xvEPA
United States
Environmental Protection
Agency
Office of Research
and Development
Washington, DC 20460
EPA/600/R-06/033
July 2006
BMP Modeling Concepts and
Simulation
-------
EPA/600/R-06/033
July 2006
BMP Modeling Concepts and Simulation
by
Wayne C. Huber
LaMarr Cannon and Matt Stouder
Oregon State University
Corvallis, Oregon 97331-2302
In support of:
EPA Contract No. 68-C-01-020
University of Colorado at Boulder
Project Officer
Dr. Fu-hsiung (Dennis) Lai
Water Supply and Water Resources Division
National Risk Management Research Laboratory
Edison, New Jersey 08837
National Risk Management Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, OH 45268
-------
Notice
The U.S. Environmental Protection Agency (EPA) through its Office of Research and Development performed and
managed the research described here. It has been subjected to the Agency's peer and administrative review and has
been approved for publication as an EPA document. Any opinions expressed in this report are those of the authors
and do not, necessarily, reflect the official positions and policies of the EPA. Any mention of products or trade names
does not constitute recommendation for use by the EPA.
-------
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the Nation's land, air, and
water resources. Under a mandate of national environmental laws, the Agency strives to formulate and implement
actions leading to a compatible balance between human activities and the ability of natural systems to support and
nurture life. To meet this mandate, EPA's research program is providing data and technical support for solving
environmental problems today and building a science knowledge base necessary to manage our ecological resources
wisely, understand how pollutants affect our health, and prevent or reduce environmental risks in the future.
The National Risk Management Research Laboratory (NRMRL) is the Agency's center for investigation of
technological and management approaches for preventing and reducing risks from pollution that threaten human
health and the environment. The focus of the Laboratory's research program is on methods and their cost-
effectiveness for prevention and control of pollution to air, land, water, and subsurface resources; protection of water
quality in public water systems; remediation of contaminated sites, sediments and ground water; prevention and
control of indoor air pollution; and restoration of ecosystems. NRMRL collaborates with both public and private
sector partners to foster technologies that reduce the cost of compliance and to anticipate emerging problems.
NRMRL's research provides solutions to environmental problems by: developing and promoting technologies that
protect and improve the environment; advancing scientific and engineering information to support regulatory and
policy decisions; and providing the technical support and information transfer to ensure implementation of
environmental regulations and strategies at the national, state, and community levels.
This document has been produced as part of the Laboratory's strategic long-term research plan. It is made available
by EPA's Office of Research and Development to assist the user community and to link researchers with their clients.
Sally Gutierrez, Director.
National Risk Management Research Laboratory
in
-------
Abstract
In order to minimize impacts of urban nonpoint source pollution and associated costs of control (storage and
treatment) associated with wet-weather flows (WWFs), storm water runoff volumes and pollutant loads must be
reduced. A number of control strategies and so-called "best management practices" (BMPs) are being used to
mitigate runoff volumes and associated nonpoint source (diffuse) pollution due to WWFs and include ponds,
bioretention facilities, infiltration trenches, grass swales, filter strips, dry wells, and cisterns. Another control option
is popularly termed "low impact development" (LID) - or hydrologic source control - and strives to retain a site's
pre-development hydrologic regime, reducing WWF and the associated nonpoint source pollution and treatment
needs.
Methodologies are needed to evaluate these BMPs, their effectiveness in attenuating flow and pollutants, and for
optimizing their cost/performance since most models only partially simulate BMP processes. Enhanced simulation
capabilities will help planners derive the least-cost combination for effectively treating WWFs. There is currently a
confusing array of options for analyzing hydrologic regimes and planning for LID. Integrating available BMP and
LID processes into one model is highly desirable.
This work analyzes several current modeling methods to evaluate BMP performance with the intention of facilitating
the integration of improved BMP modeling methods into the U.S. Environmental Protection Agency (EPA) Storm
Water Management Model (SWMM). Several other models are examined as part of this study. Options for
enhancement of SWMM's LID simulation capabilities are also presented. Two extensive case studies in Portland,
Oregon help to clarify current SWMM capabilities and needs for enhancement. The effort documented in this report
is linked to a parallel effort at the University of Colorado related to optimization strategies for WWF control.
IV
-------
Contents
Abstract iv
Contents v
List of Figures ix
List of Tables xi
List of Acronyms and Abbreviations xiii
Acknowledgments xvi
Project Publications xvii
1 Introduction 1-1
1.1 The Needs 1-1
1.2 Statement of Task 1-2
1.3 Fundamental Process Categories and Urban BMPs 1-3
1.4 Current BMP Approaches 1-7
1.5 Summary of Current SWMM BMP Simulation Capabilities 1-9
2 Study Area Options 2-12
2.1 Objectives 2-12
2.2 Portland, Oregon 2-12
2.3 Vallejo, California 2-12
2.4 Happy Acres 2-13
2.5 Wonderland Creek in Boulder, Colorado 2-13
2.6 Fair Oaks Estates in Carol Stream, Illinois 2-13
2.7 Summary 2-13
3 BMP Evaluation Options 3-14
3.1 BMP Performance Evaluation 3-14
3.2 Identification of Common EMC-Based Methods 3-15
3.3 Event Mean Concentration 3-16
3.4 Efficiency Ratio 3-16
3.5 Regression Models 3-17
3.6 Effluent Probability Method 3-18
3.7 Simulation Models 3-22
4 Current SWMM Simulation Capabilities 4-24
4.1 The Model 4-24
-------
4.2 Storage 4-24
4.3 S/T Generalized Removal Equation 4-25
4.4 Modified Streeter-Phelps Option 4-28
4.4.1 Objectives 4-28
4.4.2 Theory 4-29
4.4.3 Reaeration 4-33
4.4.4 Summary 4-33
4.5 Infiltration 4-33
4.6 Other Wet-Weather Control Options 4-34
4.7 LID Simulation Options 4-34
4.8 SWMM LID Modeling Needs 4-35
4.9 Time and Space Resolution Issues 4-36
4.10 Examples of SWMM BMP Simulation 4-36
5 Alternative Models and Approaches 5-37
6 Ponds 6-39
6.1 Introduction 6-39
6.2 Simulation of Ponds with P8 6-40
6.2.1 The Model 6-40
6.2.2 Second-Order Reactions 6-40
6.2.3 Particle Removal Scale Factor 6-42
6.2.4 Pollutant Removal 6-42
6.3 Simulation of Ponds with SLAMM 6-44
6.4 Pond Simulation with MUSIC 6-45
6.5 Pond Simulation with SWMM 6-45
6.6 Extended Detention 6-48
7 Wetlands and Bioretention Facilities 7-50
7.1 Introduction 7-50
7.2 Difficulties with Modeling Multiple Processes 7-51
7.3 Bioretention in WMM and P8 7-51
7.4 Simulation of Wetlands with the WETLAND Model 7-52
7.4.1 Introduction 7-52
7.4.2 WETLAND Cycles and Sub-models 7-52
7.4.3 WETLAND Output 7-54
7.5 Simulation of Wetlands with VAFSWM 7-55
7.5.1 Introduction 7-55
7.5.2 VAFSWM Components 7-55
7.5.3 Implications for SWMM Improvements 7-55
7.6 Simulation of Wetlands with PREWET 7-56
7.6.1 Introduction 7-56
7.6.2 PREWET Removal Mechanisms 7-56
7.6.3 PREWET Usefulness for Urban BMP Evaluation 7-57
7.7 Simulation of Wetlands with DMSTA 7-57
7.7.1 Introduction 7-57
7.7.2 DMSTA Model Features 7-58
7.7.3 DMSTA Phosphorous Cycling Model 7-58
7.7.4 DMSTA Usefulness for Urban BMP Evaluation 7-59
7.8 Simulation of Wetlands and Other BMPs with MUSIC 7-59
7.8.1 Introduction 7-59
7.8.2 MUSIC Algorithms 7-59
7.8.3 MUSIC Evaluation 7-64
7.9 SWMM Simulation of Wetlands and Bioretention Devices 7-65
VI
-------
8 Infiltration Trenches 8-66
8.1 Introduction 8-66
8.2 Simulation of Infiltration Trenches with SLAMM 8-66
8.2.1 Introduction 8-66
8.2.2 SLAMM Calculation Procedures for Infiltration Devices 8-66
8.2.3 Infiltration in Disturbed Urban Soils 8-68
8.2.4 SLAMM Procedures for SWMM 8-70
8.3 Simulation of Infiltration Trenches in SWMM 8-71
8.4 Transition to Simulation of Rain Gardens and Green Roofs 8-72
9 Grass Swales and Filter Strips 9-73
9.1 Introduction 9-73
9.2 Simulation of Grass Swales with P8 9-73
9.3 Grass Swale Performance Calculations in SLAMM 9-74
9.4 Simulation of Vegetated Filter Strips with REMM 9-75
9.5 Simulation of Grass Swales with SWMM 9-76
10 Dry Wells 10-78
10.1 Introduction 10-78
10.2 Simulation of Dry Wells with SWMM 10-78
11 Cisterns 11-79
11.1 Introduction 11-79
11.2 Simulation of Cisterns within SWMM 11-79
12 Porous pavement 12-80
12.1 Introduction 12-80
12.2 SLAMM Calculation Procedures for Porous Pavements 12-81
12.3 Simulation of Porous Pavement with SWMM 12-81
13 Hydrodynamic Devices 13-83
13.1 Introduction 13-83
13.2 Simulation of Hydrodynamic Devices with P8 13-83
13.3 Simulation of Hydrodynmic Devices with SWMM 13-83
14 Case Study: LID Simulation In Portland 14-85
14.1 Objectives 14-85
14.2 Portland Combined Sewer Study Area 14-85
14.3 Data Preparation Methods 14-88
14.3.1 Required Parameters 14-88
14.3.2 Directly Connected, orDCIA Subcatchments 14-88
14.3.3 Surface Water Subcatchments 14-91
14.4 The Models 14-92
14.4.1 Two Model Types 14-92
14.4.2 Width and Slope 14-97
14.4.3 A-Model Input Data 14-98
14.5 Modeling Results 14-98
14.6 LID Simulation 14-100
14.7 Summary and Conclusions 14-102
15 Case Study: BMP Simulation in Portland 15-103
15.1 Objectives 15-103
15.2 Lexington Hills Area Background 15-103
15.3 SWMM Modeling 15-109
15.3.1 Background Information 15-109
15.3.2 Initial Modeling 15-116
15.3.3 Quantity Modeling Results 15-117
15.3.4 Quality Simulations 15-118
vn
-------
15.4 Summary 15-120
16 Recommendations for SWMM BMP Modeling Improvements 16-121
16.1 Introduction 16-121
16.2 Ponds 16-122
16.3 Wetlands and Bioretention Facilities 16-124
16.4 Infiltration Trenches 16-124
16.5 Grass Swales 16-125
16.6 Dry Wells 16-125
16.7 Cisterns 16-125
16.8 Porous Pavement 16-126
16.9Hydrodynamic Devices 16-126
16.10 LID and Other Related Needs 16-126
16.11 Final Summary of SWMM BMP Simulation Needs 16-127
17 Conclusions 17-130
Appendix: Using REMMto Predict Riparian Buffer Performance A-132
Riparian Ecosystem Management Model (REMM) A-132
Buffer Hydrology A-133
Erosion and Sediment A-134
Nutrient Dynamics A-134
Recommendations A-135
References R-136
Vlll
-------
List of Figures
Figure 3-1. Probability plot for TSS removal at the Seton Pond facility, Austin, Texas
(Brown 2003) 3-19
Figure 3-2. Probability plot for nitrate removal at the Seton Pond facility, Austin,
Texas (Brown 2003) 3-20
Figure 3-3. Probability plot for total zinc removal at the Seton Pond facility, Austin,
Texas (Brown 2003) 3-21
Figure 4-1. Conceptual routing from the impervious sub-area of a subcatchment to the
pervious sub-area of a subcatchment (Huber 200 la) 4-35
Figure 6-1. Removal efficiency, R, as a function of dimensionless forms of rate of treatment
or loading 6-48
Figure 7-1. Relationship between the MAIN CODE and respective sub-models for an entire
simulation run (Lee et al. 2002) 7-53
Figure 7-2. Theoretical residence time distribution (Equation 7-7) for unit impulse to N
tanks in series 7-61
Figure 7-3. Pond shapes simulated by Perssonetal. (1999) 7-62
Figure 8-1. Schematic of SLAMM model area breakdown 8-67
Figure 8-2. 3-D plots showing interactions affecting infiltration rates in sandy soils
(Pitt and Voorhees 2000) 8-70
Figure 8-3. 3-D plots showing interactions affecting infiltration rates in clayey soils
(Pitt and Voorhees 2000) 8-71
Figure 9-1. TSS removal effectiveness of vegetated filter strips, based on total mass of suspended
solids entering and leaving the strip (Huber et al. 2000) 9-77
Figure 14-1. Sullivan area (Carollo Engineers, 1999) 14-86
Figure 14-2. Combined sewer study area, south of Grant Park (Carollo Engineers 1999).
