base Data Provided

No changes were made to the meteorological, point source, or land use base data.


Assumptions

Reservoirs
While there are many dams in the study area, their influence was not explicitly included in the study. The largest
of the reservoirs are on the Susquehanna River near the outlet of the study area, well below the USGS gage on the
Susquehanna River at Marietta. This USGS gage on the Susquehanna River at Marietta was selected for use in
calibration because it is the most downstream main stem gage that is still upstream of the influence of the major
reservoirs. While one would assume the major main stem reservoirs influence the flow and water quality exiting
the Susquehanna River at the outlet, for this model the impacts of these reservoirs are assumed to be implicitly
represented through the tabular representation of reach hydrologic response (FTables).

The primary intention of the 20 Watershed simulations is to examine relative changes in response of large
watersheds. Information is not available to specify future time series of operations or boundary conditions for
these reservoirs.  Therefore, representation of the reservoirs through the stage-discharge relationships expressed in
the FTables provides the most useful basis for evaluating relative changes in response.


Withdrawals and Point Sources
A variety of water withdrawals occur in the Susquehanna, but these have a relatively small effect on the overall
water balance. In addition, future changes in water withdrawals are not known. Therefore, withdrawals were not
included in the 20 Watershed model application. In contrast to withdrawals, point sources must be included for
model water quality calibration because they represent a significant fraction of nutrient loads in the system.
Existing major point source flows and loads are represented in the  model, but will be held at current levels for
simulation of future conditions to better isolate the potential direct impacts of climate and land use change.


Snow Simulation

The Susquehanna HPSF model includes snow simulation using the degree-day method for snowmelt. The initial
values extracted from the Chesapeake Bay Program Model are  assumed to be appropriate and the initial
parameterization was not adjusted.


Hydrology Calibration

As explained above, the starting parameters for the Susquehanna HSPF model came from a Chesapeake Bay
Program model of the Susquehanna River watershed. Once the starting parameters were inserted into the model
input files, average annual precipitation and potential evapotranspiration values were computed and compared to
published values. Through this process it was determined the input potential evapotranspiration time series should
be reduced by multipliers, since the computation of these time series produced more PET on an average annual
basis than the published values indicate. The multipliers used for PET were 0.75 for the Lower Susquehanna and
Western Branch, and 0.8  for the Upper Susquehanna. Calibration adjustments focused on the following
parameters:

   •   BASETP (ET by riparian vegetation): The model was significantly oversimulating the  50 percent low
       flows at the primary calibration location. Slightly increasing the BASETP value provided some ET by
       riparian vegetation and thus improved the simulation of low flows.

-------
    •   DEEPFR (fraction of groundwater inflow which will enter deep groundwater): Adding a modest amount
       of DEEPFR above the primary calibration location improved the overall water balance and improved the
       simulation of low flows.
    •   INFILT (index to mean soil infiltration rate): In the upper portions of the study area the peak flows were
       simulating too low, while the low flows were simulating too high. Decreasing INFILT for the upper
       portions of the watershed shifted flows to a faster response, increasing the peaks and reducing the low
       flows.
    •   AGWRC (Groundwater recession rate): Adjusted slightly in order to replicate groundwater recession in
       the observed data.

Initial calibration was performed at the USGS gage on the  Raystown Branch of the Juniata River at Saxton, PA
(USGS 02050303), and is summarized in Figures 6 through 12 and Tables 6 and 7. The model fit is of high
quality overall, but simulates low on the lowest 10 percent of flows. This could be due to something not
accounted for in the model, such as reservoir operations or other discharges. Given that these low flows are not
critical to the purposes of this study, the issue is being noted as an area with potential further refinement. None of
the metrics fall beyond the range of those set for the 20 Watershed study. The model calibration period was set to
the 10 water years from 10/01/1995 to 09/30/2005.
                                                F-29

-------
          I Avg Monthly Rainfall (in)
          -Avg Observed Flow (10/1/1995 to 9/30/2005 )
          -Avg Modeled Flow (Same Period)
      35000
 o
                                                                                                (0
                                                                                                «
Oct-95     Apr-97      Oct-98      Apr-00      Oct-01

                                       Date
                                                                    Apr-03
Oct-04
  Figure 6. Mean daily flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
      calibration period (HSPF).

                                                       ^•Avg Monthly Rainfall (in)
                                                       -»-Avg Observed Flow (10/1/1995 to 9/30/2005
                                                        —Avg Modeled Flow (Same Period)
    6000
    4000 -
LJ-
    2000 -
  Figure 7. Mean monthly flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
      calibration period (HSPF).
                                               F-30

-------
   6000
•s
1
   4000 -
_
-s
o
O)
o>
£
01
   2000 -
            Avg Flow (10/1/1995 to 9/30/2005
            Line of Equal Value
            Best-Fit Line
0.937x + 65.513
 R2 = 0.8993
                  2000        4000
            Average Observed Flow (cfs)
                            6000
T3
O
                                   .a
                                   g
                                   O)
                                   o
                                   ro
                                   uD
                                    _
                                   O>
        Avg Obsen/ed Flow (10/1/1995 to 9/30/2005 )
    ^Avg Modeled Flow (10/1/1995 to 9/30/2005 )
    	Line of Equal Value
100% -,
 90% -
 80% -
 70% -
 60% -
 50%
     40% -
     30% -
     20% -
     10% -
      0%
                    L  J  1  L   ,  J
                                           0-95   A-97   0-98
                          A-00   0-01
                            Month
                                   A-03  0-04
 Figure 8. Monthly flow regression and temporal variation at USGS 01562000 Raystown Branch Juniata
     River at Saxton, PA - calibration period (HSPF).
           Avg Flow (10/1/1995 to 9/30/2005)
          -Line of Equal Value
          -Best-Fit Line
 2500
            500    1000   1500   2000  2500
            Average Observed Flow (cfs)
                                             Avg Monthly Rainfall (in)
                                            -Avg Obsen/ed Flow (10/1/1995 to 9/30/2005)
                                            -Avg Modeled Flow (Same Period)
                                       2500 -i
                                             10  11  12  1  2   3   4   5   6
                                                            Month
 Figure 9. Seasonal regression and temporal aggregate at USGS 01562000 Raystown Branch Juniata
     River at Saxton, PA - calibration period (HSPF).
                                               F-31

-------
         Average Monthly Rainfall (in)
        -Median Observed Flow (10/1/1995 to 9/30/2005)
                                  I Observed (25th, 75th)
                                  I Modeled (Median, 25th, 75th)
    3000 T
    2500 -
    2000 -
    1500 -
    1000 -
     500 - --
                                                                     r 0
                                                                     - 1
                                                                     - 2
                                                                     - 3
                                                                     - 5
           10
11
12
3     4
 Month
   Figure 10. Seasonal medians and ranges at USGS 01562000 Raystown Branch Juniata River at
       Saxton, PA - calibration period (HSPF).

Table 6.     Seasonal summary at USGS 01562000 Raystown Branch Juniata River at Saxton,
            PA - calibration period (HSPF)

MON I H
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
438. 69 1 187. 00 1 119.00 530.75
1024.44 647. 50 1 149.75 1170.00
1071.221 682. 00l 280.751 1350.00
1255. 25 1 475. 00| 260.00 1180.00
1290.35 900. 00 1 470.00 1775.00
2177.77! 1690. OOl 1052.50! 2742.50
1771.52] 1180.00| 843.75) 2102.50
1310.00 739. 50 1 414.00 1777.50
895. 83 1 448. OOl 287.50! 879.00
326.40] 207. 00 1 146.00 324.00
278.62 157.50] 115.25 339.75
907.63 124. 50 1 104.00 360.25
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
599.79 301. 11 1 128.74] 788.27
1101.45 658. 92 1 138. 80 1 1324.27
1029. 71 1 830.45! 374.78! 1250.32
1394.02] 892.68] 351.51] 1730.37
1330.42 1074.05] 739. 83 ] 1743.98
1963.581 1594.05! 1008.22! 2348.98
1490.45) 989. 11 1 640.38! 1736.15
1010.40 563.51 303. 89| 1180.35
939.20I 441.19! 220.33! 910.95
407.92 216. 08 1 114. 91 1 368.66
345.53 174.23 74.24] 430.05
1120.57 133.58] 62.75| 649.56
                                              F-32

-------
               •Obsen/ed Flow Duration (10/1/1995 to 9/30/2005 )
               •Modeled Flow Duration (10/1/1995 to 9/30/2005 )
       100000
        10000
         1000
           10
                    10%     20%     30%     40%     50%     60%     70%     80%     90%     100%

                                 Percent of Time that Flow is Equaled or Exceeded
    Figure 11. Flow exceedence at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
       calibration period (HSPF).
        o
        o
        T3
        0)

        0)
         0)
         E
0)
.N
"re
E
o
                     -Observed Flow Volume (10/1/1995 to 9/30/2005 )

                     -Modeled Flow Volume (10/1/1995 to 9/30/2005 )
             120%
             100% -
              80% -
      60% -
              40% -
              20% -
                Oct-95
                           Apr-97
                                       Oct-9£
                                          Apr-00
                                                             Oct-01
Apr-03
                                                                                    Oct-04
Figure 12. Flow accumulation at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
- calibration period (HSPF).
                                                 F-33

-------
Table 7.     Summary statistics at USGS 01562000 Raystown Branch Juniata River at Saxton,
             PA - calibration period (HSPF)
HSPF Simulated Flow
REACH OUTFLOW FROM DSN 101
10-Year Analysis Period: 10/1/1995 - 9/30/2005
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9):
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3^:
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% hig_hest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Sjpring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of Efficiency, E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
19.01

8.15
2.19

2.80
4.11
6.97
5.13

7.78
1.62
Error Statistics
-0.16
6.72
-3.92
23.90
Observed Flow Gage
USGS 01562000 Raystown Branch Juniata River at Saxton, PA
Hydrologic Unit Code: 2050303
Latitude: 40.21591249
Longitude: -78.2652901
Drainage Area (sq-rri): 756
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow VolumeJ1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
7.76 » | 30
-0.84
-13.61
-0.63
39.10
0.698
0.552
0.898
30
30
20
50
Model accuracy increases
as E or E' approaches 1.0





