SERA
EPA/600/R-16/094 | July 2016 | www.epa.gov/research
United States
Environmental Protection
Agency
Hydrology and Water Quality
Modeling in the Conestoga River
TECHNICAL REPORT
Big
Spring
Run
Office of Research and Development
National Risk Management Research Laboratory | Groundwater, Watershed, and Ecosystem Restoration Division
-------
EPA/600/R-16/094
July 2016
Hydrology and Water Quality Modeling
in the Conestoga River
Prepared for
U.S. Environment tection Agency
Office of Research and Development
National Risk Management Research Laboratory
Groundwater, Watershed, and Ecosystem Restoration Division
Ada, Oklahoma
Work Assignment Manager
Timothy J. Canfield
NRMRL/GWERD/ 4
SSWR 4.02C Task Lead
Ann Keeley, Ph.D.
NRMRL/GWERD/ 4
Prepared by
5. Sarkar, H.D. Nicholas, J. B. Butcher, M.J. Paul
TetraTech
Research Triangle Park, North Carolina
1 National Risk Management Research Laboratory | Groundwater, Watershed, and Ecosystem Restoration Division
-------
Acronyms/Unitsix
Quality Assurance[[[ ix
Abstract x
1. Introduction[[[ 1
2. Watershed Model Development 3
2.1 SWAT,3
2.1.1 Topography. 4
2.1.2 Landuse and Landcover 6
2.1.3 Soils and Geology. 10
2.1.4 Hydrologic Response Unit (HRU) Development 12
2.1.5 Meteorology 13
2.1.6 Point Sources 13
2.2 HSPF 17
2.2.1 FTABLE development 18
3. Watershed Model Calibration and Validation 19
3.1 Hydrology 19
3.1.1 Methods 19
3.1.2 Results and Discussion 23
3.2 Water Quality. ..... ......... ..... ..... ..... ......... ..... 71
3.2.1 Methods[[[ 71
3.2.2 Results and Discussion 75
3.3 Plant Growth 98
4. Simulation of Restoration Scenario in SWAT[[[ 99
4.1 Approach ........... ..... ..... ..... ......... ..... ..... ..... ............. 99
4.2 Results and Discussion[[[ 100
5. Groundwater Model Development, Calibration and Application 102
5.1 GFLOW Model Development[[[ 102
5.1.1 GFLOW Model 102
5.1.2 Model Extent ..102
5.1.3 Aquifer Properties[[[ 104
5.1.4 Representation of Stream Network [[[ 104
5.1.5 Representation of Ponds and Recreation Area 105
-------
Table 1. Landuse and landcover in the Conestoga River watershed according to CBWLCD.......................... 6
Table 2. Landuse and landcover in the Conestoga River watershed according to NLCD 7
Table 3. Average area harvested by crop type in Lancaster County as per USDA-NASS Census of Agriculture
1997, 2002, 2007, 2012 8
Table 4. Crops and associated rotations in the SWAT model[[[ 8
Table 5. Crop management practices in the SWAT model 9
Table 6. Point sources in the SWAT model 14
Table 7. Geometric Parameters for FTABLE .... .... ....... .... .... 18
Table 8. Hydrology calibration and validation locations 20
Table 9. Performance targets for monthly average loads 21
Table 10. Hydrology calibration criteria 21
Table 11. Values of parameters in the calibrated and validated model .22
Table 12. Evaluation of SWAT model performance for streamflow at a monthly time-step ... 25
Table 13. Evaluation of HSPF model performance for streamflow at a monthly time-step 25
Table 14. Seasonal summary at USGS 01576754 Conestoga River at Conestoga, PA 28
Table 15. Summary statistics at USGS 01576754 Conestoga River at Conestoga, PA 30
Table 16. Seasonal summary at USGS 01576500 Conestoga River near Lancaster, PA 33
Table 17. Summary statistics at USGS 01576500 Conestoga River near Lancaster, PA............................. 35
Table 18. Seasonal summary at USGS 015767195 Big Spring Run near Mylin Corners, PA......................... 38
Table 19. Summary statistics at USGS 015767195 Big Spring Run near Mylin Corners, PA 40
Table 20. Seasonal summary at USGS 01576754 Conestoga River at Conestoga, PA............................... 43
Table 21. Summary statistics at USGS 01576754 Conestoga River at Conestoga, PA 45
Table 22. Seasonal summary at USGS 01576500 Conestoga River near Lancaster, PA 48
Table 23. Summary statistics at USGS 01576500 Conestoga River near Lancaster, PA 50
Table 24. Seasonal summary at USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon, PA 53
Table 25. Summary statistics at USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon, PA 55
Table 26. Seasonal summary at USGS 01576085 Little Conestoga Creek Churchtown, PA 58
Table 27. Summary statistics at USGS 01576085 Little Conestoga Creek Churchtown, PA .... ...60
Table 28. Seasonal summary at USGS 01576521 Big Spring Run near Willow Street, PA 63
-------
Table 30. Seasonal summary at USGS 01576529 Un-named Tributary to Big Spring Run
near Lampeter, PA 68
Table 31. Summary statistics at USGS 01576529 Un-named Tributary to Big Spring Run
near Lampeter, PA 70
Table 32. Water quality calibration and validation location 71
Table 33. Values of parameters in the calibrated and validated model 73
Table 34. SWAT model performance at a monthly time-step for Conestoga River at Conestoga 75
Table 35. Paired daily Total Suspended Solids (TSS) load (tons/day) 78
Table 36. Paired daily Total Kjeldahl Nitrogen (TKN) load (tons/day) .81
Table 37. Paired daily Nitrite + Nitrate Nitrogen (NOx) load (tons/day) 84
Table 38. Paired daily Total Nitrogen (TN) load (tons/day) 87
Table 39. Paired daily Total Phosphorus (TP) load (tons/day) ... ........ .90
Table 40. Observed and simulated biomass yields for agricultural crops 98
Table 41. Restoration scenarios simulated using SWAT 99
Table 42. Sediment and nutrient loads for the Big Spring Run .................................................. 100
Table 43. Sediment and nutrient loads for the Conestoga River................................................ 100
Table 44. Parameters associated with near-field line-sinks [[[ 105
Table 45. USGS stream flow gage test points used in GFLOW calibration ........................................ 108
Table 46. Piezometer test points used in GFLOW calibration 108
Table 47. Values of parameters associated with analytical elements in the calibrated model.................... 110
Table 48. GFLOW calibration results for piezometer test points ................................................ 112
Table 49. GFLOW calibration results for USGS gage test points ................................................. 112
Table 50. GFLOW calibration results for piezometer test points (wet season)................................... 114
Table 51. GFLOW calibration results for USGS gage test points (wet season) 114
Table 52. GFLOW calibration results for piezometer test points (dry season) .... 116
-------
BM——¦
Figure 1. Elevation in the Conestoga River watershed 4
Figure 2. Delineated subbasins and reaches for the Conestoga River watershed 5
Figure 3. Landuse and landcover in the Conestoga River watershed as per CBLWCD 2001 7
Figure 4. Soils in the Conestoga River watershed 10
Figure 5. Geology of the Conestoga River watershed 11
Figure 6. Karst features in the Conestoga River watershed 12
Figure 7. Point sources in the Conestoga River watershed 14
Figure 8. Schematic representation of SWAT simulated flow and sediment load transfer
to the HSPF model 17
Figure 9. Streamflow monitoring locations in the Conestoga River watershed 20
Figure 10. SWAT simulated hydrologic cycle of the Conestoga River watershed 23
Figure 11. Comparison of SWAT simulated evapotranspiration with satellite based estimates .................... 24
Figure 12. Mean daily flow at USGS 01576754 Conestoga River at Conestoga, PA .26
Figure 13. Mean monthly flow at USGS 01576754 Conestoga River at Conestoga, PA ............................ 26
Figure 14. Monthly flow regression and temporal variation at USGS 01576754 Conestoga River
at Conestoga, PA 27
Figure 15. Seasonal regression and temporal aggregate at USGS 01576754 Conestoga River
at Conestoga, PA 27
Figure 16. Seasonal medians and ranges at USGS 01576754 Conestoga River at Conestoga, PA 28
Figure 17. Flow exceedance at USGS 01576754 Conestoga River at Conestoga, PA 29
Figure 18. Flow accumulation at USGS 01576754 Conestoga River at Conestoga, PA 29
Figure 19. Mean daily flow at USGS 01576500 Conestoga River near Lancaster, PA 31
Figure 20. Mean monthly flow at USGS 01576500 Conestoga River near Lancaster, PA 31
Figure 21. Monthly flow regression and temporal variation at USGS 01576500 Conestoga River
near Lancaster, PA ... 32
Figure 22. Seasonal regression and temporal aggregate at USGS 01576500 Conestoga River
near Lancaster, PA 32
Figure 23. Seasonal medians and ranges at USGS 01576500 Conestoga River near Lancaster, PA 33
Figure 24. Flow exceedance at USGS 01576500 Conestoga River near Lancaster, PA 34
Figure 25. Flow accumulation at USGS 01576500 Conestoga River near Lancaster, PA 34
Figure 26. Mean daily flow at USGS 015767195 Big Spring Run near Mylin Corners, PA 36
Figure 27. Mean monthly flow at USGS 015767195 Big Spring Run near Mylin Corners, PA — 36
Figure 28. Monthly flow regression and temporal variation at USGS 015767195 Big Spring Run
near Mylin Corners, PA 37
Figure 29. Seasonal regression and temporal aggregate at USGS 015767195 Big Spring Run
near Mylin Corners, PA .37
mm
-------
Figure 30. Seasonal medians and ranges at USGS 015767195 Big Spring Run near Mylin Corners, PA 38
Figure 31. Flow exceedance at USGS 015767195 Big Spring Run near Mylin Corners, PA 39
Figure 32. Flow accumulation at USGS 015767195 Big Spring Run near Mylin Corners, PA ....................... 39
Figure 33. Mean daily flow at USGS 01576754 Conestoga River at Conestoga, PA 41
Figure 34. Mean monthly flow at USGS 01576754 Conestoga River at Conestoga, PA 41
Figure 35. Monthly flow regression and temporal variation at USGS 01576754 Conestoga River
at Conestoga, PA 42
Figure 36. Seasonal regression and temporal aggregate at USGS 01576754 Conestoga River
at Conestoga, PA 42
Figure 37. Seasonal medians and ranges at USGS 01576754 Conestoga River at Conestoga, PA 43
Figure 38. Flow exceedance at USGS 01576754 Conestoga River at Conestoga, PA 44
Figure 39. Flow accumulation at USGS 01576754 Conestoga River at Conestoga, PA ............................. 44
Figure 40. Mean daily flow at USGS 01576500 Conestoga River near Lancaster, PA 46
Figure 41. Mean monthly flow at USGS 01576500 Conestoga River near Lancaster, PA 46
Figure 42. Monthly flow regression and temporal variation at USGS 01576500 Conestoga River
near Lancaster, PA 47
Figure 43. Seasonal regression and temporal aggregate at USGS 01576500 Conestoga River
near Lancaster, PA 47
Figure 44. Seasonal medians and ranges at USGS 01576500 Conestoga River near Lancaster, PA 48
Figure 45. Flow exceedance at USGS 01576500 Conestoga River near Lancaster, PA 49
Figure 46. Flow accumulation at USGS 01576500 Conestoga River near Lancaster, PA 49
Figure 47. Mean daily flow at USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon, PA 51
Figure 48. Mean monthly flow at USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon, PA 51
Figure 49. Monthly flow regression and temporal variation at USGS 01576540 Mill Creek
at Eshelman Mill Road near Lyndon, PA 52
Figure 50. Seasonal regression and temporal aggregate at USGS 01576540 Mill Creek
at Eshelman Mill Road near Lyndon, PA 52
Figure 51. Seasonal medians and ranges at USGS 01576540 Mill Creek at Eshelman Mill Road
near Lyndon, PA 53
Figure 52. Flow exceedance at USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon, PA 54
Figure 53. Flow accumulation at USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon, PA 54
Figure 54. Mean daily flow at USGS 01576085 Little Conestoga Creek Churchtown, PA 56
Figure 55. Mean monthly flow at USGS 01576085 Little Conestoga Creek Churchtown, PA 56
Figure 56. Monthly flow regression and temporal variation at USGS 01576085 Little Conestoga Creek
Churchtown, PA .57
Figure 57. Seasonal regression and temporal aggregate at USGS 01576085 Little Conestoga Creek
Churchtown, PA 57
—
-------
m—
Figure 58. Seasonal medians and ranges at USGS 01576085 Little Conestoga Creek Churchtown, PA 58
Figure 59. Flow exceedance at USGS 01576085 Little Conestoga Creek Churchtown, PA 59
Figure 60. Flow accumulation at USGS 01576085 Little Conestoga Creek Churchtown, PA 59
Figure 61. Mean daily flow at USGS 01576521 Big Spring Run near Willow Street, PA 61
Figure 62. Mean monthly flow at USGS 01576521 Big Spring Run near Willow Street, PA 61
Figure 63. Monthly flow regression and temporal variation at USGS 01576521 Big Spring Run
near Willow Street, PA 62
Figure 64. Seasonal regression and temporal aggregate at USGS 01576521 Big Spring Run
near Willow Street, PA 62
Figure 65. Seasonal medians and ranges at USGS 01576521 Big Spring Run near Willow Street, PA 63
Figure 66. Flow exceedance at USGS 01576521 Big Spring Run near Willow Street, PA 64
Figure 67. Flow accumulation at USGS 01576521 Big Spring Run near Willow Street, PA 64
Figure 68. Mean daily flow at USGS 01576529 Un-named Tributary to Big Spring Run near Lampeter, PA ..... 66
Figure 69. Mean monthly flow at USGS 01576529 Un-named Tributary to Big Spring Run
near Lampeter, PA 66
Figure 70. Monthly flow regression and temporal variation at USGS 01576529 Un-named Tributary
to Big Spring Run near Lampeter, PA 67
Figure 71. Seasonal regression and temporal aggregate at USGS 01576529 Un-named Tributary
to Big Spring Run near Lampeter, PA ....... ....... ....... 67
Figure 72. Seasonal medians and ranges at USGS 01576529 Un-named Tributary to Big Spring Run
near Lampeter, PA 68
Figure 73. Flow exceedance at USGS 01576529 Un-named Tributary to Big Spring Run
near Lampeter, PA 69
Figure 74. Flow accumulation at USGS 01576529 Un-named Tributary to Big Spring Run
near Lampeter, PA 69
Figure 75. Distribution of shear stress (Tau) with flow for modeled reach 9 74
Figure 76. SWAT simulated sediment load proportions by landuse 75
Figure 77. SWAT simulated total nitrogen load proportions by landuse 75
Figure 78. SWAT simulated total phosphorus load proportions by landuse 75
Figure 79. Simulated bank erosion rate for each modeled reach in the SWAT model 76
Figure 80. Monthly simulated and estimated Total Suspended Solids (TSS) load at Conestoga River
near Conestoga, PA (calibration period) 77
Figure 81. Monthly simulated and estimated Total Suspended Solids (TSS) load at Conestoga River
near Conestoga, PA (validation period) 77
Figure 82. Power plot of simulated and observed Total Suspended Solids (TSS) load vs flow
at Conestoga River near Conestoga, PA (calibration period) 78
Figure 83. Power plot of simulated and observed Total Suspended Solids (TSS) load vs flow
at Conestoga River near Conestoga, PA (validation period) 78
m—
-------
m—
Figure 84. Paired simulated vs observed Total Suspended Solids (TSS) load at Conestoga River
near Conestoga, PA (calibration period) 79
Figure 85. Paired simulated vs observed Total Suspended Solids (TSS) load at Conestoga River
near Conestoga, PA (validation period) 79
Figure 86. Monthly simulated and estimated Total Kjeldahl Nitrogen (TKN) load at Conestoga River
near Conestoga, PA (calibration period) 80
Figure 87. Monthly simulated and estimated Total Kjeldahl Nitrogen (TKN) load at Conestoga River
near Conestoga, PA (validation period) 80
Figure 88. Power plot of simulated and observed Total Kjeldahl Nitrogen (TKN) load vs flow
at Conestoga River near Conestoga, PA (calibration period) 81
Figure 89. Power plot of simulated and observed Total Kjeldahl Nitrogen (TKN) load vs flow
at Conestoga River near Conestoga, PA (validation period) .81
Figure 90. Paired simulated vs observed Total Kjeldahl Nitrogen (TKN) load at Conestoga River
near Conestoga, PA (calibration period) 82
Figure 91. Paired simulated vs observed Total Kjeldahl Nitrogen (TKN) load at Conestoga River
near Conestoga, PA (validation period) 82
Figure 92. Monthly simulated and estimated Nitrite + Nitrate Nitrogen (NOx) load at Conestoga River
near Conestoga, PA (calibration period) 83
Figure 93. Monthly simulated and estimated Nitrite + Nitrate Nitrogen (NOx) load at Conestoga River
near Conestoga, PA (validation period) 83
Figure 94. Power plot of simulated and observed Nitrite + Nitrate Nitrogen (NOx) load vs flow
at Conestoga River near Conestoga, PA (calibration period) 84
Figure 95. Power plot of simulated and observed Nitrite + Nitrate Nitrogen (NOx) load vs flow
at Conestoga River near Conestoga, PA (validation period) 84
Figure 96. Paired simulated vs observed Nitrite + Nitrate Nitrogen (NOx) load at Conestoga River
near Conestoga, PA (calibration period)[[[ 85
Figure 97. Paired simulated vs observed Nitrite + Nitrate Nitrogen (NOx) load at Conestoga River
near Conestoga, PA (validation period) 85
Figure 98. Monthly simulated and estimated Total Nitrogen (TN) load at Conestoga River
near Conestoga, PA (calibration period) 86
Figure 99. Monthly simulated and estimated Total Nitrogen (TN) load at Conestoga River
near Conestoga, PA (validation period) 86
Figure 100. Power plot of simulated and observed Total Nitrogen (TN) load vs flow at Conestoga River
near Conestoga, PA (calibration period) ....... ..... ..... ......... ..... ......... 87
Figure 101. Power plot of simulated and observed Total Nitrogen (TN) load vs flow at Conestoga River
near Conestoga, PA (validation period)[[[ 87
Figure 102. Paired simulated vs observed Total Nitrogen (TN) load at Conestoga River
near Conestoga, PA (calibration period) 88
-------
m—
Figure 104. Monthly simulated and estimated Total Phosphorus (TP) load at Conestoga River
near Conestoga, PA (calibration period) 89
Figure 105. Monthly simulated and estimated Total Phosphorus (TP) load at Conestoga River
near Conestoga, PA (validation period) 89
Figure 106. Power plot of simulated and observed Total Phosphorus (TP) load vs flow
at Conestoga River near Conestoga, PA (calibration period) .90
Figure 107. Power plot of simulated and observed Total Phosphorus (TP) load vs flow
at Conestoga River near Conestoga, PA (validation period) 90
Figure 108. Paired simulated vs observed Total Phosphorus (TP) load at Conestoga River
near Conestoga, PA (calibration period) .... 91
Figure 109. Paired simulated vs observed Total Phosphorus (TP) load at Conestoga River
near Conestoga, PA (validation period) 91
Figure 110. Paired HSPF simulated and observed suspended sediment concentration 92
Figure 111. HSPF simulated and observed suspended sediment concentration vs flow .......................... 93
Figure 112. Time-series plot of HSPF simulated and observed suspended sediment concentration ............... 94
Figure 113. Paired HSPF simulated and observed suspended sediment load 95
Figure 114. HSPF simulated and observed suspended sediment load vs flow 96
Figure 115. Net scour/deposition simulated by the HSPF model for a 25 year period ............................ 97
Figure 116. Observed and simulated mean, minimum and maximum annual biomass yield for the
simulation time-period 98
Figure 117. Percent reduction in sediment and nutrient loads for the five restoration scenarios
relative to pre-restoration for the Big Spring Run .................................................. 101
Figure 118. Percent reduction in sediment and nutrient loads for the five restoration scenarios
relative to pre-restoration for the Conestoga River 101
Figure 119. GFLOW model extent [[[ 103
Figure 120. Analytical elements in the Big Spring Run GFLOW model environment ............................. 106
Figure 121. Test points in the GFLOW model 109
Figure 122. Modeled vs observed baseflow for all piezometer test points ...................................... 113
Figure 123. Modeled vs observed baseflow for all USGS gage test points ....................................... 113
Figure 124. Modeled vs observed baseflow for all piezometer test points (wet season)......................... 115
Figure 125. Modeled vs observed baseflow for all USGS gage test points (wet season).......................... 115
Figure 126. Modeled vs observed baseflow for all piezometer test points (dry season).......................... 117
Figure 127. Modeled vs observed baseflow for all USGS gage test points (dry season).......................... 117
Figure 128. GFLOW simulation of head contours for average annual recharge conditions ....................... 118
Figure 129. GFLOW simulation of line-sink exchange rates for average annual recharge conditions .............. 119
-------
Acronym
'
BMP
Best Management Practice
cfs
cubic feet per second
CBWLCD
Chesapeake Bay Watershed Land Cover Data Series
lbs
pounds
CDL
Cropland Data Layer
tons
short tons
CONUS
Continental US
mm
millimeters
DEM
Digital Elevation Model
km
kilometers
HRU
Hydrologic Response Unit
yr
year
HSPF
Hydrologic Simulation Program Fortran
hr
hour
NASS
National Agricultural Statistics Service
deg-C
degree Centigrade
NCDC
National Climatic Data Center
cu-ft
cubic feet
NED
National Elevation Dataset
mgd
million gallons per day
NHD
National Hydrography Dataset
NLCD
National Land Cover Dataset
NLDAS
National Land Data Assimilation System
NRCS
Natural Resources Conservation Service
SSURGO
Soil Survey Geographic
SWAT
Soil and Water Assessment Tool
IJSDA
United States Department of Agriculture
IJSGS
United States Geological Survey
USLE
Universal Soil Loss Equation
RMSE
Root Mean Square Error
All EPA-funded research efforts are required to participate in the U.S. EPA Quality Assurance (OA) Program, A Quality Assurance
Project Plan {QAPP) was prepared by Tetra Tech, Inc. and approved by the EPA which describes the QA process and requirements
for the watershed and groundwater models described in this report. The models were developed, calibrated, and validated using
the data and performance targets as described in the QAPP (approved 4/27/2015),
The watershed model was developed using quality checked data from several state, federal and regional agencies. The table
below lists data used is the development, and calibration and validation of the watershed and groundwater models and their
respective sources.
IL,.
Stream Layer
USGS (htlpi/Zrihcl usgs,gov/cJafa, html)
Digital Elevation Model
USGS-NED (htt|3i//nationaImap,gov/eIevation,htmI)
Landuse and landcover
USGS (ht!psi//pijbs,er,usgs,gov/publication/cls505)
Cropland Data Layer
USDA-NASS (https://nassgeodata.gmu.edu/CropScape/)
Agricultural Management Practices
Pennsylvania State Extension Service (http://exfension.psu.eclu)
Soils
Soil Survey Geographic (SSURGO) database (http://www.nrcs.usda.gov/wps/portal/nrcs/detail/soils/
su rvey/?c id=n r'cs 142 p2 J)53627)
Geology
Pennsylvania Spatial Data Access (http://www.pasda.psu.edu/)
Karst Features
Pennsylvania Spatial Data Access (http://www.pasda.Dsu.edu/)
Meteorology (precipitation and air temperature)
PRISM Climate Group (http://www.prism.oregonstate.eau/)
Meteorology (wind speed, solar radiation and
humidity)
NASA Land Data Assimilation System (http://ldas.gsfc.nasa.gov/nldas/NLDAS2forcing.php)
Evapo transpiration
NASA Moderate Resolution Imaging Spectroradiometer
(http://modis.gsfc.nasa.gov/data/dataprod/mod16.php)
Point Source Dischargers
Chesapeake Bay Program (http://www.chesapeakebay.net/)
Daily Flow Data
USGS-NWIS (http://waterclata.usgs.gov/nwis)
Water Quality Data
USGS-Chesapeake Bay Program (http://cbriin,er,usgs,gov/iiiclex,htmI)
The calibration and validation of the watershed and groundwater models were accomplished using the criteria set in the QAPP. In
addition, detailed discussion of model performances has been provided in the report.
—
-------
Watershed models using the Soil and Water
Assessment Tool (SWAT) and Hydrologic Simulation
Program FORTRAN (HSPF) were developed for the
Conestoga River watershed in Pennsylvania. The
objectives of this study are to quantify sources of
sediment and nutrient in the watershed and impacts
of stream restoration on hydrology and water quality.
The SWAT model was calibrated and validated for
streamflow, sediment and nutrients at the watershed
outlet. The daily Nash-Sutcliffe Efficiency (NSE) for
streamflow were 0.93 and 0.90 for the calibration and
validation periods, respectively. The evapotranspiration
and plant growth components of the SWAT model
were also evaluated using satellite based estimates
and county level crop yield estimates, respectively.
The monthly NSE for sediment and nutrients based
on comparison of simulated loads against regression
loads were greater than 0.65. An HSPF model for
the Conestoga River was also developed to quantify
legacy sources of sediment and parameterize the
in-stream sediment component of the SWAT modei.
The streamflow and sediment component of the
HSPF model were calibrated and validated using
the same data used for the evaluation of the SWAT
model. Simulations of stream restoration scenarios
were completed using the calibrated SWAT model
and showed large decreases in delivered sediment
and nutrient loads. The results indicate that stream
restoration has the potential to reduce sediment and
nutrient loads to the Chesapeake Bay.
A groundwater model for a smaller tributary of the
Conestoga River was also developed using GFLOW
to understand the impacts of stream restoration on
groundwater hydrology. The GFLOW model provides
a foundation for a more detailed groundwater flow
model for the Conestoga River watershed.
X
-------
The Chesapeake Bay has been listed as impaired
under the Clean Water Act since 1998. As a result,
the State of Pennsylvania through its commitment to
the Chesapeake Bay Council, set milestones in 2012
to reduce nitrogen, phosphorus and sediment loads
to the Chesapeake Bay by approximately 6.3 million,
0.2 million, and 204 million pounds, respectively, in
the year 2013 (PDEP 2012). Water quality reduction
actions for the Chesapeake Bay have attracted federal
and state government, environmentalists, academics
and others to employ their expertise in developing
and evaluating mitigation strategies to improve and
sustain the improvement of water quality of the Bay.
The Conestoga River watershed has been a focus of this
work since it contributes a significant amount of water,
nutrients, and sediment annually to the Bay.
The Conestoga River is a tributary of the Susquehanna
River and drains about 475 square miles of Lancaster,
Chester, Lebanon and Berks Counties in Pennsylvania.
Milldams constructed from the 1600s to 1900s for
power generation occurred in the high densities along
streams within the states of Maryland, Pennsylvania,
New York and central New England and resulted in the
accumulation of fine sediment over pre-settlement
wetlands (Waiter and Merritts 2008). These legacy
sediments are highly erodible and can contribute
between 50 to 80 percent of suspended sediment loads
in watersheds in Pennsylvania and Maryland (Waiter
et al. 2007). The Potomac and Susquehanna Rivers are
amongst the highest contributors of sediment loads
to the Chesapeake Bay, and one of the highest yields
in the Susquehanna River basin is from the Conestoga
River watershed (Phillips 2007).
Vegetative and structural Best Management Practices
(BMPs), and riparian management measures are
being implemented in the Conestoga River watershed
to reduce transport of sediment and nutrient. One
such research project has received national attention
on account of its comprehensive approach towards
the evaluation of stream restoration. The stream
restoration effort employs vegetative and structural
BMPs to reduce stream sediment loss and improve
water quality within the Conestoga River watershed.
The study sites include the Big Spring Run (BSR), a
tributary of the Conestoga River and the location of
a historic milldam. This site is being evaluated for the
effect of BMPs on ground water and surface water
quantity and quality, nutrient transport and speciation,
biological impacts, and physical dynamics.
The restoration work conducted at BSR provides
information on the effectiveness of restoration
of former milldam sites and associated BMPs for
reducing sediment loads and improving water quality.
Restoration efforts conducted at the BSR site contribute
to the understanding of the efficacy of vegetative and
structural BMPs. It is important to note that there
are several similar historic milldam sites scattered
throughout the Conestoga River watershed and are
likely a source of sediment and nutrient. Extending
on-site research to all such milldam sites would be
a resource intensive process. Modeling watersheds
as an approach for evaluating the impact of BMP
implementation has become increasingly relevant due
to the limitations associated with monitoring.
1
-------
The capability for mechanistically modeling
the watershed and restoration activities, and
demonstrating the impact of restoration on hydrology
and sediment delivery at a watershed scale can provide
useful information. Watershed scale models have
been applied to evaluate various aspects of non-point
source pollution and to a lesser extent the impacts of
structural BMPs. Though watershed models cannot
account for every detail, they are a good resource
for evaluating the targeted systems at work and the
dynamics between and within those systems.
Tetra Tech, Inc. was retained by the National Risk
Management Research Laboratory (NRMRL) of the
United States Environmental Protection Agency (EPA) to
provide consulting services to develop watershed and
groundwater models of the Conestoga River watershed
as part of a project on the Big Spring Run (BSR) in
Lancaster County, Pennsylvania. The objective of this
project was to employ simulation modeling to quantify
legacy sediment from historic milldam sites and study
the effects of stream restoration on hydrology and
water quality for the Conestoga River watershed.
Peer-reviewed watershed and ground water models
were used to quantify the sources of sediment
and nutrients and simulate the impacts of stream
restoration in the Conestoga River watershed. The Soil
and Water Assessment Tool (SWAT) (Neitsch et al. 2005,
Neitsch et al. 2010, Neitsch et al. 2011) and Hydrologic
Simulation Program Fortran (HSPF) (Bicknell et al. 2014)
were used to develop hydrology and water quality
models of the Conestoga River watershed. GFLOW
(Haitjema 2005) was used to develop a groundwater
model for the Big Spring Run (a small tributary of the
Conestoga River).