See color codes in caption to Figure 14-1 14-87
Figure 14-3. Aerial photo of combined sewer study area 14-87
Figure 14-4. ArcView map of study area showing individual house parcels and rooftop
imperviousness 14-90
Figure 14-5. Example slope and aspect grids from the digital terrain model 14-91
Figure 14-6. Pipe segment ID for the study area, as simulated in Extran 14-93
IX
-------
Figure 14-7. Aggregated areas for three street subcatchment includes roads, sidewalks and
grass strips 14-93
Figure 14-8. The highlighted sub-basin 28349056 (referred to as 56) has four DCIA
Subcatchments 14-94
Figure 14-9. Individual parcels for the disaggregated model 14-94
Figure 14-10. Close-up of parcels draining to pipe 62 14-95
Figure 14-11. Dark areas (blue) are DCIA and are surrounded by lighter (lavender)
pervious areas 14-96
Figure 14-12. The dark (blue) DCIA connected outside the sub-basin is not included in the
model as the runoff effects are not noticed at the monitor 14-96
Figure 14-13. Parcel land surface definitions 14-97
Figure 14-14. Five-day comparison of simulated and measured flows at the monitoring site 14-99
Figure 14-15. Simulated and measured flows for seven-hour period on January 17, 1999 14-99
Figure 14-16. Comparison between Model I and Model I-LID simulations for a seven-hour
interval, January 17, 1999 14-101
Figure 15-1. Location map for Lexington Hills BMP site (Liptan 2001) 15-104
Figure 15-2. Lexington Hills pond vicinity photo (Liptan 2001) 15-105
Figure 15-3. Lexington Hills pond site photo (Liptan 2001). Pond 3 is to left of intersection 15-105
Figure 15-4. Lexington Hills BMP site in relation to catchment (Liptan 2001) 15-106
Figure 15-5. Details of Lexington Hills extended detention Pond 3 (Liptan 2001) 15-107
Figure 15-6. Location of influent (Site 1) and effluent (Site 2) monitoring (Liptan 2001) 15-108
Figure 15-7. Regional subcatchments used in BBS modeling of Johnson Creek Watershed 15-110
Figure 15-8. Localized view of Johnson Creek Watershed subcatchments 15-111
Figure 15-9. General map of BBS raingage network for Johnson Creek near Lexington Hills
(Lexington Hills is between the Holgate and Pleasant Valley School gages) 15-112
Figure 15-10. Comparison of recorded and simulated flows vs. rainfall for the five simulated
events 15-117
-------
List of Tables
Table 1-1. Needs for publicly owned wastewater treatment facilities and other eligibilities, January
1996 dollars 1-2
Table 1-2. Structural BMPs categorized by fundamental unit processes 1-4
Table 1-2 (concluded) 1-5
Table 1-3. Representative structural stormwater BMPs 1-6
Table 1-3. (concluded) 1-7
Table 1-4. Wet-weather controls and SWMM suitability (after Huber2001b) 1-10
Table 3-1. Primary urban hydrologic, hydraulic and water quality models commonly used in the
United States 3-23
Table 4-1. Symbols and parameters for linked DO-BOD-NOD simulation, Equations 4-9-4-12 4-30
Table 5.1. Simulation models that can simulate urban BMP performance 5-38
Table 6-1. Particle class default values in P8 (Walker 1990) 6-43
Table 6-2. Example of P8 calibrated runoff concentrations (Walker 1990) 6-44
Table 7-1 Wetlands/ponds pollutant removal mechanisms 7-51
Table 7-2. Numerical results of Perssonetal. (1999) for pond shapes of Figure 7-3 7-62
Table 7-3 Calibrated k' and C* values from MUSIC based on limited model simulations 7-64
Table 7-4. First-order decay values converted from k' values for TSS in Table 7-3 for assumed depths 7-64
Table 8-1. Ranked double ring infiltration test results and observed urban soil infiltration rates from
Oconomowoc, WI (Pitt and Voorhees 2000) 8-69
Table 8-2. Categories tested for infiltration rates (Pitt and Voorhees 2000) 8-69
Table 8-3. Percolation rates for different soil texture and moisture used in SLAMM (Pitt and
Voorhees 2000) 8-69
Table 9-1 Infiltration rates used in P8 9-74
Table 14-1. Constant model parameters 14-88
Table 14-2. Extran Block conduit input data. Conduits are identified in Figure 14-6 14-89
Table 14-3. Subcatchment input data for aggregated models (A-Models) 14-98
Table 14-4. Model total flow comparison for five-day event, Jan 13-Jan 18, 1999 14-100
Table 15-1. Storm events used in SWMM simulations 15-104
Table 15-2. Runoff Block input for Lexington Hills Pond 3 simulation (Stouder 2003) 15-113
Table 15-3. Rating curve development for Pond 3 at Lexington Hills 15-116
XI
-------
Table 15-4. SWMM simulated inflows and outflows versus measured data for Pond 3 at the
Lexington Hills BMP site 15-117
Table 15-5. Measured and simulated quality results for Lexington Hills Pond 3 15-118
Table 15-6. Particle size distribution for event of October 9-10, 2000 15-119
Table 16-1. Clarification of method applicability to modeling BMPs 16-123
Table 16-2. Summary of proposed SWMM BMP/LID simulation enhancements 16-128
Table 16-2. (Continued) 16-129
Table A-l. REMM wet-weather controls program suitability 133
xn
-------
List of Acronyms and Abbreviations
AGNPS Agricultural Nonpoint Source Model
AMC antecedent moisture condition
APWA American Public Works Association
ASCE American Society of Civil Engineers
BES Bureau of Environmental Services
BMP best management practice
BOD biochemical oxygen demand
BODU ultimate carbonaceous oxygen demand
C carbon
CalTrans California Department of Transportation
CDF cumulative distribution function
cfs cubic feet per second
CBOD carbonaceous biochemical oxygen demand
CFSTR continuous-flow, stirred-tank reactor
COD chemical oxygen demand
CRCCH Cooperative Research Centre for Catchment Hydrology
CSO combined sewer overflow
CU University of Colorado
CV coefficient of variation
DCIA directly connected impervious area
DMSTA Dynamic Model for Stormwater Treatment Areas
DO dissolved oxygen
DOC dissolved organic carbon
DON dissolved organic nitrogen
DP dissolved phosphorus
DTM digital terrain model
EPA U.S. Environmental Protection Agency
ED extended detention
EMC event mean concentration
EPM effluent probability method
ER efficiency ratio
Xlll
-------
ET evapotranspiration
FPC fundamental process category
FWS free water surface
G-A Green-Ampt
GIS geographic information system
GUI graphical user interface
HRT hydraulic residence time
HSPF Hydrologic Simulation Program - Fortran
LID low-impact development
MS4 municipal separate storm sewer system
MSL mean sea level
MUSIC Model for Urban Stormwater Improvement Conceptualization
N nitrogen
NCOB nitrogen, carbon, DO, bacteria (in WETLAND model)
NO3-N nitrogen as nitrate
NH3 ammonia
NOD nitrogenous oxygen demand
NRCS Natural Resources Conservation Service
NRMRL National Risk Management Research Laboratory
NURP Nationwide Urban Runoff Program
O&M operation and maintenance
OSU Oregon State University
P phosphorus
P8 Program for Predicting Polluted Particle Passage through Pits, Ponds and Puddles
PCSWMM Personal Computer Storm Water Management Model
PFR plug flow reactor
POC particulate organic carbon
PPCC probability plot correlation coefficient
PREWET Pollutant Removal Estimates for Wetlands
PVC polyvinyl chloride
REMM Riparian Ecosystem Management Model
RTC real time control
RTD residence time distribution
S-I storage-indication
SCS Soil Conservation Service
SLAMM Source Loading and Management Model
SS suspended solids
SSF sub-surface flow
SOD sediment oxygen demand
SSO sanitary sewer overflow
S/T storage/treatment
STA stormwater treatment area
SW surface water
SWMM Storm Water Management Model
TDS total dissolved solids
TIS tanks in series
TKN total Kjeldahl nitrogen
TMDL total maximum daily load
TN total nitrogen
TOC total organic carbon
TP total phosphorus
xiv
-------
TSS total suspended solids
USDA United States Department of Agriculture
USGS United States Geological Survey
USLE Universal Soil Loss Equation
USTM Universal Stormwater Treatment Model
UWRRC Urban Water Resources Research Council
VAFSWM Virginia Field Scale Wetland Model Program
WASP Water Quality Analysis Simulation Program
WEF Water Environment Federation
WETLAND Wetland Water Balance and Nutrient Dynamics Model
WMM Watershed Management Model
WWC wet-weather control
WWF wet-weather flow
xv
-------
Acknowledgments
The author acknowledges the assistance of EPA Contract No. 68-C-01-020 in support of this research and the
assistance of Project Officer Dr. Fu-hsiung (Dennis) Lai. The opinions expressed herein do not necessarily reflect
those of the EPA.
Ms. LaMarr Clannon (who changed her name upon matrimony), a graduate research assistant employed on this
project, performed most of the Portland LID modeling work described in this report and initiated the comparative
model evaluations. Another graduate research assistant, Mr. Matt Stouder, performed most of the Portland BMP
modeling work and provided additional evaluation of alternative models. The Oregon State University team is
grateful to Dr. James Heaney and his colleagues at the University of Colorado for their partnership in this EPA
project. Chapter 2 was written in cooperation with this group. The report content was improved by the comments of
two external reviewers.
The project team is indebted to Mr. Joe Hoffman of the Portland Bureau of Environmental Services (BES) for his
considerable help in providing and explaining the study area data. Information on the subcatchments in Johnson
Creek as well as subcatchment data for the Lexington Hills BMP area came from Mr. Tom Liptan and Mr. Tim Kurtz
at the BES. Mr. Kurtz provided much additional help, and many thanks are due to him for his help in this regard.
During the course of this EPA study, the OSU investigators also participated in National Cooperative Highway
Research Program Project 25-20(01) related to evaluation of BMPs for highway applications and in Water
Environment Research Foundation Project 02-SW-l related to more general evaluation of urban BMPs. Efforts on
both of these projects contributed to an improved understanding of BMP evaluation, especially that it related to
analysis of event mean concentration data.
xvi
-------
Project Publications
Publications resulting wholly or in part from Task B project activities (documented in this report) are listed by authors
below:
Cannon, L. (2002). Urban BMPs and their Modeling Formulations, Master of Science Project Report, Dept. of Civil,
Construction, and Environmental Engineering, Oregon State University, Corvallis, August.
Huber, W.C. (2001). Wet-weather Treatment Process Simulation Using SWMM. Proc. Third International
Conference on Watershed Management. National Taiwan University, Taipei, Taiwan, pp.253-264.
Huber, W.C. and Cannon, L. (2002). "Modeling Non-Directly Connected Impervious Areas in Dense
Neighborhoods," In Global Solutions for Urban Drainage, Proc. Ninth International Conference on Urban Drainage,
E.W. Strecker and W.C. Huber, eds., Portland, OR, American Society of Civil Engineers, Reston, VA, CD-ROM,
September (2002b).