19.04

8.48
2.05

2.26
3.81
7.03
5.93

7.83
1.16





Clear [













Hydrology Validation
Validation for the Susquehanna River calibration focus was performed at the same gage (USGS 01562000
Raystown Branch Juniata River at Saxton, PA) but for the water years 10/01/1985 to 09/30/1995. Results are
presented in Figures 13 through 19 and Tables 8 and 9. Similar to the calibration years, the validation years'
model fit is of high quality, although the validation simulates the summer storm volumes somewhat high while
undersimulating the overall storm volume. This may in part reflect differences in land use and management
practices relative to the 200 NLCD. The remaining metrics fall within the acceptable range set for the 20
Watershed study.
                                              F-34

-------
           Avg Monthly Rainfall (in)
          -Avg Observed Flow (10/1/1985 to 9/30/1995
          -Avg Modeled Flow (Same Period)
      30000
  o
                                                                              Oct-94
  Figure 13. Mean daily flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
     validation period (HSPF).
                                                          Avg Monthly Rainfall (in)
                                                      -»-Avg Observed Flow (10/1/1985 to 9/30/1995 )
                                                        —Avg Modeled Flow (Same Period)
    8000
•S
    6000 -
    4000
    2000 -
0-85        A-87        0-88        A-90        O-91

                                       Month
                                                                    A-93
0-94
  Figure 14. Mean monthly flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA -
     validation period (HSPF).
                                              F-35

-------
        •   Avg Flow (10/1/1985 to 9/30/1995 )
       	Line of Equal Value
         — Best-Fit Line
  8000
                                              T3
                                               O
                                               +
                                               CO
                                              JS
                                               ro
                                              00
           100% -,
           90% -
           80% -
           70% -
           60% -
           50%
                  Avg Observed Flow (10/1/1985 to 9/30/1995 )
                  Avg Modeled Flow (10/1/1985 to 9/30/1995 )
                  -Line of Equal Value
           40% -
           30% -
           20% -
           10% -
            0%
       0      2000     4000     6000
           Average Observed Flow (cfs)
8000
               0-85   A-87   0-88
A-90  0-91
 Month
A-93  0-94
Figure 15. Monthly flow regression and temporal variation at USGS 01562000 Raystown Branch
    Juniata River at Saxton, PA - validation period (HSPF).
      •  Avg Flow (10/1/1985 to 9/30/1995)
     	Line of Equal Value
     	Best-Fit Line
                Avg Monthly Rainfall (in)
                -Avg Observed Flow (10/1/1985 to 9/30/1995)
                -Avg Modeled Flow (Same Period)
2500
          = 0,7072x+192.32
            ..R2 = 0.8986
          2500 -i
           500    1000   1500   2000   2500
           Average Observed Flow (cfs)
                10 11 12  1   23456789
                               Month
Figure 16. Seasonal regression and temporal aggregate at USGS 01562000 Raystown Branch Juniata
    River at Saxton, PA - validation period (HSPF).
                                              F-36

-------
         i Observed (25th, 75th)
         • Median Observed Flow (10/1/1985 to 9/30/1995)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
    2500
            10     11
   Figure 17. Seasonal medians and ranges at USGS 01562000 Raystown Branch Juniata River at
       Saxton, PA - validation period (HSPF).
Table 8.     Seasonal summary at USGS 01562000 Raystown Branch Juniata River at Saxton,
            PA - validation period (HSPF)
MONTH
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
351.63
128.00
104. 00 1 200.00
771.90! 411.00! 211.75! 649.00
936.58
603.50
1063.43 562.00
302.50! 1157.50
322.50! 1167.50
1214.57 754.00 410.00! 1495.00
2016.51
1240.00
670.75! 2140.00
1772. 93| 965. 00| 613. 50| 1935.00
1102.77
676.00
479.25! 1122.50
559.15! 275.50 201.00! 479.50
545.15
252.88
180.00
136.00
138.00! 418.00
108.00! 216.75
195.59! 160.00! 107.75! 233.50
MODELED FLOW (CFS)
MEAN MEDIAN
429.02
25TH 75TH
157. 78 1 101. 27 1 307.10
798.12 454.17! 239.35! 790.07
993.77
1044.74
1256.37
1591.71
1379.12
638.99
779.69
787.94
1003.08
1053.09
399.19! 1215.70
549.18! 1243.93
702.93! 1494.87
578.34! 1910.94
653. 23| 398.61 1345.43
384.81
225.71! 698.65
619.87 217.91! 135.17! 385.61
621.12
328.81
160.93
115.81
89.83! 396.62
54.58J 259.23
231.64 169.21! 67.62! 328.28
                                             F-37

-------
            -Obsen/ed Flow Duration (10/1/1985 to 9/30/1995

             Modeled Flow Duration (10/1/1985 to 9/30/1995 )
    100000
     10000


 £
 0)

 E
 T3
 O)
 ro


 o
               -Obsenyed Flow Volume (10/1/1985 to 9/30/1995


               -Modeled Flow Volume (10/1/1985 to 9/30/1995 )
      120%
         Oct-85       Apr-87
Oct-88       Apr-90       Oct-91       Apr-93       Oct-94
  Figure 19. Flow accumulation at USGS 01562000 Raystown Branch Juniata River at Saxton, PA

      validation period (HSPF).


                                                F-38

-------
Table 9.     Summary statistics at USGS 01562000 Raystown Branch Juniata River at Saxton,
             PA - validation period (HSPF)
REACH OUTFLOW FROM DSN 101
10-Year Analysis F^riod: 10/1/1985 - 9/30/1995
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total ojj3inTulatejdjTic|h^
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9):
Simulated Fall Flow Volume (months 10-12^
Simulated Winter Flow Volume (months 1-3):
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Sjpring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of EfficjencyJ^___
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
14.83

6.77
1.74

1.79
3.35
5.76
3.93

5.52
0.85
Error Statistics
-8.00
-4.61
-12.82
18.92
USGS 01562000 Raystown Branch Juniata River at Saxton, PA
Hydrologic Unit Code: 2050303
Latitude: 40.21 591 249
Longitude: -78.2652901
Drainage Area (sq-rri): 756
Total Observed In-stream Flow:

_J]oJal^f^teervedJTig_hest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow VolumeJ10-12):
Observed Winter Flow Volume (1-3):
Observed Spring Flow Volume (4-6):

Total Obsei-ved Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
7.86 » | 30
-9.69
-23.40
-14.91
56.16
0.553
0.512
0.868
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0





16.12

7.77
1.82

1.51
3.10
6.38
5.12

6.49
0.54





Clear [













Hydrology Results for Larger Watershed
Since the calibration location above represents only a small portion of the drainage area for this project, results
near the outlet of the entire watershed were examined at the Susquehanna River at Marietta, PA USGS gage
(01576000). The results at this gage look fairly good as well. The simulated output is just a little high on the low
flows and a little high on the summer volumes. A few of the metrics were exceeded, but most of the metrics fall
within the acceptable range set for the 20 Watershed study including a daily Nash-Sutcliffe of 0.77 at the Marietta
gage. The calibration results at the Susquehanna River at Marietta, PA USGS gage (01576000) are presented in
Figures 20 through 26 and Tables 10 and 11. The calibration and validation statistical measurements at all USGS
gages used in the Susquehanna River watershed for the 20 Watershed project are shown in Tables 12 and 13,
respectively.
                                               F-39

-------
     ^•Avg Monthly Rainfall (in)
     	Avg Observed Flow (10/1/1993 to 9/30/2003 )
         Avg Modeled Flow (Same Period)
    600000
t
o
                    Apr-95
Oct-96
Apr-98      Oct-99

     Date
Apr-01
Oct-02
  Figure 20. Mean daily flow at USGS 01576000 Susquehanna River at Marietta, PA - calibration period
      (HSPF).
    200000
                                                         Avg Monthly Rainfall (in)
                                                        -Avg Observed Flow (10/1/1993 to 9/30/2003
                                                        •Avg Modeled Flow (Same Period)
                                                                               0-02
  Figure 21. Mean monthly flow at USGS 01576000 Susquehanna River at Marietta, PA - calibration
      period (HSPF).
                                                 F-40

-------
            Avg Flow (10/1/1993 to 9/30/2003 )
            • Line of Equal Value
            Best-Fit Line
   200000
   150000
   100000
 o
 0)  50000
          y = 0.816x +7854.6
       0     50000   100000   150000  200000

          Average Observed Flow(cfs)
                                                  100% -i
                                                       Avg Observed Flow (10/1/1993 to 9/30/2003 )
                                                       Avg Modeled Flow (10/1/1993 to 9/30/2003 )
                                                      -Line of Equal Value
                                               O
                                               £=
                                              _
                                              ro
                                              m
                                                            A-95   O-96
                                                                        A-98   O-99
                                                                         Month
                                                                                    A-01   O-02
  Figure 22. Monthly flow regression and temporal variation at USGS 01576000 Susquehanna River at
      Marietta, PA - calibration period (HSPF).
           Avg Flow (10/1/1993 to 9/30/2003)
          • Line of Equal Value
           Best-Fit Line
  100000

   80000
   60000
40000

-------
  o
         i Observed (25th, 75th)
         • Median Observed Flow (10/1/1993 to 9/30/2003)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
    100000
     90000
     80000
     70000
     60000
     50000 -
 £  40000
     30000
     20000
     10000
         0
              10     11
   Figure 24. Seasonal medians and ranges at USGS 01576000 Susquehanna River at Marietta, PA -
       calibration period (HSPF).
Table 10.   Seasonal summary at USGS 01576000 Susquehanna River at Marietta, PA
            calibration period (HSPF)
MONTH
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
18258.39
29811.13
43879.65
46873.61
45647.16
78282.90
74228.00
46236.13
32080.10
16327.52
15446.55
16977.37
10650.00
19300.00
31150.00
20700.00
35800.00
64900.00
63000.00
33750.00
24450.00
13800.00
7330.00
7645.00
7305.00
9002.50
19700.00
14425.00
25050.00
44725.00
43400.00
23100.00
16575.00
8122.50
5292.50
4757.50
19225.00
37325.00
54050.00
45175.00
50975.00
93150.00
90750.00
54575.00
38925.00
22075.00
15175.00
21650.00
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
26039.03
33731 .91
42761 .05
50733.93
56828.16
74939.27
53288.33
36901.17
32128.54
19391.00
20184.23
26001 .69
16034.54
24238.21
32769.07
32088.13
46881 .28
64475.93
45390.82
25714.54
24520.57
14448.96
9362.29
11403.96
11020.63
10179.07
21414.73
23521.55
35098.74
47051.59
31740.68
15935.99
14349.20
8844.17
6647.42
6133.33
26062.98
40953.88
54042.56
55480.61
66429.31
90047.35
66146.48
46244.86
42832.37
24338.26
20176.83
36050.57
                                              F-42