The SWAT model for the Conestoga River was
developed at the HUC12 scale with additional
delineation for historic milldams. The model was
subsequently calibrated for streamflow, sediment
and nutrients, and stream restoration scenarios were
simulated. SWAT is an excellent tool to simulate
agricultural land uses and management practices
for sediment and nutrient source loading estimates.
However, it is not ideal for ultimate fate and transport
predictions on account of its relatively weak in-stream
sediment transport and water quality kinetics. In
addition, the groundwater component of the SWAT
model uses an empirical approach that may not be
adequate for detailed groundwater studies.
HSPF provides dynamic simulation of water, nutrients,
and sediment; including both upland and instream
sediment processes at a user-specified level of detail
and complexity. The ability of the model to simulate
the in-stream processes associated with cohesive
and non-cohesive fractions of sediment is especially
important for addressing the principle study questions.
HSPF is also supported by EPA with open source code
and has a long history of well-documented applications
for addressing hydrology and sediment management
applications. SWAT subbasin level flow and sediment
outputs (before they are routed through the stream
system) were connected to an HSPF model for the
Conestoga River. The Conestoga River HSPF model does
not have an upland component and is rather setup to
route upland loads from the SWAT model and perform
in-stream sediment processes. The HSPF model was
subsequently calibrated for streamflow and sediment
load at the watershed outlet. The erosion rates on a
reach-by-reach basis from the calibrated HSPF model
were used to inform the reach erosion rates in the
SWAT model.
GFLOW (Haitjema 2005) was used to develop a
steady state groundwater model of the Big Spring
Run tributary region of the Conestoga River. GFLOW
is a single aquifer model which solves steady state
groundwater flow using the analytic element model.
GFLOW uses the Dupuit-Forchheimer approximation
and thereby ignores resistance to vertical flow. GFLOW
also allows for areas of differing aquifer properties,
horizontal barriers with resistance to flow (slurry
walls), 3D flow near a partially penetrating well, local
transient flow near a well (Theis solution), steady state
interface flow in coastal aquifers, and UCODE support
for parameter optimization. The GFLOW model was
developed independently of the watershed models.
The GFLOW model provides detailed head contours
associated with the study area and exchanges between
the surface water and groundwater. The GFLOW model
provides useful information with regard to groundwater
behavior in a region where stream restoration activities
have been employed.
I
-------
Wr ,1
Iji tf
f .mi* I 1 Mm
• i I-In§
Spnn.&
2.1 SWAT
The Soil arid Water Assessment Too! (SWAT) was
developed by the United States Department of
Agriculture (USDA) Agricultural Research Service (ARS)
for conducting long-term, continuous, watershed level
simulations used for predicting the impact of land
management practices on water quality and quantity
for a variety of soils, land cover and management
practices (Arnold et at. 1998). SWAT is a process and
semi-physically based model with the capability for
efficiently simulating high levels of spatial detail and
requires input of weather, hydrology, soil properties,
vegetation, and land management practices (Jha et
at. 2007). SWAT has been tested extensively across
the United States and internationally for evaluating
non-point source pollution, conservation practices,
and land use management, among other applications.
The model has also been used for watershed studies
within the Chesapeake Bay area (Chu et at. 2004,
Meng et at. 2010, Sexton et at. 2010, Veith et at. 2010)
for evaluating water quality and quantity concerns,
and is part the Chesapeake Bay Operational Forecast
System (CBOFS) developed by the University of
Maryland at College Park and the National Oceanic
and Atmospheric Administration (NOAA) to provide
real time simulations of the Bay (Meng et at. 2010).
Hydrology in SWAT is based on a water balance that
includes surface runoff, precipitation, percolation,
lateral subsurface flow, groundwater return flow,
evapotranspiration, and channel transmission loss
subroutines. Surface runoff can be estimated based
on land use, antecedent moisture conditions and soil
type using the SCS curve number method. Surface
runoff can also be simulated using the Green and
Ampt infiltration method (Neitsch et at. 2011).
SWAT simulates transport of sediment through a
land component and a channel component (Neitsch
et at. 2011). Within the land component the model
estimates soil erosion and sediment from hill slope
erosion using the Modified Universal Soil Loss
Equation (MUSLE) (Williams 1975, Williams and
Berndt 1977) and transport sediments based on
particle size distributions and routes them through
surface water sources and channels (Neitsch et
at. 2011). Channel sediment routing includes in-
stream depositional and degradation processes that
are dependent on stream power, channel surface
exposure and channel bank and bed composition
(Neitsch et at. 2011). Channel routing is accomplished
in SWAT using a modification of Bagnold's sediment
transport equation (Bagnold 1977) and Stokes' law
(Chow et at. 1988) to estimate transport concentration
capacity as a function of flow velocity.
The SWAT model was setup using ArcSWAT version
2012.10_1.15 (Winchell et at. 2013). ArcSWAT is an
ArcGIS based interface that facilitates the model setup
process and generation of input files. The Conestoga
River watershed SWAT model was subsequently
parameterized, calibrated and validated using
SWAT Editor version 2012.10_2.15. SWAT Editor is a
standalone program that reads the project database
generated by ArcSWAT, allows the user to edit the
3
-------
input files and execute a SWAT run. The following
sections provide an overview of the watershed model
development and the datasets used.
2.1.1 Topography
A digital elevation model (DEM) is used by the
ArcSWAT interface for a) delineation of subbasins and
associated reaches, and b) computation of topographic
characteristics. A 1/3 arc second (approximately 10
meters) DEM acquired from United States Geological
Survey-National Elevation Dataset (USGS-NED) was
used for the development of the SWAT model. The
vertical datum of the DEM is the North American
Vertical Datum of 1988. Figure 1 shows the elevation
in the watershed, varying from 48 meters in the south-
west to 417 meters in the north-east.
2.1.1.1 Delineation of subbasins and reaches
The automatic watershed delineation feature of the
ArcSWAT interface was initially used to delineate
subbasins and reaches based on the DEM. The National
Hydrography Dataset (NHD) high-resoiution stream
centerline layer was burnt to the DEM during the
delineation process. This ensured that the model
reaches aligned correctly with the rivers/streams on
the ground. In addition, a subbasin was delineated
for each historic milldam site in the watershed.
Quantifying legacy sediment from historic milldam sites
is one of the stated objectives of this project. Most of
these milldams have been removed since the beginning
of the 20m century and these sites have been identified
as hotspots of sediment load generation (Merritts and
Walter 2003, Walter and Merritts 2007, Merritts et al.
2010). The initial delineation process resulted in 1445
subbasins and reaches.
It is important to note that the use of an overly large
number of subbasins can result in unreasonably long
model run times and large unmanageable output
files. To avoid such issues a number of subbasins in
the initial delineation were aggregated while ensuring
that a cut still existed for each historic milldam site.
The final delineation consisted of 290 subbasins and
reaches. Figure 2 shows the initial and final delineated
subbasins and reaches for the Conestoga River
watershed model.
¦
K«mhor«i
srtlUfiittt
rBfac^Cre^
Raphe
Twp
I Watershed Boundary
Elevation (m)
- High : 417
Conestoga River Watershed
Elevation
TETRA TECH
Tifnffriv
Legend
4-01 (iintnii
itmbug
CnrtTvyp
TV*|>
Rartvchijrg
HljhUniJ
Figure 1. Elevation in the Conestoga River watershed
-------
¦jfWTwj
(AjfUVlll!*
Legend
SWJ Reaches
~ SWAT Su&basins
I Watershed Boundary
Conestoga River Watershed
Watershed Delineation (initial)
10
~ Kiorreters
10
ZD Miles
TETftA TECH
— SWAT Reaches
| SVWVT Susbasins
n Watershed Boundary
Conestoga River Watershed
Watershed Delineation (final)
10
~ Kilometers
10
Hi Miles
TETRA TECH
Legend
~ Historic Dams
Figure 2. Delineated subbasins and reaches for the Conestoga River watershed ¦
-------
2.1.2 Landuse and Landcover
The Chesapeake Bay Watershed Land Cover Data Series
(CBWLCD) (USGS 1984, USGS 1992, USGS 2001, USGS
2006) was used to represent the landuse and landcover
in the watershed model. The CBWLCD was funded by
USGS and was produced to study the change in landuse
and landcover in the Chesapeake Bay watershed. The
CBLWCD was produced for 1984, 1992, 2001 and 2006,
derived from Landsat 5 Thematic Mapper and Landsat
7 Enhanced Thematic Mapper satellite imagery. Table 1
shows the area associated with each landcover class in
the Conestoga River watershed according to CBLWCD
1984, 1992, 2001 and 2006.
The CBWLCD shows that there have been small
changes in landuse areas between 1984 and 2006
with some increases in developed land. However, the
changes are not significant enough to warrant the
use of multiple layers to represent landuse change.
Therefore, CBWLCD from any year is an appropriate
representation of landuse and landcover in the
watershed. The SWAT model for the Conestoga River
watershed was run for a period of 30 years from
1/1/1985 to 12/31/2014. Since 2001 serves as the
approximate mid-point of simulation, CBWLCD 2001
was considered to be an appropriate representation of
the average landuse and landcover in the watershed
model. Figure 3 shows the landuse and landcover in
the Conestoga River watershed as per CBLWCD 2001.
National Land Cover Database (NLCD) (Jin etal. 2013,
Fry et al. 2011, Homer et al. 2007) 2001, 2006 and
2011 were also reviewed as a check on CBWLCD. NLCD
1992 was not considered on account of its different
landcover classification scheme compared to the other
years.
Table 2 shows the area associated with each landcover
class in the Conestoga River watershed according to
NLCD 2001, 2006 and 2011.
Table 1. Landuse and landcover in the Conestoga River watershed according to CBWLCD
Landuse/Landcover
Area (acres)
CBWLCD 1984 CBWLCD 1992 CBWLCD 2001 CBWLCD 2006
Open Water
887
896
896
927
Developed, Open Space
14,132
17,339
17,990
17,926
Developed, Low Intensity
22,011
26,881
27,447
27,564
Developed, Medium Intensity
9,027
10,266
10,508
10,836
Developed, High Intensity
4,841
5,273
5,548
6,092
Barren Land
1,356
1,399
1,659
1,605
Deciduous Forest
58,535
58,084
57,866
57,607
Evergreen Forest
1,193
1,162
1,148
1,134
Mixed Forest
4,262
4,263
4,272
4,257
Shrub/Scrub
14,814
14,871
15,114
14,897
Grassland/Herbaceous
3,812
3,847
4,028
4,050
Pasture/Hay
23,359
22,048
27,205
27,902
Cultivated Crops
142,427
134,329
127,038
125,975
Woody Wetlands
2,864
2,863
2,841
2,793
Emergent Herbaceous Wetlands
197
197
159
155
6
-------
10
^ Kilometers
10
^ Miles
"It
TETRA TECH
Legend
Watershed Boundary
CBWLCD 2001
Category
m Open Water
[ Perennial Ice/Snow
| Developed Open Space
| Developed low Intensity
| Developed Medium Intensity
IB Developed High Intensity
| Barren Land
Deciduous Forest
| Evergreen Forest
O Mixed Forest
| Stirvib/Scrub
| Grassland'Herbaceous
~ PastureVHay
| Cultivated Crops
| Woody WWIancs
| Emergent Herbaceous Wetlands
Conestoga River Watershed
CBLWCD 2001 Landuse/Laridcover
irr«tlr
Figure 3. Landuse and laridcover in the Conestoga River watershed as per CBLWCD 2001
Table 2, Landuse and landcover in the Conestoga River watershed according to NLCD
i_cuiuuog/i_aiiuouvci . „
|| NLCD 2001
NLCD 2006
NLCD 2011
Open Water
959
997
997
Developed, Open Space
38,346
37,486
38,753
Developed, Low Intensity
26,966
28,384
27,779
Developed, Medium Intensity
10,442
11,723
11,994
Developed, High Intensity
4,467
4,608
5,102
Barren Land
1,363
1,395
1,374
Deciduous Forest
59,274
58,933
58,527
Evergreen Forest
1,029
1,003
986
Mixed Forest
1,671
1,639
1,632
Shrub/Scrub
9,755
9,607
9,804
Grassland/Herbaceous
1,205
1,293
1,318
Pasture/Hay
61,018
60,223
59,773
Cultivated Crops
84,438
83,670
82,924
Woody Wetlands
2,496
2,473
2,470
Emergent Herbaceous Wetlands
444
438
438
-------
It is evident from the tables above that differences
exist in areas reported for different landcover classes
reported by NLCD and CBWLCD. CBWLCD is likely
more accurate since it was produced for the general
physiographic region. As a result, this dataset was
selected for landuse representation in the watershed
model.
Both NLCD and CBWLCD do not provide any
information on the types of crops or crop rotation
practices in the broader cultivated crops category.
Agricultural management practices like tillage
and fertilizer application have significant impacts
on sediment and nutrient loads generated at the
landscape level. It is therefore important to classify
the cultivated crops category to individual crops and
associated rotation practices, and parameterize the
model with their associated management operations.
Several years of Cropland Data Layer (CDL) and
National Agricultural Statistics Service (NASS) Census
of Agriculture from the US Department of Agriculture
(USDA) were used to further classify cultivated crops to
major crops and rotations in the watershed.
CDLs from 2008 to 2014 show that the major crops
in the watershed were corn and soybean. A GIS
based analysis on CDLs from 2008 to 2014 showed
that continuous corn and corn-soybean rotation are
practiced in a majority of the agricultural lands in the
watershed. CDLs however do not provide information
on the proportion of grain corn and silage corn. This
distinction is important because of the differences
in residue management and potential impacts on
sediment yield. A field with more residue cover will
likely generate less sediment per unit area compared
to one with less cover given that all other conditions
remain the same.
The NASS Census of Agriculture provides area
harvested under each crop type by county (Table 3).
Table 3 shows that approximately 54% of the corn
area is grain corn and the remaining is silage corn. It
is important to note that barley, rye and wheat are
generally grown as cover crops and are generally grown
in the same fields following corn or soybean. Based
on the information presented in Table 3 and the CDL
analysis, the following crop rotations were simulated in
the SWAT model (Table 4).
Table 3. Average area harvested by crop type in
Lancaster County as per USDA-NASS Census
of Agriculture 1997, 2002, 2007, 2012
Crop Type
Area Harvested (acres)
Percentage
Grain corn
93,705
44%
Silage corn
78,691
37%
Soybeans
31,312
15%
Tobacco
6,229
3%
Other crops
1,917
1%
Barley
8,668
4%
Rye
1,834
1%
Wheat
11,968
6%
Total
211,854
100%
Table 4. Crops and associated rotations in the
SWAT model
Crop rotation
Description
Percentage of
agricultural land
CORN
Continuous
grain corn
30%
CSIL
Continuous
silage corn
35%
COSY
Corn-soybean
rotation
15%
SYCO
Soybean-corn
rotation
15%
AGRR
Other
5%
The above crop rotations ensure that in any given year
during the simulation, 45% of agricultural lands are
under grain corn, 35% under silage corn, 15% under
soybeans and 5% under other miscellaneous crops.
These fractions are consistent with the information
from the Census of Agriculture presented in Table 3.
8
-------
The management practices
associated with these crops (and
shown in Table 5) were developed
using information published by
the Pennsylvania State Extension
Service (http://extension.psu.edu).
Tillage practices sediment and
associated nutrient loads generated
from agricultural land. A farm tilled
using conventional practices will
likely produce more sediment load
per unit area than one that is not
tilled.
Baker (2011) estimates the
area of agricultural land under
a particular tillage practice at a
HUC8 watershed scale. A feasible
way of representing variations
in tillage operations in the SWAT
model is to split the agricultural
HRUs (see section 2.1.4 for the
definition of an HRU) based on
proportions of different tillage
practices reported at the HUC8
level and simulate each split HRU
with a unique tillage operation.
The drawback of this approach is
that the number of HRUs increase
and exert a burden on model run
time. Instead we chose to represent
tillage in the model as a generic
operation which may be assumed
as an average of the different
tillage practices reported for the
watershed. The absence of specific
tillage operations may have some
impacts on sediment and nutrient
loads generated at the spatial scale
of individual HRUs. The impact at
the watershed or subbasin scale is
however expected to be small.
Background landuses (forest,
wetland and grassland) were
assigned default management
practices generated by the SWAT
model.
Table 5. Crop management practices in the SWAT model
Operation Date or Fraction of Heat Units
Grain corn (CORN)
Tillage
May 10
Planting
May 15
Fertilizer application
May 15 (22 kg-N/ha + 67 kg-P205/ha)
Fertilizer application
June 20 (157 kg-N/ha)
Harvest and Kill
October 31
Silage corn (CSIL)
Tillage
May 10
Planting
May 15
Fertilizer application
May 15 (25 kg-N/ha +140 kg-P205/ha)
Fertilizer application
June 20 (176 kg-N/ha)
Harvest and Kill
September 25
Soybeans (SOYB)
Tillage
May 25
Planting
May 31
Fertilizer application
May 31 (56 kg-P205/ha)
Harvest and Kill
October 31
Generic crop (AGRR)
Tillage
0.15
Planting
0.15
Auto fertilization initialization
0.01 (maximum of 180 kg-N/ha/yr)
Harvest and Kill
1.2
Hay (HAY)
Planting
0.15
Fertilizer application
0.6 (54.2 kg-N/ha)
Harvest
0.6
Fertilizer application
0.85 (54.2 kg-N/ha + 67 kg-P205/ha)
Harvest
0.85
Fertilizer application
1.2 (54.2 kg-N/ha)
Harvest and Kill
1.2
Bermudagrass* (BERM)
Planting
0.15
Auto fertilization initialization
0.01 (maximum of 50 kg-N/ha/yr and 17 kg-P/ha/yr)
Harvest and Kill
1.2
*Simulated on urban pervious areas and representative of urban lawns
9
-------
2.1.3 Soils and Geology
Soil Survey Geographic (SSURGO) database from the
USDA National Resource Conservation Service (NASS)
was used to represent soils and their associated
properties in the watershed model. There are 260 and
59 unique soils based on MUKEY and major component
name, respectively, in the Conestoga River watershed.
MUKEY is the unique identification number associated
with each soil survey polygon. Figure 4 shows the
soils by hydrologic soil group {HSG) in the watershed.
Based on HSG, B soils are generally dominant (found in
approximately 86 % of the watershed area).
The following properties are required for each soil in a
SWAT model,
• Number of horizons
• Hydrologic soil group
• Maximum rooting depth
• Anion exchange capacity
• Soil cracking potential
The following properties are required for the soil
horizons associated with each soil,
• Depth of horizon
• Bulk density
• Available water capacity
• Hydraulic conductivity
• Percent organic carbon
• Percent sand, silt and clay
• Percent rock
• Albedo
• USLE erosivity factor
• Electrical conductivity
£h2>b»0rtcw,»
Wftfhtivilkr
v**n«uiv
T*|>
l factor,
'»t mtHM |
Urea* t»ii
Legend
j n Watershed Boundary
SSURGO Soils
Hydrologic Soil Group
I A
HI A'D
B
B/D
HH c
I D
Conestoga River Watershed
SSURGO Soils
NAO laiwoTM Jam
N 0
5
to
TETRA TECH
A °
5 10
Un|> pnHAitnd S Sarttnr
Figure 4. Soils in the Conestoga River watershed
10
-------
All the parameters listed above were available from the
SSURGO database. A small fraction of required data
were missing, which were addressed using the following
approach,
• If values for parameters associated with a
given horizon were missing then these were
filled using data from an adjacent horizon of the
same soil.
• If data for ali horizons were missing then the
SWAT soils database was used to fill data
based upon the name of the soil.
Figure 5 shows the bedrock geology in the watershed.
Limestone and dolostone are the predominant
lithology in approximately 58% of the watershed area.
Areas under limestone and dolostone are susceptible
to the formation of karst features like sinkholes, springs
and caves. Karst topography is present in a significant
portion of the Conestoga River watershed (Figure 6)
and generally conform to the areas under limestone
and dolostone.
Bedrock geology and presence of karst features may
have significant impacts on the hydrology and water
quality, and were important considerations in the
parameterization of the model during calibration and
validation.
Legend
I' | Watershed Boundary
Geology
Predominant Lithology
| arkose
| conglomerate
| diabase
| dolostone (dolomite)
| felsic gneiss
| gneiss
| limestone
| mafic gneiss
| phyllite
| quartzite
| sandstone
| schist
¦ shale
Conestoga River Watershed
Geology
N o
5
10
TETRA TECH
A "
5 10
Uitp pmKit«d - '» Sartor
tNT M'
f U*fi T~p
Figure 5. Geology of the Conestoga River watershed
11
-------
-MWUM
to
^3 Miles
It
TETRA TECH
Legend
Karst Features
• cave
sinkhole
surface depression
• surface mine
Mode! Reaches
I | Watershed Boundary
| Modei Subbasms
Conestoga River Watershed
Karst Features
HAO_ IMSJJIMftjwtJftl
Mitp prcoecMl • S
Source: Bureau of Topographic and Geologic Survey, Department of Conservation and Natural Resources, 2007
(http://www. pasda.psu.edu/uci/Metadata Display. aspx?entry=PAS DA&file=DCNR_PAKarst,xml&dataset=3073)
Figure 6. Karst features in the Conestoga River watershed
2.1.4 Hydrologic Response Unit (HRU) Development
An HRU is the smallest physical entity in a SWAT
model for which all the land phase hydrology and
water quality processes are simulated. Each HRU in a
modeled subbasin is a unique combination of landuse/
landcover, soil and slope category. The ArcSWAT
interface generates HRUs by intersecting these layers.
Slope is calculated by the ArcSWAT interface during the
model setup process. Two slope classes were defined
for the watershed model, namely, 0 to 3% and 3% or
above. Given the absence of a general guidance on
the classification of slopes for HRU development, the
less than 3% and 3% or above slope categories were
adopted to distinguish between low and moderate to
high sloping areas, respectively. The landuse/landcover
and soil layers used in the watershed model have been
discussed in detail in the previous sections.
The HRU setup process using the ArcSWAT interface
provides a unique feature of imposing thresholds on
landuse, soil and slope to remove HRUs which occupy
small areas in a given subbasin1. The area lost from
removing such HRUs is apportioned to the remaining
HRUs in a given subbasin. The use of thresholds
is a standard (and necessary) process for SWAT
applications. Imposing thresholds ensures that an
unmanageably large number of HRUs are not produced
and that the model still represents the dominant
physical aspects of the watershed reasonably. The
presence of large numbers of HRUs has significant
impacts on model run times and often produce large
output files which makes post-processing of results for
model evaluation difficult.
'-Please refer to the ArcSWAT Interface user's guide (Wincheii et al. 2012) for details on the operations performed by the ArcSWAT
Interface when thresholds are imposed.
12
-------
Thresholds of 10% were imposed on landuse, soil
and slope for the Conestoga River watershed SWAT
model, resulting in 12,194 HRUs. If thresholds were
not imposed then the total number of HRUs would be
approximately 41,000.
Since urban and agricultural lands are generally critical
sources of sediment and nutrients, it was deemed
important that areas associated with these landuses
were not lost during the HRU process. Thresholds were
therefore not imposed on these landuses.
2.1.5 Meteorology
Meteorological data required for a SWAT application
consist of daily precipitation, maximum and minimum
daily air temperature, wind speed, relative humidity
and solar radiation. Watershed modeling applications
have historically relied on point measurements of
meteorological parameters. Point measurements
however often suffer from data gaps requiring
extensive processing to address such data gaps and do
not adequately represent spatial variations.
In recent years, several gridded meteorological
products have been made publicly available. These
products incorporate point and radar estimates of
rainfall, do not have data gaps and have generally
undergone extensive quality checks. PRISM (PRISM
Climate Group) and NLDAS-2 (http://Idas.gsfc.nasa.
gov/nIdas/NLDAS2forcing.php) are some such products
which are appropriate for watershed modeling
applications. PRISM provides daily precipitation, and
maximum and minimum temperature grids at an
approximate spatial resolution of 4 km by 4 km for
the CON US from 1981 onwards. NLDAS-2 provides
hourly precipitation, air temperature, wind speed,
longwave and shortwave radiation, specific humidity,
air pressure, and potential evapotranspiration at an
approximate spatial resolution of 12 km by 12 km for
the CONUS from 1979 onwards.
Precipitation and temperature data from PRISM
were used on account of their finer spatial resolution
compared to NLDAS-2. Solar radiation, wind speed and
relative humidity (calculated using specific humidity, air
pressure and temperature) from NLDAS-2 were used in
the watershed model. Since both PRISM and NLDAS-2
are derived from the same historical it is appropriate
to mix these sources for meteorological forcing. The
precipitation and temperature gridded data were
area weighted to have one representative station for
each modeled subbasin. The NLDAS-2 grids for solar
radiation, wind speed and relative humidity were not
aggregated and were assigned to the model subbasins
based upon proximity.
2.1.6 Point Sources
The Chesapeake Bay Program maintains monthly
flow and water quality data for all point source
dischargers in the Chesapeake Bay watershed. The
Bay Program Nutrient Point Source Database (httpi//
www.chesapeakebay.net/data/downIoads/bay_
program_nutrient_point_source_database) consists
of monitored and estimated data that are either
submitted or approved by various jurisdictions in the
Chesapeake Bay watershed. According to this database,
there are 58 municipal and 30 industrial point source
discharges in the Conestoga River watershed (Table 6).
Figure 7 shows the locations of the point sources in the
watershed. Monthly flow, sediment and nutrient loads
for all point sources were used in the Conestoga River
watershed SWAT model.