Huber, W.C., Lai, F., Clannon, L. and Stouder, M. (2004). "Modeling Concepts for BMP/LID Simulation," Best
Management Practices (BMP) Technology Symposium: Current and Future Directions, World Water and
Environmental Resources Congress, Salt Lake, City, UT, Environmental and Water Resources Institute, American
Society of Civil Engineers, Reston, VA, June, 11 pp.
Stouder, M. (2003). Simulation Methods for Wetland/Pond BMPs, Master of Science Project Report, Dept. of Civil,
Construction, and Environmental Engineering, Oregon State University, Corvallis, June.
XVII
-------
INTRODUCTION
1.1 THE NEEDS
Pollution problems stemming from combined sewer overflows (CSOs), sanitary sewer overflows (SSOs),
and stormwater discharges are extensive throughout the nation, with the Northeast, Midwest, and Far
West being the principal areas of concentration. Nationwide, approximately 1,100 municipalities have
combined sewers, 85% of which are in eleven states serving 43 million people; there are over 15,000
overflow points within these systems. SSOs occur in more than 1,000 municipalities, and stormwater
discharges occur in as many as 1.2 million municipal, industrial, commercial, institutional and retail
sources (EPANRMRL 1996).
National cost estimates have been developed to control contamination from these three sources of wet
weather flow (WWF). According to the most recent 1996 EPA Clean Water Needs Survey
(http://www.epa.gov/owmitnet/mtb/cwns/1996rtc/toc.htm), projected costs for CSO pollution abatement
totaled $44.7 billion, and stormwater contributes $7.4 billion out of total clean water needs estimate of
$139.5 billion (Table 1-1). SSO costs are included in categories I, III and IV in the table and "EPA
believes that the needs estimates in these categories related to SSOs underestimate the total costs
associated with preventing SSOs" (http://www.epa.gov/owmitnet/mtb/cwns/1996rtc/toc.htm). Indeed the
EPA Research Plan for wet-weather flows (EPA NRMRL 1996) estimates SSO costs in the "tens of
billions." The document also cites an American Public Works Association (APWA) study indicating that
costs of controlling stormwater pollution are much higher, at more than $400 billion for capital
investment and capitalized costs of $540 billion for operation and maintenance (O&M) of stormwater
control facilities, in order to meet water quality standards for stormwater discharges. These large capital
and O&M costs pose severe financial difficulties to cities and municipalities throughout the nation.
During wet weather periods, urban sewer systems often become overloaded. This can be alleviated, for
example, by providing additional storage in the system or by providing additional treatment, either
downstream at the treatment facility or upstream within the watershed. Since WWF impacts and controls
are complex and the costs of abatement alternatives are enormously large, there are tremendous
opportunities for significant cost savings as a whole if an "optimal" cost-effective combination of storage
and treatment alternatives can be objectively formulated. EPA is developing a multi-year research
program aimed at devising tools that can be used to evaluate sewerage systems and determine the optimal
combination of WWF control alternatives for the most cost-effective operation of the system.
A related need deals with EPA's total maximum daily load (TMDL) program, by which pollutant loads
are to be reduced on a watershed basis, with a goal to meeting water quality standards in receiving water
1-1
-------
systems. Management of stormwater quality is usually performed through a combination of so-called
"best management practices" (BMPs) and a form of hydrologic source control popularly known as "low-
impact development" (LID). Reliable simulation of BMPs and LIDs is needed for the modeling efforts
that usually are a part of TMDL development.
Table 1-1. Needs for publicly owned wastewater treatment facilities and other eligibilities, January
1996 dollars, (http://www.epa.gov/owmitnet/mtb/cwns/1996rtc/toc.htm)
Needs Category
Title II Eligible Projects
I
II
IIIA
IIIB
IVA
IVB
V
VI
Secondary Treatment
Advanced Treatment
Infiltration/Inflow Correction
Sewer Replacement/Rehabilitation
New Collector Sewers
New Interceptor Sewers
Combined Sewer Overflows
Stormwater*
Total Categories I-VI
Other Eligible Projects (Sections 319 and 320)
VIIA-C
VIID
VIIE-G
Nonpoint Source (agriculture and silviculture only)*
Urban Runoff
Ground Water, Estuaries, Wetlands
Total Category VII
Grand Total
Total Needs,
Billion $
26.5
17.5
3.3
7.0
10.8
10.8
44.7
7.4
128.0
9.4
1.0
1.1
11.5
139.5
*Modeled needs only. Estimated Category VI needs documented by the States are $3.2 billion.
Estimated Category VIIA-C needs documented by the States are $0.5 billion.
Costs for operation and maintenance are not eligible for federal funding and therefore are not included.
Municipal separate storm sewer system (MS4) owners and operators need to identify effective BMPs for
improving stormwater runoff water quality; owners and managers of other highly-impervious land (e.g.,
highways and industries) have similar needs. Evaluation of performance involves the use of computer
models and tools, and empirical relationships describing a quantitative estimate of pollution removed by
BMPs. This information will help planners derive an effective combination of control strategies for
WWFs. Because of the current state of the practice, however, very few sound scientific data are available
for making decisions about which structural and non-structural management practices function most
effectively under what conditions, and, within a specific category of BMPs, to what degree design and
environmental variables affect BMP efficiency.
1.2 STATEMENT OF TASK
Models for simulation of BMP and LID options must consider antecedent runoff conditions for both a
single rainfall event and continuous rainfall and simulate the physical processes of rainfall,
evapotranspiration, soil infiltration to groundwater, and overland sheet-flow processes from individual
lots to streets, then from streets to sewers and eventually to their outfalls. The EPA Stormwater
Management Model (SWMM) is a state-of-the-art urban runoff process model and may be used as the
process model of choice (Huber and Dickinson 1988, Roesner et al. 1988). It uses well-known
hydrologic and hydraulic concepts to simulate the urban watershed. Moreover, the software itself has
undergone an evolutionary upgrade to a user-friendly, object-oriented version called SWMM5
(http://www.epa.gov/ednnrmrl/models/swmm/index.htm). This program, which includes a graphical user
1-2
-------
interface (GUI), was prepared within the EPA itself and provides the framework for enhancements
necessary to better model WWF control alternatives, especially BMP and LID options. SWMM5
enhancement needs are addressed herein through a goal of better tools for quantifying the effectiveness of
various WWF control alternatives. The effectiveness of each control alternative constitutes essential
input to analysis of management alternatives.
This report is half of a larger project, "Optimization of Urban Sewer Systems during Wet-Weather
Periods" (EPA Contract No. 68-C-01-020), to the University of Colorado (CU) and a subcontract to
Oregon State University (OSU) dealing with the related issues of optimization and simulation of wet-
weather controls. Simulation can provide performance estimates of controls such as BMPs and LIDs that
can be coupled to analytical tools to optimize life-cycle costs of capital and O&M investments as well as
overall performance, e.g., as measured by effluent characteristics, amount of runoff treated, etc. The issue
of performance measures will be discussed in detail in subsequent chapters.
A companion report (Heaney and Lee 2005) deals with the optimization efforts. This report deals with
the simulation component of the project. It is based in part upon unpublished progress reports during the
course of the project as well as OSU master's project reports by Cannon (2002) and Stouder (2003).
1.3 FUNDAMENTAL PROCESS CATEGORIES AND URBAN BMPS
A fundamental process is essentially the same as a unit process in environmental engineering (Metcalf
and Eddy 2003). There is no single universal list of fundamental process categories (FPCs); Minton
(2002) provides a useful taxonomy, oriented toward stormwater treatment. Common structural BMPs are
grouped by nine FPCs in Table 1-2, but alternatives are possible, and one BMP may fall under different
FPCs.
Similarly, there are many types of BMPs. Modelers interested in describing BMP effectiveness may wish
to model BMP types (typical swale or pond), or instead model the fundamental processes that occur in a
BMP (sedimentation, infiltration, etc.). For purposes of this investigation, representative structural
stormwater BMPs are listed in Table 1-3, adapted from a taxonomy prepared by the Minnesota Pollution
Control Agency (http://www.pca.state.mn.us/water/pubs/sw-bmpmanual.html). This list is one of many
from similar sources that could be used.
If modeling BMP by type, a method for determining performance parameters must be described. This
may be done by investigating BMP effectiveness as reported, for instance, in the EPA/ASCE BMP
Database (http://www.bmpdatabase.org/) and applying similar effectiveness measures to the model. This
leads to many questions about measuring and reporting BMP performance. Modeling BMPs by
simulation of FPCs requires extensive data on stormwater treatability as well as site and design
descriptions, and raises issues about data availability. As will be described in this report, most BMPs are
modeled by a heuristic combination of FPC simulation and empirical performance measures.
1-3
-------
Table 1-2. Structural BMPs categorized by fundamental unit processes. Adapted from: BMP manual
http://www.pca.state.mn.us/water/pubs/swm
Process
Definition
BMP
Sedimentation
Gravitational settling of suspended particles
from the water column. It can be a major
mechanism of pollutant removal in BMPs
because pollutants often adsorb to particulate
matter, especially clay or organic soils.
Detention systems intercept runoff for gradual
release later. Most are designed to empty
between events and treat water quantity rather
than quality. These systems can provide
limited settling, which can often be
resuspended with a subsequent event.
Dry pond, wet pond,
other basin, small storage
devices, wetland,
underground pipes,
vaults, tanks
Flotation
Separation of particulates with a specific
gravity less than water. Trash, Styrofoam, oil
and hydrocarbons can be removed from BMPs
designed with an area for these to accumulate.
Oil-water separators,
density separators,
dissolved-air flotation
Filtration
Filtration devices remove particulates by
passing water through a porous medium like
sand, gravel, soil, peat, compost, or
combinations thereof.
Trash racks, bar racks,
screens, sand filters,
compost filters,
vegetation and soil, may
be part of filtration
systems. Modular or
drop in filter systems
Infiltration
Infiltration systems capture runoff and provide
a means of infiltration into the ground.
Infiltration is the most effective means of
controlling storm water runoff because it
reduces the volume discharged to receiving
waters. Filtration by soil removes TSS and
associated pollutants, and dissolved nutrients
are removed by adsorption.
Infiltration basins, porous
pavement, infiltration
trenches or wells, ponds,
constructed wetlands
Infiltration systems with
clay soils, adsorption to
macrophytes in
constructed wetlands,
compost filters
Adsorption
Contaminants are bound (to clay particles or
macrophytic vegetation). Adsorption is not a
common mechanism used in stormwater
BMPs, although it can occur in infiltration
systems with clayey soils, in organic filters, or
in wetland systems.
1-4
-------
Table 1-2 (concluded)
Biological Uptake and
Conversion
Biological uptake is an important nutrient
control mechanism in BMPs treating urban
runoff that typically contains high levels of
nutrients. This occurs as aquatic plants and
microorganisms utilize nutrients for growth.
Maintenance and harvest of these plants is
essential, otherwise as plants die and decay,
they re-release nutrients. Biological
conversion happens when microorganisms and
bacteria break down organic contaminants into
less harmful compounds.
Pond, bioswale, wetland
Chemical Treatment
Chemicals such as alum are added to promote
flocculation and settling. Chlorine disinfection
is sometimes used to treat combined sewer
overflows.
Precipitation,
flocculation, disinfection
Degradation (volatilization,
hydrolysis, photolysis)
Degradation happens in open pool BMPs
where contaminants volatilize, hydrolyze or
photolyze.
Pond, wetland, open pool
BMPs
Hydrodynamic Separation
Varies by device.
Swirl concentrators,
secondary current
devices, oil-water
separators
Combination (Retention)
Retention systems hold and treat runoff until
displaced by another volume of water. These
can be very effective in treating both quality
and quantity of runoff. Several processes can
occur, such as sedimentation, filtration,
infiltration, biological uptake, conversion and
degradation.
Wetlands, bioswales
1-5
-------
Table 1-3. Representative structural stormwater BMPs.*
BMP
Definition
Bioretention Facilities
(constructed wetlands,
wetland basins and
wetland channels)
Similar to other facilities (pond, basin or channel) with more than 50% of its
surface or bottom covered by emergent wetland vegetation. A wetland channel
is a channel designed to flow very slowly, probably less than 2 ft/s at the 2-year
flood peak flow rate. It has, or is designed to develop, dense wetland vegetation
on its bottom.