-------
  t
  I
  D)
  (0
  (0

  Q
              •Observed Flow Duration (10/1/1993 to 9/30/2003 )

               Modeled Flow Duration (10/1/1993 to 9/30/2003 )
      1000000
100000
        10000
         1000
                    10%
                            20%     30%     40%     50%      60%     70%     80%



                                 Percent of Time that Flow is Equaled or Exceeded
                                                                                     90%
                                                                                             100%
Figure 25. Flow exceedence at USGS 01576000 Susquehanna River at Marietta, PA - calibration

    period (HSPF).
    o
    o
    ro

    T3
    o

    I
    -a
    
-------
Table 11.   Summary statistics at USGS 01576000 Susquehanna River at Marietta, PA-
           calibration period (HSPF)
HSPF Simulated Flow
REACH OUTFLOW FROM DSN 101
10-Year Analysis Period: 10/1/1993 - 9/30/2003
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9):
Simulated Fall Flow Volume (months 10-12^
Simulated Winter Flow Volume (months 1-3^:
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Sjpring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of Efficiency, E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
20.55

6.79
3.76

2.87
4.50
7.87
5.30

7.51
1.19
Error Statistics
1.79
14.99
-7.21
34.30
Observed Flow Gage
USGS 01576000 Susquehanna River at Marietta, PA
Hydrologic Unit Code: 2050306
Latitude: 40.054541 3
Longitude: -76.5307992
Drainage Area (sq-rri): 25990
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow VolumeJ10-12):
Observed Winter Flow VolumeJ1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
11.49 »| 30
6.41
-19.82
2.72
56.64
0.771
0.582
0.861
30
30
20
50
Model accuracy increases
as E or E' approaches 1.0





20.19

7.31
3.27

2.14
4.04
7.39
6.62

7.32
0.76





Clear [













                                        F-44

-------
Table 12.    Summary statistics (percent error) for all stations - calibration period (HSPF)
Station
Error in total volume:
Error in 50% lowest
flows:
Error in 10% highest
flows:
Seasonal volume error
- Summer:
Seasonal volume error
-Fall:
Seasonal volume error
-Winter:
Seasonal volume error
- Spring:
Error in storm
volumes:
Error in summer storm
volumes:
Daily Nash-Sutcliffe
Coefficient of
Efficiency, E:
Monthly Nash-Sutcliffe
Coefficient of
Efficiency, E
01562000
Raystown Branch
Juniata River at
Saxton, PA
(1995-2005)
-0.16
6.72
-3.92
23.9
7.76
-0.84
-13.61
-0.63
39.10
0.698
0.898
01540500
Susquehanna River at
Danville, PA
(1993-2003)
6.67
25.34
-4.71
45.64
14.79
13.54
-15.66
7.37
75.91
0.786
0.837
01576000
Susquehanna River at
Marietta, PA
(1993-2003)
1.79
14.99
-7.21
34.30
11.49
6.41
-19.82
2.72
56.64
0.771
0.861
                                           F-45

-------
Table 13.    Summary statistics (percent error) for all stations - validation period (HSPF)
Station
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error -
Summer:
Seasonal volume error -
Fall:
Seasonal volume error -
Winter:
Seasonal volume error -
Spring:
Error in storm volumes:
Error in summer storm
volumes:
Nash-Sutcliffe Coefficient
of Efficiency, E:
Baseline adjusted
coefficient (Garrick), E':
01562000
Ray stow n Branch
Juniata River at Saxton,
PA
(1985-1995)
-8.00
-4.61
-12.82
18.92
7.86
-9.69
-23.40
-14.91
56.16
0.553
0.868
01540500
Susquehanna River at
Danville, PA
(1983-1993)
2.12
30.70
-13.14
54.06
13.13
8.76
-23.37
0.31
85.91
0.714
0.782
01576000
Susquehanna River at
Marietta, PA
(1983-1993)
-1.64
18.99
-13.32
35.02
15.62
2.93
-27.27
-1.26
64.86
0.665
0.777
Water Quality Calibration and Validation
The 20 Watershed models are designed to provide water quality simulation for total suspended solids (TSS), total
nitrogen, and total phosphorus. TSS is simulated with the standard HSPF approach (USEPA 2006). In contrast to
TSS, total nitrogen and total phosphorus are simulated in this application in a simplistic fashion, as HSPF general
quality constituents (GQUALs) subject to an exponential decay rate during transport.

The water quality calibration focuses on the replication of monthly loads, as specified in the project QAPP. Given
the approach to water quality simulation in the 20 Watershed model, a close match to individual concentration
observations cannot be expected. Comparison to monthly loads presents challenges, as monthly loads are not
observed. Instead, monthly loads must be estimated from scattered concentration grab samples and continuous
flow records. As a result, the monthly load calibration is inevitably based on the comparison of two uncertain
numbers. Nonetheless, calibration is able to achieve a reasonable agreement. Further, the load comparisons were
supported by detailed examinations of the relationships of flows to loads and concentrations and the distribution
of concentration prediction errors versus flow, time, and season, as well as standard time series plots.

For application on a nationwide basis, the 20 Watershed protocols assume that TSS and total phosphorus loads
will likely exhibit a strong positive correlation to flow (and associated erosive processes), while total nitrogen
loads, which often have a dominant groundwater component, will not. Accordingly, TSS and total phosphorus
loads were estimated from observations using a flow-stratified log-log regression approach, while total nitrogen
                                                F-46

-------
loads were estimated using a flow-stratified averaging estimator, consistent with the findings of Preston et al.
(1989).

Similar to hydrology, initial calibration of water quality was done on the Raystown Branch Juniata River at
Saxton, PA, comparing model results to data from USGS 01562000. The calibration used the time period 1991-
2000 and 1990 was used for validation.

Results of the TSS calibration are generally acceptable. Visually, the model is roughly simulating the trends
contained in the observed data. A variety of other diagnostics were also pursued to ensure agreement between the
model and observations. These are available in full in the calibration spreadsheets, but a few examples are
provided below. Figure 27 presents the monthly load of TSS. Load-flow power plots were compared for
individual days (Figures 28 and 29). This confirms that the relationship between flow and load is consistent across
the entire range of observed flows, for both the calibration and validation periods. Tables 14 and  15 provide
model statistics and relative errors for the TSS calibration and validation periods.
                                                TSS
     1,000,000
                                                                                     -Regression Loads
                                                                                     -Simulated Loads
             CpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpO  O

              CO —.   CD —.   CD —.   CD —.   CD  —.CD  —.   CD —.   CD  —.   CD  —.   CD  —.CD  —.
             -—i  J   —\  J   —~i  J   —~i  J  —~i   J  —~i   J  —\  J  —~i   J  —~i   J  —~i   J  —~i   J
    Figure 27. Fit for monthly load of TSS at Raystown Branch Juniata River at Saxton, PA (HSPF).
Table 14.    Model fit statistics (observed minus predicted) for monthly sediment loads using
             stratified regression - USGS 01562000 Raystown Branch Juniata River at Saxton,
             PA (HSPF)
Statistic
Relative Percent Error
Relative Average Absolute Error
Relative Median Absolute Error
Calibration period
(1991-2000)
-78.2%
146%
20.3%
Validation period
(1990)
-89.7%
124%
58.3%
                                                F-47

-------
             Raystown Branch Juniata River at Saxton, PA 1991-2000
"35
c
o
•o
ra
o
V)
                       10
100         1000


    Flow, cfs
10000
100000
            •  Simulated  A Observed ^^™Power (Simulated) ^^™Power (Observed)
Figure 28. Power plot for observed and simulated TSS at Raystown Branch Juniata River at Saxton,

    PA - calibration period (HSPF).
                Raystown Branch Juniata River at Saxton, PA 1990-1990
       10000
                                                                          10000
              •  Simulated  A Observed ^^™Power (Simulated) ^^™Power (Observed)
Figure 29. Power plot for observed and simulated TSS at Raystown Branch Juniata River at Saxton,

    PA - validation period (HSPF).
                                           F-48

-------
Standard time series plots (Figure 30) show that observed and simulated concentrations achieve good agreement,
although individual observations may deviate. Plots of concentration error versus flow and versus month (not
shown) were used to guard against hydrologic and temporal bias.
            Raystown Branch Juniata River at Saxton, PA
                                                              • Simulated  A  Observed
       10000
        1000
    O)

   eo
   CO
         100
            1993     1994      1995     1996     1997
                                            Year
1998
1999
    Figure 30. Time series plot of TSS concentration at Raystown Branch Juniata River at Saxton, PA
       calibration period (HSPF).