I
-------
Tb
10
3 Kilometers
10
ZD Mites
TETRA TECH
Legend
Point Sources
# Major
C Minor
Model Reaches
~ Model Subbasins
I | V\fetershed Boundary
Conestoga River Watershed
Point Sources
WW1 J*MJ4TU7flrt._1 l¥l
Figure 7. Point sources in the Conestoga River watershed
Table 6. Point sources in the SWAT model
NPDES ID
FACILITY NAME
TYPE
MAJOR
SUBBASIN
flowavg
(mgd)
(mg/l)
(mg/l)
T^AVG
(mg/l)
PA0007536
WILBUR CHOCOLATE CO INC
INDUSTRIAL
85
0.6000
30.0000
2.1600
1.5482
PA0008508
BURLE BUSINESS PARK LP
INDUSTRIAL
66
0.0350
30.0000
48.3514
0.2649
PA0010511
SPRING GLEN FRESH FOODS INC
INDUSTRIAL
175
0.0250
30.0000
2.2489
0.4798
PA0020222
TERREHILL BORO
MUNICIPAL
194
0.1670
30.0000
19.8714
1.0529
PA0020320
LITITZ SEWAGE AUTHORITY
MUNICIPAL
Y
83
2.6834
6.4712
20.2531
1.2931
PA0021865
ADAMSTOWN BOROAUTH OF LANCAST
MUNICIPAL
Y
183
0.3219
9.5110
15.8363
0.2772
PA0021890
NEW HOLLAND BOROUGH AUTHORITY
MUNICIPAL
Y
273
0.8821
6.2549
15.9088
2.0857
PA0026620
MILLERSVILLE BOROUGH
MUNICIPAL
Y
58
0.6761
6.2064
12.1289
2.1441
PA0026719
LANCASTER CITY
MUNICIPAL
Y
66
8.7000
0.0000
18.0000
0.0000
PA0026743
LANCASTER CITY
MUNICIPAL
Y
66
19.0231
14.3074
10.4466
1.6398
PA0027405
EPHRATA BOROUGH WWTP
MUNICIPAL
Y
131
2.8175
7.0900
15.7652
1.2924
PA0030911
EASTERN LANCASTER CO SCH DIST
MUNICIPAL
217
0.0030
30.0000
24.9880
3.9981
PA0031631
TWIN VALLEY SCHOOL DISTRICT
MUNICIPAL
226
0.0270
30.0000
37.6991
1.2161
PA0031861
ZERBE SISTERS NURSING FAC INC
MUNICIPAL
223
0.0360
30.0000
13.7184
1.3767
PA0033111
OAK CREEK CAMPGROUND INC
MUNICIPAL
193
0.0042
30.0000
86.4485
11.3845
PA0033553
GEHMANS MENNONITE SCHOOL
MUNICIPAL
183
0.0014
30.0000
27.8367
1.6792
14
-------
Table 6 (continued). Point sources in the SWAT model
NPDES ID FACILITY NAME
TYPE MAJOR SUBBASIN f\ 0V^VG
(mgd) (mg/l) (mg/l) (mg/l)
PA0035092
TYSON FOODS
INDUSTRIAL
Y
272
0.6294
8.8290
69.1018
10.4995
PA0051683
TITANIUM HEARTH TECH INC - D/B/ATIMET
INDUSTRIAL
227
0.0080
30.0000
8.5000
2.1000
PA0051764
GALEN HALL CORP
MUNICIPAL
165
0.0020
30.0000
24.9880
3.9981
PA0051861
PENN SYLVAN REALTY CORP
MUNICIPAL
193
0.0040
30.0000
16.4173
3.9981
PA0052078
ELVERSON BORO MUNI AUTH STP
MUNICIPAL
226
0.1250
30.0000
21.1532
0.5805
PA0055328
NEW MORGAN LANDFILL CO INC
INDUSTRIAL
Y
225
0.0036
5.8374
6.5306
0.3657
PA0070424
CAERNARVON TWP STP
MUNICIPAL
Y
221
0.2668
5.1873
14.4071
0.5686
PA0080331
EPHRATAAREA JOINT AUTH - FULT
INDUSTRIAL
132
0.0100
30.0000
0.2500
0.5500
PA0080438
NORTHERN LANCASTER CO AUTH
MUNICIPAL
Y
189
0.3199
12.3830
31.6251
1.0253
PA0080594
NORTH AMERICAN BRISTOL CORP/BR
INDUSTRIAL
262
0.1150
30.0000
5.6000
3.0000
PA0080918
SFS LONG, RICK
MUNICIPAL
107
0.0004
30.0000
24.9880
3.9981
PA0080926
SFS SCHROEDER, ROBERT
MUNICIPAL
115
0.0004
30.0000
24.9880
3.9981
PA0081141
LOCUST WOOD MHP
MUNICIPAL
167
0.0100
30.0000
15.1827
1.1594
PA0081213
CLAY TWP SUPERVISORS - VILLAGE
MUNICIPAL
120
0.0350
30.0000
20.2446
0.4243
PA0081515
LANCASTER CO CAREER & TECH CEN
MUNICIPAL
172
0.0043
30.0000
47.1924
6.1101
PA0081710
OUTDOOR WORLD RESORTS LLC
MUNICIPAL
59
0.0600
30.0000
69.8665
0.5795
PA0081949
WEST EARL SEW AUTH
MUNICIPAL
Y
171
0.2184
5.6725
5.6893
1.0654
PA0082023
UPPER LEACOCK TWP
INDUSTRIAL
72
0.0640
30.0000
0.1914
0.4211
PA0082066
VALLEY PROTEINS INC
INDUSTRIAL
189
0.1000
30.0000
49.8161
0.1299
PA0082333
HARNISH, ROBERT-CONESTOGAHI
MUNICIPAL
286
0.0107
30.0000
118.9413
12.1826
PA0082635
SUN VALLEY LLC C/O DIVERSIFIED
MUNICIPAL
193
0.0200
20.0000
66.9654
0.9096
PA0082791
CHESTER CO SOLID WASTE AUTH
INDUSTRIAL
229
0.0500
30.0000
7.7463
0.0131
PA0082881
ALCOA INC - LANCASTER WORKS
INDUSTRIAL
38
0.4820
30.0000
0.0000
0.0000
PA0082937
RR DONNELLEY & SONS CO
INDUSTRIAL
38
0.5180
30.0000
3.8000
0.0000
PA0083046
EASCO HAND TOOLS INC
INDUSTRIAL
26
0.2880
30.0000
0.1220
0.0031
PA0083208
REFRESHING MOUNTAIN CAMP INC
MUNICIPAL
123
0.0060
30.0000
63.1417
8.3870
PA0083429
WEST COCALICO TWP AUTH
MUNICIPAL
160
0.1500
30.0000
8.8072
3.2252
PA0083691
WEST EARL WATER AUTH - NOLT WE
INDUSTRIAL
171
0.0046
30.0000
0.5272
1.1598
PA0083844
EAST PETERSBURG BORO - GRAYSTO
INDUSTRIAL
50
0.0220
30.0000
1.0000
0.5074
PA0083909
CONESTOGO WOOD SPCLTIES CORP
MUNICIPAL
204
0.0191
30.0000
80.7630
9.5754
PA0084174
COVANCE RESEARCH PRODUCTS INC
MUNICIPAL
148
0.0300
30.0000
28.7862
0.9655
PA0084212
LEACOCK TWP SEW AUTH
MUNICIPAL
Y
281
0.1816
4.2266
5.1969
0.5424
PA0084247
NEXANS INC - BERK-TEK INC
MUNICIPAL
239
0.0075
30.0000
30.1555
3.2384
PA0084506
KITCH INC/STARLIGHT CAMPING RE
MUNICIPAL
125
0.0020
30.0000
46.7676
2.6137
PA0084581
NEW HOLLAND BORO AUTH
INDUSTRIAL
277
0.0821
30.0000
0.2500
0.5500
PA0084603
FAIRMOUNT HOMES
MUNICIPAL
175
0.0500
30.0000
11.3936
1.9044
PA0085103
DORMA DOOR CONTROLS INC
INDUSTRIAL
168
0.0111
30.0000
6.3000
0.3000
PA0085367
SILLS FAMILY CAMPGROUND
MUNICIPAL
185
0.0076
30.0000
114.6750
9.4954
PA0085448
EAST EARL SEW AUTH
MUNICIPAL
232
0.0040
30.0000
32.3845
4.2380
PA0085707
COCALICO VALLEY POULTRY FARMS
INDUSTRIAL
137
0.0160
30.0000
89.3771
12.0642
PA0086266
NORTHERN LANCASTER CO AUTH/KRA
MUNICIPAL
180
0.0070
30.0000
32.4535
4.2544
PA0086304
EARL TWP SEW AUTH
MUNICIPAL
270
0.2800
30.0000
7.4222
0.7677
PA0086541
DENVER COLD STORAGE
INDUSTRIAL
137
0.0512
30.0000
78.0876
7.1255
15
-------
Table 6 (continued). Point sources in the SWAT model
NPDES ID
FACILITY NAME
TYPE
MAJOR
SUBBASIN
flowavg
(nrigd)
tssavg
(mg/i)
tnavg
(mg/i)
TPAV3
(mg/i)
PA0086690
KALAS MANUFACTURING INC - PLAN
INDUSTRIAL
168
0.0390
30.0000
26.0500
0.0000
PA0087041
KLAAS BAKKER INC - LAKE-IN-WOO
MUNICIPAL
195
0.0150
20.0000
117.1104
1.4009
PA0087131
NORTHERN LANCASTER CO AUTH - G
MUNICIPAL
183
0.0050
30.0000
25.1525
2.7251
PA0087181
EPHRATA BORO AUTH #2
MUNICIPAL
Y
139
1.0885
4.4770
8.8649
1.3502
PA0087521
ASP REALTY INC
INDUSTRIAL
180
0.0100
30.0000
11.2000
7.0000
PA0088048
NEW MORGAN STP
MUNICIPAL
Y
224
0.0104
9.2221
11.6394
0.4440
PA0088722
JOHN F MARTIN & SONS INC
INDUSTRIAL
137
0.0900
30.0000
48.7766
0.9995
PA0088765
WENGER,NELSON
MUNICIPAL
137
0.0200
30.0000
20.2603
4.3579
PA0246395
INTEGRITY PLASTICS INC
INDUSTRIAL
168
0.0003
30.0000
11.2000
7.0000
PA0246778
MILLERS GREENHOUSE INC
INDUSTRIAL
46
0.0220
30.0000
11.2000
7.0000
PA0247197
LANCASTER CITY - CONESTOGA WTP
INDUSTRIAL
66
1.1100
30.0000
0.2500
0.5500
PA0247537
CHERRY PLACE PROPERTIES - FOUR
INDUSTRIAL
141
0.0025
30.0000
11.2000
7.0000
PA0247596
EPHRATAAREAJTAUTH
INDUSTRIAL
130
0.1560
30.0000
0.2500
0.5500
PA0247642
CLAY MANOR HOMEOWNERS ASSOC
MUNICIPAL
115
0.0110
30.0000
11.5854
1.6210
PA0247731
DS WATERS OF AMERICA LP
INDUSTRIAL
94
0.0600
30.0000
7.6463
0.6517
PA0247804
PRECISION MEDICAL PRODS INC -
INDUSTRIAL
168
0.0022
30.0000
0.6009
0.1324
PA0261360
MOUNTAIN VIEW MHP
MUNICIPAL
118
0.0020
30.0000
25.5278
1.9291
PA0261556
WHISPERING HOPE EAST STP
MUNICIPAL
174
0.0007
30.0000
24.9880
3.9981
PAG043623
SFS KNISLEYL0T2
MUNICIPAL
198
0.0004
30.0000
24.6260
4.1043
PAG043893
SFS SHORT, BARTON
MUNICIPAL
200
0.0004
30.0000
24.6260
4.1043
PAG043895
SFS PEIFER BROTHERS
MUNICIPAL
94
0.0004
30.0000
24.6260
4.1043
PAG044892
SARAH E EDGE SFTF
MUNICIPAL
149
0.0004
30.0000
24.6260
4.1043
PAG044966
WHITE, JAMES R. & SUSAN L.
MUNICIPAL
159
0.0004
30.0000
24.6260
4.1043
PAG045066
KENNETH KLEIN SEW SYS
MUNICIPAL
81
0.0004
30.0000
24.6260
4.1043
PAG045071
JOSEPH GEUNOT SFTF
MUNICIPAL
91
0.0004
30.0000
24.6260
4.1043
PAG045150
STANKO RESIDENCE SFTF
MUNICIPAL
226
0.0004
30.0000
24.6260
4.1043
PAG045169
CURTCHAFFE SRSTP
MUNICIPAL
226
0.0004
30.0000
24.6260
4.1043
PAG045229
SMALL FLOW TREATMENT FACILITY
MUNICIPAL
226
0.0004
30.0000
24.6260
4.1043
PAG046279
RUSSELL BURK RESIDENCE
MUNICIPAL
99
0.0004
30.0000
24.6260
4.1043
It is evident from the table above that a large number
of dischargers in the Conestoga River watershed have
TSS discharge at 30 mg/L. As noted in the earlier, point
source discharge data compiled by the Chesapeake Bay
Program for the Chesapeake Bay models and located
within the Conestoga River watershed were used in the
watershed model. The 30 mg/L is likely an assumption
that is used by the Bay Program for point sources
that did not have TSS discharge data in the PCS and is
often the permitted limit. The 30 mg/L concentration
also seems to apply mostly to minor dischargers in
the watershed. This assumption is not likely to have a
significant impact on the model outcome since the TSS
non-point source load in this watershed is much larger
than the point source load.
16
-------
2.2 HSPF
HSPF is a result of continued efforts of the USEPA,
USGS and other public and private entities over a
span of approximately 55 years and has been widely
used to support development of TMDLs, water supply
protection plans, and other watershed modeling efforts
throughout the US. HSPF was originally developed
as the Stanford Watershed Model in the 1960's. The
model has undergone several revisions since then
including the addition of water-quality processes,
algorithm enhancement, and development of
processing software. The model is in the public domain,
with current release Version 12.4, and model code is
stable and freely available for review.
An HSPF model was developed for the Conestoga River
watershed for the all the modeled reaches in the SWAT
model. The input to the HSPF model consisted of
a) SWAT simulated subbasin levei flow and sediment
loads, and b) point source loads. Time-series of daily
flow and sediment load were input to the HSPF model
as point sources to each respective modeled reach.
Flow and sediment loads were then routed through
the stream network using HSPF's internal routing
algorithms. Figure 8 shows a schematic representation
of the load transfer from the SWAT model to the HSPF
model.
SWAT Subbasin 1 point + non-point source flow and sediment load to HSPF Reach 1
HSPF Reach 1
SWAT Subbasin 2 point + non-point source flow and sediment load to HSPF Reach 2
HSPF Reach 2
Figure 8. Schematic representation of SWAT simulated flow and sediment load transfer to the
HSPF model
The parameters required for each modeled reach
in the HSPF model are:
• LEN (length of the reach)-
The ArcSWAT interface calculates the length
of each modeled reach during the SWAT model
setup. The reach lengths reported by SWAT model
were used.
• DELTH (drop in water elevation from the upstream
to the downstream end of the reach) -
The ArcSWAT interface reports the maximum
and minimum elevation associated with each
modeled reach. DELTH was calculated as the
difference between these reported elevations.
• VOL (initial volume of water in the reach) -
This was calculated as the product of reach length
(LEN), and mean reach depth and width calculated
using regional regression equations discussed
below.
In addition, parameters associated with scour and
deposition of sand, silt and clay were varied and
finalized during the calibration process. These are
discussed in section 3.2.1.2.
The storage and discharge associated with each reach
in an HSPF model is simulated based on user defined
stage-area-volume-discharge relationships or FTABLES.
The development of FTABLES are discussed below.
17
-------
2.2.1 FTABLE Development
An FTABLE is a piecewise linear function table with
each row consisting of values for surface area, volume
and outflow rate for a specified depth of water.
FTABLE development followed the methodology
outlined in BASINS Technical Note 2 (USEPA 2007).
Trapezoidal geometry is assumed for the channel
and the floodplain, and Manning's equation is used
for the calculation of outflow at various user-defined
depths. The geometric parameters required and their
respective sources are shown in Table 7.
Table 7. Geometric parameters for FTABLE
Parameter
Description
Source
Mean flow depth
Bieger et al. (2015)
Mean flow width
Bieger et al. (2015)
L
Length of reach
Conestoga River SWAT model
S
Longitudinal slope
reach dataset
m
c
Channel side slope
1.0 (assumed)
n
Manning's roughness coefficient
0.03 (assumed)
yc
Bankfull depth
1.25Ym (assumed)
n,
Floodplain split depth
1,5Yc (assumed)
Maximum floodplain depth
62.5Ym (assumed)
wF
Floodplain width
Wm (assumed)
mF
Floodplain side slope
2.0 (assumed)
The mean flow depth (V ) and mean flow width (W ) were determined based on regional regression equation
(shown below) for the Appalachian Highlands as reported by Bieger et al. (2015).
Ym = 0.26 * (Drainage Area)0 '287
Wm = 3.12 * (Drainage Area)0 415
For a given depth of water (y), surface area, volume and outflow are determined using the equations below.
Surface Area (Sj4) = L(b + 2 * mc * y)
Volume (F) = L(b * y 4- mc* y2)
-2j.
Outflow (Q) — — (b * y + mc * y2)5/3 + 2 * yjm2 +
Where, b = Wm — 2 * mc * Ym
18
-------
3.1 Hydrology
3.1.1 Methods
The calibration and validation process in the SWAT
model consisted of a systematic adjustment of
parameters generally geared towards getting the
closest match between simulated and observed daily
flows. Water years 2005 to 2014 was adopted as
the calibration period. Model validation consisted of
comparing simulated flows to observed flows at either
the same location as calibration for water years 1995 to
2004 or at locations not considered during calibration
with data prior to water year 2005. Table 8 and Figure 9
show the streamflow monitoring locations used in the
calibration and validation process.
The calibration process also consisted of comparing
simulated evapotranspiration to satellite based
estimates at a monthly time-step from MODIS Global
Evapotranspiration Project (MOD16) (http://ntsg.umt.
edu/project/modl6). MOD16 is part of a National
Aeronautics and Space Agency (NASA) project to
estimate global terrestrial evapotranspiration using
remote sensing data. The MOD16 evapotranspiration
datasets are estimated using an algorithm based on
the Penman-Monteith equation as outlined in Mu et al.
(2011).
As discussed in the Introduction section, a GFLOW
model was constructed for the Big Spring Run tributary
region of the Conestoga River watershed. The GFLOW
model was developed to understand the groundwater
dynamics in the region and to overcome some of
the limitations of the SWAT model associated with
simulating groundwater fluxes. The recharge simulated
by the SWAT modei was validated against the recharge
used in the GFLOW model.
Models are deemed acceptable when they can
simulate field data within predetermined statistical
measures. Model performance was generally assessed
at a monthly time-step using model evaluation criteria
suggested by Moriasi etal. (2007) (Table 9). For the
simulation of flow and pollutant loads, Moriasi et
al. summarized recent research and recommended
performance targets in terms of the Nash-Sutcliffe
coefficient (NSE), RMSE-observations standard
deviation ratio (RSR) and the magnitude of the relative
average error (RE, which Moriasi refers to as PBIAS).
The RE is a measure of the average tendency of the
simulated data to be larger or smaller than observed.
The NSE is an indicator of a model's ability to predict
the timing and magnitude of observed data. Values
may vary from -°° to 1.0. A value of NSE = 1.0 indicates
a perfect fit between modeled and observed data,
while values equal to or less than 0 indicate the
model's predictions are no better than using the
average of observed data.
19
-------
Table 8. Hydrology calibration and validation locations
USGS Id
Name
Calibration
Period
Validation
Period
Subbasin Id
01576754
Conestoga River at Conestoga, PA
2005-2014
1995-2004
9
01576500
Conestoga River near Lancaster, PA
2005-2014
1995-2004
68
01576540
Mill Creek at Eshelman Mill Road near Lyndon, PA
no data
1993-1998
255
01576085
Little Conestoga Creek near Churchtown, PA
no data
1991-1995
208
01576521
Big Spring Run near Willow Street, PA
no data
1994-2001
289
015765195
Big Spring Run near Mylin Corners, PA
2009-2014
no data
289
01576529
Unnamed Tributary to Big Spring Run near Lampeter, PA
no data
1994-2001
290
Figure 9. Streamflow monitoring locations in the Conestoga River watershed
* fUpno
V. / TWp
s ManMNi
n»TUv-.* *
¦MllHU
TW
I-:'
(Ml
Legend
USGS Flow Gages
Model Reaches
I! ^ Watershed Boundaiv
~ Model Stibbasins
Cones roga Creek near Churchtown, PA
Conestoga River a I Lancaster, PA
Unnamed Trib to Big Spring Run near Lampeter PA
ItrialMHc T«j.
Big Spring Run near Willow Street, PA
Big Spring Run near Mylin Comers, PA
Mill Creek at Eshelman Mill Road near Lyndon, PA
at Conestoga, PA
Conestoga River Watershed
Streamflow Monitoring Locations
NAO_1MO_IJTM _/**1
20
-------
Table 9. Performance targets for monthly average loads
Constituent
Very Good
Good
Satisfactory
Unsatisfactory
Flow, Sediment and Nutrients (NSE)
>0.75
>0.65
>0.5
<0.5
Flow, Sediment and Nutrients (RSR)
<0.5
<0.6
<0.7
>0.7
Flow (RE)
<±10%
<±15%
< ±25%
> ±25%
Sediment (RE)
<±15%
< ±30%
< ±55%
> ±55%
Nutrients (RE)
< ±25%
< ±40%
< ±70%
> ±70%
RSR is the ratio of RMSE and standard deviation
of measured data. RSR may be considered as a
normalized error index statistic with values ranging
from 0 to °°. An RSR of 0 indicates a zero RMSE and
hence a perfect model.
In addition, the model performance was also
evaluated using baseline adjusted coefficient of
model fit efficiency (Garrick et al., 1978), which
depends on absolute differences rather than
squared differences and is less sensitive to extreme
values.
A hydrologic calibration spreadsheet was also
used to determine the acceptability of modeling
results on the basis of statistical criteria in Table 10.
The spreadsheet computes the relative error for
various aspects of the hydrologic system. Statistical
targets developed and implemented in previous
studies (Lumb et al. 1994, Duda et al. 2012) were
defined and met for each aspect of the system
before accepting the model.
Graphical comparisons of observed and simulated
flows were performed in addition to the
statistical evaluation. Graphical comparisons are
extremely useful forjudging the results of model
calibration; time-variable plots of observed versus
modeled flow provide insight into the model's
representation of storm hydrographs, baseflow
recession, time distributions, and other pertinent
factors often overlooked by statistical comparisons.
The model's accuracy was assessed by interpreting
these time-variable plots.
Table 10. Hydrology calibration criteria
Statistic Criteria
Error in total volume
< 10%
Error in 50% lowest flows
< 10%
Error in 10% highest flows
< 15%
Seasonal volume error (summer)
< 30%
Seasonal volume error (fall)
< 30%
Seasonal volume error (winter)
< 30%
Seasonal volume error (spring)
< 30%
Error in storm volumes
< 20%
Error in summer storm volumes
< 50%
21
-------
The parameters adjusted during the calibration and
validation process are listed below. Table 11 shows
the values of these parameters in the calibrated and
validated model.
a) Subbasin level parameters
• CH_N2 - Manning's n value for main channel
• CH_N1 - Manning's n value for tributary channels
• CH_K1 - effective hydraulic conductivity in
tributary channel alluvium (mm/hr)
• SFTMP - snowfall temperature (deg-C)
• SMTMP - snowmelt temperature (deg-C)
• SMFMX - maximum melt rate for snow during
the year
• SMFMN - minimum melt rate for snow during
the year
b) HRU level parameters
• ESCO - soil evaporation compensation factor
• SURLAG - surface runoff lag time (days)
• GW_DELAY - groundwater delay (days)
• ALPHA_BF - baseflow alpha factor (days)
• GWQMN - threshold depth of water in shallow
aquifer required for return flow to occur (mm)
• REVAPMN - threshold depth of water in shallow
aquifer required for revap to occur (mm)
Consistent with the methodology adopted by
Baffaut and Benson (2009), areas under karst were
parameterized with higher values of CH_K1 compared
to the rest of the watershed.
A formal calibration and validation exercise was not
required for the HSPF model since SWAT simulated
flows at the subbasin level were used. The HSPF
model was however evaluated at USGS 01576754
Conestoga River at Conestoga for the calibration and
validation periods using the same statistical criteria
used for the SWAT model.
Table 11. Values of
parameters in the calibrated
and validated model
Parameter
Value
CH_N2
0.03
CH_N1
0.3
CH_K1*
0,75-300
SFTMP
0
SMTMP
0
SMFMX
3
SMFMN
1
ESCO
0.7
SURLAG
1
GW_DELAY**
10,75
ALPHA_BF
0.95
GWQMN
0
REVAPMN
0
*CH_K1 was assigned a value of 75 to 300 for
subbasins in karstified areas. For subbasins
in non-karst areas, the value of CH_N1 was
set to 0.
The value of GW_DELAY was set to 10
days for HRUs in subbasins 197 to 240. For
HRUs in all the other subbasins, the value of
GW_DELAY was set to 75 days. The value
of GW_DELAY was primarily based on the
behavior of the recession limb of observed
hydrographs.
22
-------
3.1.2 Results and Discussion
Setup : Hydrology; Sediment Nitrogen Cycle Phosphorus Cycle Plant Growth Landscape Nutrient Losses Land Use Summary j Instream Processes j Point Sources ' Reservoirs j About
». A A A A A A
PET
987.4
Vadose (unsaturated)
Zone
Shallow (unconflned)
Aquifer
Deep (confined)
Aquifer
Evaporation and
Transpiration
b»4.4
/ /
/ Precipitation
t J 1,111.3
t/iti
/ / /
/ / / / i
Average Curve Number
7227
F V> V>
r I
Inhltratoon/piant uptake/
- U!
boii moisture reostnbu ion
185.06
'-.'I |r.i":"T
80.17
Return Flow
Revep from shadow aquifer
19.65
Perception to shallow aquifer
283.1
283.09
Flow out of watershed
Recharge to deep aquifer
0
Realistic hydrology is the foundation of any model. Pay particular attention to
evapotranspiration, baseflowand surface runoff ratios. Basefiow/streamflow
ratios for the US are provided by the USGS. these data are accessible via
the button below. The ranges specified here are general guidelines only,
and may not apply to your simulation area.
Show Avg. Monthly Basin Values
Show US Baseflow Map
Messages and Warnings
Water yield may be excessive
Water Balance Ratios
Streamflow/Precip 0.49
Baseflow/Total Flow 0.66
Surface Runoff/Total Flow 0.34
Perc/Precip 0.25
Deep Recharge/Precip 0
ET/Precipitation
0.53
Figure 10. SWAT simulated hydrologic cycle of the Conestoga River watershed
Figure 10 shows the simulated hydrologic cycle of the
watershed evaluated using the SWAT Error Checker
(http://swat.tamu.edu/software/swat-check/).
Sanford and Selnick (2013) have estimated
evapotranspiration as a ratio of precipitation for the
Continental US (CONUS) using regression relationships.
The ratio of SWAT simulated evapotranspiration
and precipitation is 0.53, which is in the same range
reported for this region in Figure 13 of Sanford and
Selnick (2013).
The average annual recharge (or percolation to shallow
aquifer) simulated by SWAT is 283.1 mm/yr (~0.0008
m/day). This is comparable to the steady state recharge
rate of 0.0009 m/day in the GFLOW modei (see section
5 for details). It is important to note that the GFLOW
model was developed for the Big Spring Run, a very
small tributary of the Conestoga River. The average
recharge rate for the entire Conestoga River watershed
may likely be different from Big Spring Run.
The Error Checker did issue a warning stating that the
water yield may be excessive. The SWAT Error Checker
provides general guidelines only and this particular
warning may be disregarded because the total flow
was calibrated to observed streamflow at multiple
locations in the watershed. The model performance
pertaining to the various aspects of the hydrologic
cycle are discussed in the following sections.
23
-------
3.1,2,1 Evapotranspiration
The monthly simulated actual evapotranspiration
matches well with the satellite based estimates as
shown in Figure 11.
The statistical measures suggested by Moriasi et
al. (2007) for flow and water quality were used to
compare SWAT simulated evapotranspiration with the
MODIS estimates. The NSE, RSR and RE are 0.92, 0.28
and 1.01, respectively, indicating that SWAT simulates
evapotranspiration well for the watershed on a
monthly time-step.
•MODIS
•SWAT
c
o
E
c
o
v>
ro
i—
'CL
(/I
C
ro
L_
-t-J
o
Q.
ro
>
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month
SWAT vs MODIS ET
c
o
E
on
6
5
4
3
2
1
0
y = l.OOx - 0.09
R2 = 0.92
(2v 00
2 3 4
MODIS ET (in/month)
Figure 11. Comparison of SWAT simulated evapotranspiration with satellite based estimates
-------
3.1.2.2 Streamflow
The performance of the SWAT model is very good
at flow monitoring locations on the main stem
Conestoga River and is generally good to very good at
tributary monitoring locations. The summary of model
performance for streamflow for the calibration and
validation periods are provided in Table 12.
The SWAT model was not able to capture some of the
peak flows observed at the Big Spring Run and Little
Conestoga Creek flow monitoring locations. Since both
NSE and RSR are very sensitive to extreme events, the
inability of the model to capture some of these peaks
is reflected in the relatively lower values for these
statistical measures. It is also important to note that
the objective of this watershed model is to reasonably
represent long term average loads and simulate the
impacts of restoration BMPs. A few missed events in
smaller tributary areas is likely to have minimal impact
on the overall objectives of the project. Detailed
calibration and validation results for each location are
provided in sections 3.1.2.2.1 to 3.1.2.2.9.
The performance of the HSPF model was only
evaluated on the main stem Conestoga River at
Conestoga and was very good for both the calibration
and validation periods (Table 13).
Table 12. Evaluation of SWAT model performance for streamflow at a monthly time-step
USGS Id
Name
NSE
RSR
RE
Value
Performance
Value
Performance
Value
Performance
Calibration Period
1576754
Conestoga River at Conestoga, PA
0.93
Very Good
0.26
Very Good
7.04
Very Good
1576500
Conestoga River near Lancaster, PA
0.91
Very Good
0.30
Very Good
8.63
Very Good
15765195
Big Spring Run near Mylin Corners, PA
0.68
Good
0.57
Good
-9.24
Very Good
1576540
Mill Creek at Eshelman Mill Road near Lyndon, PA
no data
1576085
Little Conestoga Creek near Churchtown, PA
1576521
Big Spring Run near Willow Street, PA
1576529
Unnamed Tributary to Big Spring Run near Lampeter, PA
Validation Period
1576754
Conestoga River at Conestoga, PA
0.90
Very Good
0.32
Very Good
0.35
Very Good
1576500
Conestoga River near Lancaster, PA
0.89
Very Good
0.33
Very Good
4.28
Very Good
15765195
Big Spring Run near Mylin Corners, PA
no data
1576540
Mill Creek at Eshelman Mill Road near Lyndon, PA
0.85
Very Good
0.39
Very Good
4.14
Very Good
1576085
Little Conestoga Creek near Churchtown, PA
0.62
Satisfactory
0.62
Satisfactory
11.78
Good
1576521
Big Spring Run near Willow Street, PA
0.80
Very Good
0.45
Very Good
-12.25
Good
1576529
Unnamed Tributary to Big Spring Run near Lampeter, PA
0.82
Very Good
0.42
Very Good
-5.58
Very Good
Table 13. Evaluation of HSPF model performance for streamflow at a monthly time-step
USGS Id
Name
NSE
RSR
RE
Value
Performance
Value
Performance
Value
Performance
01576754
Conestoga River at Conestoga, PA (calibration)
0.93
Very Good
0.27
Very Good
7.99
Very Good
01576754
Conestoga River at Conestoga, PA (validation)
0.90
Very Good
0.31
Very Good
-1.29
Very Good
As evident from the results above, the performance
of both the SWAT and HSPF models are comparable
to each other primarily because both are configured
to use upland flow simulated by SWAT. Minor
discrepancies exist on account of the differences in
stream routing algorithms in the SWAT and HSPF
models.