Dry Wells
Dry wells are drilled often through impervious layers to reach lower pervious
layers, filled with porous media designed to percolate surface water to
groundwater.
Filter Strips
Grass filter strips, sometimes called bio-filters, bioswales, or buffer strips, are
vegetated areas designed to accept sheet flow provided by flow spreaders, which
accept flow from an upstream development. Vegetation may take the form of
grasses, meadows, forests, etc. The primary mechanisms for pollutant removal
are filtration, infiltration, and settling.
Grass Swales
A swale, sometimes also called a bio-filter or bioswale, is a shallow grass-lined
channel with little bottom width, designed for shallow flow near the source of
storm runoff.
Ponds
Retention ponds are also commonly known as "wet ponds" because they have a
permanent pool of water, unlike detention basins, which dry out between storms.
The permanent pool of water is replaced in part or in total by stormwater during
a storm event. The design is such that any runoff captured during a storm event
is released over time. The hydraulic residence time (HRT) for the permanent
pool over time can provide biological treatment. A dry-weather base flow, pond
liner and/or high groundwater table are required to maintain the permanent pool.
Cisterns
Rain barrels or cisterns act as storage devices and retain a portion of runoff for
later use, such as irrigation or use as grey water.
Infiltration Trenches
Percolation or infiltration trenches can generally be described as ditches filled
with porous media designed to encourage rapid percolation of runoff to the
groundwater. An infiltration basin can capture a given stormwater runoff
volume and infiltrate it into the ground, transferring this volume from surface
flow to groundwater flow.
Extended Detention
(ED)
ED dry basins are designed to completely empty at some time after stormwater
runoff ends. These are adaptations of the detention basins used for flood
control. The primary difference is in outlet design; the extended detention basin
uses a much smaller outlet that extends the detention time for more frequent
events in order to facilitate pollutant removal. The term "dry" implies that there
is no significant permanent water pool between storm runoff events. Multiple
uses (e.g., recreation) are possible for land occupied by ED facilities.
1-6
-------
Porous Pavement
Table 1-3. (concluded)
Modular-block porous pavement is composed of perforated concrete slab units
underlain with gravel. The surface perforations are filled with coarse sand or
sandy turf. It is used in low traffic areas to accommodate vehicles while
facilitating storm water runoff at the source. The units should be placed in a
concrete grid which restricts horizontal movement of infiltrated water through
the underlying gravels.
Poured-in-place porous concrete or asphalt is generally placed over a substantial
layer of granular base. The pavement is similar to conventional materials, except
for the elimination of sand and fines from the mix. If infiltration to ground water
is not desired, a liner may be used below the porous media along with a
perforated pipe and a flow regulator to slowly drain the water stored in the media
over a 6 to 12 hour period.
lydrodynamic Device
These devices are BMPs such as oil-water separators, sand interceptors, swirl-
type concentrators, sedimentation vaults, and other prefabricated and package-
type treatment devices.
*BMP definitions are from the ASCE BMP database: website, www.bmpdatabase.org
1.4 CURRENT BMP APPROACHES
Traditional stormwater controls have strongly emphasized large spatial scales for flood and water-quality
analyses. Runoff volume is usually the most important hydrologic parameter in water quality studies,
while peak flow rate and time of concentration are usually the most important hydrologic parameters for
flooding and drainage studies. The relationships between these different hydrologic parameters and storm
parameters are significantly different for different classes of rains. Runoff models for water quality
investigations should therefore have additional capabilities beyond those of runoff models for flooding
and drainage investigations.
Common, small storms (also termed "micro-storms" in the literature) are responsible for most of the
annual urban runoff discharge quantities throughout North America (Heaney et al. 1977, EPA 1983, Pitt
1987, WEF and ASCE 1998, Wright et al. 2000). However, some existing urban runoff models originate
from drainage and flooding evaluation procedures that emphasize very large design storms (several inches
in depth). These large storms only contribute small portions of the annual average discharges. Several
authors have suggested that stormwater quality can be managed by treating the flows associated with
small storms (Heaney et al. 1977, EPA 1983, WEF and ASCE 1998, Wright et al. 2000, Pitt and
Voorhees 2000, Sample et al. 2001). These storms occur many times a year and are responsible for the
majority of the pollutant discharges. Runoff from about 70 to 80% comprises these frequent discharges
of the annual precipitation onto urban areas, the effects of which are mostly chronic in nature (such as
contaminated sediment and frequent high flow rates). Pitt and Voorhees (2000) use the breakdown
bulleted below for various storms. It should be noted that relationships between depth and cumulative
percentages vary regionally (Heaney et al. 1977, WEF and ASCE 1998).
Frequent storms having relatively low pollutant discharges are associated with depths of less than 0.5
in. (12 mm). These are key events in which runoff-associated water quality violations, such as for
bacteria, are of concern. In most areas, runoff from these rain storms should be totally captured and either
1-7
-------
re-used for on-site beneficial uses or infiltrated in upland areas. For most areas, the runoff from these
rains is relatively easy to remove from the surface drainage system.
Rains between 0.5 and 1.5 in. (12 and 38 mm) are responsible for about 75% of the nonpoint source
pollutant discharges and are key rains in terms of addressing mass pollutant discharges. The small rains
in this category can also be removed from the drainage system and the runoff re-used on site for
beneficial means or infiltrated to replenish the lost groundwater infiltration associated with urbanization.
The runoff from the larger rains should be treated to prevent pollutant discharges from entering the
receiving waters.
Rains greater than 1.5 in. (38 mm) are associated with drainage design and are only responsible for
relatively small portions of the annual pollutant discharges. Typical storm drainage design events fall in
the upper portion of this category. Extensive pollution control designed for these events would be very
costly, especially considering the relatively small portion of the annual runoff associated with the events.
However, discharge rate reductions are important to reduce habitat problems in the receiving waters. The
infiltration and other treatment controls used to handle the smaller storms in the above categories would
have some benefit in reducing pollutant discharges during these larger, rarer storms.
In addition, extremely large rains also occur infrequently that exceed the capacity of the drainage
system and cause local flooding. Two of these extreme events were monitored in Milwaukee during the
Nationwide Urban Runoff Program, NURP (EPA 1983). Such storms, while very destructive, are
sufficiently rare that the resulting environmental problems do not justify the massive stormwater quality
controls that would be necessary for their reduction. The problem during these events is massive property
damage and possible loss of life. These rains typically greatly exceed the capacities of the storm drainage
systems, causing extensive flooding.
This interest in smaller storms requires modeling approaches that can properly account for losses
(hydrologic abstractions such as infiltration, depression storage, and evapotranspiration) for low rainfall
depths. SWMM is one such model with this capability. However, as noted elsewhere in this report, even
though SWMM provides the option for vadose zone simulation, interaction with subsurface hydrologic
processes could be improved with respect to the overall water balance. Alternatively, better opportunities
to interface with subsurface models could be provided.
The micro-scale (spatially) is defined as a fundamental hydrologic unit, that is, a sub-parcel of an
individual tract of land. A sub-parcel could be a roof, sidewalk, grass lawn, driveway, garden, or other
landscape area. LID, also known as hydrological source control, is based on reducing hydrologic impacts
and incorporating micro-scale BMPs throughout the subcatchment (Prince George's County 2000).
Reducing runoff by incorporating storage and infiltration onsite improves the quality and reduces the
quantity of runoff produced by a developed site.
The LID philosophy is that runoff controls based on the micro-scale may be an important component of a
stormwater management plan, and stormwater control at the micro-scale may effectively mitigate the
effect of urban areas within the long-term hydrologic cycle. While not a panacea (Strecker 2001), this
volume-based approach to management of runoff has great potential to reduce the runoff volume,
sediment loads, and floatables that can reach receiving waters. By reducing runoff (and hence the
associated pollutants) municipalities stand to substantially reduce the estimated billions of dollars
associated with pollution abatement from WWF.
One question is whether the existing stormwater models designed for drainage and flood control can be
adapted for evaluating these small-storm events at the sub-parcel scale. Flood design is usually based on
1-8
-------
a single design event. However, there are no accepted criteria for selecting a design small-storm. Rather,
the appropriate strategy should incorporate continuous simulation over multiple years so that the
integrated behavior of the system can be evaluated over all of the storms that can be the dominant source
of water onto pervious areas. SWMM is one such model with this capability; others are discussed in
Chapter 5.
1.5 SUMMARY OF CURRENT SWMM BMP SIMULATION CAPABILITIES
The heart of the research described herein has been to document SWMM capabilities needs with regard to
BMP and LID simulation. The current SWMM (version 4.4h) as well as previous versions are each
divided into four primary computational "blocks" or "modules": Runoff (converting rainfall to runoff and
generating nonpoint source runoff quality), Transport (kinematic wave flow routing and water quality
routing through conveyances and storage), Extran (dynamic wave flow routing), and Storage/Treatment
(S/T) (treatment and storage devices). The object-oriented SWMM5 will not refer to "blocks" but rather
to hydrologic, hydraulic, and other descriptive "objects" such as watersheds, channels, storages, land use,
etc. To set the stage for the report that follows, capabilities and limitations of the current (SWMM4.4h
and SWMM5) model have been identified and are summarized in Table 1-4 (Huber 200Ib). Reference is
made in Table 1-4 to the four SWMM blocks just described.
The implications of Table 1-4 are:
1. Storage is well simulated in any of the four current SWMM blocks, although the Runoff Block offers
the least flexibility.
2. The S/T Block offers the most flexibility in terms of mimicking conventional and high-rate treatment
devices, e.g., for combined sewers.
3. Removal fractions (applied to incoming loads) may be used for Runoff Block overland flow segments
and Transport Block channel/pipes.
4. First-order decay may be applied in the Runoff, Transport, and S/T Blocks.
5. Settling velocities may be used in the Transport Block to simulate sedimentation (but no related
effects, such as build-up of solids or resuspension).
6. Particle size ranges (or settling velocity ranges) can be tracked through the current Runoff and
Transport Blocks as separate constituents. However, there is no linkage with the S/T Block, which
provides the most sophisticated sedimentation routines, including application of Camp's (1946)
sedimentation theory to up to five settling velocity ranges.
7. Overland flow rerouting options inherent in LID can be simulated in the Runoff Block, although one
subcatchment per surface type might be necessary. SWMM can be applied at the parcel (individual
lot) level on the basis of the modeling shown in this report.
8. Infiltration from channels is not simulated, except artificially, e.g., by an imposed "negative
hydrograph" or through monthly evaporation.
9. Biological interactions (e.g., in bioswales, wetlands) may be simulated only through first-order decay
(Runoff, Transport, S/T) and/or removal equations in the S/T Block.
10. There are few mechanistic, fundamental treatment processes included in the SWMM model, except
for sedimentation in the S/T Block.
1-9
-------
Table 1-4. Wet-weather controls and SWMM suitability (after Huber 2001b).
WWC Option
Storage and Associated
Treatment, e.g., Ponds
Screening and Filtration
Chemical Treatment,
Chlorination
Wetlands, Bioretention
Source Controls
Street Cleaning
Cisterns
Dry Wells
Overland Flow, Swales,
Infiltration, Porous Pavement,
Filter Strips
Infiltration Trench
Maintenance, e.g., Sewer
Flushing
Illicit Connection Removal
Inlet Constrictions
Real-Time Control
SWMM Suitability
Simulation of most storage options, with several options for
hydraulic controls. Treatment by removal equations, first-order
decay, or sedimentation. Removal in Transport Block
channel/pipes by first-order decay or sedimentation (constant
settling velocity) or removal fraction applied to incoming loads.
Simulation by removal equations.
Simulation by removal equations.
Simulation to the extent that wetland or biological channel behaves
like a storage device.
Simulated by reduced loadings.
Simulated directly through removal fractions or by reduced
loadings.
Can divert water to storage but cannot arbitrarily retrieve it.
Can divert water to well, but it then goes "out of simulation"
unless tracked using groundwater options.
Simulation in Runoff Block only. Optional quality removal by
first-order decay or removal fraction applied to incoming load.
Infiltrating water carries pollutants with it.
Can simulate infiltration from overland flow "trench" in Runoff.