Table 15.    Relative errors,(observed minus simulated) for TSS concentrations at USGS
             01562000 Raystown Branch Juniata River at Saxton, PA (HSPF)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1991-2000)
106
-138.4%
-29.73%
Validation period
(1990)
15
-30.82%
24.63%
The total phosphorus load calibration performed well for the Susquehanna River watershed calibration focus area.
Adjustments were made to the accumulation rate and storage limits for the impervious surfaces. In general, the
observed and simulated total phosphorus loads attained an acceptable match for the simulation period (Figure 31).
As with TSS, additional diagnostics for total phosphorus included flow-load power plots (Figures 32 and 33) and
time series plots (Figure 34). All show acceptable agreement. Tables 16 and 17 provide model statistics and
relative errors for the total phosphorus calibration and validation periods. In contrast to load, phosphorus
concentrations are generally over-estimated (observed minus simulated concentration less than zero). This is due
to an over-estimation of observed phosphorus concentrations at low flows that may be due to the simplistic
representation of point source discharges in the model.
                                               F-49

-------
   o
   E
   In
   I
                                          Total P
      100
-Regression Loads
-Simulated Loads
   Figure 31. Fit for monthly load of total phosphorus at Raystown Branch Juniata River at Saxton, PA
      (HSPF).
Table 16.    Model fit statistics (observed minus predicted) for monthly total phosphorus loads
            using stratified regression - USGS 01562000 Raystown Branch Juniata River at
            Saxton,  PA (HSPF)
Statistic
Relative Percent Error
Relative Average Absolute Error
Relative Median Absolute Error
Calibration period
(1991-2000)
26.0%
49%
22.7%
Validation period
(1990)
21.5%
45%
35.4%
                                           F-50

-------
             Raystown Branch Juniata River at Saxton, PA 1991-2000
                      10
100         1000


    Flow, cfs
10000
                                100000
           •  Simulated  A  Observed ^^~Power (Simulated) ^^™ Power (Observed)
Figure 32. Power plot for observed and simulated total phosphorus at Raystown Branch Juniata

    River at Saxton, PA - calibration period (HSPF).
             Raystown Branch Juniata River at Saxton, PA 1990-1990
      10
>
ra
•o

«
c
o
•o
re
o
     0.1
    0.01
   0.001
                        10
   100


Flow, cfs
                     1000
             10000
          »  Simulated  A Observed ^^~ Power (Simulated) ^^~ Power (Observed)
Figure 33. Power plot for observed and simulated total phosphorus at Raystown Branch Juniata

    River at Saxton, PA - validation period (HSPF).
                                           F-51

-------
                    Raystown Branch Juniata River at Saxton, PA
1994
1995
1996      1997
    Year
                                                        1998
1999
    Figure 34.  Time series plot of total phosphorus concentration, at Raystown Branch Juniata River at
       Saxton, PA (HSPF).

Table 17.   Relative errors (observed minus simulated) for total phosphorus concentrations at
            USGS 01562000 Raystown Branch Juniata River at Saxton, PA (HSPF)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1990-1990)
122
-121.72%
-50.63%
Validation period
(1990)
18
-0.88%
-12.24%
Results for total nitrogen are summarized in Figures 35 through 38. The results are acceptable, and generally
better than those for total phosphorus. This is due to total nitrogen not being sediment associated, therefore,
problems with sediment are not reflected in the calibration for total nitrogen. Tables 18 and 19 provide model
statistics and relative errors for the total nitrogen calibration and validation periods.
                                              F-52

-------
                                          Total N
     1,400
                                                                            -Averaging Loads
                                                                            -Simulated Loads
   Figure 35.  Fit for monthly load of total nitrogen at Raystown Branch Juniata River at Saxton, PA
      (HSPF).

Table 18.   Model fit statistics (observed minus predicted) for monthly total nitrogen loads
           using averaging estimator- USGS 01562000 Raystown Branch Juniata River at
           Saxton,  PA (HSPF)
Statistic
Relative Percent Error
Relative Average Absolute Error
Relative Median Absolute Error
Calibration period
(1991-2000)
7.0%
34%
16.8%
Validation period
(1990)
17.2%
29%
26.2%
                                           F-53

-------
             Raystown Branch Juniata River at Saxton, PA 1991-2000
   1000
   0.01
                    10
100         1000

    Flow, cfs
10000       100000
          •  Simulated  A  Observed ^^™Power (Simulated) ^^™Power (Observed)
Figure 36. Power plot for observed and simulated total nitrogen at Raystown Branch Juniata River at
   Saxton, PA - calibration period (HSPF).
            Raystown Branch Juniata River at Saxton, PA 1990-1990
   100
                                                                      10000
           Simulated  A  Observed ^^™Power (Simulated) ^^™Power (Observed)
Figure 37. Power plot for observed and simulated total nitrogen at Raystown Branch Juniata River at
   Saxton, PA - validation period (HSPF).
                                           F-54

-------
                 Raystown Branch Juniata River at Saxton, PA
                                                         • Simulated  A  Observed
   O)
20
18
16
14
12
10
 8
 6
 4
 2
        1986     1987     1988     1989      1990
                                        Year
                                                1991
1992
   Figure 38. Time series plot of total nitrogen concentration, at Raystown Branch Juniata River at
       Saxton, PA (HSPF).
Table 19.   Relative errors (observed minus simulated) for total nitrogen concentration at
            USGS 01562000 Raystown Branch Juniata River at Saxton, PA (HSPF)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1991-2000)
13
-22.89%
-22.60%
Validation period
(1990)
6
11.10%
7.42%
Water Quality Results for Larger Watershed
As with hydrology, the Raystown Branch Juniata River (USGS 01562000) watershed parameters for water quality
were directly transferred to other portions of the watershed. Summary statistics for the water quality calibration
and validation at other stations in the watershed are provided in Tables 20 and 21, respectively.
                                             F-55

-------
Table 20.   Summary statistics for water quality for all stations - calibration period (HSPF)
Station
Relative Percent Error TSS Load
TSS Concentration Median Error
Relative Percent Error TP Load
TP Concentration Median Error
Relative Percent Error TN Load
TN Concentration Median Error
01576000
Susquehanna
River at Marietta,
PA
(1991-1995)
26.5%
-24.8%
44.0%
-1.0%
-14.4%
-34.8%
01540500
Susquehanna River at
Danville, PA
(1991-1994)
27.5%
-6.3%
50.0%
24.6%
6.2%
-4.8%
01562000
Raystown Branch
Juniata River at Saxton,
PA
(1991-2000)
-78.2%
-29.7%
26.0%
-50.6%
7.0%
-22.6%
Table 21.   Summary statistics for water quality for all stations - validation period (HSPF)
Station
Relative Percent Error TSS Load
TSS Concentration Median Error
Relative Percent Error TP Load
TP Concentration Median Error
Relative Percent Error TN Load
TN Concentration Median Error
01576000
Susquehanna
River at Marietta,
PA
(1980-90)
-0.6%
-27.3%
38.8%
5.1%
-7.1%
-17.2%
01540500
Susquehanna River at
Danville, PA
(1986-90)
-11.1%
-1 1 .2%
40.5%
20.7%
10.0%
5.9%
01562000
Raystown Branch
Juniata River at Saxton,
PA
(1990)
-89.7%
24.6%
21.5%
-12.2%
17.2%
7.4%
                                           F-56

-------
 a a a a a a a a  a a a a a a a a a a a a a  a a a a a a a a a a a a a  a a a a a a  a a a a a a a a a a a a a  a a a a a a a a a a a a a  a a a a a a a a a a a a a  a a a a a a  a si
The SWAT model for the Susquehanna River watershed was set up with the ArcSWAT Version 2.3.3 interface
using the subwatersheds and stream network layers obtained from CBP, and other geospatial coverages described
above for the HSPF model. The precipitation and temperature data were preprocessed from BASINS Weather
Data Management (WDM) files to obtain the daily values.

The SWAT modeling process started with hydrology calibration, followed by calibration of sediment, and then
calibration of nitrogen and phosphorous. The USGS gage on the Raystown Branch of the Juniata River at Saxton,
PA (USGS 02050303) was used as the initial calibration location. The parameters were then transferred to the
entire Susquehanna River watershed and results were evaluated at the gages on the Susquehanna River at
Danville, PA (USGS 01540500) and on the Susquehanna River at Marietta, PA (USGS 01576000). While
hydrology parameters were readily transferable, water quality parameters, especially those related to sediment
needed some adjustment at the larger watershed scale.
No changes were made to the meteorological or land use base data for the SWAT model.
Though there are a number of reservoirs present in Susquehanna River watershed, they are located below the
calibration and validation locations. Hence the information regarding these reservoirs was not specifically
addressed in the SWAT model, consistent with the approach used for HSPF.

The point source data were specified for all the major active point sources in the Susquehanna study area. The
point source flows and concentrations for each facility in the watershed were assumed to be constant throughout
the simulation period. The data from the time period from 1991-2006 were compiled from the PCS database and
the median values were estimated. The facilities that were missing a total nitrogen, total phosphorus, and TSS
concentration value were filled with a typical pollutant concentration value from literature (Typical Pollutant
Concentration for NCPDI Discharge Categories -Improving Point Source Loading  Data for Reporting National
Water Quality Indicators) prepared for Jim Home, EPA/OWM (Tetra Tech 1990) based on the SIC
classification. All POTWs were assumed to have secondary treatment. The median concentrations for the nutrient
species were estimated based on the values reported in Cheaspake Bay Phase 5 Model report for species
relationship for point sources and used in the model.
Similar to HSPF, hydrology calibration was performed at the USGS gage on the Raystown Branch of the Juniata
River at Saxton, PA (USGS 01562000). Though some adjustments are made at the major watershed level, a
spatial calibration approach was not adopted for Susquehanna River watershed SWAT modeling. The calibration
efforts were geared toward getting a closer match between simulated and observed flows at the outlet of the
calibration focus area and to limit the error statistics within the acceptable ranges listed in the QAPP.
                                                F-57

-------
Land Use/Soil/Slope definition
A 5/10/5 percent threshold was used for land use/soil/slope in the SWAT model while defining the HRUs. Urban
(including current and future urban class types) classes were exempt from applying the thresholds.

Elevation Bands
The topographical analysis of Susquehanna River watersheds showed a significant range of elevations within
some individual modeling subwatersheds. This is likely to result in orographic variability in precipitation. Eight
elevation bands were used to account for the orographic effects on temperature and precipitation.