25
-------
3.1.2.2.1 USGS 01576754 Conestocia River at Conestoaa. PA (calibration period)
Avg Monthly Rainfall (in)
Avg Observed Flow (10/1/2004 to 9/30/2014 )
Avg Modeled Flow (Same Period)
30000
25000
_ 20000
V)
^ 15000
o
"" 10000
5000
0
Oct-04 Apr-06 Oct-07
Apr-09 Oct-10 Apr-12 Oct-13
Date
Figure 12. Mean daily flow at USGS 01576754 Conestoga River at Conestoga, PA
Figure 13. Mean monthly flow at USGS 01576754 Conestoga River at Conestoga, PA
O-04
Avg Monthly Rainfall (in)
- Avg Observed Flow (10/1/2004 to 9/30/2014 )
Avg Modeled Flow {Same Period)
A-06
O-07
A-09
0-10
A-12
0-13
Month
26
-------
Avg Flow (10/1/2004 to 9/30/2014 )
¦ Line of Equal Value
Best-Fit Line
4000
5 3000
o
¦a
o
2000
"a
o
CD
§>1000
L_
CD
>
<
y = 0.9331 x + 104.49
R* = 0.9499
*
*
*
*
/y
7^
1000 2000 3000 4000
Average Observed Flow (cfs)
Avg Observed Flow (10/1/2004 to 9/30/2014 )
Avg Modeled Flow (10/1/2004 to 9/30/2014 )
- Line of Equal Value
100%
90%
80°/t
70%
60%
50%
40%
30%
20%
10%
O-04 A-06 O-07 A-09 O-10 A-12 0-13
Month
Figure 14. Monthly flow regression and temporal variation at USGS 01576754 Conestoga River at
Conestoga, PA
1500
V)
o
<:
E 1000
"O
0)
CD
"D
O
® 500 -
cn
CD
i—
<
• Avg Flow (10/1 /2004 to 9/30/2014)
Line of Equal Value
Best-Fit Line
y = 0.80E
R2 =
(6x+ 199.32
0.9679
~
~
*
/
*
/
~
*
/
>7
• /
*
s
*
*
*
500 1000
Average Observed Flow (cfs)
1500
1000
a)
I
500
1500
Avg Monthly Rainfall (in)
—*-Avg Observed Flow (10/1/2004 to 9/30/2014)
Avg Modeled Flow (Same Period)
Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep
I I I 1 1 1 1 1 1 T 1
10 11 12 1 2 3 4 5 6 7 8 9
Month
0
1
2 =
ro
c
Q CO
6 m
c
o
Figure 15. Seasonal regression and temporal aggregate at USGS 01576754 Conestoga River at
Conestoga, PA
27
-------
¦ Observed (25th, 75th)
- Median Observed Flow (10/1/2004 to 9/30/2014)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
1500
1000 -
500
Apr May Jun Jul Aug Sep
-i r
10 11 12 1
3 4
Month
Figure 16. Seasonal medians and ranges at USGS 01576754 Conestoga River at Conestoga, PA
Table 14. Seasonal summary at USGS 01576754 Conestoga River at Conestoga, PA
MONTH
OBSERVED FLOW (CFS)
MODELED FLOW (CFS)
MEAN
MEDIAN
25TH
75TH
MEAN
MEDIAN
25TH
75TH
Oct
785.81
499.50
297.50
755.00
874.20
572.63
345.31
853.29
Nov
684.82
509.50
373.00
726.00
735.48
577.40
410.36
795.82
Dec
931.41
796.00
481.50
1107.50
978.66
789.99
533.08
1150.73
Jan
896.46
758.50
560.25
1010.00
938.81
770.57
594.43
1055.29
Feb
862.34
748.00
591.25
981.00
909.05
700.29
545.70
961.44
Mar
1059.72
865.50
569.25
1187.50
1004.53
841.02
536.08
1166.09
Apr
945.34
766.50
522.50
1052.50
933.90
797.05
551.97
1042.67
May
740.48
549.00
431.25
830.00
827.84
666.57
525.57
952.00
Jun
705.57
489.50
348.75
673.25
781.79
611.12
451.68
844.64
Jul
528.60
391.50
293.25
571.50
633.88
509.77
411.59
704.00
Aug
387.74
289.50
228.25
398.75
478.83
385.64
279.97
494.05
Sep
618.09
266,00
180.751
438.75
690.96
313.26
215.60
541.02
28
-------
^—Observed Flow Duration (10/1/2004 to 9/30/2014 )
—Modeled Flow Duration (10/1/2004 to 9/30/2014 )
Figure 17. Flow exceedance at USGS 01576754 Conestoga River at Conestoga, PA
^—Observed Flow Volume (10/1/2004 to 9/30/2014 )
•^—Modeled Flow Volume (10/1/2004 to 9/30/2014 )
11200
-------
Table 15. Summary statistics at USGS 01576754 Conestoga River at Coriestoga, PA
SWAT Simulated Flow
Observed Flow Gage
REACH OUTFLOW FROM OUTLET 9
10-Year Analysis Period; 10/1/2004 - 9/30/2014
Flow volumes are (inches/year) for upstream drainage area
USGS 01576754 Conestoga River at Conestoga, PA
Hydrologic Unit Code 2050306
Latitude: 39.9464886
Longitude: -76.3677388
Drainage Area (sq-mi): 470
Total Simulated In-stream Flow:
23.56
Total Observed In-stream Flow:
22.01
Total of simulated highest 10% flows:
7.39
Total of Observed highest 10% flows:
7.24
Total of Simulated lowest 50% flows:
5.98
Total of Observed Lowest 50% flows:
5.13
Simulated Summer Flow Volume {months 7-9):
4.37
Observed Summer Flow Volume (7-9):
3.72
Simulated Fall Flow Volume (months 10-12):
6.29
Observed Fall Flow Volume (10-12):
5.84
Simulated Winter Flow Volume (months 1-3):
6.80
Obsened Winter Flow Volume (1-3):
6.72
Simulated Spring Flow Volume (months 4-6):
6.10
Observed Spring Flow Volume (4-6):
5.74
Total Simulated Storm Volume:
7.48
Total Observed Storm Volume:
6.99
Simulated Summer Storm Volume (7-9):
1.50
Observed Summer Storm Volume (7-9):
1.45
Errors (Simulated-Observed)
Error Statistics
Recommended Criteria
Error in total volume:
7.04
10
Error in 50% lowest flows:
16.41
10
Error in 10% highest flows:
2.11
15
Seasonal volume error - Summer:
17.62
30
Seasonal volume error - Fall:
7.76 » 30
Clear
Seasonal volume error - Winter:
1.08
30
Seasonal volume error - Spring:
6.42
30
Error in storm volumes:
7.00
20
Error in summer storm volumes:
2.99
50
Nash-Sutcliffe Coefficient of Efficiency. E:
0.862
Model accuracy increases as
E or E' approaches 1.0
Baseline adjusted coefficient (Garrick), E':
0.628
Monthly NSE
0.934
30
-------
3.1.2.2.2 USGS 01576500 Conestoaa River near Lancaster. PA (calibration period)
Avg Monthly Rainfall (in)
• Avg Observed Flow (10/1/2004 to 9/30/2014 )
Avg Modeled Flow (Same Period)
o
30000
25000
20000
15000
10000
10 to
Oct-04 Apr-06 Oct-07 Apr-09 Oct-10
Date
Apr-12
Oct-13
Hgure 19. Mean daily flow at USGS 01576500 Conestoga River near Lancaster, PA
Avg Monthly Rainfall (in)
• Avg Observed Flow (10/1/2004 to 9/30/2014 )
Avg Modeled Flow (Same Period)
!i
EE 1000
O-04
A-06
O-07
A-09
0-10
A-12
0-13
Month
Figure 20. Mean monthly flow at USGS 01576500 Conestoga River near Lancaster, PA
31
-------
• Avg Flow (10/1/2004 to 9/30/2014 )
Line of Equal Value
Best-Fit Line
2500
if?
5
o
2000
-a 1500
-o
° 1000
cu
en
2
QJ
>
<
500
y = 0.8383x + 128.16
Rz = 0.9389
/
/ •
/ /
/
*
y
* ^
* ^
* /
< •
J?
•
r
0 500 1000 1500 2000 2500
Average Observed Flow (cfs)
100%
"D
90% -
O
80% -
+
if)
70% -
JZ1
O
60% -
0
O
/—
50% -
L_
03
40% -
03
CD
30% -
L_
0)
•*->
03
20% -
5
10% -
Avg Observed Flow (10/1/2004 to 9/30/2014 )
Avg Modeled Flow (10/1/2004 to 9/30/2014 )
Line of Equal Value
0%
.
-
1 r-
0-04 A-06 O-07 A-09 0-10 A-12 0-13
Month
Figure 21. Monthly flow regression and temporal variation at USGS 01576500 Conestoga River near
Lancaster, PA
800
CO
o
| 600
T3
1)
a> 400
o
0)
2 200
$>
>
<
Avg Flow (10/1/2004 to 9/30/2014)
• Line of Equal Value
Best-Fit Line
y = 0.7376X + 180.35
R2 = 0.9666
V/ /
JB *
*
*
V\
\
N
\
>
\
/
*
*
*
/
200 400 600
Average Observed Flow (cfs)
800
Avg Monthly Rainfall (in)
-Avg Observed Flow (10/1/2004 to 9/30/2014)
Avg Modeled Flow (Same Period)
800 :
Ocf Nov Dec Jan Feb Mer Apr May Jun Jul Aug Sep
600
w
5
o
400
200
10 11 12 1 2 3 4 5 6 7 8 9
Month
Figure 22. Seasonal regression and temporal aggregate at USGS 01576500 Conestoga River near
Lancaster, PA
32
-------
¦ Observed (25th, 75th) Average Monthly Rainfall (in)
- Median Observed Flow (10/1/2004 to 9/30/2014) » Modeled (Median, 25th, 75th)
10 11 12 1 2 3 4 5 6 7 8 9
Month
Figure 23. Seasonal medians and ranges at USGS 01576500 Conestoga River near Lancaster, PA
Table 16. Seasonal summary at USGS 01576500 Conestoga River near Lancaster, PA
MONTH
OBSERVED FLOW (CFS)
MODELED FLOW fCFS)
MEAN
MEDIAN
25TH 75TH
MEAN
MEDIAN 25TH
75TH
Oct
543.14
311.50
174.001 526.25
600.82
413.181 227.18
: \
611.47
Nov
458.50
331.50
219.25 484.25
517.32
413.36 283.13
569.10
Dec
678.50
548.50
316.50 829.50
687.08
553.381 373.631
825.22
Jan
621.57
478.50
352.00: 724.00
651.38
530.071 393.76
745.67
Feb
598.30
486.501
351.75! 694.75
616.20
467.921 353.68
665.51
Mar
737.25
560.50
374.00 837.50
702.17
581.99; 352.86
854.70
Apr
657.11
466.00
327.25 720.00
643.81
548.08! 380.96 i
i :
738.61
May
496.13
325.50
259.00 533.75
576.47
458.391 357.74
672.66
Jun
444.24
296.50
199.75 434.25
529,89
419.01 302.42
! 579.60
Jul
327.61
235.00
168.00| 358.75
435.83
343.82! 269.56
494.94
Aug
233.99
165.00
125.25| 244.25
327.22
255.80 183.76
345.99
Sep
417.97
153.00
94.00! 247.25
459.43
202.94 133.65!
384.58
33
-------
Figure 24. Flow exceedance at USGS 01576500 Conestoga River near Lancaster, PA
^^"Observed Flow Duration (10/1/2004 to 9/30/2014 )
— Modeled Flow Duration (10/1/2004 to 9/30/2014 )
55000
5
° 5500
^—Observed Flow Volume (10/1/2004 to 9/30/2014 )
—Modeled Flow Volume (10/1/2004 to 9/30/2014 )
20% 30% 40% 50% 60% 70% 80%
Percent of Time that Flow is Equaled or Exceeded
100%
Oct-07 Apr-09 Oct-10
•-12
1-13
120%
100%
80%
60%
40%
20%
0%
Oct-04
Apr-06
Figure 25. Flow accumulation at USGS 01576500 Conestoga River near Lancaster, PA
34
-------
Table 1/.Summary statistics at USGS 01576500 Conestoga River near Lancaster, PA
SWAT Simulated Flow
Observed Flow Gage
REACH OUmOW FROM OUTLET 88
10-Year Analysis Period: 10/1/2004 - 9OT2014
Flow volumes are (inches/year) for upstream drainage area
USGS 01576500 Conestoga River at Lancaster. PA
Hydrologic Unit Code: 2050306
Latitude 40.05009748
Longitude. -76,2771798
Drainage Area (sq-mi)' 324
Total Simulated In-stream Flow:
23.57
Total Observed In-stream Flow.
21.70
Total of simulated highest 10% flows;
7.38
Total of Observed highest 10% flows:
7.88
Total of Simulated lowest 50% flows:
5.82
Total of Observed Lowest 50% flows:
4.42
Simulated Summer Flow Volume (months 7-9):
4.30
Observed Summer Flow Volume (7-9):
3.44
Simulated Fall Flow Volume (months 10-12):
6.37
Observed Fall Flow Volume (10-12):
5.93
Simulated Winter Flow Volume (months 1-3):
6.81
Observed Winter Flow Volume (1-3):
6.77
Simulated Spring Flow Volume (months 4-6):
6.09
Observed Spring Flow Volume (4-6):
5.56
Total Simulated Storm Volume:
7.41
Total Observed Storm Volume:
7.83
Simulated Summer Storm Volume (7-9):
1.45
Observed Summer Storm Volume (7-9):
1.49
Errors (Simulated-Observed)
Error Statistics
Recommended Criteria
Error in total volume:
8.63
10
Error in 50% lowest flows:
31.74
10
Error in 10% highest flows:
-6.33
15
Seasonal volume error - Summer:
25.01
30
- . 1
Seasonal volume error - Fall:
7.40 » 30
Clear
Seasonal volume error - Winter:
0.58
30
Seasonal volume error - Spring:
9.63
30
Error in storm volumes:
-5.26
20
Error in summer storm volumes:
-2.77
50
Nash-Sutcliffe Coefficient of Efficiency, E:
0.857
Model accuracy increases as
E or E' approaches 1.0
Baseline adjusted coefficient (Garrick), E':
0.604
Monthly NSE
0 911
35
-------
3.1.2.2.3 USGS 015767195 Big Spring Run near Mvlin Corners. PA (calibration period')
f
1
LL
Avg Monthly Rainfall (in)
- Avg Observed Flow (10/1/2008 to 9/30/2014 )
Avg Modeled Flow (Same Period)
140
120
100
80
60
40
20
0
Oct-08
| rip-
nnrTfr
inprrlTPT'
TWij
" I|T 1
p1 "ii
T|TirP|ifp
pn
r
r
i
I.
li
h
|
¦
,1
LiiUi
till i
ulll
jLj!
yu
! ii
Oct-09
Oct-10
Oct-11
Date
Oct-12
Oct-13
0
2
4
6
£
c
'ro
10 "5
~
12
14
Figure 26. Mean daily flow at USGS 015767195 Big Spring Run near Mylin Corners, PA
Figure 27. Mean monthly flow at USGS 015767195 Big Spring Run near Mylin Corners, PA
14
12
10 =§•
8 f
'ra
CC
6 >,
.c
4 o
2
2
0
Avg Monthly Rainfall (in)
—~—Avg Observed Flow (10/1/2008 to 9/30/2014 )
Avg Modeled Flow (Same Period)
O-09
0-10 0-11 0-12 0-13
Month
0
O-08
36
-------
£
o
$
o
TJ
Q)
a)
•a
o
'X
• P9J
••
*
* •
0
1
5 10
100%
"O
90%
o
80%
+
(fl
70%
_Q
o
60%
8
50%
C
CO
40%
CO
CD
30%
u.
-------
¦ Observed (25th, 75th)
- Median Observed Flow (10/1/2008 to 9/30/2014)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
6
Oct
f
«
Nov Dec Jan Feb
May
*
Jun Jul Aug Sep
f
I
¦ 1
- 2
- 3
4
- 5
- 6
CD
ra
01
c
o
10
11
12
3 4
Month
Hgure 30. Seasonal medians and ranges at USGS 015767195 Big Spring Run near Mylin Corners, PA
Table 18.Seasonal summary at USGS 015767195 Big Spring Run near Mylin Corners, PA
MONTH
OBSERVED FLOW (CFS)
MODELED FLOW (CFS)
MEAN
MEDIAN
25TTH
75TH
MEAN
MEDIAN
25TH
75TH
Oct
3.70
1.80
1301
2.50
3.50
1.57
0.89;
2.36
Nov
2.42
1.901
1.48
2.43
2.22
1.66
1.351
2.33
Dec
4.00
2.90!
1.831
3.78
3.70
2.59
1.691
3.27
Jan
3.62
2.90
2.10]
3.60
3.48
2.35
1.921
2.83
Feb
3.20
2.901
2.20
3.50
3.21
2.22
1.811
2.77
Mar
3.75
3 20
2.23!
3.98
3.06
2.12
1.64
3.13
Apr
3.64
2.90
2.10;
3.90
2.86
2.52
1.641
3.17
May
3.32
2.60
2.10
3.50
3.04
2.22
1.55
3.52
Jun
3.47
2.50
2.20i
3.20
2.90
2.46
1.771
3.05
Jul
2.28
1.80;
1.501
2.10
2.32
1.83
1.44
2.17
Aug
2.04
1.40
1.10
1.60
1.95
1.40
1.12
1.74
Sep
3.43
1.301
0.95!
2.50
3.09
1.18
0.87;
1.90
38
-------
42
o
0
O)
ro
s
<
ro
O
¦Observed Flow Duration {10/1/2008 to 9/30/2014 )
Modeled Row Duration {10/1/2008 to 9/30/2014 )
1000
100
L—J
20% 30% 40% 50% 60% 70% 80%
Percent of Time that Flow is Equaled or Exceeded
Figure 31. Flow exceedance at USGS 015767195 Big Spring Run near Mylin Corners, PA
¦Observed Flow Volume (10/1/2008 to 9/30/2014 )
Modeled Flow Volume (10/1/2008 to 9/30/2014 )
100%
o
O
w
CO
•o
CD
£
CD
in
_Q
o
0
E
3
§
T3
a)
N
ro
E
120%
100%
80% -
60% •
40%
20% -
0%
Oct-08
_,
»
L .
OcM)9
Oct-10
Oct-11
Oct-12
Oct-13
Figure 32. Flow accumulation at USGS 015767195 Big Spring Run near Mylin Corners, PA
39
-------
Table 19.Summary statistics at USGS 015767195 Big Spring Run near Mylin Corners, PA
SWAT Simulated Flow
Observed Flow Gage
REACH OUTFLOW FROM OUTLET 188
6-Year Analysis Period 10/1/2008 - 9OT2014
Flow \olumes are (inches/year) for upstream drainage area
USGS 015765195 Big Spring Run near Mylin Corners, PA
Hydrologic Unit Code: 2050306
Latitude. 39 9959361
Longitude -76.26403889
Drainage Area (sq-mi) 1 68
Total Simulated In-stream Flow:
23.79
Total Observed In-stream Flow
26.21
- , ' *
Total of simulated highest 10% flows;
9.31
Total of Observed highest 10% flows;
9.64
Total of Simulated lowest 50% flows:
5.63
Total of Observed Lowest 50% flows:
6.36
Simulated Summer Flow Volume (months 7-9):
4.98
Observed Summer Flow Volume (7-9):
5.25
Simulated Fall FlowVolume (months 10-12):
6.42
Observed Fall Flow Volume (10-12):
6.89
Simulated Winter FlowVolume (months 1-3):
6.49
Observed Winter FlowVolume (1-3):
7.06
Simulated Spring FlowVolume {months4-6):
5.91
Observed Spring FlowVolume (4-6):
7.01
Total Simulated Storm Volume:
6.13
Total Observed Storm Volume:
6.36
Simulated Summer Storm Volume (7-9):
1.53
Observed Summer Storm Volume (7-9):
1.61
Errors (Simulated-Observed)
Error Statistics
Recommended Criteria
Error in total volume:
-9.24
10
Error in 50% lowest flows:
-11.48
10
Error in 10% highest flows:
-3.45
15
Seasonal volume error - Summer:
-5.20
30
Seasonal volume error - Fall:
-6.88 » | 30
Clear
Seasonal volume error - Winter:
-8.12
30
Seasonal volume error - Spring:
-15.72
30
Error in storm volumes:
-3.53
20
Error in summer storm volumes:
-5.11
50
Nash-Sutcliffe Coefficient of Efficiency, E:
-0.005
Model accuracy increases as
E or E' approaches 1.0
Baseline adjusted coefficient (Garrick), E':
0.236
Monthly NSE
0.676
40
-------
3.1.2.2.4 USGS 01576754 Conestoaa River at Conestoaa. PA (validation period)
Avg Monthly Rainfall {in)
- Avg Observed Flow (10/1/1994 to 9/30/2004 )
Avg Modeled Flow (Same Period)
16000
14000
12000
10000
6000 -
10 io
Oct-94
Apr-96
Oct-97
Apr-99
Oct-OO
Apr-02
Oct-03
Date
Hgure 33. Mean daily flow at USGS 01576754 Conestoga River at Conestoga, PA
Avg Monthly Rainfall (in)
-Avg Observed Flow (10/1/1994 to 9/30/2004 )
Avg Modeled Flow (Same Period)
A-99 O-OO
Month
Figure 34. Mean monthly flow at USGS 01576754 Conestoga River at Conestoga, PA
41
-------
2500 t
f
I
2000 -
-a 1500
Q)
Q)
-o
¦§ 1000 -]
<
500
• Avg Flow (10/1/1994 to 9/30/2004 )
Line of Equal Value
Best-Fit Line
y =
0.9425X + 42.506
R2 = 0.9055
•
*
s
* X
*
* s
•
* s
s 4
s 4
Ssm.%
y •
•
•
gLjf •
^ *
m r #
•
•
•
MS~m
500 1000 1500 2000 2500
Average Observed Flow (cfs)
Avg Observed Flow (10/1/1994 to 9/30/2004 )
Avg Modeled Flow (10/1/1994 to 9/30/2004 )
- Line of Equal Value
100%
70% -
40% -
20% -
0-94 A-96 0-97 A-99 O-OO A-02 O-03
Month
Figure 35. Monthly flow regression and temporal variation at USGS 01576754 Conestoga River at
Conestoga, PA
1500
o
:iooo
"O
0
CD
"O
O
0 500
U)
CD
s
>
<
Avg Flow (10/1 /1994 to 9/30/2004)
•Line of Equal Value
Best-Fit Line
y = 0.838
R2
5x + 114.24
= 0.906
*
s
/
*
*
*
A *
9 9
. - *
/
*
*
*
*
s
500 1000
Average Observed Flow (cfs)
1500
1000
w
o
500 -
1500
Avg Monthly Rainfall (in)
-Avg Observed Flow (10/1/1994 to 9/30/2004)
Avg Modeled Flow (Same Period)
Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep
i i i i
10 11 12 1 2 3 4 5 6 7 8 9
Month
ro
1
ro
_>%
£
c
o
Figure 36. Seasonal regression and temporal aggregate at USGS 01576754 Conestoga River at
Conestoga, PA
42
-------
i Observed (25th, 75th)
¦ Median Observed Flow (10/1/1994 to 9/30/2004)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
May Jun
Month
Figure 37. Seasonal medians and ranges at USGS 01576754 Conestoga River at Conestoga, PA
Table 20. Seasonal summary at USGS 01576754 Conestoga River at Conestoga, PA
MONTH
OBSERVED FLOW (CFS)
MODELED FLOW (CFS)
MEAN
MEDIAN 25TH
75TH
MEAN MEDIAN
25TH
75TH
Oct
499.48
301.50: 193.25!
535.00
546.801 292.21
156.27
611.03
Nov
592.72
346.00 192.75
753.75
625.26 331.27
138.28
756.26
Dec
792.30
458.501 176.00
1047.50
767.95 i 396.76
I 111.14
975.39
Jan
856,41
574,00 295.50
909.75
696.741 510.30
236.55
813.74
Feb
812.02
667.00! 477.50
S I
949.50
849.801 618.36
445.85
893.46
Mar
1066.04
836.50: 626.25
1210.00
1044.48 i 761,21
j569.71
1147.90
Apr
818.04
752.00; 549.75
989.75
812.52! 692.70
!
j 530.16
937.78
May
658.94
517.00 385.75
769.25
676 62; 570.16
429.16
770.04
Jun
675.06
470.50; 281.75
744.50
707.20) 498.29
[ 345.06
742.49
Jul
533.65
389.00! 184.00
j |
556.00
523.37] 344.28
235.08
590.20
Aug
416.01
243.50 148.75
415.75
441.091 237.28
174.43
486.11
Sep
544.33
235.00! 153.75
542.25
609.12! 248.14
152.89
545.52
43
-------
•Observed Flow Duration (10/1/1994 to 9/30/2004 )
• Modeled Flow Duration (10/1/1994 to 9/30/2004 )
-
...
¦¦¦¦ "
-
r
—
1
i
1
1
1 i
!
~i
*j
~i
¦
:
;
1
'
_
..
. __
¦
52000
5200
520
52
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Percent of Time that Flow is Equaled or Exceeded
Figure 38. Flow exceedance at USGS 01576754 Conestoga River at Conestoga, PA
Oct-OO Apr-02 Oct-03
"^"Observed Flow Volume (10/1/1994 to 9/30/2004 )
—Modeled Flow Volume (10/1/1994 to 9/30/2004 )
120%
100%
80%
® 20%
Oct-94
Apr-96
Oct-97
Apr-99
Figure 39. Flow accumulation at USGS 01576754 Conestoga River at Conestoga, PA
44
-------
Table 21. Summary statistics at USGS 01576754 Conestoga River at Conestoga, PA
SWAT Simulated Flow
Observed Flow Gage
REACH OUTFLOW FROM OUTLET 9
10-Year Analysis Period: 10/1/1994 - 9/30/2004
Flow volumes are (inches/year) for upstream drainage area
USGS 01576754 Conestoga River at Conestoga, PA
Hydrologic Unil Code. 205Q306
Latitude: 39 9464BB6
Longitude: -76,3677388
Drainage Area (sq-mi): 470
Total Simulated In-stream Flow:
19.96
Total Observed In-stream Flow.
19.89
Total of simulated highest 10% flows:
7.39
Total of Observed highest 10% flows:
7.06
Total of Simulated lowest 50% flows:
3.71
Total of Observed Lowest 50% flows:
3.79
Simulated Summer Flow Volume (months 7-9):
3.81
Observed Summer Flow Volume (7-9):
3.62
Simulated Fall FlowVolume (months 10-12):
4.71
Observed Fall Flow Volume (10-12):
4.58
Simulated Winter FlowVolume (months 1-3):
6.17
Observed Winter FlowVolume (1-3):
6.53
Simulated Spring FlowVolume (months 4-6):
5.27
Observed Spring Flow Volume (4-6):
5.16
Total Simulated Storm Volume:
7.56
Total Observed Storm Volume:
6.52
Simulated Summer Storm Volume (7-9):
1.49
Observed Summer Storm Volume (7-9):
1.46
Errors (Simulated-Observed)
Error Statistics
Recommended Criteria
Error in total volume:
0.35
10
Error in 50% lowest flows:
-2.07
10
Error in 10% highest flows:
4.70
15
Seasonal volume error - Summer:
5.25
30
Seasonal volume error - Fall:
2.92 » 30
Clear
Seasonal volume error - Winter:
-5.51
30
Seasonal volume error - Spring:
2.06
30
Error in storm volumes:
15.86
20
Error in summer storm volumes:
2.13
50
Nash-Sutcliffe Coefficient of Efficiency, E:
0.760
Model accuracy increases as
E or E' approaches 1.0
Baseline adjusted coefficient (Garrick), E':
0.619
Monthly NSE
0 904
45
-------
3.1.2.2.5 USGS 01576500 Conestoaa River near Lancaster PA (validation period')
Figure 40. Mean daily flow at USGS 01576500 Conestoga River near Lancaster, PA
Avg Monthly Rainfall (in)
- Avg Observed Flow (10/1/1994 to 9/30/2004 )
Avg Modeled Flow (Same Period)
2000 12
10
8
10000
8000
6000
4000
2000
12000
Avg Monthly Rainfall (in)
Avg Observed Flow (10/1/1994 to 9/30/2004 )
Avg Modeled Flow (Same Period)
0
Oct-94
Apr-96
Oct-97
Apr-99
Oct-OO
Apr-02
Oct-03
1500
1000
0
0-94
A-96
0-97
A-99 O-OO
Month
A-02
O-03
Figure 41. Mean monthly flow at USGS 01576500 Conestoga River near Lancaster, PA
46
-------
Avg Flow (10/1/1994 to 9/30/2004 )
¦ Line of Equal Value
Best-Fit Line
2000
ST
o,
11500
x>
0
-$1000
Q)
OJ
ro 500
CD
>
<
y = 0.
F
9701x + 33.045
i2 = 0.9019
•
*
s
s
/
•
•
•
• v •
tVr .
hv
•
500 1000 1500
Average Observed Flow (cfs)
2000
Avg Observed Flow (10/1/1994 to 9/30/2004 )
Avg Modeled Flow (10/1/1994 to 9/30/2004 )
- Line of Equal Value
100% -
90% -
0-94 A-96 0-97 A-99 O-OO A-02 O-03
Month
Figure 42. Monthly flow regression and temporal variation at USGS 01576500 Conestoga River near
Lancaster, PA
to
§600
o
"O
a>400
o
0
co 200
0)
>
<
Nov Dec Jan Feb h\ar Apr May Jun
Avg Monthly Rainfall (in)
—~— Avg Observed Flow (10/1/1994 to 9/30/2004)
Avg Modeled Flow (Same Period)
• Avg Flow (10/1 /1994 to 9/30/2004)
Line of Equal Value
Best-Fit Line
y = 0.7837x + 11
R2 = 0.9128
600
200
0.5
1
c
1.55-"
3 >•
-C
3.5c
o
4 2
4.5
Average Observed Flow (cfs) Month
Figure 43. Seasonal regression and temporal aggregate at USGS 01576500 Conestoga River near
Lancaster, PA
47
-------
Hgure 44. Seasonal medians and ranges at USGS 01576500 Conestoga River near Lancaster, PA
900
800
700
^ 600
*t/T
500
| 400
LL
300
200
100
0
0
1
1
2
2
3
3
4
4
5
5
¦ Observed (25th, 75th) Average Monthly Rainfall (in)
- Median Observed Flow (10/1/1994 to 9/30/2004) ¦ Modeled (Median, 25th, 75th)
10 11 12 1 2 3 4 5 6 7 8 9
Month
Table 22. Seasonal summary at USGS 01576500 Conestoga River near Lancaster, PA
MONTH
OBSERVED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
MODELED FLOW (CFS)
MEAN MEDIAN 25TH 75TH
Oct
Nov
Dec
317.25) 175.00
370.75 208.50
533.65? 290.50
93.25
116.00
104.00
354.25
467.25
674.00
362.59
418.13
525.74
184.111 89.36
|
214.13 73.46
249.36 58.70
424.22
524.60
679.37
Jan
Feb
Mar
554.19; 360.00
540.37; 450.00
731.93 i 564.00
192.25
320.00
424.25
599.25
640.50
831.50
467.65
574.19
710.02
332.36 136.81
425.901 294.95
523.01; 384.22
568.83
608.30
776.92
Apr
May
Jun
561.97 j 520.50
424.95} 327.00
436.59; 292.00
364.00
242.50
172.75
670.00
483.50
459.00
555.55
467.54
488.76
477.98 359.06
392.35 290.07
341.12; 247.51
645.29
545.26
496.97
Jul
Aug
Sep
348.34) 231.50
264.00 151.50
323.68 135.50
104.00
80.00
79.75
346.75
245.50
322.00
370.88
299.01
403.16
226.74 155.62
147.77 107.38
151.89 88.68
399.76
305.48
382.11
48
-------
—^Observed Flow Duration (10/1/1994 to 9/30/2004 )
—Modeled Flow Duration (10/1/1994 to 9/3^2004 )
1300
LL
O)
130
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Percent of Time that Flow is Equated or Exceeded
Figure 45. Flow exceedance at USGS 01576500 Conestoga River near Lancaster, PA
—Observed Flow Volume (10/1/1994 to 9/30/2004 )
^—Modeled Flow Volume (10/1/1994 to 9/30/2004 )
120%
100%
80%
60%
40%
20%
0%
Oct-94 Apr-96
t-97
Apr-99
Oct-OO
Apr-02
Oct-03
Figure 46. Flow accumulation at USGS 01576500 Conestoga River near Lancaster, PA
49
-------
Table 23. Summary statistics at USGS 01576500 Conestoga River near Lancaster, PA
SWAT Simulated Flow
Observed Flow Gage
REACH OUTFLOW FROM OUTLET 68
10-Year Analysis Period: 10/1/1994 - 9/30/2004
Flow volumes are (inches/year) for upstream drainage area
USGS 01576500 Conestoga River at Lancaster. PA
Hydrologic Unit Code: 2050306
Latitude: 40 05009748
Longitude: -76.2771798
Drainage Area (sq-mi): 324
Total Simulated In-stream Flow:
19.69
Total Observed In-stream Flow
18.88
Total of simulated highest 10% flows:
7.44
Total of Observed highest 10% flows:
6.98
Total of Simulated lowest 50% flows:
3.41
Total of Observed Lowest 50% flows:
3.28
Simulated Summer Flow Volume (months 7-9):
3.77
Observed Summer Flow Volume (7-9):
3.29
Simulated Fall Flow Volume (months 10-12):
4.60
Observed Fall Flow Volume (10-12):
4.30
Simulated Winter Flow Volume (months 1-3):
6.05
Observed Winter Flow Volume (1-3):
6.33
Simulated Spring Flow Volume (months 4-6):
5.26
Observed Spring Flow Volume (4-6):
4.95
Total Simulated Storm Volume:
7.51
Total Observed Storm Volume:
6.51
Simulated Summer Storm Volume (7-9):
1.49
Observed Summer Storm Volume (7-9):
1.34
Errors (Simulated-Observed)
Error Statistics
Recommended Criteria
Error in total volume:
4.28
10
Error in 50% lowest flows:
3.80
10
Error in 10% highest flows:
6.58
15
Seasonal volume error - Summer:
14.53
30
Seasonal volume error - Fall:
6. 88 » | 30
Clear
Seasonal volume error - Winter:
-4.36
30
Seasonal volume error - Spring:
6.24
30
Error in storm volumes:
15.41
20
Error in summer storm volumes:
11.12
50
Nash-Sutcliffe Coefficient of Efficiency, E:
0.781
Model accuracy increases as
E or E' approaches 1.0
Baseline adjusted coefficient (Garrick), E1:
0.619
Monthly NSE
0.893
50
-------
3.1.2.2.6 USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon. PA (validation period)
Avg Monthly Rainfall (in)
¦ Avg Observed Flow (10/1/1992 to 9/30/1998 )
Avg Modeled Flow (Same Period)
3000
2500
2000
£
o
>
1500
o
Ll.