Simulation by reduced loading, with simplistic option for sediment
scour/deposition in Transport Block.
Simulation through modification of drainage system connectivity.
Hydraulic control, simulated in Extran Block.
Orifice and weir settings in Extran can be controlled as function of
time and/or head changes.
Within this report, modeling concepts and, where appropriate, mathematical formulations are identified
for viable BMP/LID alternatives that have been identified as missing from SWMM (Table 1-4). The
effort to prepare this information is based on the assumption that such formulations are, in fact, available,
which has not always been born out. Generally, BMP alternatives include (Table 1-3): storage (ponds,
dry detention), bioretention facilities, infiltration trenches, grassed swales, filter strips, dry wells, cisterns,
and hydrodynamic devices. Some theoretical and/or heuristic processes/equations that are based on state-
of-the-engineering knowledge and information are proposed for possible inclusion in SWMM - probably
in version 5 (SWMM5). The use of some BMPs to attenuate stormwater in upland locations as LID
options, including rain gardens and roof-top green areas, is also gaining much attention. Necessary
modeling parameters, e.g., soil characteristics and antecedent moisture content, which reflect local site-
specific conditions, are identified in the discussion and in the examples of Chapters 14 and 15. SWMM
modeling limitations for various BMP/LID alternatives are summarized in Chapter 16. The areas that
have been addressed include:
Spatial resolution to allow application of micro-management of flows in a residential lot (Section
4.9).
Overland flow movement in pervious and impervious surfaces within a residential lot, and from lots
to street gutters, swales, buffer strips, channels, and sewers (Sections 4.7 and 4.8).
1-10
-------
Subsurface flow movement of rainwater infiltration through unsaturated (aeration) zone and
interaction of surface and ground waters (Chapters 8 - 11).
Routing and attenuation of pollutants in overland and subsurface flow movement considering
subsurface adsorption, absorption, and dispersion processes (needs identified in Chapter 16).
Routing of flows and pollutants from lots to swales/street gutters/inlets/sewers (Sections 4.7 and 4.8).
Hydraulic efficiency of storage-routing for pollutant removal (Sections 6.5 and 7.8).
Runoff/storage/infiltration/treatment BMP/LID process designs (Chapters 4, 6-8).
Generally, the first part of this report deals with BMPs and their simulation options, followed by Chapters
14 and 15, which are devoted to two detailed case studies. Note: generic reference to "SWMM
capabilities " will refer to version 4.4h (April 2002) since SWMM5, which was officially released in
October 2004, will continue to be enhanced in the future.
1-11
-------
STUDY AREA OPTIONS
2.1 OBJECTIVES
BMP and LID simulation may be more easily grasped in the context of examples. A component of this
project is to provide locations for which case studies may be possible. Case studies with enough intra-
event data for BMP simulation are just emerging (July 2003) from examination of the EPA/ASCE BMP
Database (http://www.bmpdatabase.org/), and no attempt has been made to simulate only "real" BMPs for
this reason. However, regarding LID, are there hypothetical residential lots (either conceptual or an
actual residential lot) for which alternative BMP/LID stormwater management techniques could be
demonstrated? The SWMM model could then be applied to these hypothetical lots to demonstrate how
alternative scenarios would affect the model's WWF control simulation capabilities. Case studies based
on these locations should include typical urban land uses and include a mix of combined and separate
sewer systems. The locations should have detailed data on soils, land use, precipitation, infrastructure
components, etc., and ideally should have been modeled (perhaps by the agency supplying the data) using
SWMM or a similar program. Cost and performance data are also important - if unlikely to be available.
Five locations were identified during the course of this study that might be suitable for LID evaluation
and/or simulation. They are summarized briefly below.
2.2 PORTLAND, OREGON
Portland, Oregon is a city containing many CSO areas with serious problems in terms of flooded
basements and overflows to receiving waters. The Portland Bureau of Environmental Services (BES) is
using a wide variety of BMPs and has been a leader in applying simulation models to their problems.
Their CSO control program is embedded in a watershed plan for the entire area. Good cost data and
excellent internet links for accessing data are available (http://www.cleanrivers-pdx.org/). Portland can
be studied at micro and macro scales. Additional information is deferred to the extensive case studies
presented in Chapters 14 and 15.
2.3 VALLEJO, CALIFORNIA
Vallejo, California is an SSO area with serious basement flooding and overflow problems. An excellent
database on flows, sewer infrastructure, failure rates, control costs, and SWMM modeling results is
available. A good general description of the Vallejo Sanitation and Flood Control District SSO
elimination program can be found on their consultant's web site, http://www.carollo.com/vsfcd/.
Technical information and data may be found in Carollo Engineers and CH2M Hill (2000). Results of
risk optimization for Vallejo can be found in Wright et al. (200 la, b).
2-12
-------
2.4 HAPPY ACRES
A hypothetical 105-ac study area called Happy Acres has been used extensively for research and teaching
at the University of Colorado. This study area is comprised of low and medium density residential land
use, a shopping center, and a school. It has been used as a single study area to evaluate water supply,
wastewater, and stormwater options. Optimized designs have been done for the water, wastewater, and
stormwater infrastructure. Arc View files have been generated to summarize the geographic information
system (GIS) information. An Access database has also been developed for each parcel, and a facilities
database is being developed. Results of analyses have been published in previous reports and papers
dealing with costs (Heaney et al. 1999a, Fan et al. 2000, Sample et al. 2001), GIS (Heaney et al. 1999b,
Sample et al. 2001), and optimization (Heaney et al. 1999c,d; Heaney et al. 1999e). A good benchmark
of alternative designs is available for Happy Acres based on its extensive use in both research and
teaching.
2.5 WONDERLAND CREEK IN BOULDER, COLORADO
Wonderland Creek in Boulder is a 14-ac residential neighborhood that has been evaluated in great detail
regarding the nature of the imperviousness and its effect on runoff. Various levels of spatial detail may
be obtained from the GIS coverage for this study area. SWMM has been set up and run for this
neighborhood using 1-, 15-, and 60-minute rainfall data for Boulder (Lee 2003). The results of an
accurate determination of the nature of imperviousness in urban areas indicate that existing estimates can
be very inaccurate. The error in estimating imperviousness causes major changes in the accuracy of the
predicted hydrographs (Lee 2003).
2.6 FAIR OAKS ESTATES IN CAROL STREAM, ILLINOIS
The Fair Oaks storm sewer and detention pond is used as a design example in the second edition of the
Hydrology Handbook (ASCE 1996). Fair Oaks Estate is a subdivision of Carol Stream, Illinois, a suburb
west of Chicago. The total area of the subdivision, which includes 13 lots, is 13.4 ac. Details of the
detention pond design are shown in Burke (1979), Burke and Gray (1979), Burke and Burke (1994), and
ASCE (1996). The design was done in the 1970s and the system has been in operation since that time. It
is a conventional design and appears to be working fine after 20 years of operation. In his original MS
thesis, Burke (1979) evaluated the impact of using the Rational Method (Dooge 1973) vs. the ILLUDAS
(Terstriep and Stall 1974) model to solve the Fair Oaks problems. The use of ILLUDAS showed the
hydraulic inadequacy of some of the pipes selected based on the Rational Method. Similar comparisons
were done for other hydrologic and hydraulic models.
2.7 SUMMARY
Two locations in Portland were selected for SWMM simulations in this study for several reasons:
Proximity to OSU: the project team performed most of the SWMM simulations.
Existence of good flow monitoring data, rainfall data, and catchment characterization data for
multiple sites.
Willingness on the part of the Portland BES to share data and simulation results.
Lack of some key information from the other sites, plus the fact that Happy Acres is hypothetical.
Using the two Portland locations, SWMM simulation examples are presented in Chapter 14 for LID and
Chapter 15 for a pond.
2-13
-------
BMP EVALUATION OPTIONS
3.1 BMP PERFORMANCE EVALUATION
If a model such as SWMM is used to simulate BMP performance, on what basis will the BMP's
effectiveness be measured? A simulation model has the capability to produce simulated influent and
effluent concentrations for thousands of storm events. Statistical methods used to evaluate monitored
BMPs might also be used to evaluate simulated BMPs. If so, what conclusions can be drawn from
evaluation efforts to date of monitored BMPs? This chapter discusses options for BMP evaluation based
on measured concentrations and flows. Some of these options could also be applied to simulation model
output.
A fundamental statistic of a quality monitoring program is the event mean concentration, or EMC.
Although it is defined as the total constituent mass for an event divided by the total flow volume for the
event, it is usually computed by preparing one flow-weighted composite sample from the several quality
samples taken during a storm, and sending the one sample to the laboratory for analysis. The BMP
"performance" is then usually - but not always - computed on the basis of the relationship between the
influent and effluent EMCs, hopefully for many monitored storm events. In essence, the BMP is
considered to be a "black box," to which several statistical and mathematical procedures might be applied
to deduce the "transfer function" (Chapra 1997) that relates input to output. The most widely used
statistical methods will be listed in the next section.
The issues involved in selecting methods for quantifying BMP efficiency, performance, and effectiveness
are complex. It is also important to appreciate that the reliability and performance of many of these
controls have not been well established. Accurate reporting of BMP effectiveness is important to
modelers wishing to calibrate to actual practices.
The EPA/ASCE BMP Database (http://www.bmpdatabase.org/) currently characterizes BMP
performance as EMC (i.e., a single representative concentration). Although some groups believe that a
better method for measuring performance is amount of flow treated and effluent quality, (Strecker et al.
2001), an outlet EMC value is probably the simplest single measure of BMP performance.
While analyzing the EPA/ASCE BMP Database, Strecker et al. (2001) observed a trend indicating higher
removal efficiency in BMPs with higher influent concentrations. Strecker et al. (2001) aver that several
BMPs can be characterized by the effluent EMC distribution, that is, the effluent probability method
(EPM), rather than a heuristic or process model of the system. However, current data available are
insufficient to tie BMP design to effluent quality using the EPM, and it would be difficult to model BMPs
3-14
-------
in series (treatment train) with this method. As the EPA/ASCE BMP Database evolves, the EPM method
(discussed below) may be enhanced.
Before discussing some EMC methods in more detail, it is important to note that quality is not the only
performance measure. Strecker et al. (2001) proposed that there should be a threefold metric for BMP
characterization:
How much catchment runoff is averted by a control measure (hydrologic source control), e.g., by
infiltration or evapotranspiration?
Of the runoff that enters the BMP, how much is treated and how much is bypassed or routed through
the BMP at rates exceeding treatment rates?
What are the quality characteristics of the treated BMP effluent?
A fourth metric can be added:
How much downstream flow management is provided by the BMP?
Metrics one, two, and four are hydrologic and depend strongly on regional patterns of storm events and
dry-weather intervals (and of course, local catchment's characteristics such as soils). It will be seen in
this report that SWMM is capable of good short-term and long-term hydrologic performance
characterization - except for infiltration from channels. SWMM can also produce the event EMCs that
can be used to characterize effluent quality. Finally, the viability of BMPs such as constructed wetlands
and biofilters depends in part upon the nature of the entering stormwater. Whether or not the BMP can
support healthy and diverse vegetation may depend on issues related to nutrient dynamics and subsurface
water quality, which SWMM may or may not be able to address. The bulk of this report identifies ways
in which SWMM deficiencies can be overcome in order to enhance the model's already powerful
capabilities.
3.2 IDENTIFICATION OF COMMON EMC-BASED METHODS
The difficulty in selecting measures of efficiency stems from the desire to compare a wide range of BMPs
and the large number of methods currently in use. There is much variation and disagreement in the
literature about what measure of efficiency is best applied in specific situations; however, it is generally
accepted that the EMC and long-term loading provide the best data for observing the effects of the BMP
on acute and chronic pollution, respectively (GeoSyntec et al. 2002).
Ten methods for evaluating BMP performance using EMC data are summarized by GeoSyntec et al.
(2002) in documentation for the ASCE/EPA BMP Database. These methods are used when the input
database consists of event mean influent and effluent concentrations.