Calibration Parameters
The initial values of the parameters were set by ArcGIS based on various geospatial datasets and the defaults set
in the SWAT database. During the calibration process, adjustments were focused on the following parameters:

    •   ICN (Daily curve number calculation method) - In order to make the CN less dependent on the soil
       moisture content and more dependent on the antecedent conditions, the ICN was set to 1 (CN as function
       orET)
    •   FFCB (Initial soil water storage expressed as a fraction of field capacity water content)
    •   CN_FROZ (Frozen curve number active)
    •   Tlaps (Temperature laps rate)
    •   Flaps (Preciptation laps rate)
    •   ALAI_MIN (Minimum leaf area index for plant during dormant period)
    •   CurYr_Mat (Current age of trees)
    •   LAI_Ini (Initial leaf area index)
    •   EPCO (Plant uptake compensation factor)
    •   CN (Curve Number)
    •   TIMP (Snow pack temperature lag factor)
    •   ESCO (Soil evaporation compensation factor)
    •   CANMX (Maximum canopy storage)
    •   Alpha_bf (Baseflow alpha factor)
    •   GW_Delay (Groundwater delay)
    •   GWQMN (Threshold depth of water in the shallow aquifer for return flow to occur)
    •   REVAPMN (Threshold depth of water in the shallow aquifer for revap to occur)
    •   GW_REVAP (Groundwater revap coefficient)
    •   SURLAG (Surface runoff lag time)
    •   CH_K1 (Effective hydraulic conductivity in tributary channel alluvium)
    •   CH_K2 (Effective hydraulic conductivity in main channel alluvium)
    •   SFTMP (Snowfall temperature)
    •   SMTMP (Snow melt base temperature)
    •   SMFMX (Maximum melt rate for snow during year)
    •   SMFMN (Minimum melt rate for snow during year)
    •   CNCOEFF (Plant ET curve number coefficient)
    •   CH_N2 (Manning's n value for the main channel)
    •   CH_N1 (Manning's "n" value for the tributary channels)
    •   HRU_Slope (Average slope steepness)
    •   Slsubbsn (Average slope length)
                                               F-58

-------
Initial calibrations were performed for the Raystown Branch of the Juniata River, comparing model results to data
from USGS 01562000 (Raystown Branch Juniata River At Saxton, PA), and are summarized in Figures 39
through 45 and Tables 22 and 23. The model fit is of good quality, but summer volumes are over estimated. The
model calibration period was set to the 10 water years from 10/01/1995 to 09/30/2005.
          Avg Monthly Rainfall (in)
         -Avg Observed Flow (10/1/1995 to 9/30/2005 )
         •Avg Modeled Flow (Same Period)
    35000

    30000

_  2500°
o,  20000

1   15000
LJ_
    10000
       5000 4-

                                                                           IWT
                                                                                   - 2
                                                                           Oct-04
    Figure 39. Mean daily flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
       calibration period (SWAT).
      5000
                                                        Avg Monthly Rainfall (in)
                                                        -Avg Observed Flow (10/1/1995 to 9/30/2005 )
                                                        •Avg Modeled Flow (Same Period)
         0-95
                   A-97
0-04
    Figure 40. Mean monthly flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
       calibration period (SWAT).
                                                F-59

-------
           Avg Flow (10/1/1995 to 9/30/2005 )
           • Line of Equal Value
           Best-Fit Line
   5000
                                         100% -,
                                      "g  90% -
                                      5  80% -
                                      en  70% -
                                      O^  60% -
                                      g  50% -
                                      j|  40% -
                                      ,S  30% -
                                      jy  20% -
                                      ;>  10% -
                                           0%
                                                Avg Observed Flow (10/1/1995 to 9/30/2005 )
                                                Avg Modeled Flow (10/1/1995 to 9/30/2005 )
                                               -Line of Equal Value
                                                             IJ      i
                                                                 mil
       0   1000  2000   3000   4000  5000
          Average Observed Flow (cfs)
                                             O-95  A-97  O-98  A-00   O-01
                                                             Month
                                                                       A-03  O-04
  Figure 41. Monthly flow regression and temporal variation at USGS 01562000 Raystown Branch
     Juniata River at Saxton, PA - calibration period (SWAT).
      •  Avg Flow (10/1 /1995 to 9/30/2005)
      • - - - • Line of Equal Value
         Best-Fit Line
                                              Avg Monthly Rainfall (in)
                                             -Avg Observed Flow (10/1/1995 to 9/30/2005)
                                             •Avg Modeled Flow (Same Period)
  2500
I
o
2000
-o 1500
J5

-------
          Average Monthly Rainfall (in)
         - Median Observed Flow (10/1/1995 to 9/30/2005)
[Observed (25th, 75th)
 Modeled (Median, 25th, 75th)
    3000 i
    2500
    2000
    1500 -
    1000 - —
     500 - —
           10     11
   Figure 43. Seasonal medians and ranges at USGS 01562000 Raystown Branch Juniata River at
       Saxton, PA - calibration period (SWAT).
Table 22.   Seasonal summary at USGS 01562000 Raystown Branch Juniata River at Saxton,
            PA - calibration period (SWAT)
MONTH

Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
438.69
1024.44
1071.22
1255.25
1290.35
2177.77
1771.52
1310.00
895.83
326.40
278.62
907.63
187.00
647.50
682.00
475.00
900.00
1690.00
1180.00
739.50
448.00
207.00
157.50
124.50
119.00
149.75
280.75
260.00
470.00
1052.50
843.75
414.00
287.50
146.00
115.25
104.00
530.75
1170.00
1350.00
1180.00
1775.00
2742.50
2102.50
1777.50
879.00
324.00
339.75
360.25
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
692.82
1034.60
969.20
998.54
1120.65
1804.98
1425.47
1035.99
999.51
440.54
404.37
1129.77
371.51
754.67
765.27
505.18
539.96
1195.40
986.87
630.37
535.90
222.43
249.52
239.93
162.66
141.71
263.19
171.86
290.45
691.99
613.33
283.51
258.19
144.38
100.82
73.81
959.94
1346.72
1190.02
1224.18
935.31
2568.17
1666.15
1281.83
1082.22
426.51
519.83
1015.30
                                              F-61

-------
t
I
D)
(0
            •Observed Flow Duration (10/1/1995 to 9/30/2005 )

            •Modeled Flow Duration (10/1/1995 to 9/30/2005 )
    100000
     10000
      1000 -
       100 -T
                 10%
                         20%      30%     40%     50%     60%     70%     80%



                              Percent of Time that Flow is Equaled or Exceeded
                                                                                  90%
                                                                                          100%
  Figure 44.  Flow exceedence at USGS 01562000 Raystown Branch Juniata River at Saxton, PA

      calibration period (SWAT).
  o
  o
  ro

  T3
  o

  I
  -a
  
-------
Table 23.    Summary statistics at USGS 01562000 Raystown Branch Juniata River at Saxton,
             PA - calibration period (SWAT)

REACH OUTFLOW FROM OUTLET 26
10-Year Analysis F^riod: 10/1/1995 - 9/30/2005
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9)
Simulated Fall Flow Volume (months 10-12^
Simulated Winter Flow Volume (months 1-3):
Simulated Srjring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error -Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
Nash^3ut£liffeJDpj5fficje^^
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
18.01

7.87
1.94

2.96
4.06
5.83
5.16

6.86
1.30
Error Statistics
-5.41
-5.67
-7.15
30.66
Observed Flow Gage
USGS 01562000 Raystown Bran
Hydrologic Unit Code: 2050303
Latitude: 40.21591249
Longitude: -78.2652901
Drainage Area (sq-rri): 756

ch Juniata River at Saxton, PA
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow VolumeJ10-12):
Observed Winter Flow Volume (1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
6.47 » | 30
-17.01
-13.07
-12.43
11.80
0.294
0.395
0.669
30
30
20
50
Model accuracy increases
as E or E' approaches 1.0





19.04

8.48
2.05

2.26
3.81
7.03
5.93

7.83
1.16





Clear [













Hydrology Validation
Hydrology validation for the Susquehanna River watershed model was performed at the same gage location but
for the period 10/1/1985 through 9/30/1995. Results are presented in Figure 49 through 52 and Tables 24 and 25.
The validation achieves a reasonable coefficient of model fit efficiency, but many of statistics show that simulated
values were underestimated compared to the observed values.

In general, the sign of the errors in the validation period are similar to those in the calibration period, but the
discrepancies are larger. Additional factors that may have contributed to the difference in the flows between the
calibration and validation period are:
    •   Drainage area of the observed USGS gage is about 6% higher than that of the calibration watershed.
    •   Increase in urban impervious surface areas between the 1980s and present.
                                               F-63

-------
      Avg Monthly Rainfall (in)
      -Avg Observed Flow (10/1/1985 to 9/30/1995 )
      •Avg Modeled Flow (Same Period)
  30000

                 Apr-87      Oct-88      Apr-90      Oct-91

                                            Date
                                             Apr-93      Oct-94
Figure 46. Mean daily flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
    validation period (SWAT).
  8000
                                                      Avg Monthly Rainfall (in)
                                                     -Avg Observed Flow (10/1/1985 to 9/30/1995 )
                                                     •Avg Modeled Flow (Same Period)
     O-85
A-87
O-94
Figure 47. Mean monthly flow at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
    validation period (SWAT).
                                              F-64

-------
   8000
             Avg Flow (10/1/1985 to 9/30/1995 )
            • Line of Equal Value
             Best-Fit Line
             2000     4000    6000     8000
          Average Observed Flow (cfs)
                                                  100% -,
                                              +
                                              &
                                               0)
                                               O
                                               ro
                                              m
                                                       Avg Observed Flow (10/1/1985 to 9/30/1995 )
                                                       Avg Modeled Flow (10/1/1985 to 9/30/1995 )
                                                      -Line of Equal Value
                                                            A-87   O-88
                                                                        A-90   O-91
                                                                         Month
                                                                                    A-93   O-94
   Figure 48. Monthly flow regression and temporal variation at USGS 01562000 Raystown Branch
      Juniata River at Saxton, PA - validation period (SWAT).
       •   Avg Flow (10/1 /1985 to 9/30/1995)
       • - - - • Line of Equal Value
           Best-Fit Line
  2500
I
o
2000
-o 1500

-------
      • Observed (25th, 75th)  Average Monthly Rainfall (in) -Median Observed Flow (10/1/1985 to 9/30/1995)  Modeled (Median, 25th, 75th)
    2500
            10     11     12      1
8      9
   Figure 50. Seasonal medians and ranges at USGS 01562000 Raystown Branch Juniata River at
       Saxton, PA - validation period (SWAT).
Table 24.   Seasonal summary at USGS 01562000 Raystown Branch Juniata River at Saxton,
            PA - validation period (SWAT)
MONTH

Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
351 .63
771 .90
936.58
1063.43
1214.57
2016.51
1772.93
1102.77
559.15
545.15
252.88
195.59
128.00
41 1 .00
603.50
562.00
754.00
1240.00
965.00
676.00
275.50
180.00
136.00
160.00
104.00
211.75
302.50
322.50
410.00
670.75
613.50
479.25
201.00
138.00
108.00
107.75
200.00
649.00
1157.50
1167.50
1495.00
2140.00
1935.00
1122.50
479.50
418.00
216.75
233.50
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
479.66
769.29
790.68
731 .64
615.95
1497.03
1417.56
652.39
552.94
739.08
441 .53
310.53
229.97
500.41
506.59
553.91
454.68
838.37
554.79
355.80
232.64
192.13
223.70
246.34
83.68
209.87
271.91
238.23
238.94
522.30
298.14
230.15
136.75
108.06
89.30
56.63
417.42
857.53
1027.22
916.06
827.25
1559.50
1296.67
793.79
476.75
694.82
404.09
481 .69
                                             F-66

-------
t
I
D)
(0
(0
Q
            •Observed Flow Duration (10/1/1985 to 9/30/1995 )

            •Modeled Flow Duration (10/1/1985 to 9/30/1995 )
    100000
     10000
1000
 100 -E
                 10%
                         20%
                                 30%
                                          40%
                                                  50%
                                                          60%
                                                                  70%
                                                                          80%
                                                                                  90%
                                                                                          100%
                              Percent of Time that Flow is Equaled or Exceeded
  Figure 51.  Flow exceedence at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
      validation period (SWAT).
                •Observed Flow Volume (10/1/1985 to 9/30/1995 )

                Modeled Flow Volume (10/1/1985 to 9/30/1995 )
  o
  o
  ro
  T3
  _a
  O
  _
  o


  o
  ro
       120%
       100%
        80%
        60%
  40%
        20%
          Oct-85
                      Apr-87
                            Oct-88        Apr-90
                                                          Oct-91
Apr-93       Oct-94
  Figure 52.  Flow accumulation at USGS 01562000 Raystown Branch Juniata River at Saxton, PA
      validation period (SWAT).
                                                F-67

-------
Table 25.    Summary statistics at USGS 01562000 Raystown Branch Juniata River at Saxton,
             PA - validation period (SWAT)
SWAT Simulated Flow
REACH OUTFLOW FROM OUTLET 26
10-Year Analysis F^riod: 10/1/1985 - 9/30/1995
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9)
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3):
Simulated Spring Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error -Summer:
Seasonal volume error - Fall:
Seasonal volume error- Winter:
Seasonal volume error - Sjpring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of Efficiency, E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
13.49

6.35
1.59

2.26
3.07
4.25
3.90

4.60
0.77
Error Statistics
-16.30
-12.87
-18.30
50.02
Observed Flow Gage
USGS 01562000 Raystown Branch Juniata River at Saxton, PA
Hydrologic Unit Code: 2050303
Latitude: 40.21591249
Longitude: -78.2652901
Drainage Area (sq-rri): 756
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow VolumeJ1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
-1.00 » | 30
-33.35
-23.82
-29.17
42.83
0.415
0.383
0.664
30
30
20
50
Model accuracy increases
as E or E' approaches 1.0




16.12

7.77
1.82

1.51
3.10
6.38
5.12

6.49
0.54




Clear [













Hydrology Results for Larger Watershed
The parameters determined for the Raystown Branch of the Juniata River gage were transferred to the remainder
of the watershed without detailed spatial calibration. Tests of calibration and validation were pursued at a total of
three gages throughout the watershed, all of them at the outlet of 8-digit HUCs. Calibration results were generally
acceptable at all gages, as summarized in Table 26. The match between observed and predicted flow
corresponding to the largest watershed of the three gages (USGS 01576000, Susquehanna River at Marietta) is
shown in Figures 53 through 59 and Tables 27 and 28. Validation results were also generally in the acceptable
range for all the gages, as summarized in Table 29. It appears, however, that there are some systematic biases in
the model, including under-prediction of the 10 percent highest flows and winter flows, coupled with over-
prediction of summer flows
                                              F-68

-------
Table 26.    Summary statistics (percent error) for all stations - calibration period (SWAT)
Station
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
Daily Nash-Sutcliffe Coefficient of
Efficiency, E:
Monthly Nash-Sutcliffe Coefficient of
Efficiency,
01576000
Susquehanna River
at Marietta, PA
(1993-2003)
-9.74
2.03
-19.80
15.58
-2.46
-31.54
1.99
-22.08
9.52
0.451
0.669
01540500
Susquehanna River
at Danville, PA
(1993-2003)
-4.51
5.90
-11.24
11.96
-4.24
-38.92
26.18
-29.54
-12.84
0.327
0.573
01562000 Raystown
Branch Juniata River
at Saxton, PA
(1995-2005)
-5.41
-5.67
-7.15
30.66
6.47
-17.01
-13.07
-12.43
11.80
0.294
0.669
 I
 I
          Avg Monthly Rainfall (in)
         -Avg Observed Flow (10/1/1993 to 9/30/2003 )
         •Avg Modeled Flow (Same Period)
     600000
      500000
400000

300000

200000

100000

                        T

                                    TTNiT1
                                    	T	1—1___^	r—_|___T	,_
Oct-93      Apr-95
                                Oct-96      Apr-98      Oct-99

                                               Date
Apr-01
                                                                      Oct-02
    Figure 53. Mean daily flow at USGS 01576000 Susquehanna River At Marietta, PA - calibration period
       (SWAT).
                                                F-69

-------
o
    200000
    150000 --
    100000 -----
     50000 -
          O-93
                                                          Avg Monthly Rainfall (in)
                                                         -Avg Observed Flow (10/1/1993 to 9/30/2003
                                                         •Avg Modeled Flow (Same Period)
            A-95
                                                                             O-02
  Figure 54. Mean monthly flow at USGS 01576000 Susquehanna River at Marietta, PA - calibration
      period (SWAT).
            Avg Flow (10/1/1993 to 9/30/2003 )
            • Line of Equal Value
            Best-Fit Line
   200000
|
150000 -
T3
O
D)
   100000 -
    50000
          y = 0.7524X + 5781.3
              R2 = 0.7671
0     50000   100000   150000  200000
   Average Observed Flow(cfs)
                                              £
                                              O
                                              _
                                              ro
                                              m
                                              o>
                                              -t-j
                                              TO
                                        100% -,
                                         90%
                                         80%
                                         70% -
                                         60%
                                         50%
                                                Avg Observed Flow (10/1/1993 to 9/30/2003 )
                                                Avg Modeled Flow (10/1/1993 to 9/30/2003 )
                                               -Line of Equal Value
                                                     O-93  A-95  O-96
                                                                       A-98  O-99
                                                                        Month
                                                                                    A-01   O-02
  Figure 55. Monthly flow regression and temporal variation at USGS 01576000 Susquehanna River at
      Marietta, PA - calibration period (SWAT).
                                                  F-70

-------
           Avg Flow (10/1/1993 to 9/30/2003)
          • Line of Equal Value
           Best-Fit Line
  100000
•6
             20000  40000  60000  80000 10000
                                       0
             Average Observed Flow (cfs)
 Avg Monthly Rainfall (in)
-Avg Observed Flow (10/1/1993 to 9/30/2003)
 Avg Modeled Flow (Same Period)
                                              o
   10 11 12  1  2 3  4  5  6  7  8  9
                Month
  Figure 56. Seasonal regression and temporal aggregate at USGS 01576000 Susquehanna River at
      Marietta, PA - calibration period (SWAT).
          • Observed (25th, 75th)
          •Median Observed Flow (10/1/1993 to 9/30/2003)
       Average Monthly Rainfall (in)
       Modeled (Median, 25th, 75th)
    120000
o
              10      11      12     1
  Figure 57. Seasonal medians and ranges at USGS 01576000 Susquehanna River at Marietta, PA-
      calibration period (SWAT).
                                                 F-71

-------
Table 27. Seasonal summary at USGS 01576000 Susquehanna River at Marietta, PA - calibration
             period (SWAT)
MONTH
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
18258.39
29811.13
43879.65
46873.61
45647.16
78282.90
74228.00
46236.13
32080.10
16327.52
15446.55
16977.37
10650.00
19300.00
31150.00
20700.00
35800.00
64900.00
63000.00
33750.00
24450.00
13800.00
7330.00
7645.00
7305.00
9002.50
19700.00
14425.00
25050.00
44725.00
43400.00
23100.00
16575.00
8122.50
5292.50
4757.50
19225.00
37325.00
54050.00
45175.00
50975.00
93150.00
90750.00
54575.00
38925.00
22075.00
15175.00
21650.00
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
22700.72
32093.73
34985.85
27590.78
29342.81
59826.28
74819.70
47504.67
33247.00
16522.99
17697.63
22211.19
16857.46
22972.19
28272.92
19412.47
19732.07
54243.33
68934.23
39005.05
26980.41
14343.05
9907.53
12176.50
11114.41
12402.51
19425.72
14285.67
13347.18
29365.91
41733.11
21727.35
14564.65
8986.70
6522.62
7015.26
28029.25
41600.68
39499.45
32425.04
30222.29
81082.47
95561 .49
63884.23
43437.04
20941 .60
17877.17
30420.94
             •Observed Flow Duration (10/1/1993 to 9/30/2003 )
             Modeled Flow Duration (10/1/1993 to 9/30/2003 )
      1000000
  o
  D)
  ro
  (D
  ro
  Q
      100000
       10000
        1000
                  10%
                          20%    30%     40%     50%    60%     70%    80%

                              Percent of Time that Flow is Equaled or Exceeded
                                                                              90%
                                                                                     100%
   Figure 58.  Flow exceedence at USGS 01576000 Susquehanna River at Marietta, PA - calibration
       period  (SWAT).
                                              F-72

-------
o
o
ro
T3
o

^3
LL
T3
N
15
              •Observed Flow Volume (10/1/1993 to 9/30/2003 )
               Modeled Flow Volume (10/1/1993 to 9/30/2003 )
     120%
     100% -
      80%
      60%
      40% - -
      20%  —
                    Apr-95
                                Oct-96
                                             Apr-98
                                                         Oct-99
Apr-01
                                                                                 Oct-02
Figure 59. Flow accumulation at USGS 01576000 Susquehanna River at Marietta, PA - calibration
    period (SWAT).
                                               F-73