1000
500
0
Oct-92
Oct-93
Oct-94
Oct-95
Date
Oct-96
|1P|
rw
r i'|if(
uprnT]
H'r|"ji |j
rr
¦|pf^TT
f|r|T|
MIT" ff|jj|1 '[ri •"
-
r
•
i
1
i
0
2
4 £
6 I
c
n 03
8 q;
10 "to
12
14
Oct-97
Hgure 47. Mean daily flow at USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon, PA
§
400
300 -
200
100
0-92
Avg Monthly Rainfall (in)
-Avg Observed Flow (10/1/1992 to 9/30/1998 )
-Avg Modeled Flow (Same Period)
Month
.2
c
'ro
CC
>
jc.
•*—>
C
o
Figure 48. Mean monthly flow at USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon, PA
51
-------
o
o
• Avg Flow (10/1/1992 to 9/30/1998)
Line of Equal Value
Best-Fit Line
400
300 -
-o
CD
a) 200
"O
o
CD
g> 100
I
y = 0.8916x + 12.
R2 = 0.8569
178
*
S
*
* M
•
•
—
s
/'>
* /
*
* ^
•
/,
* s*"
I • /
•
jtiSi •
100 200 300
Average Observed Flow (cfs)
100%
TD
90%
O
80%
+
CO
70%
JD
o
60%
0
(J
50%
c
CD
40%
CO
CD
30%
L_
0>
-*—*
20%
§
10%
0%
400
Avg Observed Flow (10/1/1992 to 9/30/1998 )
Avg Modeled Flow (10/1/1992 to 9/30/1998 )
- Line of Equal Value
f ' p IT
0-92 0-93
0-94
0-95
Month
0-96 0-97
Figure 49. Monthly flow regression and temporal variation at USGS 01576540 Mill Creek at
Eshelman Mill Road near Lyndon, PA
• Avg Flow (10/1 /1992to 9/30/1998)
Line of Equal Value
Best-Fit Line
42
I
Avg Monthly Rainfall (in)
-Avg Observed Flow (10/1/1992 to 9/30/1998)
Avg Modeled Flow (Same Period)
200
150
100
50
0
0.5
1 ^
c
l-5ir
r\ CD
g
2.5«
3 >>
sz
3.5c
o
4 2
4.5
5
Oct Nov Dec Jan Feb Mar Apr May Jun
10 11 12 1 2 3 4 5 6 7 8 9
Month
200
_o.
|150-|
0)
® 100
y = 1.1753x-10.642
R2 = 0.9281
Average Observed Flow (cfs)
Figure 50. Seasonal regression and temporal aggregate at USGS 01576540 Mill Creek at Eshelman
Mill Road near Lyndon, PA
52
-------
i Observed (25th, 75th)
¦ Median Observed Flow (10/1/1992 to 9/30/1998)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
200 -r
Nov Dec Jan Feb
Apr _ j May Jun Jul Aug
80 -
Month
CD
to
cc
Figure 51. Seasonal medians and ranges at USGS 01576540 Mill Creek at Esheiman Miil Road
near Lyndon, PA
I able 24. Seasonal summary at USGS 01576540 Mill Creek at Esheiman Mill Road near Lyndon, PA
MONTH
OBSERVED FLOW fCFS)
MODELED FLOW (CFS)
MEAN
MEDIAN
257H
75TH
MEAN
MEDIAN
25TH
75TH
Oct
59.60
35.00
22.00
53.00
56.22
32.521
14.78
56.75
Nov
73.37
40.00
32.00
75.25
76.15
40.59
28.48
77.13
Dec
98.69
52.00
38.00
100.25
100.33
61.02
43.68
! 118.18
Jan
108.22
55.00
41.00
89.00
92.64
62.30
46.47
102.02
Feb
83.84
65.00
45.00
96.00
100.81
69.99
46.97
111.35
Mar
147.26
93.50
77.25
146.75
163.94
114.10;
75.05
188.54
Apr
107.92
93.00
63.75
125.00
132.24
108.24 J
65.44
158.63
May
79.38
75.00
52.25
101.25
88.89
82.65
55.32
105.34
Jun
62.01
49.00
36.00
64.00
62.25
58.501
39.24
73.17
Jul
58.36
42.50
31.00
62.75
52.68
44.73
31.83
62.19
Aug
43.97
35.00
21.25
49.75
43.33
30.69:
20.83
47.10
Sep
43,10
30,50
20.00
52.00
37.82
24.881
16,69
42.58
53
-------
^""Observed Flow Duration (10/1/1992 to 9/30/1998 )
—Modeled Flow Duration (10/1/1992 to 9/30/1998 )
6000
600
Li-
ra)
60
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Percent of Time that Flow is Equaled or Exceeded
Figure 52. Flow exceedance at USGS 01576540 Mill Creek at Eshelmari Mill Road near Lyndon, PA
"^—Observed Row Volume (10/1/1992 to 9/30/1998 )
— Modeled Flow Volume (10/1/1992 to 9/30/1998 )
120%
100%
80%
60%
40%
20%
0%
Oct-92 Oct-93 Oct-94 Oct-95 Oct-96 Oct-97
Figure 53. Flow accumulation at USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon, PA
54
-------
I able 25. Summary statistics at USGS 01576540
/lill Creek at Eshelman l\
/lill Road near Lyndon, PA
SWAT Simulated Flow
Observed Flow Gage
REACH OUTFLOW FROM O ITT LET 255
6-Year Analysis Period. 10/1/1992 - 9/30/1998
Flow volumes are (inchas/year) for upstream drainage area
USGS 01576540 Mill Creek at Eshelman Mill Road near Lyndon. PA
Hydrologic Unit Code. 2050306
Latitude 40 0100985
Longitude, -76.2771806
Drainage Area (sq-mi): 54 2
Total Simulated In-stream Flow:
21.02
Total Observed In-stream Flow
20.19
Total of simulated highest 10% flows:
7.57
Total of Observed highest 10% flows:
7.66
Total of Simulated lowest 50% flows:
4.24
Total of Observed Lowest 50% flows:
4.27
Simulated Summer Flow Volume (months 7-9):
2.82
Observed Summer Flow Volume (7-9):
3.06
Simulated Fall Flow Volume (months 10-12):
4.90
Observed Fall Flow Volume (10-12):
4.88
Simulated Winter Flow Volume (months 1-3):
7.41
Observed Winter Flow Volume (1-3):
7.06
Simulated Spring Flow Volume (months 4-6):
5.90
Observed Spring Flow Volume (4-6):
5.19
Total Simulated Storm Volume:
6.47
Total Observed Storm Volume:
6.19
Simulated Summer Storm Volume (7-9):
0.71
Observed Summer Storm Volume (7-9):
0.93
Errors (Simulated-Observed)
Error Statistics
Recommended Criteria
Error in total volume:
4.14
10
Error in 50% lowest flows:
-0.82
10
Error in 10% highest flows:
-1.17
15
Seasonal volume error - Summer:
-7.94
" 30
Seasonal volume error - Fall:
0.42 » 30
Clear
Seasonal volume error - Winter:
4.98
30
Seasonal volume error - Spring:
13.65
30
Error in storm volumes.
4.49
20
Error in summer storm volumes:
-24.43
50
Nash-Sutcliffe Coefficient of Efficiency, E:
0.518
Model accuracy increases as
E or E' approaches 1.0
Baseline adjusted coefficient (Garrick), E':
0.440
Monthly NSE
0.852
55
-------
3.1.2.2.7 USGS 01576085 Little Conestoga Creek Churchtown, PA (validation period)
£
I
Avg Monthly Rainfall (in)
- Avg Observed Flow {10/1/1990 to 9/30/1995 )
Avg Modeled Row (Same Period)
450
400
350
300
150
100
0
Oct-90
'] 'jjf 1 r"IJ
IjpnrjTf" |
Hip
l|S|i"riTf'm^ir
Ifl
iifprnn
1 i" i*|i" '
--—
i
i
-
i.
L\
,
i. i
iAi
Al
A-JLvi. i ^
jLikLil
3
c
'to
a:
10 "t
12
14
Oct-91
Oct-92 Oct-93
Date
Oct-94
Figure 54. Mean daily flow at USGS 01576085 Little Conestoga Creek Churchtown, PA
I
§
Avg Monthly Rainfall (in)
-Avg Observed Flow (10/1/1990 to 9/30/1995 )
Avg Modeled Flow (Same Period)
£
c
nj
a:
>>
c
o
O-90
0-91
0-92
0-93
0-94
Month
Figure 55. Mean monthly flow at USGS 01576085 Little Conestoga Creek Churchtown, PA
56
-------
100
Avg Flow (10/1/1990 to 9/30/1995 }
• Line of Equal Value
Best-Fit Line
y = 0.5428x +4.2194
R2 = 0.6433
¦o 60 •
20 40 60 80
Average Observed Flow (cfs)
Avg Observed Flow (10/1/1990 to 9/30/1995 )
Avg Modeled Flow (10/1/1990 to 9/30/1995 )
Line of Equal Value
100%
g 50%
Jj 40% H
CD
CO 30% H
0)
-t—•
03
100
20% -\\
10%
0% -V
O-90
0-91
0-92 0-93
Month
0-94
Figure 56. Monthly flow regression and temporal variation at USGS 01576085 Little Conestoga
Creek Churchtown, PA
40
5^ 30
o
T3
0>
T3
O
0)
O)
5 10
a>
• Avg Flow (10/1 /1990 to 9/30/1995)
Line of Equal Value
Best-Fit Line
s-
II
>*
7261 x + 2.8753
!2 = 0.7865
*
*
' t
*
/
*
/
*
1
•V
/
10 20 30
Average Observed Flow (cfs)
40
in
t5
I
Avg Monthly Rainfall (in)
-Avg Observed Flow (10/1/1990 to 9/30/1995)
Avg Modeled Flow (Same Period)
Oct Nov Dec Jan Feb Mar Apr MayJun Jul
-i r
10 11 12 1
Month
Figure 57. Seasonal regression and temporal aggregate at USGS 01576085 Little Conestoga Creek
Churchtown, PA
-------
¦ Observed (25th, 75th)
- Median Observed Flow (10/1/1990 to 9/30/1995)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
£
i
40
30
20
10
Oct
Nov
Dec
Jan
Feb
_ f
j
i
i
i
4
t
Apr May Jun Jul Aug Sep
10 11
12
3 4
Month
- 0
- 1
- 2
r J cc
_>*
b4 f
o
5
6
Hgure 58. Seasonal medians and ranges at USGS 01576085 Little Conestoga Creek Churchtown, PA
Table 26. Seasonal summary at USGS 01576085 Little Conestoga Creek Churchtown, PA
MONTH
OBSERVED FLOW (CFS)
MODELED FLOW (CFS)
MEAN
MEDIAN
25TH
75TH
MEAN
MEDIAN
25TH
75TH
Oct
2.21
1.30
0.991
1.90
6.36
4.22
1.95
8.50
Nov
4.57
2.10
1.30
3.58
8.98
6.72
3.01
11.56
Dec
9.97
5.90
3,70]
8.80
15.70
13.59
8.60
18.24
Jan
8.56
5.30
3.051
7.75
9.27
6.51
3.84
11.05
Feb
6.67
4.80
3.501
5.80
5.96
3.44
1.11
7.01
Mar
29.70
9.00
5.901
28.50
21.69
11.99
7.68
29.52
Apr
11.12
6.20
4.501
14.00
15.06
9.48
4.40
25.02
May
4.87
4.10
3.10]
5.60
5.95
4.67
3.41
7.09
Jun
3.15
2.40
1.90!
3.18
3.23
1.74
0.81
3.77
Jul
2.17
1.40
1.10]
1.80
1.81
1.14
0.39
2.49
Aug
2.43
1.20
0.611
2.05
1.89
0.87
0.44
1.70
Sep
2.55
0.89
0.561
1.80
2.47
0.69
0.26
3.37
58
-------
(/)
I
100
CD
CD
cc
I
10
CD
o
0.1
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Observed Flow Duration (10/1/1990 to 9/30/1995 )
—Modeled Flow Duration (10/1/1990 to 9/30/1995 )
Percent of Time that Flow is Equaled or Exceeded
Figure 59. Flow exceedance at USGS 01576085 Little Conestoga Creek Churchtown, PA
Observed Flow Volume (10/1/1990 to 9/30/1995 )
— Modeled Flow Volume (10/1/1990 to 9/30/1995 )
120%
100%
80%
60%
40%
20%
0%
Oct-90 Oct-91
Oct-92 Oct-93 Oct-94
Figure 60. Flow accumulation at USGS 01576085 Little Conestoga Creek Churchtown, PA
59
-------
Table 27. Summary statistics at USGS 01576085 Little Conestoga Creek Churchtowri, PA
SWAT Simulated Flow
Observed Flow Gage
REACH OUTFLOW FROM OUTLET 208
5-Year Analysis Period; 10/1/1990 - 9/30/1995
Flow volumes are (inches/year) for upstream drainage area
USGS 01576085 Little Conestoga Creek near Churchtown, PA
Hydralogic Unit Code: 2050306
Latitude: 40 14481885
Longitude: -75.9885539
Drainage Area (sq-mi): 5.82
Total Simulated In-stream Flow:
19.19
Total Observed In-stream Flow
17.17
Total of simulated highest 10% flows:
8.39
Total of Observed highest 10% flows.
9.82
Total of Simulated lowest 50% flows:
2.05
Total of Observed Lowest 50% flows:
1.90
Simulated Summer Flow Volume (months 7-9):
1,21
Observed Summer Flow Volume (7-9):
1.40
Simulated Fall Flow Volume (months 10-12):
6.09
Observed Fall FlowVolume (10-12):
3.29
Simulated Winter FlowVolume (months 1-3):
7.21
Observed Winter FlowVolume (1-3):
8.78
Simulated Spring Flow Volume (months 4-6):
4.69
Observed Spring Flow Volume (4-6):
3.70
Total Simulated Storm Volume:
2.71
Total Observed Storm Volume:
4.63
Simulated Summer Storm Volume (7-9):
0.21
Observed Summer Storm Volume (7-9):
0.55
Errors (Simulated-Observed)
Error Statistics
Recommended Criteria
Error in total volume:
11.78
10
Error in 50% lowest flows:
7.86
10
Error in 10% highest flows:
-14 56
15
Seasonal volume error - Summer:
-13.80
30
Seasonal volume error - Fall:
85.24 » , 30
Clear
Seasonal volume error - Winter:
-17.91
30
Seasonal volume error - Spring:
26.59
30
Error in storm volumes:
-41.40
20
Error in summer storm volumes:
-61.08
50
Nash-Sutcliffe Coefficient of Efficiency, E:
0.271
Model accuracy increases as
E or E' approaches 1.0
Baseline adjusted coefficient (Garrick), E':
0.222
Monthly NSE
0.623
60
-------
3.1.2.2.8 USGS 01576521 Big Spring Run near Willow Street. PA (validation period')
J2
5
o
Avg Monthly Rainfall (in)
- Avg Observed Flow (10/1/1993 to 7/31/2001 )
Avg Modeled Flow (Same Period)
100 -
>.
10 ro
Q
Oct-93
Oct-94
Oct-95
Oct-96
Oct-97
Oct-98
Oct-99
Oct-OO
Hgure 61. Mean daily flow at USGS 01576521 Big Spring Run near Willow Street, PA
15 n
.£
o
§
Avg Monthly Rainfall (in)
Avg Observed Flow (10/1/1993 to 7/31/2001 )
Avg Modeled Flow (Same Period)
r 12
0-96
Month
0-93 0-94
Figure 62. Mean monthly flow at USGS 01576521 Big Spring Run near Willow Street, PA
61
-------
Avg Flow (10/1/1993 to 7/31/2001 )
• Line of Equal Value
Best-Fit Line
15
-f2
<:
o
[E 10
T3
0)
0)
T3
O
0
O)
CD
1
y = 0.8717X + 0.0203
/
0.
*
R2 =
18448
S
/
.A*
•
•
ri
5 10
Average Observed Flow (cfs)
15
Avg Observed Flow (10/1/1993 to 7/31/2001 )
Avg Modeled Flow (10/1/1993 to 7/31/2001 )
- Line of Equal Value
-o
O
+
i/J
.a
O
CJ
c
_CD
CD
CD
&
J9
100% -
80% -
70% -
60% -
30% -
20% -
10% -
0-93 0-94 0-95 0-96 0-97 0-98 0-99 O-OO
Month
Figure 63. Monthly flow regression and temporal variation at USGS 01576521 Big Spring Run near
Willow Street, PA
CO
t5
-2 4
LL- H
"O
0)
0)
~o
O
Q) 2
O)
CD
5
I
• Avg Flow (10/1 /1993 to 7/31 /2001)
Line of Equal Value
Best-Fit Line
y = 1.1336x- 0.7304
R2 = 0.925
*
SS
/
*
#
*
' •
* z
,-y
SjT
/
* *
¦
*
s
s
*
*
0
I 4
6
Average Observed Flow (cfs)
I
I
Avg Monthly Rainfall (in)
-Avg Observed Flow (10/1/1993 to 7/31/2001)
Avg Modeled Flow (Same Period)
Jan Feb Mar Apr May Jun Jul Aug Sep
3.5 c
10 11 12 1 2 3 4 5 6 7 8 9
Month
Figure 64. Seasonal regression and temporal aggregate at USGS 01576521 Big Spring Run near
Willow Street, PA
62
-------
i Observed (25th, 75th)
¦Median Observed Flow (10/1/1993 to 7/31/2001)
Average Monthly Rainfall (in)
Modeled (Median, 25th, 75th)
2
Oct Nov Dec
t
Jan Feb Mar Apr May Jun \ Jul Aug Sep
M"
•IT
*tl
¦
* ill
t cb
3 £
I-3 £
¦4—»
4 c
4 O
10
11
12
3 4
Month
Figure 65. Seasonal medians and ranges at USGS 01576521 Big Spring Run near Wiiiow Street, PA
Table 28. Seasonal summary at USGS 01576521 Big Spring Run near Willow Street, PA
MONTH
OBSERVED FLOW (CFS)
MODELED FLOW (CFS)
MEAN
MEDIAN
25TH
75TH
MEAN
MEDIAN
25TH 75TH
Oct
2.24
1.50
0.97
2.00
1.77
1.02
0.691
1.49
Nov
2.47
1.40
1.10
2.50
2.12
1.09
0.60
2.00
Dec
2.94
2.00
1.08
2.90
2.76
1.52
0.52:
2,40
Jan
3.72}
2.10
1.70
3.20
3.11
1.66
1.08
3.02
Feb
3.15;
2.70
2.10
3.38
3.51
1.91
1.39
2.93
Mar
4.83:
3.00
2.50
4.20
4.70
2.46
1.95 j
3.84
Apr
3.231
2.90
2.38
3.60
3.04
2.24
1.98 j
3.00
May
2.81
2.30
1.70
3.13
2.37
1.89
1.571
2.64
Jun
2.511
1.90
1.40
2.50
1.81
1.57
1.241
2.06
Jul
2.15!
1.60
1.00
2.20
1.57
1.21
0.95:
1.65
Aug
1.80
1.10
0.69
1.70
1.40
0.93
0.78
1.38
Sep
2.38i
1.10
0.69
1.90
1.88
0.91
0.68
1.42
63
-------
Figure 66. Flow exceedance at USGS 01576521 Big Spring Run near Willow Street, PA
Observed Flow Duration (10/1/1993 to 7/31/2001 )
—Modeled Flow Duration (10/1/1993 to 7/31/2001 )
20% 30% 40% 50% 60% 70% 80%
Percent of Time that Flow is Equaled or Exceeded
100%
^—Observed Flow Volume (10/1/1993 to 7/31/2001 )
— Modeled Flow Volume (10/1/1993 to 7/31/2001 )
120%
100%
80%
60%
40%
20%
0%
Oct-93
>94 Oct-95 Oct-96 Oct-97 Oct-98 Oct-99 Oct-OO
Figure 67. Flow accumulation at USGS 01576521 Big Spring Run near Willow Street, PA
64
-------
Table 29. Summary statistics at USGS 01576521 Big Spring Run near Willow Street, PA
SWAT Simulated Flow
Observed Flow Gage
REACH OUTFLOW PROM OUTLET 289
7-83-Year Analysis Period: 10/1/1993 - 7/31/2001
Flow volumes are (Iriches/year) for upstream drainage area
USGS 01576521 Big Spring Run near Willow Street, PA
Hydrologic Unit Code 2050306
Latitude: 39 9961544
Longitude- -76.2649563
Drainage Area (sq-mi): 1.77
Total Simulated In-stream Flow:
19.32
Total Observed In-stream Flow:
22.02
Total of simulated highest 10% flows:
8.29
Total of Observed highest 10% flows:
8.45
Total of Simulated lowest 50% flows:
3.88
Total of Observed Lowest 50% flows:
4.81
Simulated Summer Flow Volume (months 7-9):
2.92
Observed Summer Flow Volume (7-9):
3.82
Simulated Fall Flow Volume (months 10-12):
4.38
Observed Fall Flow Volume (10-12):
5.04
Simulated Winter Flow Volume (months 1-3):
7.32
Observed Winter Flow Volume (1-3):
7.59
Simulated Spring Flow Volume (months 4-6):
4.70
Observed Spring Flow Volume (4-6):
5.57
Total Simulated Storm Volume:
5.27
Total Observed Storm Volume:
5.57
Simulated Summer Storm Volume (7-9):
0.79
Observed Summer Storm Volume (7-9):
1.27
Errors (Simulated-Observed)
Error Statistics
Recommended Criteria
Error in total volume:
-12.25
10
Error in 50% lowest flows:
-19.46
10
Error in 10% highest flows:
-1.87
15
Seasonal volume error - Summer:
-23.60
30
Seasonal volume error - Fall:
-13.00 » 30
Clear
Seasonal volume error - Winter:
-3.63
30
Seasonal volume error - Spring:
-15.53
30
Error in storm volumes
-5.51
20
Error in summer storm volumes:
-37.68
50
Nash-Sutcliffe Coefficient of Efficiency, E:
0.137
Model accuracy increases as
E or E1 approaches 1.0
Baseline adjusted coefficient (Garrick), E':
0.283
Monthly NSE
0.804
65
-------
3.1.2.2.9 USGS 01576529 Un-named Tributary to Big Spring Run near Lampeter. PA (validation period')
Avg Monthly Rainfall (in)
- Avg Observed Flow (10/1/1993 to 7/31/2001 )
Avg Modeled Flow (Same Period)
60
50
40
•B
5
30
o
LL.
20
10
0
™lmHn«T P
10 to
Oct-95
Oct-96
Oct-97
Oct-98
Oct-99
Oct-OO
Oct-93
Oct-94
Figure 68. Mean daily flow at USGS 01576529 Un-named Tributary to Big Spring Run near
Lampeter, PA
¦2
o,
5
o
Avg Monthly Rainfall (in)
#- Avg Observed Flow (10/1/1993 to 7/31/2001 )
— Avg Modeled Flow (Same Period)
Month
*2
c
RJ
on
2?
Jz
Figure 69. Mean monthly flow at USGS 01576529 Un-named Tributary to Big Spring Run near
Lampeter, PA
66
-------
CO
*~—
5
o
"O
J)
0)
~0
O
a)
CD
ro
I
• Avg Flow {10/1/1993 to 7/31/2001 )
Line of Equal Value
Best-Fit Line
y = 0.852x +0.1552
R2 = 0.8209
100% -I
-O
90% -
o
80% ¦
+
•
sz
3.5c
o
4 2
Average Observed Flow (cfs) Month
Figure 71. Seasonal regression and temporal aggregate at USGS 01576529 Un-named Tributary to
Big Spring Run near Lampeter, PA
67
-------
¦ Observed (25th, 75th) Average Monthly Rainfal) {in)
-Median Observed Flow (10/1/1993to 7/31/2001) ¦ Modeled (Median, 25th, 75th)
10 11 12 1 2 3 4 5 6 7 8 9
Month
Hgure 72. Seasonal medians and ranges at USGS 01576529 Un-named Tributary to Big Spring Run
near Lampeter, PA
Table 30. Seasonal summary at USGS 01576529 Un-named Tributary to Big Spring Run
near Lampeter, PA
MONTH
OBSERVED FLOW (CFS)
MODELED FLOW (CFS)
MEAN
MEDIAN
25TH
75TH
MEAN
MEDIAN
25TH
75TH
Oct
1.17
0.61
0.35
0.99
0.971
0.46
0.25
0.77
Nov
1.42
0.65
0.39
1.50
1.25]
0.55
0.18
1.15
Dec
1.78
1.00
0.37
1.80
1.821
1.03
0.13
1.72
Jan
2.02
1.10
0.73
1.80
2.111
1.15
0.64
2.33
Feb
1,93
1.70
1.30
2.20
2.451
1.28
0.89
2.12
Mar
3.26
1.90
1.40
2.70
3.13]
1.76
1.26
2.81
Apr
1.93
1.70
1.28
2.20
2.001
1.45
1.30
2.10
May
1.46
1.10
0.83
1.80
1.441
1.11
0.95
1.72
Jun
1.54
0.92
0.60
1.50
1.00:
0.86
0.67
1.24
Jul
1.13
0.76
0.52
1.20
0.791
0.57
0.46
0.88
Aug
0.86
0.41
0.22
0.70
0.701
0.40
0.31
0.68
Sep
1.19
0.39
0.18
1.20
0.92!