1. Efficiency ratio
2. Summation of loads
3. Regression of loads
4. Event mean concentration
5. Efficiency of individual storm
6. "Irreducible concentration" and "achievable efficiency"
7. Percent removal relative to water quality standards
8. Lines of comparative performance
9. Multi-variate and nonlinear models
10. Effluent probability method
All of these methods appear to be statistical models that are appropriate to use when the only input data
that are available are storm event mean influent and effluent concentrations. They do not provide a
process level characterization of performance. Four methods are described below, however, the ASCE
3-15
-------
study team (Strecker et al, 2001) does not recommend any of the first three (EMC, efficiency ratio,
regression). In fact, the team recommends that a better measure of BMP performance is the amount of
runoff prevented, the amount of flow treated (and bypassed) and the effluent quality of the treated flow.
They have observed that BMPs tend to produce a fairly narrow range in effluent quality, and, therefore,
percent removal is merely a function of how polluted the inflow is. Characterization of effluent quality is
one feature of the EPM, also discussed below, and recommended by the team.
Additional work to standardize BMP monitoring protocols and calculations is needed to make monitoring
data comparable from site to site (GeoSyntec et al. 2002). In addition to the EPA/ASCE BMP Database
reports, documents that summarize BMP efficiency information include the National Pollutant Removal
Performance Database (Brown and Schueler 1997), the Terrene Institute's report The Use of Wetlands for
Controlling Stormwater Pollution (Strecker at al. 1992), and the recent text by Minton (2002).
3.3 EVENT MEAN CONCENTRATION
The term event mean concentration (EMC) is a statistical parameter used to represent the flow-
proportional average concentration of a given parameter during a storm event. It is defined as the total
constituent mass divided by the total runoff volume, although it is usually computed in practice by
compositing multiple samples on the basis of the flow rate at the time the sample was taken, prior to
sending the samples to the laboratory. The single flow-weighted composite then provides EMCs upon
chemical analysis. When combined with flow measurements, the EMC can be used to estimate the
pollutant loading from a given storm. Under most circumstances, the EMC provides the most useful
means for quantifying the level of pollution resulting from a runoff event. Collection of EMC data has
been the primary focus of the EPA/ASCE BMP database project.
The EMC for an individual event or set of field measurements, where discrete samples have been
collected, is defined as:
EMC = Jl (3-1)
where
Vj = volume of flow during the period j,
Q = average concentration associated with the period and volume Vj, and
n = total number of measurements taken during event.
Currently, SWMM can easily produce a constant concentration (EMC value) in the Runoff Block and as
the output of a treatment device in the S/T Block. However, storage and quality control devices in Runoff
and Transport cannot easily (i.e., without odd data manipulations) produce a constant outflow
concentration different from the inflow concentration. That is, it is not easy to have a constant inflow
concentration of, say, 20 mg/L and a constant outflow concentration of, say, 8 mg/L from a Runoff or
Transport flow element.
3.4 EFFICIENCY RATIO
The efficiency ratio (ER) is most often used to report BMP efficiency. While this may be appropriate for
determining the reduction in pollution for an event, it may not indicate performance of a BMP over time,
or for events of varying intensity and volume (GeoSyntec et al. 2002). Another reason is because the ER
approach does not consider the statistical significance of the result. Most researchers assume that ER has
the meaning of "percent removal." While this method is not recommended for monitoring the
3-16
-------
effectiveness of BMPs, because of the data availability and the wide spread analysis with this method, ER
values are often reported in the literature. If modeling with a "black box" approach, average EMC and
ER values are simple to implement, e.g., in spreadsheets. The efficiency ratio is defined in terms of the
average EMC of pollutants over some time period:
average outlet EMC average inlet EMC - average outlet EMC
ER. = 1 = (3-2)
average inlet EMC averageinletEMC
This method (GeoSyntec et al. 2002):
Weights EMCs from all storms equally regardless of relative magnitude of storm. For example, a
high concentration/high volume event has equal weight in the average EMC as a low
concentration/low volume event.
Minimizes the potential impacts of "smaller/cleaner" storm events on actual performance
calculations. For example, in a storm-by-storm efficiency approach, a low removal value for such an
event is weighted equally to a larger value.
Allows for the use of data where portions of the inflow or outflow data are missing, based on the
assumption that the inclusion of the missing data points would not significantly affect the calculated
average EMC.
Comments:
Many studies use the ER method to characterize performance, but it fails to take into account some of
the complexities of BMP design, especially media filters and other BMPs that treat to relatively
constant levels and are independent of inflow concentrations.
This method also assumes that if all storms at the site had been monitored, the average inlet and outlet
EMCs would be similar to those that were monitored.
Under all circumstances this method should be supplemented with an appropriate non-parametric (or,
if applicable, a parametric) statistical test indicating whether the differences in mean EMCs are
statistically significant (it is better to show the actual level of significance found, rather than just
noting whether the result was significant, assuming a 0.05 level).
Currently, SWMM can use an efficiency ratio (based on inflow concentration) to simulate BMP
performance in the S/T Block. The Runoff and Transport Blocks allow application of a removal fraction
to the incoming load, which is similar, but not the same. (It would be the same if the outflow equaled the
inflow at the particular time step.)
3.5 REGRESSION MODELS
Regression methods may also be used to evaluate BMPs. These include regression of outlet load vs. inlet
load ("load regression") and regression of ER values vs. inflow concentration. The former method has
been used with some success to characterize performance (GeoSyntec et al. 2002), although there are
many assumptions involved. The current version of SWMM provides this capability in the S/T Block
through the use of its "universal removal equation" (Section 4.3). However, regression of ER values
against inflow concentration produces spurious correlation (Benson 1965) since it amounts to regression
of EMCout/EMCm vs. EMCjn. A seemingly significant regression is inevitable (Benson 1965). That is, the
dependent variable includes as a divisor the independent variable EMCm. The resulting relationship
implies that removal efficiencies depend on influent concentration, but such a relationship may or may
not exist. It always indicates a spurious relationship between percent removal and influent quality, and
can lead to misguided "lines of comparative performance" (GeoSyntec et al. 2002, Minton 2002).
Regression models are useful in doing preliminary investigations into cause-effect relationships, but they
should be restricted to functional forms of how the BMP should work.
3-17
-------
3.6 EFFLUENT PROBABILITY METHOD
The effluent probability method (EPM) is straightforward and provides a clear, but qualitative picture of
BMP effectiveness. The EPM consists of a lognormal probability plot (although any distribution could be
used for which probability paper exists, including normal) of EMC vs. either probability of occurrence or
percent exceedance (equivalent to the cumulative distribution function, or CDF). Probability plots are
among the most useful pieces of information that can result from a BMP evaluation study (Burton and Pitt
2001). The authors of the EPA/ASCE BMP Database Monitoring Manual strongly recommend that the
stormwater industry accept this approach as a standard "rating curve" for BMP evaluation studies, as it
provides a visual representation of the frequency distribution of both influent and effluent quality
(GeoSyntecetal. 2002).
Lognormal plots are ordinarily used because the lognormal distribution has been found to be a good fit to
most stormwater EMC data (USEPA 1983, Driscoll 1986, Van Buren et al. 1997). One advantage of
normality, either of the logs or of the untransformed data, is that parametric statistical tests can be applied,
such as the t-test, chi-square test, and analysis of variance. Statistical tests used to compare the difference
between data sets typically require normality among the data sets, and some also require the sets to have
the equal variances. Lognormal (and normal) probability plots can be used for qualitative guidance, since
the same slope of two such curves indicates the same variance of the data.
The most basic test for normality is whether or not the data plot is a straight line on normal probability
paper (or vs. equivalent values of the standard normal variate, z) (Burton and Pitt 2001, Bedient and
Huber 2002). Tests for normality itself include tests directly related to probability plots, such as the
probability plot correlation coefficient (PPCC) (Vogel 1986) and the Shapiro-Wilk test (Helsel and Hirsch
1992). Tests not related to probability plots include the Kolomogorov-Smirnov test and the chi-square
test (Benjamin and Cornell 1970, Helsel and Hirsch 1992). However, the latter are less powerful, in a
statistical sense, than tests that use probability plots, not to mention that the plots themselves yield great
qualitative information, as discussed below. It is of interest that if ranked data are plotted against
standard normal variates ("z-values") obtained from the inverse of the plotting position probability
(Bedient and Huber 2002), using Excel, the linear fit of data (untransformed or logarithms) obtained using
Excel's "trend line" option provides the required PPCC. The PPCC can then be tested for statistical
significance (Vogel 1986, Helsel and Hirsch 1992). The critical values of the correlation coefficient
account for the inherent correlation between two ranked data sets (order statistics). If normality, and in
some cases equal variance, is ensured among the respective data sets, parametric tests can be employed to
test the difference between the means/medians of the data sets.
Parametric and nonparametric statistical tests should be conducted after the probability plots are
generated to indicate whether perceived differences in influent and effluent mean EMCs are statistically
significant (it is preferable to provide the level of significance, instead of just noting whether the result
was significant, such as at a 95% confidence level). Helsel and Hirsch (1992) provide an excellent primer
on such methods, with applications to water resources and water quality. Many parametric and non-
parametric tests are included in standard statistical software.
Regarding the effluent probability plots, there are limited quantitative assumptions that can be made
simply on the basis of the plots themselves. As the data rarely enter into the plots as matched pairs (if the
quantiles did occur as storm-event pairs, it would be sheer coincidence), there are limited inferences that
can be made regarding BMP effectiveness over a particular concentration range. That is, one cannot
make such inferences simply on a comparison of quantiles. Only when the data sets are entered as
matched pairs, as for instance, in a simple scatter plot of EMCout vs. EMCm, can a quantitative assessment
of the removal over specified concentration ranges be conducted. Other authors (Burton and Pitt 2001, p.
584) have discussed the EPM plots, drawing such inferences as "shows that SS are highly removed over
3-18
-------
influent concentrations ranging from 20 to over 1,000 mg/L." Unless the data are entered as matched
pairs, such statements cannot be justified. For example, there may be one particular storm event in which
no removal was observed (the influent and effluent EMC values are the same), but by plotting quantiles,
the outcome of any single storm event may not be readily observed. There may be some effluent EMCs
that are greater than the effluent data points for storm events, even though the generated trend line for the
influent data set is still higher than the trend line generated for the effluent data set. Therefore, removal is
not always guaranteed.
However, the relative range of both influent and effluent quality can be determined based upon the
concentration range between given percentile values. In addition, the normality and equal variance
among the data sets can be qualitatively observed, although any inferences about variance must be
confirmed through quantitative statistical testing. Separation between the trend lines, an indication of
removal, may be tested parametrically with a t-test if both lines are lognormal, or by using the non-
parametric Kruskal-Wallis or Hodges-Lehmann tests, for example (Helsel and Hirsch 1992). Examples
(Brown 2003) of effluent probability plots are shown as Figures 3-1 to 3-3, using data for an extended dry
detention facility (Seton Pond) in Austin, Texas (Keblin et al. 1997).
The effluent probability plot for TSS at the Seton Pond facility is shown in Figure 3-1. From the figure,
one can gage the relative range of influent and effluent EMCs, as signified by the concentration range
between the 10th and 90th percentile. The slopes of the influent and effluent trend lines are similar, thus
indicating similar variance between the logs of the two data sets. Both slopes appear somewhat flat,
indicative of a relatively low coefficient of variation. But of considerable interest, contrary to what might
be expected from a visual evaluation, the logs of the effluent TSS EMCs fail two normality tests (Brown
2003). Hence, the parametric t-test cannot be used to compare means. Finally, by comparing the data
points plotted, it appears the lowest influent data point is still greater than the highest effluent point.
Therefore, this BMP is projected to achieve removal, based upon the relative distance between the
influent and effluent trend lines and the fact that there is no overlap among influent and effluent data
points. This may be confirmed through the non-parametric Kruskal-Wallis test.
1000
» Influent
Effluent
* Method of moments (influent)
Method of moments (effluent)
Least squares fit (effluent)
Least Squares fit (influent)
-2 -1.75 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
Frequency factor (z)
Figure 3-1. Probability plot for TSS removal at the Seton Pond facility, Austin, Texas (Brown 2003).