-------
Table 28. Summary statistics at USGS 01576000 Susquehanna River at Marietta, PA - calibration
           period (SWAT)
SWAT Simulated Flow
REACH OUTFLOW FROM OUTLET 75
10-Year Analysis F^riod: 10/1/1993 - 9/30/2003
Flow volumes are (inches/year) for upstream drainage area
Total Simulated In-stream Flow:

Total of simulated highest 10% flows:
Total of Simulated lowest 50% flows:

Simulated Summer Flow Volume (months 7-9):
Simulated Fall Flow Volume (months 10-12):
Simulated Winter Flow Volume (months 1-3):
Simulated Spjinj Flow Volume (months 4-6):

Total Simulated Storm Volume:
Simulated Summer Storm Volume (7-9):
Errors (Simulated-Observed)
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error - Summer:
Seasonal volume error - Fall:
Seasonal volume error -Winter:
Seasonal volume error - Sjpring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of Efficiency, E:
Baseline adjusted coefficient (Garrick), E':
Monthly NSE
18.22

5.86
3.34

2.47
3.94
5.06
6.75

5.70
0.84
Error Statistics
-9.74
* 2.03

15.58
Observed Flow Gage
USGS 01576000 Susquehanna River at Marietta, PA
Hydrologic Unit Code: 2050306
Latitude: 40.0545413
Longitude: -76.5307992
Drainage Area (sq-rri): 25990
Total Observed In-stream Flow:

Total of Observed highest 10% flows:
Total of Observed Lowest 50% flows:

Observed Summer Flow Volume (7-9):
Observed Fall Flow Volume (10-12):
Observed Winter Flow VolumeJ1-3):
Observed Spring Flow Volume_(4-6):

Total Observed Storm Volume:
Observed Summer Storm Volume (7-9):
Recommended Criteria
10
10
15
30
-2.46 » | 30
-31.54
1.99
-22.08
9.52
0.451
0.443
0.641
30
30
20
50
Model accuracy increases
as E or E' approaches 1 .0





20.19

7.31
3.27

2.14
4.04
7.39
6.62

7.32
0.76





Clear [













                                          F-74

-------
Table 29.    Summary statistics (percent error) for all stations - validation period (SWAT)
Station
Error in total volume:
Error in 50% lowest flows:
Error in 10% highest flows:
Seasonal volume error -
Summer:
Seasonal volume error - Fall:
Seasonal volume error - Winter:
Seasonal volume error - Spring:
Error in storm volumes:
Error in summer storm volumes:
Nash-Sutcliffe Coefficient of
Efficiency, E:
Baseline adjusted coefficient
(Garrick), E':
01576000 Susquehanna
River at Marietta, PA
(1983-1993)
-15.17
1.37
-24.62
22.63
-8.86
-34.95
-12.26
-28.54
29.42
0.485
0.657
01540500
Susquehanna River
at Danville, PA
(1983-1993)
-10.17
1.84
-17.11
17.20
-11.29
-40.55
9.55
-36.20
-15.18
0.372
0.573
01562000 Raystown
Branch Juniata River
at Saxton, PA
(1985-1995)
-16.30
-12.87
-18.30
50.02
-1.00
-33.35
-23.82
-29.17
42.83
0.415
0.664
Water Quality Calibration and Validation
Initial calibration and validation of water quality was done on data from the Raystown Branch Juniata River at
Saxton (USGS 01562000), using 1991-2000 for calibration and 1990 for validation. As with hydrology,
calibration was performed on the later period as this better reflects the land use included in the model. The start of
the validation period is constrained by data availability.

Calibration adjustments for sediment focused on the following parameters:
    •   SPCON (Linear parameters for estimating maximum amount of sediment that can be re-entrained during
       channel  sediment routing)
    •   SPEXP  (Exponent parameter for calculating sediment re-entrained in channel sediment routing)
    •   CH_COV (Channel cover factor)
    •   CH_EROD (Channel credibility factor)

Various plots that compare TSS simulated by SWAT against the observed data are shown in Figures 60 through
63. The comparison statistics are provided in Tables 30 and 31. The fit to monthly sediment loads is generally
better than that obtained with the HSPF model. However, the correlation between observed and predicted
concentrations is weak - in part because many of the observed data are reported as less than a detection limit of 2
mg/L (plotted at 1 mg/L in Figure 63).
                                               F-75

-------
                                               TSS
        100,000
              CpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpCpOO
                                                                                 -Regression Loads
                                                                                 -Simulated Loads
   Figure 60. Fit for monthly load of TSS at USGS 01562000 Raystown Branch Juniata River at Saxton,
       PA (SWAT).
Table 30.    Model fit statistics (observed minus predicted) for monthly sediment loads using
            stratified regression - USGS 01562000 Raystown Branch Juniata River at Saxton,
            PA (SWAT)
Statistic
Relative Percent Error
Relative Average Absolute Error
Relative Median Absolute Error
Calibration period
(1991-2000)
-10.1%
80%
11.1%
Validation period
(1990)
-33.6%
67%
41.1%
                                            F-76

-------
            Raystown Branch Juniata River at Saxton, PA 1991-2002
•o

«
c
o
•o
ra
(0
(0
     0


    0.0001


   0.00001
                                  100000
                                      Flow, cfs
              Simulated  A Observed'
  Fbw er (Simulated)
  Power (Observed)
Figure 61. Power plot for observed and simulated TSS at USGS 01562000 Raystown Branch Juniata

    River at Saxton, PA - calibration period (SWAT).
            Raystown Branch Juniata River at Saxton, PA 1986-1990
                                                                      10000
          »  Simulated  A Observed
Power (Simulated)
Power (Observed)
Figure 62. Power plot for observed and simulated TSS at USGS 01562000 Raystown Branch Juniata

    River at Saxton, PA - validation period (SWAT).
                                           F-77

-------
                    Raystown Branch Juniata River at Saxton, PA 1990-2002
                                      Paired data ^-^ Equal fit
       1000
   T   100
       0.01
          0.01
0.1            1            10
          Observed TSS (mg/L)
100
1000
    Figure 63. Correlation between observed and predicted TSS concentration at USGS 01562000
       Raystown Branch Juniata River at Saxton, PA (SWAT).

Table 31.    Relative errors (observed minus simulated) for TSS concentrations at USGS
             01562000 Raystown Branch Juniata River at Saxton, PA (SWAT)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1991-2000)
106
-26.49%
12.04%
Validation period
(1990)
14
73.88%
73.84%
Calibration adjustments for the simulation of total phosphorus and total nitrogen focused on the following
parameters:
    •   PHOSKD (phosphorus soil partitioning coefficient)
    •   NPERCO (nitrogen percolation coefficient)
    •   PPERCO (phosphorus percolation coefficient)
    •   SOL_NO3 (initial nitrate concentration in soil layers)
    •   SOL_ORGN (initial organic nitrogen concentration in soil layers)
    •   SOL_SOLP (initial soluble phosphorus concentration in soil layers)
    •   SOL_ORGP (initial organic phosphorus concentration in soil layers)

Various plots that compare total phosphorous simulated by SWAT against the observed data are shown in Figures
64 through 67. The comparison statistics are provided in Tables 32 and 33. Similarly, the results corresponding to
total nitrogen are shown in Figures 68 though 71 and Tables 34 and  35. The model representation of total load is
generally acceptable, although better for phosphorus than for nitrogen. As with the HSPF application, phosphorus
                                               F-78

-------
loads and concentrations tend to be overestimated at lower flows, likely as a result of the simplified representation
of point sources.
                                            Total P
       100
   o
   E
   In
   I
-Regression Loads
-Simulated Loads
   Figure 64. Fit for monthly load of total phosphorous at USGS 01562000 Raystown Branch Juniata
       River at Saxton, PA (SWAT).
Table 32.    Model fit statistics (observed minus predicted) for monthly phosphorus loads using
            stratified regression - USGS 01562000 Raystown Branch Juniata River at Saxton,
            PA (SWAT)
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1991-2000)
-0.5%
73%
48.6%
Validation period
(1990)
9.2%
54%
44.6%
                                             F-79

-------
             Raystown Branch Juniata River at Saxton, PA 1991-2002
      100
                      10
100         1000


    Flow, cfs
     10000
100000
           •  Simulated  A  Observed ^^~Power (Simulated) ^^"Fbwer (Observed)
Figure 65. Power plot for observed and simulated total phosphorus at USGS 01562000 Raystown

    Branch Juniata River at Saxton, PA - calibration period (SWAT)
             Raystown Branch Juniata River at Saxton, PA 1986-1990
•o
ra
o
    0.01
   0.001
                                                                     10000
         »  Simulated  A  Observed
 Power (Simulated)
Power (Observed)
Figure 66. Power plot for observed and simulated total phosphorus at USGS 01562000 Raystown

    Branch Juniata River at Saxton, PA - validation period (SWAT)
                                          F-80

-------
                  Raystown Branch Juniata River at Saxton, PA
        1993     1994     1995     1996     1997

                                      Year
1998
1999
   Figure 67. Time series plot of total phosphorus concentration at USGS 01562000 Raystown Branch
      Juniata River at Saxton, PA (SWAT).
Table 33.   Relative errors (observed minus predicted), total phosphorus concentration, USGS
           01562000 Raystown Branch Juniata River at Saxton, PA (SWAT)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1991-2000)
122
-377.89%
-71.74%
Validation period
(1990)
18
-58.65%
-9.69%
                                          F-81

-------
                                          Total N
     1,400
                                                                            -Averaging Loads
                                                                            -Simulated Loads
   Figure 68. Fit for monthly load of total nitrogen at USGS 01562000 Raystown Branch Juniata River at
      Saxton, PA (SWAT)
Table 34.    Model fit statistics (observed minus predicted) for monthly total nitrogen loads
            using averaging estimator- USGS 01562000 Raystown Branch Juniata River at
            Saxton,  PA (SWAT)
Statistic
Relative Percent Error
Average Absolute Error
Median Absolute Error
Calibration period
(1991-2000)
28.6%
45%
19.7%
Validation period
(1990)
43.9%
53%
58.3%
                                           F-82

-------
             Raystown Branch Juniata River at Saxton, PA 1991-2002
   1000
                    10
100         1000