0.36
0.25
0.70
68
-------
Figure 73. Flow exceedance at USGS 01576529 Un-named Tributary to Big Spring Run near
Lampeter, PA
Observed Flow Duration (10/1/1993 to 7/31/2001 )
^—Modeled Flow Duration (10/1/1993 to 7/31/2001 )
^—Observed Flow Volume (10/1/1993 to 7/31/2001 )
^—Modeled Flow Volume (10/1/1993 to 7/31/2001 )
20% 30% 40% 50% 60% 70% 80%
Percent of Time that Flow is Equaled or Exceeded
100%
120%
100%
80%
60%
40%
20%
0%
Oct-93 Oct-94 Oct-95 Oct-96
Oct-97 Oct-98 Oct-99 Oct-OG
Figure 74. Flow accumulation at USGS 01576529 Un-named Tributary to Big Spring Run near
Lampeter, PA
69
-------
Table 31. Summary statistics at USGS 01576529 Un-named Tributary to Big Spring Run
near Lampeter, PA
SWAT Simulated Flow
Observed Flow Gage
REACH OUTFLOW FROM OUTLET 290
7.83-Year Analysis Period: 10/1/1993 - 7/31/2001
Flow volumes are (inches/year) for upstream drainage area
USGS 0157S529 Unnamed Trib to Big Spring Run near Lampeter, PA
Hydrologic Unit Code 2050306
Latitude: 39.9992655
Longitude: -76 2641249
Drainage Area (sq-ml): 1.42
Total Simulated In-stream Flow:
14.92
Total Observed In-stream Flow
15.80
Total of simulated highest 10% flows:
6.81
Total of Observed highest 10% flows:
6.73
Total of Simulated lowest 50% flows:
2.35
Total of Observed Lowest 50% flows:
2.62
Simulated Summer Flow Volume {months 7-9):
1.81
Observed Summer Flow Volume (7-9):
2.40
Simulated Fall Flow Volume (months 10-12):
3.31
Observed Fall Flow Volume (10-12):
3.58
Simulated Winter Flow Volume (months 1-3):
6.20
Observed Winter Flow Volume (1-3):
5.84
Simulated Spring Flow Volume (months 4-6):
3.60
Observed Spring Flow Volume (4-6):
3.99
Total Simulated Storm Volume:
3.76
Total Observed Storm Volume:
4.16
Simulated Summer Storm Volume (7-9):
0.46
Observed Summer Storm Volume (7-9):
0.87
Errors (Simulated-Observed)
Error Statistics
Recommended Criteria
Error in total volume:
-5.58
10
Error in 50% lowest flows:
-10.58
10
Error in 10% highest flows:
1.22
15
Seasonal volume error - Summer:
-24.63
. 30
Seasonal volume error - Fall:
-7.33 » 30
Clear
Seasonal volume error - Winter:
6.19
30
j
Seasonal volume error - Spring:
-9.79
30
Error in storm volumes:
-9.51
20
Error in summer storm volumes:
-46.78
50
Nash-Sutcliffe Coefficient of Efficiency, E:
0.164
Model accuracy increases as
E or E' approaches 1.0
Baseline adjusted coefficient (Garrick), E:
0.320
Monthly NSE
0.815
70
-------
3.2 Water Quality
3.2.1 Methods
3.2.1.1 SWAT
Water quality calibration and validation were
conducted for total suspended solids (TSS), total
phosphorus (TP) and soluble reactive phosphorus
(SRP), total nitrogen (TN), and species of nitrogen,
namely, total Kjeldahl nitrogen (TKN) and
nitrate+nitrite nitrogen (NOx). Water quality calibration
and validation focused on water years 2005 to 2014
and 1995 to 2004, respectively.
The focus of the calibration and validation was the
monitoring location on the Conestoga River near
Conestoga. This is the only location in the watershed
that has long term water quality observations. Table 32
lists the water quality station used for calibration and
validation.
Table 32. Water quality calibration
and validation location
USGS Id
Name
Calibration
Period
Validation
Period
Subbasin
Id
01576754
Conestoga River
at Conestoga, PA
2005-2012
1995-2004
9
The USGS water quality monitoring station on the
Conestoga River at Conestoga has a long period of
consistent data for sediment and nutrients, and was
used for the evaluation of the SWAT model. This station
is one of the non-tidal monitoring stations in the
Chesapeake Bay Program (http://cbrim.er.usgs.gov/
index.html).
Comparison of model results to monthly loads presents
challenges because 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 comparing two uncertain numbers.
Flow stratified log-log regression and averaging
approaches were used to estimate constituent loads.
A flow stratified approach uses different functions to
fit constituent load over varying ranges of flow. The
change from low to high rates of constituent transport
occurs at a breakpoint, which is defined as the flow
where the fitted functions intersect. The model used
here fits linear segments on a log-log scale. Identifying
the transition phase is done by visually inspecting
the plot of constituent load against flow on a log-log
scale. The regression approach is adopted when the
constituent has a strong correlation with flow. An
averaging approach is adopted when there is little
or no correlation between the constituent load and
flow. The averaging approach consists of estimating
loads using observed flow and average observed
concentration in a stratum. The regression models used
for load estimation are provided in Appendix A.
Consistent with recommendations of Moriasi et al.
(2007), water quality calibration focused on replicating
monthly loads (Table 9). Model performance was
deemed acceptable where a performance evaluation
of "good" or "very good" was attained. Moriasi et al.
note that these comparisons are most appropriate for
evaluation of the quality of water quality simulations
when a nearly complete measured time series
exists, and, when only scattered grab samples are
available, "comparison of frequency distributions and/
or percentiles...may be more appropriate than the
quantitative statistics guidelines."
The load comparisons were also supported by detailed
examinations of the relationships of flows to loads
and concentrations, the distribution of concentration
prediction errors versus flow, time, and season, and
standard time-series plots. The key statistic is the
relative percent error, which shows the total error
in the prediction of monthly load normalized to the
estimated load. Relative average absolute error was
also calculated, which is the average of the relative
magnitude of errors in individual monthly load
predictions. That number is inflated by outlier months
in which the simulated and estimated loads differ by
large amounts. Outliers may be a result of uncertainty
in the estimated load because of limited data or on
account of problems with the model. The third statistic,
the relative median absolute error, is likely more
relevant and shows better agreement because it is not
influenced by outlier months.
Non-point sediment and nutrient loads are rarely
available due to a lack of observed data at the field
scale. The Chesapeake Bay Phase 5.3 model provides
estimates of edge-of-stream (EoS) sediment and
nutrient loads (http://www.chesapeakebay.net/about/
programs/modeling/53/). The non-point sediment
and nutrient loading rates by landuse in the Conestoga
71
-------
SWAT model were generally based on those reported
by the Chesapeake Bay model. The Phase 5.3 model
however, does not provide direct estimates of EoS
sediment loads but that of edge-of-field (EoF) loads
and an equation to estimate sediment delivery fraction
(SDF). The EoS sediment load is calculated as the
product of EoF loads and SDF.
The SDF accounts for the depositional losses between a
field and the stream. The Phase 5.3 Bay model uses the
following equation to estimate SDF,
SDF = 0.4177620T0134958) - 0.12 7 0 9 7
where,
A = drainage area in square miles
The average value of SDF determined for the
Conestoga River watershed using the above equation
is approximately 0.25. The area-based method is
however subject to large errors as it does not take into
account either the topography of the watershed or the
connectivity between source areas and ultimate sinks.
Further, the empirical data comparing basin outlet
data to field-scale soil loss estimates on which the
relationship is based does not account for additional
sediment sources such as channel degradation, gully
formation, or soil creep. This results in a potential
high bias in which the area-based SDR over-estimates
the fraction of upland sheet and rill erosion that
is delivered to the basin mouth. The initial target
EoS sediment loads for the watershed model were
generally adjusted downward during the calibration
and validation process to account for the uncertainties
in SDF and potential for over-prediction.
The parameters adjusted during the water quality
calibration and validation process of the SWAT model
are listed below.
Table 33 shows the values of these parameters in the
calibrated and validated model. The values for some of
these parameters were determined from experimental
work conducted in the basin or the general
physiographic region. The sources of the parameter
values (wherever applicable) are also listed in the table.
a) Basin level parameters
• SPCON -
Linear parameter for calculating the
maximum amount of sediment that can be
re-entrained during channel sediment routing
• NPERCO -
Nitrate percolation coefficient
• CDN -
Denitrification exponential rate coefficient
• SDNCO -
Denitrification threshold water content
• PPERCO -
Phosphorus percolation coefficient
• PHOSKD-
Phosphorus soil partitioning coefficient
b) Subbasin and HRU level parameters
• PRF -
Peak rate adjustment factor for sediment
routing in the main channel
• CH_BNK_KD-
Erodibility of channel bank sediment (cm3/N-s)
• CH_BNK_BD-
Bulk density of channel bank sediment (g/cm3)
• ERORGN-
Organic nitrogen enrichment ratio
• SHALLST_N -
Nitrate concentration in shallow aquifer (mg-N/L)
• HLIFE_NGW -
Half-life of nitrate in the shallow aquifer (days)
• CH_ONCO-
Organic nitrogen concentration in the channel
(ppm)
• ERORGP-
Organic phosphorus enrichment ratio
• CH_OPCO - Organic phosphorus concentration in
the channel (ppm)
72
-------
Table 33. Values of parameters in the calibrated and validated model
Parameter
Historic Dam Site
No Dams
Source
SPCON
0.001
0.001
-
PRF
2
0.3
-
CH_BNK_KD
0-3.5
0
-
CH_BNK_BD
1.04
1.04
-
NPERCO
0.5
0.5
-
CDN
1
1
-
SDNCO
1
1
-
PPERCO
10
10
-
PHOSKD
200
200
-
ERORGN
3
3
-
SHALLST_N
15
15
-
HLIFE_NGW
36,500
36,500
-
CH_0NC0
1160
1160
Walter and Merritts 2007, Merritts et al. 2010
ERORGP
1.5
1.5
-
CH_0PC0
556
556
Walter and Merritts 2007, Merritts et al. 2010
3.2.1.2 In-stream Sediment Simulation in HSPF and SWAT
Bank erosion has been identified as a major source
of sediment load in the Conestoga River watershed
(Merritts and Walter 2003, Walter and Merritts 2007,
Gellis et al. 2009, Merritts et al. 2010). Walter and
Merritts (2007) reported erosion rates of 0.2 to 0.9
tons/ft/yr with an average of 0.3 tons/ft/yr at six sites
in Lancaster and York counties in Pennsylvania. Walter
and Merritts (2007) estimate an average annual stream
sediment load of approximately 102,000 tons/yr for
a period of 100 years since early 20th century from
the Conestoga River. Extensive work carried out by
Merritts et al. (2010) on stream corridor erosion along
breached millponds suggest average bank erosion rates
ranging from 0.2 tons/ft/yr to 0.7 tons/ft/yr in selected
tributaries (Big Spring Run, Hammer Creek and West
Branch Little Conestoga Creek) of the Conestoga
River. Based upon their observations at breached
millpond sites, Merritts etal. (2010) constructed a dam
breach model and ran several scenarios. The range of
estimates of sediment load from channel erosion highly
variable, and were dependent upon the time of dam
breach and the time elapsed since dam breach. The
in-stream sediment component of the SWAT and HSPF
models were configured to be generally consistent with
the estimates reported by these studies.
Merritts et al. (2010) however acknowledge that not
all of the eroded sediment is transported and that
deposition likely occurs between the site of erosion
and the mouth of the river. Sediment fingerprinting
using radio isotopes has recently gained application
in sediment source identification and apportionment.
Gellis et al. (2009) have attributed approximately 23%
of the sediment loads to channel banks in the Little
Conestoga Creek using sediment fingerprinting.
Bank and bed erosion can be simulated in a SWAT
model using either a simplified version of Bagnold
equation or a physics-based approach. In the simplified
Bagnold model erosion rates are specified by the user
and erosion occurs as long as the channel has transport
capacity. In other words, the sediment supply from
channel erosion is unlimited. In the physics-based
approach, bank and bed erosion is simulated only
when the total shear stress exceeds the critical shear
stress. The total shear stresses are calculated based on
channel geometry and wetted perimeter. The critical
shear stress calculations are based on percent silt and
clay content, and channel cover. If the physics-based
approach is selected then transport capacity can be
simulated using one of four stream power models -
73
-------
1) the simplified Bagnold model, 2) Kodatie model,
3) Molinas and Wu model, and 4) Yang sand and gravel
model. The physics-based approach for bank and bed
erosion along with the simplified Bagnold model is
used in the Conestoga River watershed model.
In the physics-based approach implemented in SWAT,
the total and critical shear stresses are applied to the
entire sediment fraction rather than just the cohesive
particles. As a result, there can be large errors in SWAT
simulated bank and bed erosion rates. To overcome
this limitation, an HSPF model was developed for the
stream network in the Conestoga River watershed
which allows separate simulations of sand, silt and
clay fractions. Complex cycles of deposition and scour
occur in stream reaches, determined by the shear
stress exerted on the bed material and the external
sediment supply are simulated by HSPF. The model
simulates deposition and scour (aggradation and
deposition) of silt and clay in stream reaches based on
exerted shear stress relative to critical shear stresses
for deposition and scour (xCD and xcs) for each sediment
size class, particle deposition velocities, and a limiting
maximum potential rate of scour (IV, lb/ft2/d). The
parameters tcd, tcs, and W are site-specific and vary by
reach. (HSPF is a spatially lumped model, with one-
dimensional representation of reaches. Further, the
exerted shear stress, based on simulation of reach-
average conditions, varies continuously based on local
characteristics of the channel. Thus a single set of
parameters will not adequately represent the behavior
of bed sediment in all reaches.)
For the non-cohesive, sand fraction of sediment,
HSPF provides several options, including the
Toffaletti method, the Colby method, and a simplified
exponential relationship to flow. Sufficient information
is not available to implement the first two options,
which additionally can cause stability problems in the
model, so the third approach is used. In this approach
sand transport capacity is a function of
KSAND ¦ AVVELE EXPSND; where AVVELE is the average
velocity and KSAND and EXPSND are user-specified
parameters.
In-stream sediment calibration was conducted
consistent with BASINS Technical Note 8 (USEPA
2008). Critical shear stresses for scour and deposition
of cohesive sediment in channels are set equal to
percentiles of the simulated distribution. Silt was set
to deposit below the 20th percentile and scour above
the 90th percentile daily shear stress, while clay was
set to deposit below the 15th and scour above the 85th
percentile. An example of the shear stress distribution
is shown in Figure 75.
A reach-by-reach mass balance analysis of channel
sediment was conducted to ensure reasonable
sediment accumulation/loss behavior.
The net scour simulated by the HSPF model on a reach
by reach basis was used to parameterize the bank
erosion component of the SWAT model. Bank erosion
in the SWAT model was adjusted to generally match
the long term average net scour simulated by the HSPF
model on a reach by reach basis.
• tau 20th percentile 90th percentile
0.7
0.6
0.5
£ 0.4
ra 0.3
0.2
0.1
1 10 100 1,000 10,000 100,000
REACH 9 Flow (cfs)
Figure 75. Distribution of shear stress (lau) with flow for modeled reach 9
74
-------
3.2.2 Results and Discussion
3.2,2.1 SWAT Results
Tabie 34 shows the performance of the SWAT model at
a monthly time-step. The performance of the model for
sediment and nutrient loads generally varies between
good and very good. Detailed calibration and validation
results by constituent are provided in sections 3.2.2.1.1
to 3.2.2.1.5.
It is important to note that the model predicts a higher
than observed sediment concentration and load for
low to mid-range flows. This is likely on account of the
The channel erosion component of the SWAT model
was configured using the net scour results on a reach
by reach basis from the HSPF model. HSPF simulation
of in-stream sediment is discussed in detail in section
3.2.2.2. The average annual net bank erosion simulated
by the SWAT model was approximately 114,000 tons/yr
and is in close agreement with the net scour simulated
uncertainties associated with the in-stream sediment
of the SWAT model. Uncertainties in point source
loads are also a likely cause for biases in sediment and
nutrient loads during low flow periods.
Figure 76 to Figure 78 show the non-point source
sediment and nutrient load fractions by landuse
simulated by the SWAT model. The results show that
agriculture and developed land are the major upland
sources of sediment and nutrient in the watershed.
by the HSPF model. The average and median erosion
rates simulated by the SWAT model are 0.10 and 0.07
tons/ft/yr, respectively. Figure 79 shows the mean,
median, rnin, max, and the 25th and 75th percentile
bank erosion rates simulated by the SWAT model for
each modeled reach.
Table 34. SWAT model performance at a monthly time-step for Conestoga River at Conestoga
Statistic
Calibration
Validation
TSS
TKN
NOx
TN
TP
TSS
TKN
NOx
TN
TP
Average absolute error
49.7%
20.7%
22,7%
16.9%
33.9%
45.0%
26.4%
29.9%
24.3%
34.8%
Median absolute error
17.4%
10.2%
13.3%
10.8%
19.1%
11,5%
13.0%
21.1%
18,1%
13.6%
RE
16.9%
-3.0%
11.4%
6,6%
21.5%
6,4%
-9.8%
3.7%
-1.1%
1.5%
NSE
0.790
0.734
0.722
0.873
0.897
0.693
0.711
0.726
0.820
0.637
RSR
0.458
0.515
0.527
0.356
0.321
0.554
0,538
0.523
0.424
0.603
Barren 0.5%
Natural Grass Wetland
Agriculture
79.2%
Sediment
Figure 76. SWAT simulated
sediment load proportions by
landuse
Forest
28.3% _
Barren _
0.4%
Developed
15.6%
Agriculture
523%
Total Nitrogen
Figure 77. SWAT simulated
total nitrogen load proportions
by landuse
Forest
Natural Grass
v.'^ti,Ki d
Barren
O.l
Agriculture
72.5%
Total Phosphorus
Figure 78. SWAT simulated
total phosphorus load proportions
by landuse
75
-------
a 25th to 75th Percentile - Min/Median/Max ~ Mean
3
2.5
c ,
O 2
4)
5 1.5
0.5
mJJlLj
'~±W!
Reach
^ 'f ^ ^ ^ 41, ^ ^ «f tf
25th to 75th Percentile - Min/Median/Max ~ Mean
J LiLJ
o
=>% ^ ^ ^ 4 4 ^ *f & 4> ¦$• ¦;? $ ^ <£> -{? ^ ^ ^ ^ T>% ^ ^ ¦$ & -& $ 4" 4 S~ -f ^ a%<0 4
Reach
Figure 79. Simulated bank erosion rate for each modeled reach in the SWAT model
76
-------
3.2.2.1.1 Total Suspended Solids (TSS)
TSS (calibration)
¦ Regression Loads
- Simulated Loads
o
E
v>
c
o
1000000
100000
10000
1000
100
10
1
m
lO
CD
CO
h-
r*-
CO
CO
o>
as
o
O
T—
T-
CM
o
o
O
o
o
o
o
o
o
o
¦*-
t—
¦*-
¦*—
T—
u.
t.
tL
Q.
Q
a
o
Q.
o
Q.
o
a.
o
CL
o
a
o
CL
<
O
<
O
<
O
<
O
<
O
<
o
<
O
<
Figure 80. Monthly simulated and estimated Total Suspended Solids (TSS) load at Conestoga River
near Conestoga, PA (calibration period)
TSS (validation)
1000000
100000
10000
o
E
a
o 1000
100
—•—Regression Loads Simulated Loads
10
1
"
a>
o
o
5—
CN
OJ
CO
CO
0)
O)
o
05
0)
a>
O)
a>
0>
O)
a>
o
o
o
o
o
o
o
o
o
wL
iL
\L
lL
lL
tL
iL
tL
tL
tL
o
a
CL
O
a
o
CL
o
a
o
a
o
a.
CL
a
o
CL
O
<
o
<
O
<
O
<
O
<
O
<
O
<
o
<
o
<
O
<
Figure 81. Monthly simulated and estimated Total Suspended Solids (TSS) load at Conestoga River
near Conestoga, PA (validation period)
-------
Table 35. Paired daily Total Suspended Solids (TSS) load (tons/day)
Period
1994-2004
2004-2012
1994-2012
Statistic
Ave
Median
Ave
Median
Ave
Median
Simulated
1021.813
134.702
1202.427
146.608
1089.506
140.876
Observed
1362.106
42.800
916.251
25.453
1195.003
37.282
Conestoga River near Conestoga, PA 2004-2012
5
To
c
2
"O
03
cn
U)
100000
10000
1000
100
10
1
0.1
0.01
0.001
1
10
100 1000
Flow, cfs
10000
100000
• Simulated a Observed
Power (Simulated) -Power (Observed)
Figure 82. Power plot of simulated and observed Total Suspended Solids (TSS) load vs flow at
Conestoga River near Conestoga, PA (calibration period)
Conestoga River near Conestoga, PA 1994-2004
100000
10000
0.1 .
1 10 100 1000 10000 100000
Flow, cfs
~ Simulated a Observed
Power (Simulated) Power (Observed)
Figure 83. Power plot of simulated and observed Total Suspended Solids (TSS) load vs flow at
Conestoga River near Conestoga, PA (validation period)
78
-------
Conestoga River near Conestoga, PA 2004-2012
100000
10000
re
5
c 1000
100
10
1
0.1
CO
CO
t-
"O
_ro
3
E
in
0.1
~ Paired data Equal fit
~ *1
~
~
~ ~ ~ >4
5^ <*~
MT ~
~
~
10 100 1000
Observed TSS (tons/day)
10000
100000
Figure 84. Paired simulated vs observed Total Suspended Solids (TSS) load at Conestoga River
near Conestoga, PA (calibration period)
Conestoga River near Conestoga, PA 1994-2004
>%
re
5
"en
CO
CO
h-
"O
JS
3
E
CO
Paired data
Equal fit
100000
10000
1000
10 100 1000
Observed TSS (tons/day)
10000 100000
Figure 85. Paired simulated vs observed Total Suspended Solids (TSS) load at Conestoga River
near Conestoga, PA (validation period)
79
-------
3.2.2.1.2 Total Kieldahl Nitrogen (TKN-)
TKN (calibration)
¦ Regression Loads
- Simulated Loads
1000
o
E
w
c
o
100
O
.i,
u
O
in
o
CL
<
to
o
I
u
O
<0
o
w
Q_
<
CD
O
u
o
r*-
o
k_
CL
<
h-
o
\
u
O
00
O
ii.
CL
<
00
o
I
o
O
OJ
o
CL
<
05
o
o
o
Q,
<
u
o
CL
<
o
o
CL
<
Figure 86. Monthly simulated and estimated Total Kjeldahl Nitrogen (TKN) load at Conestoga River
near Conestoga, PA (calibration period)
TKN (validation)
- Regression Loads
- Simulated Loads
1000
o
E
05
c
o
100
Figure 87. Monthly simulated and estimated Total Kjeldahl Nitrogen (TKN) load at Conestoga River
near Conestoga, PA (validation period)
-------
Table 36. Paired daily Total Kjeldahl Nitrogen (TKN) load (tons/day)
Period
1994-2004
2004-2012
1994-2012
Statistic
Ave
Median
Ave
Median
Ave
Median
Simulated
7.989
2.083
7.479
2.013
7.781
2.043
Observed
6.674
2.421
4.970
1.360
5.977
1.847
Conestoga River near Conestoga, PA 2004-2012
>»
CO
5
o>
c
o
-a
TO
o
1000
100
10
0.1
0.01
A/A A
100 1000
Flow, cfs
10000
100000
• Simulated a Observed
Power (Simulated) Power (Observed)
Figure 88. Power plot of simulated and observed Total Kjeldahl Nitrogen (TKN) load vs flow at
Conestoga River near Conestoga, PA (calibration period)
Conestoga River near Conestoga, PA 1994-2004
1000 -T
0.01 . I
1 10 100 1000 10000 100000
Flow, cfs
• Simulated a Observed Power (Simulated) — ¦ Power (Observed)
Figure 89. Power plot of simulated and observed Total Kjeldahl Nitrogen (TKN) load vs flow at
Conestoga River near Conestoga, PA (validation period)
81
-------
Conestoga River near Conestoga, PA 2004-2012
>.
TO
"55
c
o
1000
100
SZ 10
-a
B
n
3
E
CO
0.1
0.01
0,01
~ Paired data Equal fit
1 10
Observed TKN (tons/day)
1000
Figure 90. Paired simulated vs observed lotal Kjeldhal Nitrogen (TKN) load at Conestoga River
near Conestoga, PA (calibration period)
>•
TO
~5>
¦o
3
to
3
E
in
1000
100
10
0.1
0.01
Conestoga River near Conestoga, PA 1994-2004
0.01
Paired data
¦Equal fit
0.1 1 10
Observed TKN (tons/day)
100
1000
Figure 91. Paired simulated vs observed Total Kjeldhal Nitrogen (TKN) load at Conestoga River
near Conestoga, PA (validation period)
82
-------
3.2.2.1.3 Nitrite + Nitrate Nitrogen (NOx)
NOx (calibration)
• Regression Loads
- Simulated Loads
o
E
In
c
o
10000
1000
100
Figure 92. Monthly simulated and estimated Total Nitrite + Nitrate Nitrogen (NOx) load at Conestoga
River near Conestoga, PA (calibration period)
NOx (validation)
¦ Regression Loads
¦ Simulated Loads
10000
1000
o
E
w
c
o
Figure 93. Monthly simulated and estimated Total Nitrite + Nitrate Nitrogen (NOx) load at Conestoga
River near Conestoga, PA (validation period)
83
-------
Table 37. Paired daily Nitrite + Nitrate Nitrogen (NOx) loac
i (tons/day)
Period
1994
-2004
2004
-2012
1994
-2012
Statistic
Ave
Median
Ave
Median
Ave
Median
Simulated
14.624
11.797
12.561
11.089
13.845
11.510
Observed
18.670
12.637
16.815
11.978
17.970
12.356
Conestoga River near Conestoga, PA 2004-2012
1000
1 .
1 10 100 1000 10000 100000
Flow, cfs
~ Simulated a Observed
Power (Simulated) Power (Observed)
Figure 94. Power plot of simulated and observed Nitrite + Nitrate Nitrogen (NOx) load vs flow at
Conestoga River near Conestoga, PA (calibration period)
Conestoga River near Conestoga, PA 1994-2004
>*
CO
5
(A
c
o
"O
ro
O
_J
x
O
1000
100
10
0.1
100 1000
Flow, cfs
10000
100000
~ Simulated a Observed Power (Simulated) ¦ ¦ Power (Observed)
Figure 95. Power plot of simulated and observed Nitrite + Nitrate Nitrogen (NOx) load vs flow at
Conestoga River near Conestoga, PA (validation period)
84
-------
Conestoga River near Conestoga, PA 2004-2012
~ Paired data Equal fit
1000
>.
re
TJ
"S>
100
x
o
T3
2
n
3
E
55
10
~
0.1
0.1
1 10
Observed NOx (tons/day)
100
1000
Figure 96. Paired simulated vs observed Nitrite + Nitrate Nitrogen (NOx) load at Conestoga River
near Conestoga, PA (calibration period)
Conestoga River near Conestoga, PA 1994-2004
~ Paired data Equal fit
1000
33 100
CO
c
2
X
g 10
"O
3
JH
2
E 1
CO
0.1
0.1 1 10 100 1000
Observed NOx (tons/day)
Figure 97. Paired simulated vs observed Nitrite + Nitrate Nitrogen (NOx) load at Conestoga River
near Conestoga, PA (validation period)
-------
3.2.2.1.4 Total Nitroaeri (TN)
TN (calibration)
¦ Regression Loads
• Simulated Loads
10000
1000
2 100
o
t)
O
in
o
I
a
<
in
o
o
O
CD
O
CL
<
CD
O
O
O
r-
o
lL
Ql
<
r>-
o
Q
O
00
o
Q.
<
CO
o
o
O
o
o
I
Q.
<
a>
o
t5
O
CL
<
o
O
a.
<
o
o
CL
<
Figure 98. Monthly simulated and estimated Total Nitrogen (TN) load at Conestoga River near
Conestoga, PA (calibration period)
TN (validation)
10000
1000
¦ Regression Loads
Simulated Loads
o
|
"5
c
o
Figure 99. Monthly simulated and estimated Total Nitrogen (TN) load at Conestoga River near
Conestoga, PA (validation period)
86
-------
Table 38. Paired daily Total Nitrogen (TN) load (tons/day)
Period
1994-2004
2004-2012
1994-2012
Statistic
Ave
Median
Ave
Median
Ave
Median
Simulated
22.990
15.204
19.910
13.656
21.730
14.397
Observed
26.436
17.015
21.665
12.988
24.484
15.322
Conestoga River near Conestoga, PA 2004-2012
>»
re
"5
-o
re
O
1000
100
10
100 1000
Flow, cfs
10000 100000
• Simulated a Observed
Power (Simulated) Power (Observed)
Figure 100, Power plot of simulated and observed Total Nitrogen (TN) load vs flow at Conestoga
River near Conestoga, PA (calibration period)
Conestoga River near Conestoga, PA 1994-2004
w
T3
T3
re
o
1000
100
10
100 1000
Flow, cfs
10000 100000
~ Simulated A Observed
Power (Simulated) Power (Observed)
Figure 101. Power plot of simulated and observed Total Nitrogen (I N) load vs flow at Conestoga
River near Conestoga, PA (validation period)
87
-------
Conestoga River near Conestoga, PA 2004-2012
1000
>.
re
33
~vi
c
o
13
o
JS
3
E
<75
100
10
~ Paired data Equal fit
10 100
Observed TN (tons/day)
Figure 102. Paired simulated vs observed Total Nitrogen (TN) load at Conestoga River
near Conestoga, PA (calibration period)
1000
Conestoga River near Conestoga, PA 1994-2004
>s
re
5
w
c
o
T3
B
re
E
CO
~ Paired data
1000
100
10
¦ Equal fit
10 100
Observed TN (tons/day)
Figure 103. Paired simulated vs observed Total Nitrogen (TN) load at Conestoga River near
Conestoga, PA (validation period)
1000
88
-------
3.2.2.1.5 Total Phosphorus (TP)
TP (calibration)
- Regression Loads
- Simulated Loads
1000
o
|
c
o
100
Figure 104. Monthly simulated and estimated Total Phosphorus (TP) load at Conestoga River near
Conestoga, PA (calibration period)
TP (validation)
1000
- Regression Loads
- Simulated Loads
o
E
~5>
c
o
100
•sr
m
ir>
CO
CO
h-
r-
oo
oo
CD
CD
o
o
x—
V"
CN
CM
CO
CO
o>
o>
CD
CD
CD
CD
CT)
O)
O)
CD
CD
o
o
O
O
O
O
O
o
o
o
i_
Q_
o
li
CL
T5
iL.
Q.
a
%L
CL
i_
Q_
is
1_
CL
o
L_
CL
-6
CL
"6
L_
CL
is
Q.
O
<
o
<
O
<
O
<
O
<
o
<
O
<
o
<
o
<
o
<
Figure 105. Monthly simulated and estimated Total Phosphorus (TP) load at Conestoga River near
Conestoga, PA (validation period)
89
-------
Table 39. Paired daily Total Phosphorus (TP) load (tons/day)
Period
1994-2004
2004-2012
1994-2012
Statistic
Ave
Median
Ave
Median
Ave
Median
Simulated
2.325
0.239
2.459
0.203
2.375
0.228
Observed
2.314
0.350
1.959
0.235
2.181
0.291
Conestoga River near Conestoga, PA 2004-2012
1000
100
10000
100000
Flow, cfs
• Simulated A Observed
Power (Simulated) —— Power (Observed)
Figure 106. Power plot of simulated arid observed Total Phosphorus (TP) load vs flow at Conestoga
River near Conestoga, PA (calibration period)
Conestoga River near Conestoga, PA 1994-2004
100
0.01 T
1 10 100 1000 10000 100000
Flow, cfs
« Simulated a Observed
Power (Simulated) Power (Observed)
Figure 107. Power plot of simulated and observed Total Phosphorus (TP) load vs flow at Conestoga
River near Conestoga, PA (validation period)
90
-------
>s
re
5
"5>
e
o
¦4-"
Q.