In comparison, Figure 3-2 shows the effluent probability plot for nitrate at the Seton Pond facility. Both
the influent and effluent EMCs are confirmed to be lognormal (Brown 2003). The range of influent and
3-19
-------
effluent concentrations can still be observed, but the range of effluent concentrations actually exceeds the
range of influent concentrations so little can be said about the system performance. The slopes of the
influent and effluent generated trend lines are quite dissimilar, indicating differing variance among the
logs of the nitrate data sets, and as trend lines intersect, there is little to be said about the removal at any
range. With the trend line intersection, it is easily observed that the lowest influent data point is much
lower than the highest effluent data point. Therefore this BMP is not predicted to achieve removal of
nitrate, based upon the overlap and intersection of data points and generated trend lines. Indeed, the
parametric t-test indicates that equality of the means of the logs cannot be rejected as does the non-
parametric Kruskal-Wallis test (Brown 2003).
Influent
Effluent
- -*- - Method of moments (influent)
Method of moments (effluent)
Least squares fit (effluent)
Least squares fit (influent)
0.1 -I
-2 -1.75 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
Frequency factor (z)
Figure 3-2. Probability plot for nitrate removal at the Seton Pond facility, Austin, Texas (Brown 2003).
The final example in Figure 3-3 shows the effluent probability plot for total zinc at the Seton Pond
facility. The lognormality of both the influent and effluent EMCs may be confirmed by statistical tests.
Like the previous plots, the range of influent and effluent concentrations is easily observed, with only a
minimal overlap between influent and effluent EMCs. The slopes of the trend lines appear dissimilar,
indicating differing variance among the data sets. But appearances can be deceiving! The hypothesis of
equality of variance between influent and effluent EMCs is not rejected by three statistical tests (Brown
2003). The magnitude of the effluent slope is greater than the magnitude of the influent slope; therefore,
the effluent data set is predicted to have a larger coefficient of variation (because the effluent mean of
logs is also lower than the influent mean of logs). Comparing the data points themselves, it appears that
the lowest influent data point is slightly lower than the highest effluent data point. Therefore, removal for
this BMP can only be ensured through appropriate statistical examination, to determine whether a
significant difference does exist between the two data sets, as it can not readily be shown through the
plots. In fact, removal is confirmed by both the parametric t-test and non-parametric Kruskal-Wallis test
(Brown 2003).
3-20
-------
i °'1-
E
o
m
LLJ
£ 0.01 -
N
0 001
P 1-
1.15
/6 * :
'- = 0.8828 _ ^^ 1
Influ
Eft I
- -»- - Met
- *- - Met
Lea
*^-^^ I
*~! -^,^**^=»=
^ ^^^t :
^Jff^
^^A 1"^ y = 0 024e0-9639x I
^-^**"" R2 = 0.9332 |
L^**"*"^ ! ^^
-f^ i i
ii
-2 -1.75 -1.5 -1.25 -1 -0.75
1 |
ent
jent
lod of moments (influent)
lod of moments (effluent)
st squares fit (effluent)
st squares fit (influent)
-0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
Frequency factor (z)
Figure 3-3. Probability plot for total zinc removal at the Seton Pond facility, Austin, Texas (Brown 2003).
In many ways, the qualitative inferences from the EPM may also be obtained from other descriptive
statistics, such as box plots, as well as quantitatively through the parametric t-test and non-parametric
comparisons of medians. However, the EPM has the advantage of illustrating the lognormal (or other
distribution) fit of the data, rather than simply certain quantiles, as with box plots. The primary problem
with the EPM is that certain quantitative assumptions, such as removal and performance at or around a
certain concentration value, cannot be made unless data points are entered as matched pairs (e.g., as in
scatter plots of effluent vs. influent EMC). This discrepancy was noted in the CalTrans BMP study and
assessment (CalTrans 2003), indicating that interpretation of these plots should be performed in
conjunction with related analyses, such as scatter plots. Another concern raised with the EPM is its
ability to provide sufficient information regarding BMP selection. In areas requiring a set removal
percentage, use of the EPM may not adequately portray whether a BMP is capable of meeting that
performance standard (CalTrans 2003). However, the EPM retains the advantage of being able to deal
easily with an unequal number of influent and effluent EMC data points.
Currently, SWMM is not capable of simulating a prescribed lognormal (or any) frequency distribution,
either for influent concentrations or for BMP outflows. That is, it is not possible to enter parameters of a
lognormal distribution as a Runoff Block option instead of, say, a buildup-washoff formulation. In
principle, such a distribution could be observed over the period of a long-term continuous simulation
from constituent quality generated by buildup-washoff or another mechanism (Huber et al. 1987), but the
need here is to specify a distribution a priori. Equally useful would be a method to simulate the water
quality of runoff on the basis of a specified frequency distribution. This would be an option to be used
instead of buildup-washoff, constant concentration, etc. Likewise, the frequency distribution of BMP
effluent EMCs could be prescribed, on the basis of observed data. However, it should be noted that the
SWMM Statistics Block is capable of analyzing any flow or quality time series that can be placed on an
"interface file" as the output of a SWMM Block, to generate frequency distributions and identify
lognormal parameters through the method of moments. This could easily be enhanced through the
SWMM5 GUI.
3-21
-------
3.7 SIMULATION MODELS
Simulation models allow evaluation of long-term rainfall to assess volumes of stormwater treated by
BMPs and how much is bypassed, or processed at rates that result in non-effective treatment. Thus, in
principle, a simulation model can provide all the information needed to evaluate a BMP that would be
prohibitively expensive to collect in a monitoring program - if the model is to be believed. This includes
information such as much of the rainfall record is treated or controlled. The dynamics of the filling and
emptying of these BMPs is vital to understanding treatment efficiencies and runoff removal rates. The
bulk of this report will be devoted to evaluating the effectiveness of SWMM in this regard and examining
alternative simulation algorithms from other models.
In the event that a model such as SWMM has not been specified a priori, searching for a model that
characterizes BMP effectiveness is not simple. The lack of consistent and concise evaluation of modeling
abilities makes choosing a model quite complex and time consuming. Just as BMP effectiveness
monitoring is currently being standardized, so is model evaluation. Model descriptions may be found in
texts and reports (e.g., Singh 1995, Field et al. 2001, Debo and Reese 2002, Field and Sullivan 2003,
Field et al. 2004), but anything printed may soon lose currency. One such listing is shown in Table 3-1.
Web pages, such as a circa-1998 site provided by the Great Lakes Sediment Management Program
(http://www.glc.org/tributary/pdf/allweb32.pdf), can provide lengthier and timelier model information - if
they are updated!
Some models listed in Table 3-1 do not perform quality calculations; they are included in this listing
because 1) they are commonly used in urban areas of the U.S. and Canada, and 2) quality loads are often
computed after an accurate hydrologic and hydraulic model is run, simply by multiplying flows by EMCs.
Additional, less common models evaluated as part of this study are listed in Chapter 5.
3-22
-------
Table 3-1. Primary urban hydrologic, hydraulic and water quality models commonly used in the United States.
Source: W.C. Huber, class notes. Updated March 25, 2003.
Model
DR3M-QUAI/
FEQab
HEC-1/HMSC
HEC-2/RASc
HSPFad
InfoWorks CSe
MIKE 1 lf
MOUSEf
P88
Santa Barbara
scsh
*SewerCAT
SLAMM1
*STORMC
SWMMd'!'k
UNETC
Agency/Source
USGS
USGS
HEC/Vendors
HEC/Vendors
EPA
HRWallingfordinUK,
Montgom. Watson in US
Danish Hydraulics Inst.
Danish Hydraulics Inst.
Wm. W. Walker, Jr.
Vendors
NRCS/Vendors
Reid Crowther Consult.
R. Pitt
HEC/Vendors
EPA/OSU
HEC/Vendors
Primarily
Hydrology/
Hydraulics
Hydrology
Hydraulics
Hydrology
Hydraul. (backwater)
Hydrology
Hydrology/Hydraulics
Hydraulics (open
channels)
Hydrology/Hydraulics
Hydrology
Hydrology
Hydrology
Hydraulics
Hydrology
Hydrology
Hydrology/Hydraulics
Hydraulics
Continuous
Simulation or
Storm Event
CS/SE
SE
SE
Steady state
CS/SE
CS/SE
SE
CS/SE
CS/SE
SE
SE
SE
CS
CS
CS/SE
SE
Complete
Dynamic Flow
Routing?
No
Yes
No
No
No
Yes
Yes
Yes
No
No
No
Yes
No
No
Yes
Yes
Quality
Simulation?
Yes
No
No
No
Yes
Yes
Yes
Yes
Yes
No
No
No
Yes
Yes
Yes
No
Graphical
User
Interface1
ANNIEa'd
No
Yes
Yes
ANNIEa'd,
EPA BASINS
Yes
Yes
Yes
Menu
3rd party
3rd party
Yes
No
No
3rd party
HECDSS
Web addresses for models begin with prefix: http://
a. h2o.usgs.gov/software/surface_water.html b. www.dilurb.er.usgs.gov/proj/feq/
c. www.hec.usace.army.mil/ d. www.epa.gov/ceampubl/softwdos.htm
e. www.hrwallingford.co.uk/ f www.dhi.dk
g. www.wwwalker.net/p8/ h. www.ncg.nrcs.usda.gov/tech_tools.html
i. www.ccee.orst.edu/swmm (not currently available, contact Dr. C. Vitasovic at DHI)
j. 3p = 3rd party (e.g., an outside vendor) k. www.epa.gov/ednnrmrl/swmm/index.htm (for SWMM5)
1. www.eng.ua.edu/~rpitt/SLAMMDETPOND/WinSlamm/WINSLAMM.shtml *May not be currently available.
3-23
-------
CURRENT SWMM SIMULATION CAPABILITIES
4.1 THE MODEL
The EPA SWMM is a dynamic rainfall-runoff simulation model, primarily but not exclusively for urban
areas, for single-event or long-term (continuous) simulation. Flow routing is performed for surface and
sub-surface conveyance and groundwater systems, including the option of fully dynamic hydraulic
routing in the Extran Block. Nonpoint source runoff quality and routing may also be simulated, as well as
storage, treatment and other BMPs.
Version 4 of the SWMM evolution is currently (March 2004) at version 4.4h (Huber and Dickinson 1988,
Roesner et al. 1988). The most current release of version 4.4h can be downloaded from the Oregon State
University website (www.ccee .orst.edu/swmm). Simulation of storage and treatment was generalized in
Version III of the model (Nix et al. 1978) so that algorithms were not specific to individual devices.
SWMM has been through many modifications since Version II, and since the first version 4 release in
1988, it now includes several options that enhance the model's ability to simulate BMPs and general
control options for management of stormwater and combined sewers (Huber 1996, Huber 2001b).
Version 4.4h will gradually be replaced by SWMM5, developed by the EPA, as that object-oriented
model and its GUI gain functionality (http://www.epa.gov/ednnrmrl/swmm/index.htm). However, most
of the BMP and LID simulation capabilities described for version 4.4h also apply to SWMM5.
SWMM capabilities regarding simulation of wet-weather control (WWC) alternatives have been
discussed in Chapter 1 and summarized in Table 1-4. The following text discusses significant control
options and consequent SWMM simulation capabilities.
4.2 STORAGE
Storage in an urban catchment may occur on the ground surface, in the drainage system, and in specific
storage devices (e.g., ponds, tanks, secondary flow removal devices). Pollutant removal occurs primarily
through sedimentation and decay. SWMM is most effective at simulating storage-type BMPs because
such devices have been extensively studied and information about them is widely available. Flow routing
through storage is easily performed using a variety of methods, and multiple outlets (and bypass) may
readily be simulated. SWMM also has the ability to simulate hydraulic controls and time-dependant
regulators (such as weir and orifice settings that depend on stages in specified locations), although these
options are primarily located within the Extran block, which does not model water quality. Storage
options are available in the Runoff, Transport, Extran and Storage/Treatment (S/T) Blocks.
4-24
-------
S/T quality removal may be based on a first-order process, for which simple finite difference formulations
exist for well-mixed (i.e., continuous-flow, stirred-tank reactors or CFSTRs) or plug-flow systems.