    Flow, cfs
10000
100000
          •  Simulated A Observed ^^™Power (Simulated) ^^™Power (Observed)
Figure 69. Power plot for observed and simulated total nitrogen at USGS 01562000 Raystown Branch
   Juniata River at Saxton, PA - calibration period (SWAT).
100 -
| 10

-------
      30
      25
      20
      15
      10
                 Raystown Branch Juniata River at Saxton, PA
                                                          • Simulated  A  Observed
        1986     1987      1988     1989     1990     1991     1992
                                        Year
   Figure 71. Time series plot of total nitrogen concentration at USGS 01562000 Raystown Branch
       Juniata River at Saxton, PA (SWAT).
Table 35.   Relative errors (observed minus predicted), total nitrogen concentration, USGS
            01562000 Raystown Branch Juniata River at Saxton, PA (SWAT)
Statistic
Count
Concentration Average Error
Concentration Median Error
Calibration period
(1991-2002)
13
-13.82%
21.58%
Validation period
(1990)
6
32.07%
36.92%
Water Quality Results for Larger Watershed
Similar to hydrology calibration, water quality results were compared at other gages. Note that in contrast to the
HSPF model, water quality for the SWAT model for Susquehanna River at Marietta, PA was calibrated to the
stratified regression monthly load estimates for the entire 1991-2005 period, although the observed data stop with
1995. Summary statistics for the water quality calibration and validation at other stations in the watershed are
provided in Tables 36 and 37.
                                             F-84

-------
Table 36.   Summary statistics for water quality for all stations - calibration period (SWAT)
Station
Relative Percent Error TSS
Load
TSS Concentration Median
Error
Relative Percent Error TP
Load
TP Concentration Median
Error
Relative Percent Error TN
Load
TN Concentration Median
Error
01576000
Susquehanna River at
Marietta, PA
(1991-1995)
25.2%
-34.0%
-1 1 .4%
-23.2%
-14.0%
-39.2%
01540500 Susquehanna
River at Danville, PA
(1991-1994)
28.4%
-19.9%
22.6%
-2.4%
-1.6%
-7.8%
01562000 Raystown
Branch Juniata River at
Saxton, PA (1991 -2000)
-10.1%
12.04%
-0.5%
71.74%
28.6%
21.6%
Table 37.   Summary statistics for water quality for all stations - validation period (SWAT)
Station
Relative Percent Error TSS
Load
TSS Concentration Median
Error
Relative Percent Error TP
Load
TP Concentration Median
Error
Relative Percent Error TN
Load
TN Concentration Median
Error
01576000
Susquehanna River at
Marietta, PA
(1980-1990)
15.2%
-22.8%
0.9%
-21.2%
-0.1%
-16.6%
01540500 Susquehanna
River at Danville, PA
(1986-1990)
17.1%
-3.4%
10.9%
-21.9%
15.7%
16.0%
01562000 Raystown
Branch Juniata River at
Saxton, PA (1990)
-33.6%
-73.8%
9.2%
9.7%
43.9%
36.9%
                                          F-85

-------

Preston, S.D., V.J. Bierman, Jr., and S.E. Silliman. 1989. An evaluation of methods for the estimation of tributary
mass loads. Water Resources Research. 25(6): 1379-1389.

SRBC (Susquehanna River Basin Commission). 2008. Comprehensive Plan for the Water Resources of the
Susquehanna River Basin. Harrisburg, PA.
http://www.srbc.net/planning/ComprehensivePlanwithoutAppendices62012.pdf (Accessed June, 2012)

Tetra Tech. 1990. Typical Pollutant Concentration for NCPDI Discharge Categories -Improving Point Source
Loading Data for Reporting National Water Quality Indicators. Prepared for Jim Home, EPA/OWM.

USEPA (United States Environmental Protection Agency). 2010. Chesapeake Bay Phase 5 Community
Watershed Model. (In preparation). U.S. Environmental Protection Agency, Chesapeake Bay Program Office,
Annapolis, MD. http://www.chesapeakebay.net/modeljhase5.aspx7menuitenrf6169.
                                                F-86

-------
Appendix G
Model Configuration, Calibration and
Validation

Basin: Minnesota River (Minn)
                   G-l

-------
Contents
Watershed Background	G-7
  Water Body Characteristics	G-7
  Soil Characteristics	G-9
  Land Use Representation	G-9
  Point Sources	G-13
  Meteorological Data	G-15
  Watershed Segmentation	G-18
  Calibration Data and Locations	G-20
HSPF Modeling	G-21
  Changes Made to Base Data Provided	G-21
  Assumptions	G-22
  Hydrology Calibration	G-23
  Hydrology Validation	G-28
  Hydrology Results for Larger Watershed	G-34
  Water Quality Calibration and Validation	G-40
  Water Quality Results for Larger Watershed	G-51
SWAT Modeling	G-53
  Changes Made to Base Data Provided	G-53
  Assumptions	G-53
  Hydrology Calibration	G-53
  Hydrology Validation	G-59
  Hydrology Results for Larger Watershed	G-64
  Water Quality Calibration and Validation	G-71
  Water Quality Results for Larger Watershed	G-80
References	G-83
                                               G-2

-------
Tables
Table 1.       Aggregation of NLCD land cover classes	G-ll
Table 2.       Land use distribution for the Minnesota River watershed (2001 NLCD) (mi2)	G-12
Table 3.       Major point source discharges in the Minnesota River watershed	G-13
Table 4.       Precipitation stations for the Minnesota River watershed model	G-15
Table 5.       Calibration and validation locations in the Minnesota River basin	G-20
Table 6.       Seasonal summary at USGS 05317000 Cottonwood River near New Ulm, MN - calibration
              period (HSPF)	G-26
Table 7.       Summary statistics at USGS 05317000 Cottonwood River near New Ulm, MN - calibration
              period (HSPF)	G-28
Table 8.       Seasonal summary at USGS 05317000 Cottonwood River near New Ulm, MN - validation
              period (HSPF)	G-31
Table 9.       Summary statistics at USGS 05317000 Cottonwood River near New Ulm, MN - validation
              period (HSPF)	G-34
Table 10.      Summary statistics (percent error): all stations - calibration period (HSPF)	G-3 5
Table 11.      Seasonal summary at USGS 05330000 Minnesota River near Jordan, MN - calibration period
              (HSPF)	G-38
Table 12.      Summary statistics: all stations - validation period (HSPF)	G-40
Table 13.      Model fit statistics (observed minus predicted) for monthly TSS loads using stratified regression
              (HSPF)	G-42
Table 14.      Relative errors (observed minus predicted), TSS concentration at USGS 05317000 Cottonwood
              River (HSPF)	G-44
Table 15.      Model fit statistics (observed minus predicted) for monthly phosphorus loads using stratified
              regression (HSPF)	G-45
Table 16.      Relative errors (observed minus predicted), total phosphorus concentration at USGS 05317000
              Cottonwood River (HSPF)	G-48
Table 17.      Model fit statistics (observed minus predicted) for monthly total nitrogen loads using averaging
              estimator (HSPF)	G-49
Table 18.      Relative errors (observed minus predicted), total nitrogen concentration at USGS 05317000
              Cottonwood River (HSPF)	G-51
Table 19.      Summary statistics for water quality: all stations - calibration period 1993-2002 (HSPF)	G-51
Table 20.      Summary statistics for water quality: all stations - validation period 1986-1992 (HSPF)	G-52
Table 21.      Seasonal summary at USGS 05317000 Cottonwood River near New Ulm, MN - calibration
              period (SWAT)	G-57
Table 22.      Summary statistics at USGS 05317000 Cottonwood River near New Ulm, MN - calibration
              period (SWAT)	G-59
Table 23.      Seasonal summary at USGS 05317000 Cottonwood River near New Ulm, MN - validation
              period (SWAT)	G-62
Table 24.      Summary statistics at USGS 05317000 Cottonwood River near New Ulm, MN - validation
              period (SWAT)	G-64
Table 25.      Summary statistics (percent error): all stations - calibration period (SWAT)	G-65
Table 26.      Seasonal summary at USGS 05330000 Minnesota River near Jordan, MN - calibration period
              (SWAT)	G-68
Table 27.      Summary statistics at USGS 05330000 Minnesota River near Jordan, MN - calibration period
              (SWAT)	G-70
Table 28.      Summary statistics: all stations - validation period (SWAT)	G-71
Table 29.      Model fit statistics (observed minus predicted) for monthly sediment loads using stratified
              regression at USGS 05317000 Cottonwood River (SWAT)	G-72
Table 30.      Relative errors (observed minus predicted), TSS concentration at USGS 05317000
              Cottonwood River (SWAT)	G-74
Table 31.      Model fit statistics (observed minus predicted) for monthly phosphorus loads using stratified
              regression at USGS 05317000 Cottonwood River (SWAT)	G-75

-------
Table 32.      Relative errors (observed minus predicted), total phosphorus concentration at USGS
              05 317000 Cottonwood River (SWAT)	G-77
Table 33.      Model fit statistics (observed minus predicted) for monthly total nitrogen loads using
              averaging estimator at USGS 05317000 Cottonwood River (SWAT)	G-78
Table 34.      Relative errors (observed minus predicted), total nitrogen concentration at USGS 05317000
              Cottonwood River (SWAT)	G-80
Table 35.      Summary statistics for water quality at all stations - calibration period 1993-2002 (SWAT).. G-82
Table 36.      Summary statistics for water quality at all stations - validation period 1986-1992 (SWAT)... G-82
Figures
Figure 1. Location of the Minnesota River watershed	G-8
Figure 2. Land use in the Minnesota River watershed	G-10
Figure 3. Major point sources in the Minnesota River watershed	G-15
Figure 4. Weatherstations for the Minnesota River watershed model	G-17
Figure 5. Model segmentation and USGS stations utilized for the Minnesota River watershed	G-19
Figure 6. Mean daily flow at USGS 05317000 Cottonwood River near New Ulm, MN - calibration period
         (HSPF)	G-24
Figure 7. Mean monthly flow at USGS 05317000 Cottonwood River near New Ulm, MN - calibration period
         (HSPF)	G-24
Figure 8. Mean monthly flow regression and temporal variation at USGS 05317000 Cottonwood River near
         New Ulm, MN - calibration period (HSPF)	G-25
Figure 9. Seasonal regression and temporal aggregate at USGS 053170