I—
T3
£3
n
3
£
CO
100
10
0.1
0.01
Conestoga River near Conestoga, PA 2004-2012
0.01
~ Paired data Equal fit
~ ~
~ ~~
~ /
~
V -
~
~ ~
Kr* *
~
0.1 1
Observed TP (tons/day)
10
100
Hgure 108. Paired simulated vs observed Total Phosphorus (TP) load at Conestoga River near
Conestoga, PA (calibration period)
Conestoga River near Conestoga, PA 1994-2004
>.
re
"35
c
o
T3
V
•+-*
TO
E
CO
100
10
0.1
0.01
~ Paired data
Equal fit
0.01
0.1 1 10
Observed TP (tons/day)
100
Figure 109. Paired simulated vs observed Total Phosphorus (TP) load at Conestoga River near
Conestoga, PA (validation period)
91
-------
3.2.2,2 Channel Erosion in HSPF
As evident from Figure 112, the HSPF model is
able to capture the trends in suspended sediment
concentration well. The average concentration
errors for the calibration and validation period are
-86% and 16%, respectively (negative indicates over-
prediction). The average error is likely biased by
outliers and a median concentration error is a more
appropriate measure of model performance. The
median concentration errors are -0.67% and -5.67%
for the calibration and validation periods, respectively.
A closer examination of simulated versus observed
concentrations (Figure 110, Figure 111, Figure 112,
Figure 113 and Figure 114) show under-prediction and
over-prediction at high and mid-range concentrations,
respectively. The model is especially deficient in
predicting the highest concentrations. These biases are
likely a result of the uncertainties associated with the
transformation of a time-series at a daily step in SWAT
to an hourly time-step in HSPF. A uniform distribution
was assumed when disaggregating daily flow and
sediment load to an hourly time-step. This could likely
be revised based on Soil Conservation Service (SCS)
rainfall distribution and has the potential to improve
the ability of the model to simulate extreme events.
Conestoga River near Conestoga, PA 1994-2004
~ Paired data Equal fit
10000
1000
100
w
OT
I-
TJ
I
0)
II)
-B
o
0.1
0.1
1
10000
10
100
1000
Observed TSS (mg/L)
Conestoga River near Conestoga, PA 2004-2012
~ Paired data — Equal ft
10000
1000
O)
E
100
~~ A
¦a
ai
_ra
3
E
«
~~
100
0.1
1
10
1000
10000
Observed TSS (mg/L)
Figure 110. Paired HSPF simulated and observed suspended sediment concentration
92
-------
Conestoga River near Conestoga, PA 1994-2004
~ Simulated a Observed
10000
1000
d 100
o>
E
CO
<0 10
1
0.1
1 10 100 1000 10000 100000
Flow, cfs
Conestoga River near Conestoga, PA 2004-2012
~ Simulated a Observed
10000
1000
d 100
O)
E
1
0.1
1 10 100 1000 10000 100000
Flow, cfs
Figure 111. HSPF simulated and observed suspended sediment concentration vs flow
93
-------
Conestoga River near Conestoga, PA
10000
- Simulated & Observed
cn
E
to"
1000
100
10
0-1
of" Jp & °P <£> J#5 pip & fb1 ^ J? <#> f? f? J? J?
-------
Conestoga River near Conestoga, PA 1994-2004
~ Paired data Equal fit
>.
CO
5
»
c
o
to
to
I-
¦D
0)
_re
3
E
to
100000
~~
10000
10 100 1000
Observed TSS (tons/day)
10000 100000
Conestoga River near Conestoga, PA 2004-2012
100000
— 10000
«
i 1000
CO
W 100
"D
»
<*9
n
3
E
to
~ Paired data Equal fit
10 100 1000
Observed TSS (tons/day)
10000 100000
Figure 113. Paired HSPF simulated and observed suspended sediment load
95
-------
1000000
100000
10000
1000
100
10
1
0.1
0.01
100 1000
Flow, cfs
10000
100000
~ Simulated a Observed
Power (Simulated)
Power (Observed)
* Simulated a Observed Power (Simulated) —» Power (Observed)
Conestoga River near Conestoga, PA 1994-2004
Conestoga River near Conestoga, PA 2004-2012
10000000
1000000
>, 100000
OJ
TJ
I 10000
o
1000
100
10
1
0.1
—' 1
100 1000
Flow, cfs
1
10000 100000
Figure 114. HSPF simulated and observed suspended sediment load vs flow
96
-------
Figure 115 shows the net scour and deposition
simulated by the HSPF mode! for a 25 year period. The
average annual net scour simulated by the HSPF model
is approximately 114,000 tons/year, consistent
with the range estimated by studies conducted in
the watershed. The scour simulated for reaches with
historic milldam sites is more than 99% of the total
scour simulated by the HSPF model.
rA
"V"
'II
Reach number
omuscnrMincor-j^rv
m^DciNLncoHt
r-.r-sr-cocococr>cri
Reach number
a
o
3
o
oo*H^is*ocnioo«Ni/)oo*H«a-r-*omi£)a>LnLnuDfsHrv.r-.r-*ooooco
HfNfMlN(NrgN(MNtNNN(NfNfNNfS(NOJ(NN(NNtN(NN(NN(NfM(N
Reach number
Figure 115. Net scour/depositiori simulated by the HSPF model for a 25 year period2
2Red bars represent scour and gray bars represent deposition.
97
-------
3.3 Plant Growth
Plant growth is an important component of the
nutrient balance of a SWAT model. Yield biomass
associated with agricultural crops were compared to
average annual yields published by the USDA-NASS
(http://quickstats.nass.usda.gov/) to ensure that
biomass simulated by the SWAT model are reasonable.
The yields reported by USDA-NASS for grain com and
soybean are in bushels per acre. These were converted
to tons per acre assuming using standard assumptions
of 56 pounds per bushel at 15.5% moisture content for
grain corn and 60 pounds per bushel at 13.5% moisture
content for soybean (http://extension.missouri.edu/
publications/DisplayPub.aspx?P=G4020). Moisture
content for silage corn and hay were assumed at
65% and 10%, respectively (http://extension.psu.
edu/agronomy-guide/cm/sec2/sec24e4). Table 40
shows the average biomass yield for agricultural
crops simulated by the SWAT model is comparable to
observed yields.
Figure 116 shows the observed and simulated biomass
yields for the simulation time-period. It is evident
from the figure that the model simulates the average
yield well but shows less variability from year to year
compared to observed data.
For most non-agricultural crops it was ensured that
the SWAT Error Checker did not raise any errors or
warnings associated with simulated biomass. The SWAT
Error Checker however did raise warnings associated
with forest (FRSD) and barren (SWRN) landuses,
• Crop FRSD: biomass may be too high 99.62 mg/ha
• Crop SWRN: biomass may be too low 0.40 mg/ha
Forest biomass simulated by the SWAT model is in
the same order of magnitude as reported by Blackard
etal. (2008) for this region of the United States. Low
biomass is expected for barren lands.
Table 40. Observed and simulated
biomass yields for agricultural crops
Crop
Average Yield (tons/ac/yr)
Observed
Simulated
Corn (grain)
3.3
3.8
Corn (silage)
7.0
6,9
Soybean
1.2
1.1
Hay
2.9
3.2
— Observed —Simulated
15
OJ
>
10
9
8
7
6
5
4
3
2
1
0
Corn (grain)
Corn (silage)
Soybean
Hay
Figure 116. Observed and simulated mean, minimum and maximum annual biomass
yield for the simulation time period
98
-------
V » V
enario in SWAT
4 Simulation of Restoration
4.1 Approach
The stream restoration efforts in the Big Spring Run
consisted of removing legacy sediments from historic
milldam sites and in the process reconnect floodplain
hydrology and expose buried wetlands (Hartranft
et al. 2011). There is limited guidance on modeling
the hydrological and water quality effects of stream
restoration at historic milldam sites. Changes were
made to both parameters in the SWAT model and to
the structure of the watershed by adding wetland
functions to historic milldam reaches. The Conservation
Practice Modeling Guide for SWAT and APEX (Waidler
et al. 2009) provides recommendations for simulating
stream restoration in the Agricultural Policy/
Environmental extender (APEX) model. The inputs to
the model to simulate stream restoration include,
• geometry of restored channel,
• channel vegetation or cover,
• reduced channel erodibility factor and
• Manning's constant reflecting the new roughness.
Sediment routing in APEX is based on Bagnold's
sediment transport equation (Williams et al. 2012).
As noted in section 2.1, sediment routing in SWAT is
also based on Bagnold's equation. While Waidler et al.
(2009) do not provide specific recommendations for
SWAT, we believe that the above recommendations
for stream restoration are also applicable to SWAT
on account of the similarities in sediment routing
methods. The following changes were made to the
reaches associated with former milldams in the SWAT
model to simulate the impacts of stream restoration.
• The channel longitudinal slopes were reduced
by 50% to promote sediment settling. Additional
changes to the channel geometry were not carried
out on account of a general lack of literature on
bankfull geometry pre- and post-restoration.
• The channel erodibility factors were reduced from
their calibrated values.
• Manning's constant of former milldam reaches
were increased from 0.03 to 0.06 to represent
increased roughness.
As mentioned above, removal of legacy sediment
also has the potential to expose buried wetlands
which may have additional water quality benefits. In
addition to the stream restoration scenario, wetlands
were simulated for reaches with historic milldams to
intercept runoff generated from upland sources. Berg
et al. (2013) suggest a maximum ponded volume of
1 foot in a floodpiain wetland to ensure interaction
between runoff and wetland plants, and a minimum
wetland to watershed surface area ratio of 1% to
ensure adequate retention times. The wetlands
simulated in the SWAT model were parameterized
along these guidelines.
Several restoration scenarios were simulated consisting
of varying reach parameters individually and in
combination with each other. Wetlands were simulated
as a standalone BMP and in conjunction with channel
restoration. The scenarios simulated are listed in
Table 41.
Table 41. F
restoration scenarios simulated using SWAT
Restoration
Scenario #
Longitudinal
Slope
Erodibility
Factor
Manning's n
Wetlands
1
Reduced
by 50%
Reduced
by 50%
0.06
Not
simulated
2
Reduced
by 50%
Reduced
by 75%
0.06
Not
simulated
3
No
Change
No
change
0.06
Simulated
4
Reduced
by 50%
Reduced
by 75%
0.06
Simulated
99
-------
4.2 Results and Discussion
The restoration simulations show decreases in
sediment and nutrient loads at both small and large
spatial scales. The net bank erosion decreases in the
Big Spring Run and the Conestoga River, respectively.
The net sediment and nutrient load transported also
reduce suggesting that settling losses occur in the
wetland and channel system in the post-restoration
scenarios. Table 42 and Table 43 summarize the
sediment and nutrient loads for pre- and post-
restoration scenarios for the Big Spring Run and the
Conestoga River.
For the Big Spring Run the sediment and nutrient
loads for scenarios 1 and 2 are identical to each other
suggesting that reducing the erodibility factor by
50% reduces erosion from channel sources by 100%.
Scenario 3 shows a larger reduction in both sediment
and nutrients. For the Conestoga River scenario 2
and 1 show an incremental reduction in sediment
and nutrient loads with an increase in erodibility
factor. Scenario 3 shows a slightly lower reduction in
sediment load compared to scenario 2 but a larger
reduction in nutrient loads. For both the Big Spring
Run and the Conestoga River scenario 4 shows the
highest reduction in sediment and nutrient loads.
Figure 117 and Figure 118 show the percent reduction
in simulated sediment and nutrient loads for the Big
Spring Run and Conestoga River for the four restoration
scenarios relative to pre-restoration.
Table 42. Sediment and nutrient loads for the Big Spring Run
Constituent Pre-restoration Scenario 1 Scenario 2 Scenario 3 Scenario 4
Sediment (tons/yr)
19,288
15,124
15,124
12,579
11,726
Total Phosphorus (Ibs/yr)
106,209
104,111
104,111
101,651
103,265
Total Nitrogen (Ibs/yr)
2,759,783
2,755,219
2,755,219
2,692,167
2,695,144
Table 43. Sediment and nutrient loads for the Conestoga River
Constituent
Pre-restoration
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Sediment (tons/yr)
2,630,919
1,909,688
1,512,173
1,576,699
938,495
Total Phosphorus (Ibs/yr)
12,356,851
11,710,346
11,051,473
10,950,105
10,282,921
Total Nitrogen (Ibs/yr)
267,330,347
264,823,364
263,452,090
260,943,232
258,652,632
100
-------
Sediment *10131 Phosphorus "Total Nitrogen
45%
40%
35%
g 30%
| 25%
0
1 20%
o:
15%
10%
5%
0%
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Figure 117. Percent reduction in sediment and nutrient loads for the five restoration scenarios relative
to pre-restoration for the Big Spring Run
70%
60%
50%
T
Sediment "Total Phosphorus "Total Nitrogen
o 40%
o
"8 30%
IT
20%
10%
0%
I
I
Scenario 1
I
Scenario 2
Scenario 3
Scenario 4
Figure 118. Percent reduction in sediment and nutrient loads for the five restoration scenarios relative
to pre-restoration for the Conestoga River
101
-------
5 GroundwaterV ilodel Development
Calibration an\J Application
5.1 GFLOW Model Development
5.1.1 GFLOW Model
Simulation modeling was conducted to evaluate
surface water-groundwater interactions in the Big
Spring Run (BSR) watershed using GFLOW (http://www.
haitjema.com/). GFLOW is a two-dimensional, steady-
state, analytical element model of groundwater flow
that is strongly focused on the evaluation of surface
and groundwater interactions. The results of the
GFLOW model were used to validate the groundwater
component of the SWAT model being developed for
the Conestoga River watershed under the same work
assignment.
It should be noted that the GFLOW model application
is a simplified representation of the groundwater
flow system that represents average conditions and is
based on limited data, both in regards to stratigraphy
and measured head elevations. The results reported
here are thus only a preliminary insight into the
groundwater and surface water interactions in the BSR
watershed. More refined results will be forthcoming if
and when a dynamic MODFLOW mode! of the area is
constructed. Developing and calibrating such a model
will, however, require extensive data collection and a
program of regular monitoring of water elevations in
wells throughout the area of interest.
GFLOW models steady state flow in a single
aquifer using the Dupuit-Forchheimer assumption:
groundwater flows horizontally in an unconfined
aquifer, and groundwater discharge is proportional to
saturated aquifer thickness. The "conjunctive surface
water-groundwater" solution option within the
model allows for the simulation of these interactions
by calculating average baseflows and limiting the
infiltration rates of losing streams. The conjunctive
solution allows for the routing stream network and
groundwater to be modeled simultaneously with
a dictated number of iterations. Model baseflow is
calculated during the solution process by accumulating
groundwater inflows to the stream network from
upstream to downstream. In cases where baseflow is
negative, a reach is considered losing at that point.
5.1.2 Model Extent
The objective of this study is to develop a better
understanding of the small-scale groundwater and
surface water interactions in the BSR. In addition, the
groundwater model also provides information to the
larger scale SWAT model of the entire Conestoga River.
To facilitate model inter-comparison, sub-watershed
boundaries consistent with the SWAT surface water
model are used to define the area of interest for the
groundwater model. Specifically, the GFLOW model
extent aligns with the SWAT model subbasins 282, 283,
289 and 290.
102
-------
The GFLOW model area of interest encompasses
the BSR watershed. The focus area is bounded by
the Paquea Creek in the south, and Mill Creek and
Conestoga River in the north (Figure 119), This area
was chosen because it contains a stretch of stream
which has undergone wetland restoration in the past
decade (Hartranft et al. 2011). The BSR site was chosen
for restoration initially by the Pennsylvania Department
of Environmental Protection as a test site to explore
a new approach to aquatic ecosystem restoration.
Restoration at this site entailed removal of legacy
sediment which was built up historically due to the
presence of a milldam, and thereby restoring natural
valley morphology. Hartranft etal. (2011) characterizes
the post-restoration ecosystem as a successfully
restored native plant-populated palustrine emergent
wetland.
Conestoga River
Watershed
Susquehanna
River )
Big Spring Run
Legend
River/Stream
~ Big Spring Run Watershed
~ Conestoga
~ County Boundary
River Watershed
Mill Creeks
Paquea
WltKW
Stieet
Big Spring Run
Location Map
NADJ 9S3_UTM_Z«ie_17N
Map produced - H. Nicholas
N 0 0.2 0 4 0.8 Kilometers
^ 0 0.2 0.4 0.8 Miles
It
TETRA TECH
Figure 119. GFLOW model extent
103
-------
5.1.3 Aquifer Properties
The basic aquifer parameters required by GFLOW are
aquifer base elevation, aquifer thickness, hydraulic
conductivity, and recharge rates. Areas with properties
significantly different from the main aquifer are
modeled as inhomogeneities. Inhomogeneities are
subset domains where the aquifer properties are
redefined.
The BSR area is underlain almost entirely by the
limestone Conestoga Formation with the exception
of a narrow ridge that includes areas of the Vintage
Formation, and Antietam and Harpers Formation
(Berg et al. 1980). These formations are generally
expected to have a lower hydraulic conductivity than
the Conestoga Formation. According to Galeone et al.
(2006), this narrow ridge was found to have "little or
no" impact on flow patterns and directions within the
drainage. As a result, this relatively small geological
difference was not explicitly represented as an
inhomogeneity in the GFLOW model.
Presence of carbonate rocks results in karst terrain
within the Conestoga River watershed. The BSR area
contains no sinkholes or caves based on inspection
of aerial imagery products and GIS maps. As a result,
it is likely that the karst behavior is diffuse infiltration
dominated. Thus the aquifer system is assumed to
behave as an isotropic porous medium, which is
consistent with GFLOW model assumptions.
The datum or base elevation of the model was set at
72.18 m based on Berg et al. (1980), and Miles and
Whitfield (2001). The aquifer thickness was set at an
artificial value of 10,000 meters based on the model
developer's suggestion that an unconfined aquifer
should be specified with a "sufficiently large thickness...
to ensure that the water table will never reach the
aquifer top, except where it intercepts streams"
(Haitjema 1995).
Based on results from tests of 77 wells in the region,
Low etal., (2002) determined the median value
of hydraulic conductivity (Kjat) of the Conestoga
Formation as 0.579 m/d. Gerhart and Lazorchick
(1984) found that hydraulic conductivity in the lower
Susquehanna increases from hilltops to valley bottoms
and can be generally estimated by using multipliers
on the average. The multipliers proposed by Gerhart
and Lazorchick (1984) were 0.4 and 3.0 for steeper
areas and valleys, respectively. The regional Kjat
parameter was manually adjusted during calibration,
and the appropriate range of values was estimated
using the above mentioned coefficients to account
for topographic effect. The Kjat values used in the
calibrated model ranges between 0.232 and 1.737 m/d.
The USGS gages, namely, Little Conestoga River near
Churchtown (USGS 01576085) and Conestoga River at
Conestoga (01576754), were used to estimate recharge
using a recession-curve-displacement method (Risser
et al. 2005). USGS used the RORA program (named
for the Rorabaugh Method) to estimate recharge at
these gage locations based on the change in total
potential groundwater discharge that is caused by each
recharge event (Rorabaugh 1964, Daniel 1976). The
long-term recharge estimated at USGS 01576085 and
USGS 01576754 are 12.56 and 15.41 in/yr, respectively,
with an average of 13.985 in/yr (or 0.00097 m/d). A
net areal recharge rate of 0.00097 m/d was initially
specified for the GFLOW model and adjusted during
calibration.
5.1.4 Representation of Stream Network
Rivers and streams, and lakes connected to the stream
network, are represented as line-sinks in the GFLOW
model. Line-sinks are connected to the aquifer and may
gain from or lose to groundwater depending on the
water table elevation relative to line-sink head.
NHD high resolution stream centerline layer was
used to provide locations of all waterways. Line-sinks
may be specified as near-field or far-field. Near-field
line-sinks are assigned specific routing and hydraulic
properties while far-field line-sinks provide constant
boundary conditions. The near-field line-sinks were
represented by simplifying the NHD high resolution
stream centerline layer so as to reduce the number
of line-sink segments while maintaining the general
shape and character of each stream. Far-field line-sinks
were represented using coarser approximations. By
surrounding the area of interest with far-field line-sinks
outside of the stream network, natural hydrogeological
boundaries develop along the natural divide outside
the area-of-interest.
The BSR and its tributaries were represented as
near-field line-sinks. Paquea Creek on the south, and
Conestoga River and Mill Creek on the north generally
served as the far-field line-sinks. The sections of
104
-------
5.1.5.2 Restored Wetland
In 2009, the EPA began a restoration initiative to assess
the potential benefits of riparian wetland restoration
best management practices (Hartranft etal. 2011).
The project consisted of restoring an incised, single-
thread, sinuous modern channel into a natural, wider,
wetland/wet-meadow morphology (Hartranft, 2013).
The restoration process largely entailed the removal of
historical milldam legacy sediment which had built up
over time and buried the traditional wetland soils. The
restoration area is approximately 7 acres and lies on an
unnamed tributary which runs parallel to the Gypsy Hill
Road in the south-west portion of the watershed. The
area is downstream of the perennial spring for which
Big Spring Run is named.
The restoration area was represented as a head-
specified near-field line-sink in the GFLOW model.
Piezometer data in the restoration area suggest a
constant head of approximately 94.5 meters which
does not vary seasonally. Since the hydraulic head
in the restoration area is generally static, the use of
a head-specified line-sink for representation in the
model is appropriate.
from the dry season model.
Table 44. Parameters associated with near-field line-sinks
Input Parameter
Definition
Model Input
Resistance
Thickness of resistance layer between surface
water feature and the aquifer divided by the
average vertical hydraulic conductivity of the
resistance layer. [T1]
An initial value of 1 day was used for all reaches
in the network to represent the riverbed resistance
to groundwater flow. This is one of the primary
parameters adjusted during model calibration. The
final values vary by reach.
Width
When choosing "along surface water boundary",
the width provided is the actual stream width. [L]
Stream widths were estimated using aerial imagery.
Widths range from 1 to 5 meters in the area-of-
interest.
Depth
Distance between the surface water elevation
and bottom of the resistance layer, and used to
determine the aquifer thickness underneath the
surface water. [L]
Depth was initially estimated at 1 meter for all
reaches in the network, and was adjusted for
individual reaches during calibration.
Mill Creek immediately above and below the BSR
confluence were however, modeled as head-specified
near-field line-sinks.
Hydraulic properties associated with near-field line-
sinks required as inputs to the model are resistance,
width and depth. Definitions of these parameters
and their associated values are provided in Table
44. All line-sinks in the model require upstream and
downstream head elevations, and were assigned values
based on a 1-meter resolution digital elevation model
(DEM).
5.1.5 Representation of Ponds and
Restoration Area
5.1.5.1 Ponds
There are several small ponds present in the study
area which are not connected to the surface stream
network. These ponds are modeled as near-field head-
specified line-sinks with water elevations applied
"along the surface water boundary". The head values
assigned to each pond were estimated using a 1-meter
resolution DEM. These line-sinks were not included
in the stream network for purposes of routing, which
allows for these ponds to act as infinite water sources
to the system. These ponds were completely removed
105
-------
5.1.6 Combined Model Elements and Model Environment
All of the physical features (rivers, ponds, aquifer parameters) in the model are shown within the model
environment in Figure 120.
Legend
Restoration Area
Near Field Linesink (routing)
Ponds or Near Field Linesink (non-routing)
Far Field Linesink (non-routing)
t^ShPP*
t»p
BSR Boundary
Creek
Conestoga
River
Big
Spring
Run
K
Paquea
Creek
Irwol
Big Spring Run
GFLOW Model Elements
N 0 0.3 0 6 1.2 Kilometers
~ 0.3 0.6 1.2 Miles
TETRA TECH
NAD _1983_UTM_Zane_17N
Map produced - H Nicholas
Figure 120. Analytical elements in the Big Spring Run GFLOW model environment
106
-------
5.2 GFLOW Model Calibration
5.2.1 Calibration Approach
Calibration of dynamic models generally consists of
comparing observed and simulated time-series. As
a steady-state model, this approach is however not
applicable. Instead, the objective is to make the model
as realistic as possible in representing long-term,
annual and seasonally averaged water potentiometric
surface elevations and fluxes between surface and
groundwater. The primary test of model realism is
comparison to potentiometric heads and observed
flows. Some continuous piezometer data series are
available in this region, especially the restoration area.
Therefore, the comparison relies primarily on static
water levels recorded in the wells and the baseflow
determined from stream flow time-series at USGS
monitoring locations, identified in the model as
"test points".
Relative Percent Difference (RPD)
Calculated heads are expected to differ from observed
heads for many reasons, most importantly because
the model aquifer is merely an abstraction of the
real aquifer system. A successful model will show
deviations relative to observed heads that are both
positive and negative with a spatial distribution of
deviations that is not strongly clustered. Statistics used
for model calibration are relative percent difference
(RPD) and root mean squared error (RMSE), along with
the slope and squared correlation coefficient (R2) for a
linear regression between the simulated and observed
data.
{Model Result - Observed Result)
Average(Model Result, Observed Result)
Root Mean Squared Error (RMSE) =
j(Observed Result — Model Result)-
number of observations
5.2.1.1 Test Points
Test points used for model calibration consisted of
stream flow gages and groundwater piezometers
operated by USGS, and piezometers installed by EPA in
the Big Spring Run watershed restoration area (Table
45 and Table 46). Median of observed time-series
associated with the piezometers were compared to
the potentiometric head simulated by GFLOW. Note
that the piezometers in the study area are frequently
present in clusters, with each cluster generally
consisting of two to three individual piezometers. The
heads associated with each piezometer in a cluster
were generally found to be similar to one another and
therefore each such cluster was treated as a single test
point in the model. The average of the potentiometric
heads associated with each piezometer in a cluster was
used in model calibration. The existence of comparable
potentiometric heads in a relatively small region
also validates the assumption of a single phreatic
aquifer adopted for the model environment. The EPA
piezometer time-series data extends back to the pre-
restoration time in 2009, however data only after May
2010 was used for model calibration. As stated earlier,
the hydraulic behavior of the restoration area
is configured to have a constant head representative of
post-restoration behavior.
Long-term median baseflow associated with the stream
flow gages were compared to the simulated outflow
from relevant line-sinks. The baseflow associated
with each USGS flow gage were determined using
the sliding-interval baseflow separation method on
observed stream flow time-series. Baseflow separation
was accomplished by using the HYSEP program. HYSEP
is a hydrograph separation program developed by
USGS that provides techniques to determine baseflow
from a continuous streamflow time-series (Sloto and
Crouse 1996). For line-sinks with more than one USGS
gage, the average of the two baseflow time-series was
adopted. Time-series of gage height was reported for
USGS gages at Beaver Valley and Willow Street instead
of flow, therefore flows were estimated for these gages
using USGS rating curves. HYSEP was subsequently run
on the stream flow time-series to determine baseflow.
Figure 121 shows the locations of the test points in the
GFLOW model environment.
I
-------
Table 45. USGS stream flow gage test points used in GFLOW calibration
Assigned
Reference
Name
Assigned
Baseflow
(m/d)
Data Type
Gage ID
Location Name
Median
Observed
Baseflow
(m3/d)
Data
Range
Gypsy Hill
4403.83
Daily
Discharge
01576521
Big Spring Run near Willow Street
4403.83
1993-2001
Lampeter
2201.91
Daily
Discharge
01576529
Unnamed Tributary to Big Spring Run
near Lampeter
2201.91
1993-2001
Beaver Valley
2691.23
Daily Gage
Height
01576516
Big Spring Run about Tributary near
Willow Street
2691.23
2012-2015
Willow Street
2446.57
Daily Gage
Height
015765185
Unnamed Tributary to Big Spring Run
near Willow Street
2446.57
2012-2015
North Fork
305.82
Daily
01576527
North Fork Unnamed Tributary to
Big Spring Run at Lampeter
391.45
1993-2001
Discharge
015765265
North Fork Unnamed Tributary to
Big Spring Run near Lampeter
220.19
1995-2001
Table 46. Piezometer test points used in GFLOW calibration
Assigned
Reference
Name
Assigned Head
(m)
Agency
Date Type
Well ID
Well
Name
Average
Observed Head
(m)
Data
Range
Half-Hour
395936076154404
LN 2130
94.49
2010-2011
Nest 2
94.30
EPA
Water Table
395936076154405
LN 2131
93.96
2010-2011
Elevation
395936076154406
LN 2132
94.43
2010-2011
Half-Hour
395933076154404
LN 2133
95.69
2010-2011
Nest 3
95.28
EPA
Water Table
395933076154405
LN 2134
94.78
2010-2011
Elevation
395933076154406
LN 2135
95.36
2010-2011
Half-Hour
395934076154604
LN 2136
96.08
2010-2011
Nest 4
95.22
EPA
Water Table
395934076154605
LN 2137
94.42
2010-2011
Elevation
395934076154606
LN 2138
95.18
2010-2011
Daily Water
Table Elevation
395957076155203
LN 2043
88.97
1993-2000
Nest BSR
89.08
USGS
395957076155204
LN 2044
89.12
1993-2000
395957076155201
LN 2041
89.14
1993-1999
Nest North
105.45
USGS
Daily Water
395947076145901
LN 2037
105.37
1994-2001
Fork
Table Elevation
395947076145904
LN 2040
105.52
1994-2001
108
-------
Gypsy Hill
01576521
Lampeter
01576529
North Fork
01576527
015765265
\ ^ LN 2037-
Beaver Valley: |_N 20401
01576516\
¦Willow Street
015765185
Big Spring Run
GFLOW Model Test Points
NAD_1983_UTM_Zone_1 7N
Map produced - H. Nicholas
N 0 0.2 0.4 0.8 Kilometers
i\ 0 0.2 0.4 0.8 Miles
TETRATECH
Legend
A Test Point: Stream Gage
O Test Point: Piezometer
Willow
Street
i BSR Boundary
I :
Near Field Linesink (routing)
Far Field Linesink (non-routing)
Near Field Linesink (non-routing)
Figure 121. Test points in the GFLOW model
109
-------
5.2.1.2 Seasonal Model Setup and Calibration
In addition to a model representing long-term average
conditions, seasonal models were also configured to
simulate the hydraulic behavior of the region under
wet and dry conditions. An analysis of precipitation
shows that April, May and June are the wettest
months, and December, January and February are
the driest in the watershed (PRISM Climate Group
2004). The hydrologic behavior of the system however
exhibits a pattern different from the precipitation. The
difference is most likely on account of the interaction
between snowmelt and evaporation. Snowmelt often
peaks in March and April, so as a result, the seasonality
observed in the baseflow at the USGS gages were used
to drive the wet and dry season models.
The baseflow was generally highest during February,
March and April, and lowest during August, September
and October at all the USGS gages except Beaver
Valley. The baseflow at Beaver Valley was found to be
relatively stable probably because it is spring-fed. This
gage is included in the wet season model but excluded
from the dry season model setup. Using the same
test point locations and evaluation criteria as before,
median baseflow and hydraulic head were estimated
for the wet and dry seasons and the performance of
the GFLOW models were assessed.
Setup for the two seasonal models were identical to
the setup for the long-term average conditions model
with a few exceptions. All model parameters were held
constant except for the recharge rate and the line-sink
heads which were used as calibration parameters.
Recharge for the dry season model was set to zero, and
recharge for the wet season model was increased from
0.00097 m/d to 0.0012 m/d. As mentioned previously
as well, the dry season model does not include farm
ponds as they are assumed to dry up during this
period.
It is important to note that while there are clear
seasonal differences in baseflow at the stream flow
gages, there is not a clear seasonal pattern in the
piezometers. Piezometer Nests 2, 3, and 4 (located in
the wetland restoration area) show very little variability
in observed heads. The range of average daily heads for
Nest BSR and Nest North Fork differ by approximately
1 meter for the period of record. Highest and lowest
heads are generally around March and August,
respectively, which mirrors the seasonal behavior of
the observed baseflow.
5.2.2 Calibration Results
Model calibration consisted of a systematic adjustment
of parameters generally geared towards achieving
the closest match between simulated and observed
potentiometric heads and baseflow. The parameters
adjusted during the calibration consisted of stream
resistance and depth, aquifer recharge rate and aquifer
hydraulic conductivity. The values of these parameters
in the calibrated model are listed below.
• Recharge Rate - 0.0009 meters per day
• Hydraulic Conductivity - 1.737 meters per day
• Line-sink Depth - 0 to 5 meters
• Line-sink Resistance - 0 to 12 days
It is important to note that the line-sink associated
with the Beaver Valley gage is the only tributary with
zero depth and zero resistance. The use of zero depth
and resistance is appropriate because this line-sink is
believed to be spring-fed (Galeone et al. 2006). The
Galeone report identified the spring along this reach to
have an approximate discharge of 50 gallons/minute.
Table 47 shows the values of the parameters associated
with each analytical element in the calibrated model.
Table 47. Values of parameters associated with analytical elements in the calibrated model
Label
Field
Feature Type
Streamflow
Routing
Width (m)
Depth (m)
Resistance (m)
Ending Head (m)
BSR_1
Near
Stream
Yes
1
1
2
85.71
BSR_2
Near
Stream
Yes
5
5
12
81.00
BSR_3
Near
Stream
Yes
5
5
12
79.95
BSR_4
Near
Stream
Yes
2
1
2
81.02
BSR_5
Near
Stream
Yes
1
1
4
102.00
BSR_6
Near
Stream
Yes
1
1
2
102.00
110
-------
Table 47 (continued). Values of parameters associated with analytical elements in the calibrated model
Label
Field
Feature Type
Streamflow
Routing
Width (m)
Depth (m)
Resistance (m)
Ending Head (m)
BSR_7
Near
Stream
Yes
6
5
4
88.30
BSR_8
Near
Stream
Yes
5
5
12
86.01
BSR_9
Near
Stream
Yes
5
5
12
89.55
BSR_10
Near
Stream
Yes
3
1
2
96.38
BSR_12
Near
Stream
Yes
1
1
1
102.96
BSR_13
Near
Stream
Yes
1
1
1
102.00
FF_1
Far
Stream
No
0
0
0
75.00
FF_2
Far
Stream
No
0
0
0
67.00
FF_3
Far
Stream
No
0
0
0
56.00
FF_4
Far
Stream
No
0
0
0
90.00
FF_5
Far
Stream
No
0
0
0
95.00
FF_6
Far
Stream
No
0
0
0
95.00
FF_7
Far
Stream
No
0
0
0
80.00
FF_8
Far
Stream
No
0
0
0
90.00
FF_9
Far
Stream
No
0
0
0
100.00
FFJO
Far
Stream
No
0
0
0
100.00
FF_13
Far
Stream
No
0
0
0
105.00
FF_14
Far
Stream
No
0
0
0
97.00
Lakel
Near
Pond
No
5
0
0
108.00
Lake2
Near
Pond
No
5
0
0
121.00
Lake3
Near
Pond
No
5
0
0
99.70
Conestoga
Far
Stream
No
0
0
0
56.00
Paquea
Far
Stream
No
0
0
0
47.24
Conestoga_
FarField
Far
Stream
No
0
0
0
67.03
MillCreekUpper
Near
Stream
Yes
12
0
0
79.95
MillCreekLower
Near
Stream
Yes
12
0
0
75.00
MillCreekFarLower
Far
Stream
No
0
0
0
67.03
MillCreekTop
Far
Stream
No
0
0
0
79.93
FarField JO
Far
Stream
No
0
0
0
89.02
FarField_Con1
Far
Stream
No
0
0
0
73.80
FarField_Con2
Far
Stream
No
0
0
0
73.48
FF_Top
Far
Stream
No
0
0
0
76.89
BSRJ1
Near
Stream
Yes
1
0
0
97.60
Waterbody
Near
Pond
No
5
0
0
124.00
Waterbody2
Near
Pond
No
5
0
0
115.00
Restoration
Near
Restoration
Wetland
No
5
0
0
94.70
111
-------
Table 48 and Table 49 show the observed and
simulated head and baseflow at the test points for
the long-term average conditions model. The RPD and
RMSE for piezometer and USGS gage test points are
generally low indicating very good model performance.
Figure 123 and Figure 122 show the simulated heads
and baseflow against observed for all test points in the
watershed. Both the slope and R2 of the regression
lines are close to 1, indicating excellent model
performance.
Table 48. GFLOW calibration results for piezometer test points
Test Point
Observed Head (m)
Modeled Head (m)
RPD
RMSE (m)
Nest North Fork
105.45
106.58
1.07%
1.13
Nest BSR
89.08
90.43
1.51%
1.35
Nest 2
94.30
94.98
0.71%
0.68
Nest 3
95.28
95.18
-0.11%
0.10
Nest 4
95.22
95.40
0.19%
0.18
Table 49. GFLOW calibration results for USGS gage test points
Test Point
Observed Baseflow (cfs)
Modeled Baseflow (cfs)
RPD
RMSE (cfs)
Lampeter
0.90
1.06
16.46%
0.16
Beaver Valley
1.10
1.09
-0.83%
0.01
Willow Street
1.00
0.98
-1.99%
0.02
North Fork
0.12
0.13
5.89%
0.01
Gypsy Hill
1.80
1.91
5.89%
0.11
112
-------
1 1 1 1
85 90 95 100 105 110
Observed Head(m)
Figure 122. Modeled vs observed baseflow for all piezometer test points
H 1 h-
0.5 1.0 1.5
Observed Baseflow (cfs)
Figure 123, Modeled vs observed baseflow for all USGS gage test points
113
-------
The model performance for the wet season model
was also very good. RPD and RMSE were low for all
piezometers and USGS gage test points (Table 50
and Table 51). The linear regression of simulated vs
observed heads and baseflow had slopes and R2 close
to 1 indicating very good model performance (Figure
124 and Figure 125).
Table 50. GFLOW calibration results for piezometer test points (wet season)
Test Point
Observed Head (m)
Modeled Head (m)
RPD
RMSE (m)
Nest North Fork
105.55
106.39
0.80%
0.84
Nest BSR
89.15
90.13
1.10%
0.98
Nest 2
95.28
95.18
-0.10%
0.10
Nest 3
95.22
95.38
0.16%
0.16
Nest 4
94.30
94.96
0.69%
0.66
Table 51. GFLOW calibration results for USGS gage test points (wet season)
Test Point
Observed Baseflow (cfs)
Modeled Baseflow (cfs)
RPD
RMSE (cfs)
Lampeter
1.60
1.57
-2.02%
0.03
Beaver Valley
0.95
0.93
-1.89%
0.02
Willow Street
1.95
1.90
-2.42%
0.05
North Fork
0.28
0.28
0.17%
<0.01
Gypsy Hill
2.60
2.75
5.58%
0.15
114
-------
110
105
y = 1.003x + 0.232
R2 = 0.994
100
95
90
85
80
80
85
90 95 100
Observed Head(m)
105
110
Figure 124. Modeled vs observed baseflow for all piezometer test points (wet season)
1.0 1.5 2.0
Observed Baseflow (cfs)
Figure 125, Modeled vs observed baseflow for all USGS gage test points (wet season)
115
-------
The dry season model was generally able to simulate
the observed potentiometric heads very well (Table
52). The model performance was generally good for
baseflow as well with the exception of the test point
at Lampeter (Table 53). The RPD at this location is low
indicating under-prediction, while the RMSE is low
indicating good model performance. Since GFLOW is a
steady state model and statistical evaluation consists
of comparing one simulated value to one observed
value at a test point, the RPD is likely a better measure
of model performance than RMSE. The slopes of the
regression lines for simulated heads and baseflow
against observed are generally less than 1 and the R2 is
greater than 0.95 (Figure 126 and Figure 127).
Table 52. GFLOW calibration results for piezometer test points (dry season)
Test Point
Observed Head (m)
Modeled Head (m)
RPD
RMSE (m)
Nest North Fork
105.30
104.98
-0.31%
0.32
Nest BSR
89.03
89.85
0.92%
0.82
Nest 2
95.28
95.09
-0.20%
0.19
Nest 3
95.22
95.30
0.09%
0.08
Nest 4
94.30
94.77
0.50%
0.47
Table 53. GFLOW calibration results for USGS gage test points (dry season)
Test Point
Observed Baseflow (cfs)
Modeled Baseflow (cfs)
RPD
RMSE (cfs)
Lampeter
0.39
0.26
-41.81%
0.13
Willow Street
0.60
0.66
10.16%
0.06
North Fork
0.03
0.04
16.14%
0.01
Gypsy Hill
1.10
1.03
-6.51%
0.07
116
-------
110
105
100
-o 95
o>
90 +
85
80
y = 0.933x + 6.581
R2 = 0.998
+
80
85
90 95 100
Observed Head(m)
105
110
Figure 126. Modeled vs observed baseflow for all piezometer test points (dry season)
1.5
T 1.0
2 0.5
y = 0.969x - 0.017
R2 = 0.962
0.0
0.0
0.5 1.0
Observed Baseflow (cfs)
1.5
Figure 127, Modeled vs observed baseflow for all USGS gage test points (dry season)
117
-------
5.3 GFLOW Model Applications
5.3.1 Head Contours and Flow Lines
The calibrated model produces estimates of groundwater head and stream lines in the study area. The stream lines
are norma! to the head contours. Figure 128 shows the head contours for average annua! recharge conditions.
Big Spring Run
10-m Head Contours
NAD J983JJTM_Zone_17N
Map~produced - H Nicholas
N 0 0 3 0.6 1.2 Kilometers
0 0.3 0.6 1.2 Miles
TETRATECH
Legend
Restoration Area
— Near Field Linesink (routing)
Ponds or Near Field Linesink (non-routing)
Far Field Linesink (non-routing)
f—T
i BSR Boundary
Figure 128. GFLOW simulation of head contours for average annual recharge conditions
118
-------
5.3.2 Surface Water Exchanges
GFLOW provides estimates of exchange rates along each line-sink segment in units of m2/d. The exchange rate for
each routing near-field line-sink is shown in Figure 129. It is evident from the results that some of the headwater
line-sink segments have essentially zero exchanges with groundwater (shown using orange circles). The restoration
area and line-sinks in the restoration area are generally losing to groundwater (shown using red circles). All the
other line-sink segments are gaining.
Legend
; i BSR Boundary
Stream Discharge (m2/d)
• <0
Big Spring Run
Line Sink Exchange Rates
N AD_1983_UTM_Zon e_17N
Map produced - H Nicholas
N 0 0.3 0.6 1.2 Kilometers
0 0.3 0.6 1.2 Miles
It
TETRA TECH
Figure 129. GFLOW simulation of line-sink exchange rates for average annual recharge conditions
119
-------
5.3.3 Baseflow
GFLOW provides estimates of streamflow for each line-sink segment (Figure 130). The baseflow generally increases
from the headwaters to the downstream line-link segments. Some of the headwater reaches essentially show
zero discharge (shown using red circles), which is probably on account of a net zero exchange with groundwater as
shown in the previous figure.
Legend
i BSR Boundary
Baseflow (cfs)
Big Spring Run
Line Sink Baseflow
1.2 Kilometers
1.2 Miles
NAD J 983_UTM_Zone_17N
Map produced -H Nicholas
Figure 130. GFLOW simulation of line-sink baseflow for average annual recharge conditions
120
-------
Arnold, J.G., Srinivasan, R., Muttiah, R.S., and Allen, P.M. 1998. Large Area Hydrologic Modeling and Assessment Part I:
Model Development. Journal of the American Water Resources Association, 34(1): 73-89.
Baffaut, C. and Benson, V.W. 2009. Modeling Flow and Pollutant Transport in a Karst Watershed with SWAT.
Transactions of the ASABE, 52(2): 469-479.
Bagnold, R. A. 1977. Bed Load Transport by Natural Rivers. Water Resources Research, 13(2), 303-312.
Baker, NT. 2011. Tillage Practices in the Counterminous United States, 1989-2004-Datasets Aggregated by Watershed:
U.S. Geological Survey Data Series 573.
Berg, J., Burch, J., Cappuccitti, D., Filoso, S., Fraley-McNeal, L., Goerman, D,, Hardman, N., Kaushal, S., Medina, D.,
Meyers, M., Kerr, B,, Stewart, S,, Sullivan, B,, Walter, R. and Winters, J. 2013. Recommendations of the Expert Panel
to Define Removal Rates for individual Stream Restoration Projects. Chesapeake Stormwater Network and Center for
Watershed Protection.
Berg, T. M., Edmunds, W.E., Geyer, A.R. and others, compilers. 1980. Geologic Map of Pennsylvania. Pennsylvania
Geological Survey. 4th ser., Map 1, 2nd ed., 3 sheets, scale 1:250,000.
Bicknell, B.R., Imhoff, J.C., Kittle Jr., J.L., Jobes, T.H., Duda, P.B. and Donigian Jr., A.S. 2014. HSPF Version 12.4 User's
Manual. Aqua Terra Consultants, Mountain View, CA.
Bicknell, B.R., Imhoff, J.C., Kittle Jr., J.L., Jobes, T.H., Duda, P.B. and Donigian Jr., A.S. 2014. HSPF Version 12.4 User's
Manual. Aqua Terra Consultants, Mountain View, CA.
Bieger, K., Rathjens, H,, Allen, P.M. and Arnold, J.G. 2015. Development and Evaluation of Bankfull Hydraulic
Geometry Relationships for the Physiographic Regions on the United States. Journal of the American Water Resources
Association, 51(3): 842-858.
Blackard, J.A., Finco, M.V., Helmer, E.H., Holde, G.R., Hoppus, M.L., Jacobs, D.M., Lister, A.J., Moisen, G.G., Nelson,
M.D., Riemann, R., Ruenfenacht, B., Salajanu, D., Weyermann, D.L., Winterberger, K.C., Brandeis, T.J., Czaplewski, R.L.,
McRoberts, R.E., Patterson, P.L. and Tymcio, R.P. 2008. Mapping U.S. Forest Biomass using Nationwide Forest Inventory
Data and Moderate Resolution Information. Remote Sensing of Environment, 112:1658-1677.
Chow, V.T., Maidment, D.R., and Mays, L.W. 1988. Applied Hydrology. McGraw-Hill, New York, New York.
Chu, T.W., Shirmohammadi, A., Montas, H., and Sadeghi, A. 2004. Evaluation of the SWAT Model's Sediment and
Nutrient Components in the Piedmont Physiographic Region of Maryland. Transactions of the ASABE,
47(5): 1523-1538.
121
-------
Daniel, J.F, 1976. Estimating Groundwater Evapotranspiration from Streamflow Records. Water Resources Research.
12(3): 360-364.
Duda, P.B., Hummel, P.R., Donigian Jr., A.S. and Imhoff, J.C. 2012. BASINS/HSPF: Model Use, Calibration and Validation.
Transactions of the ASABE. 55(4): 1523-1547.
Fry, J., Xian, G,, Jin, 5,, Dewitz, J., Homer, C,, Yang, L, Barnes, C,, Herold, N. and Wickham, J. 2011. Completion of the
2006 National Land Cover Database for the Conterminous United States, Photogrammetric Engineering and Remote
Sensing, 77(9): 858-864.
Galeone, D.G., Brightbill, R.A., Low, D.J. and O'Brien, D.L. 2006. Effects of Streambank Fencing of Pasture Land on
Benthic Macroinvertebrates and the Quality of Surface Water and Shallow Ground Water in the Big Spring Run Basin of
Mill Creek Watershed, Lancaster County, Pennsylvania, 1993-2001. USGS Report 2006-5141.
Garrick, M,, Cunnane, C. and Nash, J.E. (1978) A criterion of efficiency for rainfall-runoff models. Journal of Hydrology,
36(3-4): 375-381.
Gellis, A.C., Hupp, C.R., Pavich, M.J., Landwehr, J.M., Banks, W.S.L., Hubbard, B.E., Langland, M.J., Ritchie, J.C. and
Reuter, J.M. 2009. Sources, Transport, and Storage of Sediment in the Chesapeake Bay Watershed: U.S. Geological
Survey Scientific Investigations Report 2008-5186, 95p.
Gerhart, J.M. and Lazorchick, G.L. 1984. Evaluation of the Ground-Water Resources of Parts of Lancaster and Berks
Counties, Pennsylvania. US Geological Survey Water-Resources Investigations Report 84-4327.
Haitjema, H.M. 2005. Analytic Element Modeling of Groundwater Flow. Academic Press. ISBN 1-59533-803-9.
Hartranft, J.L. 2013. Applying Principles for Ecological Restoration of Aquatic Resources to Legacy Sediment Problems
in Pennsylvania. Bureau of Waterways Engineering and Wetlands.
Hartranft, J.L., Merrits, D.J., Walter, R.C. and M. Rahnis. 2011. The Big Spring Run Restoration Experiment: Policy,
Geomorphology, and Aquatic Ecosystems in the Big Spring Run Watershed, Lancaster County, PA. Sustain: Stream
Restoration. The Kentucky Institute for the Environment and Sustainable Development, 24: 24-30.
Homer, C,, Dewitz, J., Fry, J., Coan, M,, Hossain, N,, Larson, C,, Herold, N,, McKerrow, A., VanDriel, J.N. and Wickham,
J. 2007. Completion of the 2001 National Land Cover Database for the Conterminous United States. Photogrammetric
Engineering and Remote Sensing, 73(4): 337-341.
Jha, M. K,, Gassman, P. W. and Arnold, J. G. 2007. Water Quality Modeling for the Raccoon River Watershed Using
SWAT. Transactions of ASABE, 50(2), 479-493.
Jin, S,, Yang, L,, Danielson, P., Homer, C,, Fry, J. and Xian, G. 2013. A Comprehensive Change Detection Method for
Updating the National Land Cover Database to Circa 2011. Remote Sensing of Environment, 132:159-175.
Low, D.J., Hippe, D.J. and Yannacci, D. 2002. Geohydrology of Southeastern Pennsylvania. USGS in cooperation with
Pennsylvania Department of Conservation and Natural Resources, Bureau of Topographic and Geologic Survey. Water
Resources Investigations Report 00-4166. New Cumberland, Pennsylvania.
Lumb, A.M., McCammon, R.B. and Kittle Jr., J.L. 1994. User's Manual for an Expert System (HSPEXP) for Calibration of
the Hydrological Simulation Program- FORTRAN. USGS Water Resources Investigation Report 94-4168. U.S. Geological
Survey, Reston, VA.
Meng, H,, Sexton, A.M., Maddox, M.C., Sood, A., Brown, C.W., Ferrara, R.R., and Murtugudde, R. 2010. Modeling
Rappahannock River Basin Using SWAT - Pilot for Chesapeake Bay Watershed. Applied Engineering in Agriculture,
26(5): 795.
Merritts, D. and Walter, R. 2003. Colonial Millponds of Lancaster County Pennsylvania as a Major Source of Sediment
Pollution to the Susquehanna River and Chesapeake Bay. Southeast Friends of the Pleistocene (SEFOP) Guidebook, Fall
2003 Fieldtrip.
Merritts, D,, Walter, R. and Rahnis, M.A. 2010. Sediment and Nutrient Loads from Stream Corridor Erosion along
Breached Millponds. Franklin and Marshall College.
I
-------
Miles, C, E, and Whitfield, T.G, compilers. 2001, Bedrock geology of Pennsylvania, Pennsylvania Geological Survey.
4th ser,, dataset, scale 1:250,000.
Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Bingner, R.L., Harmel, R.D, and Veith, T.L, 2007. Model evaluation guidelines
for systematic quantification of accuracy in watershed simulations. Transactions of the ASABE 50(3): 885-900.
Mu, Q,, Zhao, M, and Running, S.W. 2011, Improvements to a MODIS Global Terrestrial Evapotranspiration Algorithm.
Remote Sensing and Environment, 115:1781-1800.
Nash, J.E. and Sutcliffe, J.V. 1970. River flow forecasting through conceptual models: Part 1, A Discussion of Principles.
J. Hydrology 10(3): 282-290.
Neitsch, S.L., Arnold, J.G., Kiniry, J.R, and Williams, J.R. 2010. Soil and Water Assessment Tool, Input/Output File
Documentation, Version 2009. TR-365, Texas Water Resources Institute, College Station, TX.
Neitsch, S.L., Arnold, J.G., Kiniry, J.R. and Williams, J.R. 2011. Soil and Water Assessment Tool Theoretical
Documentation Version 2009. Texas Water Resources Institute Technical Report No. 406, College Station, TX.
Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Srinavasan, R. and Williams, J.R. 2005. Soil and Water Assessment Tool,
Theoretical Documentation, Version 2005. Grassland, Soil and Water Research Laboratory, Agricultural Research
Service, Temple, TX.
Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Srinavasan, R. and Williams, J.R. 2010. Soil and Water Assessment Tool, Input/
Output File Documentation, Version 2009. TR-365. Texas Water Resources Institute, College Station, TX.
Pennsylvania Department of Environmental Protection (PDEP). 2012. PA Final 2012-2013 Milestones.
http://www.portal.5tate.pa.us/portal/server.pt/communitv/chesapeake_b3y_program/10513.
Phillips, S.W., ed. 2007. Synthesis of U.S. Geological Survey Science for the Chesapeake Bay Ecosystem and Implications
for Environmental Management: U.S. Geological Survey Circular 1316. U.S. Geological Survey, Reston, VA.
PRISM Climate Group. Oregon State University, http://prisrii.oregofistate.edu
Risser, D.W., Conger, R.W., Ulrich, J.E. and Asmussen, M.P. 2005. Estimates of Ground-Water Recharge Based on
Streamflow-Hydrograph Methods. Pennsylvania: U.S. Geological Survey Open-File Report 2005-1333.
Rorabaugh, M.I. 1964. Estimating Changes in Bank Storage and Ground-Water Contribution to Streamflow.
International Association of Scientific Hydrology, 63: 432-441.
Sanford, W.E. and Selnick, D.L. 2013. Estimation of Evapotranspi ration across the Conterminous United States
using a Regression with Climate and Land-Cover Data. Journal of the American Water Resources Association.
49(1): 217-230.
Sexton, A.M., Sadeghi, A.M., Zhang, X., Srinivasan, R,, and Shirmohammadi, A. 2010. Using NEXRAD and Rain Gauge
Precipitation Data for Hydrologic Calibration of SWAT in a Northeastern Watershed. Transactions of the ASBAE, 53(5):
1501-1510.
Sloto, R.A. and Crouse, M.Y. 1996. HYSEP: A Computer Program for Streamflow Hydrograph Separation and Analysis.
U.S. Geological Survey Water-Resources Investigations Report 96-4040.
United States Environmental Protection Agency (USEPA). 2006. BASINS Technical Note 8: Sediment Parameter and
Calibration Guidance for HSPF. Office of Water 4305.
United States Environmental Protection Agency (USEPA). 2007. BASINS Technical Note 2: Two Automated Methods for
creating Hydraulic Function Tables (FTABLES). Office of Water 4305.
USGS 1984. Chesapeake Bay Watershed Land Cover Data Series 1984.
ft p: //ft p. ci h e s a p e a ke b ay. n et/Gis/ C B LC D_5 erie s/
USGS 1992. Chesapeake Bay Watershed Land Cover Data Series 1992.
ftp://ft p.chesapea kebay. net/Gis/CBLCD_5eries/
I
-------
USGS 2001, Chesapeake Bay Watershed Land Cover Data Series 2001,
ftp://ftp.chesapeakebay.net/Gis/CBLCD_Series/
USGS 2006. Chesapeake Bay Watershed Land Cover Data Series 2006.
ftp://ftp.chesapeakebay.net/C3is/CBLCD_Series/
Veith, T.L., Van Liew, M.W., Bosch, D.D., and Arnold, J.G, 2010. Parameter sensitivity and uncertainty in SWAT: A
comparison across five USDA-ARS watersheds. Transactions of the ASABE, 53(5): 1477-1485.
Waidler, D,, White, M,, Steglich, E,, Wang, S,, Williams, J., Jones, C.A. and Srinivasan, R. 2009. Conservation Practice
Modeling Guide for SWAT and APEX. Conservation Practice Modeling Guide.
Walter, R,, Merritts, D. and Rahnis, M. 2007. Estimating Volume, Nutrient Content, and Rates of Stream Bank Erosion of
Legacy Sediment in the Piedmont Valley and Ridge Physiographic Provinces, Southeastern and Central PA. A Report to
the Pennsylvania Department of Environmental Protection.
Walter, R.C., and Merritts, D.J. 2008. Natural Streams and the Legacy of Water-Powered Mills. Science, 319(5861): 299-
304.
Williams, J.R. 1975. Sediment Routing for Agricultural Watersheds. Journal of the American Water Resources
Association, 11(5): 965-974.
Williams, J.R., and Berndt, H.D. 1977. Sediment Yield Prediction Based on Watershed Hydrology. Transactions of the
American Society of Agricultural Engineers, 20(6).
Winchell, M,, Srinivasan, R,, Di Luzio, M. and Arnold, J. 2013. ArcSWAT Interface for SWAT2012 User's Guide.
Texas AgriLife Research and USDA Agricultural Research Service, Temple, TX.
Williams, J.R., Izaurralde, R.C. and Steglich, E.M. 2012. Agricultural Policy/Environmental Extender Model Theoretical
Documentation, Version 0806.
Photo Credits:
Pg. x - Big Conestoga Creek Bridge No. 12, Spanning Conestoga River at Farmersville Road State Route 1010,
Brownstown, Lancaster County, PA. Photograph by Jet Lowe. Historic American Engineering Record, National Park
Service, U.S. Department of the Interior. Retrieved from the Library of Congress.
Pg. 1 - Cascades on the Conestoga River in Lancaster County Central Park. Photograph by AppalachianViews/
Thinkstock.
Pg. 99 - USDA Agricultural Stabilization and Conservation Service Survey Photograph 1971, Conestoga, PA.
Pg. 121 - Ducks wading on wintry Conestoga River. Photograph by mrmikemtr26/Thinkstock.
Pg. 125 - Pennsylvania Railroad, Safe Harbor Bridge, Spanning mouth of Conestoga River, Safe Harbor, Lancaster
County, PA. Photograph by Jet Lowe. Historic American Engineering Record, National Park Service, U.S.
Department of the Interior Retrieved from the Library of Congress.
Back cover - USGS and Pennsylvania Department of Environmental Protection streamgage webcam, 01531500
Susquehanna River.
I
-------
The regression models used for the estimation of a continuous load time-series based on observed daily flow and
grab samples are of the form,
In C = a + b # In
where, a and b are regression coefficients, C is the concentration of the constituent and Q is the flow.
A breakpoint regression technique was used such that two regression models were used for load estimation below
and above a user defined breakpoint flow. The breakpoint flow was determined by visual inspection of observed
load and flow relationships. The breakpoint flow used for the regression models is 1922 cfs. The table below shows
the values of the regression coefficients for each constituent for each flow stratum.
Stratum
Coefficients TSS
TKN
NOx
TN
TP
1
a
-2.988
-1.127
1.810
1.872
-1.281
b
0,925
0.141
0.008
0.023
-0,067
2
a
-1.944
-0.931
5.094
3.933
-5.463
b
0.920
0.177
-0.446
-0.257
0,610
125
-------
SERA
United States
Environmental Protection
Agency
Office of Research arid Development (8101R)
1200 Pennsylvania Ave. NW
Washington, DC 20460
epa.gov/research
E PA/600/R-16/094
July 2016
------- |