(However, see discussion of possible errors in the manner in which the S/T Block simulates well-mixed,
first-order decay, in Section 4.3 below.) A second mechanism consists of a generalized removal equation
designed such that most empirical or other results can be mimicked (Section 4.3). For example, removal
of a given pollutant can be computed as a function of concentration of any pollutant, removal fraction of
any other pollutant, and/or detention time. A third removal option is by sedimentation using Camp's
(1946) theory for quiescent conditions, modified for turbulence by Chen (1975). For this formulation up
to five settling velocity ranges must be defined and then routed through a progression of up to five S/T
units. While this third option is ideal for simulation of S/T units in series (e.g., a treatment train),
unfortunately, such pollutant characterization is not generated elsewhere (upstream) in the model. That
is, treatability data are typically not generated by the model upstream of the S/T Block and cannot be
linked - unless the user specifically simulates discrete particle size ranges as separate constituents in
upstream blocks, which certainly is an option.
Removal by first-order decay may also be simulated in the Transport, and Runoff Blocks, assuming
complete mixing (in accordance with their quality routing methods) within conveyance and storage
elements. Constituents in Transport and Runoff Block channels, pipes, and storages are also subject to
removal based on constant removal fractions applied to incoming loads and settling with a constant
settling velocity (Huber 200 Ib).
Interactions among water quality constituents are simulated minimally in the S/T Block (e.g., removal of
one pollutant can depend upon removal of another, in the manner of sorption). The Transport Block has a
relatively untested capability to simulate linked BOD-NOD-DO dynamics, essentially using a modified
Streeter-Phelps analysis (Huber 200 Ib) in the manner of WASP (Wool et al. 2001). There is only limited
capability in SWMM to simulate combined physical-biological removal processes in wetlands and
bioswales except to the extent that such removal can be characterized by the processes to be described.
To summarize, fundamental process categories of sedimentation, biological removal, sorption, filtration,
flotation, chemical treatment, high-rate biological treatment, degradation, hydrodynamic separation may
all be mimicked by the S/T Block as long as removal characterization relations are available. However,
the only FPCs simulated explicitly in storage are first-order decay (e.g., for degradation) and
sedimentation.
4.3 S/T GENERALIZED REMOVAL EQUATION
Because the generalized removal equation or "universal removal equation" or "one-equation-fits-all" may
be used for several purposes, additional information is provided here.
The "universal removal equation" within the S/T Block is
R = (a12
= a12e''x*
where
x; = removal equation variables (model state variables),
^ = coefficients,
R = removal fraction, 0 < R < RMX < 1 .0, and
RMX = maximum removal.
4-25
-------
The removal fraction is applied to the load (mass) in a plug (for plug flow), at each time step. The
removal fraction can thus be made a function of selected state variables, x;. Each removal equation
variable (selected state variable), xi; may represent one of several variables available in the program at
each time step, such as an inflow concentration or residence time, described further below. With these
variables and the coefficients, al5 the user can develop the desired removal equation.
An example of the application of Equation 4-1 for plug flow is provided by the common exponential
removal equation characteristic of tank reactors and ponds observed in EPA Nationwide Urban Runoff
Program (NRUP) studies (USEPA 1983). For suspended solids (SS) this might be:
where
Rss = suspended solids removal fraction, 0 < Rss < Rmax < RMX,
Rnax = maximum removal fraction in fitted equation,
RMX = maximum removal, an S/T input variable, < 1.0.,
td = detention time (sec), and
k = first order decay coefficient (I/sec).
This equation can be constructed from Equation 4-1 by setting au = Rmax, ai3 = -Rmax, a3 = -k, ai6 = 1.0,
and letting x3 = detention time, td (by setting INPUT(I,IP,3) = 1 on data group G2 in SWMM S/T). All
other coefficients, aj would equal zero. RMX would not be necessary, as Rmax limits the value of R.
We 11 -mixed storage units are simulated by a finite-difference solution to the conservation of mass
equation for a CFSTR (Chapra 1997):
IC: -QC-kCV (4-3)
dt
where
C = well-mixed concentration in storage unit,
V = volume in storage unit,
I = inflow to storage unit,
C1 =inflow concentration in inflow to storage unit,
Q = outflow from storage unit,
t = time, and
k = first-order decay coefficient, I/time.
All variables except k are functions of time for a variable -volume storage device. Since storage routing is
performed before the quality computations, inflows, outflows, and volumes are known at the beginning
and end of each time step, leading to a finite difference formulation used in SWMM S/T,
4-26
-------
Change in
Mass entering Mass leaving
mass in basin = - - Decay during At
during At during At
during At
r:T +r: T r n +T O C V +C V
(~< Y C V = ^n+l-Ln+l A j _ ^nVn ^^n+lVn+1 A f _ V " " n+1 v n+1 A ^
n+1 n+1 n n ~ ~ ~
(4-4)
where subscripts n and n+1 refer to the beginning and end of a time step, At. Equation 4-4 is then solved
for the unknown CFSTR (and effluent) concentration Cn+i,
1 T +C1 T r n kC V
C V | n n "+1 n+1^At At At
r -
~~
n+1
When a CFSTR is simulated, the equation (Subroutine EQUATE, called from Subroutine UNIT in the
S/T Block) returns the product of k-At, where k = first order decay coefficient (I/time) and At = time step.
This is done mechanistically by setting the x2 variable = 1 on the S/T G2 line such that it will equal At
when Equation 4-1 is evaluated. Then set a2 and ai6 = 1.0, and a]2 = k.
However, there are some undocumented time step limitations to this process with the current Fortran
coding. The first is that the maximum value of the product k-At is input parameter RMX (maximum
removal per time step, useful for plug flow but also used for complete mixing), which in turn has a
maximum value of 1.0. Hence, the maximum value of k that the program will end up using is RMX/At or
I/At. Thus, if k-At > 1.0, an effective k value will be used, I/At, that is smaller than that input by the user.
The other consideration is the nature of the finite difference form of Equation 4-3, namely Equation 4-4.
For constant inflow, outflow and, therefore, constant volume, the finite difference solution (Equation 4-5)
can be reduced to:
-At 1-( + k)
^ + Cn %-Z (4-6)
2V 2V
where
n, n+1 = previous and new time steps, respectively,
C = concentration in outflow and in storage unit,
C: = average inflow concentration over the time step,
At = time step,
Q = constant inflow and outflow, and
V = constant volume.
It can be seen that if 1 ( h k) < 0, negative concentrations can result, and if the term is
4-27
-------
< -1, unstable computations can result. Because Q and V vary during a real S/T simulation, it is not
necessarily easy to predict whether or when this will occur.
The conclusion is that Equations 4-4 and 4-5 are a poor method for solving Equation 4-3. Better would
be to use a standard 4th order Runge-Kutta (RK4) numerical solver for the ordinary differential Equation
4-3. This is the method that will be implemented in SWMM5.
In passing, it should be observed that the current SWMM solution for complete mixing in Runoff and
Transport channels, pipes, and storage elements does not use the S/T finite difference method. Instead, an
integrated form of Equation 4-3 is used, wherein the analytical solution is obtained for a duration of one
time step, as explained in Appendix IX of the User's Manual (Huber and Dickinson 1988). The method
has proven stable and non-negative with respect to the time step. However, the only removal option is
first-order decay except for the largely untested implementation of a settling velocity and constant
removal fraction, provided to SWMM version 4.4h in spring 2002 and described in the next subsection.
Considerable additional concepts for simulating storage in SWMM are provided during the discussion of
ponds, in Section 6.4. The transition between plug flow and complete mixing is also discussed therein.
4.4 MODIFIED STREETER-PHELPS OPTION
4.4.1 Objectives
The Transport Block is typically applied for urban drainage systems consisting of conduits and open
channels, but there are often instances in which it would be useful to have a simplified receiving water
quality simulation capability available. This is available with the modified Streeter-Phelps option, for
which parameters are input in data groups F3 - F6 (with reference to SWMM version 4.4h). These
options permit dissolved oxygen (DO) to be linked to decay of carbonaceous biochemical oxygen demand
(CBOD) and nitrogenous oxygen demand (NOD). This option is probably warranted only when
Transport is used to simulate a stream extending several miles downstream from the urban loading. Most
sewer systems remain saturated with DO because of their high turbulence, in spite of the heavy oxygen
demand. Nonetheless, this modified Streeter-Phelps option may be run for any Transport channel,
conduit, or storage unit. In fact, quality routing is performed for any element containing a volume of
water, including wet wells and manholes containing surcharge. As explained earlier, use of SWMM
version 4.4h will diminish with the release of SWMM5, hence, the descriptions that follow may be
considered as candidate algorithms for inclusion in SWMM5 flow objects.
The modified Streeter-Phelps method is included as a WASP model option (Wool et al. 2001), and the
theoretical formulation (described below) in SWMM is very similar. The WASP model itself is another
(much more comprehensive) option for simulation of receiving water quality. A special hydrodynamic
link from the version 4.4h Transport Block allows Transport to "drive" the WASP water quality model by
providing the needed file of flows, velocities, and storages. A similar option is available in the Extran
Block. Although the SWMM Transport formulation is based on the WASP model, classic DO kinetics
apply (e.g., Thomann and Mueller 1987, Chapra 1997).
The presentation in this subsection also serves the purpose of introducing a more general form of CFSTR
kinetics than is provided by Equation 4-3. Although the kinetics to be demonstrated in Equations 4-9 - 4-
12 are specifically for the Modified Streeter-Phelps formulation, it will be seen that variants on Equation
4-9 (for just one constituent) are used in the model algorithms of several models that are discussed in
Chapters 6 and 7.
4-28
-------
4.4.2 Theory
Classic Streeter-Phelps simulation of water quality applies to steady-state flows and considers only input
of ultimate biological oxygen demand (BODU or CBOD in notation that follows), linked to DO through its
first-order decay. In turn, DO is depleted by BOD and replenished by reaeration. The so-called modified
Streeter-Phelps includes an additional nitrogenous oxygen demand, wherein total Kjeldahl nitrogen
(TKN) decays to nitrate-nitrogen (NO3-N). Formally,
Organic-N > Ammonia > Nitrite > Nitrate (4-7)
during which Nitrosomonas bacteria convert ammonia to nitrite and Nitrobacter bacteria convert nitrite to
nitrate. Hence, there are three reactions to consider. The reactions are usually considered to be first-
order, but Michaelis-Menten (or Monod) kinetics may also be employed (Chapra 1997), an option in
WASP (Wool et al. 2001) but not SWMM. The modified Streeter-Phelps method simplifies the process
by ignoring the conversion of organic-N to ammonia and using their sum, TKN = organic -N + ammonia-
N, as total nitrogenous oxygen demand, or NOD. The NOD is decayed directly to nitrate, without the
typically-brief intervening nitrite step. Hence, the true three-reaction process is reduced to one. Chapra
(1997, p. 425) points out the shortcomings of this method, including the fact that because of the delay in
converting organic-N to ammonia, there is usually a delay in the appearance of a nitrogenous oxygen
demand relative to CBOD. Nonetheless, the objective was not to develop the Transport Block as a
sophisticated receiving water quality simulation model, but rather to provide for at least a minimal DO
simulation capability, similar to this intermediate (modified Streeter-Phelps) option in the WASP model.
When ammonia is oxidized to nitrate, oxygen is used according to
NH3+ + 2 O2 -> NO3 + H2O + H+ (4-8)
From the ratio of molecular weights, 64/14 = 4.57 mg of oxygen are needed to convert one mg of NH3-N
to NO3, where 64 = 2 x molecular weight of O2 and 14 is the atomic weight of N. This stoichiometric
factor is included in the coupled conservation equations.
The SWMM Transport Block simulates the linked DO-BOD-CBOD process in four steps for every
element containing storage (including pump wet wells and surcharged manholes), as follows (symbols are
defined in Table 4-1 following the four equations):
Step 1, for CBOD = d:
V^L + C1^ = (QIC;n+L1)(l-R1)-QC1-k1C1V-vsl(l-F1)C1As (4-9)
(Equation 4-9 will also serve as a more general reference equation for CFSTR kinetics in sections that
follow, wherein Cj would be any constituent of interest.)
Step 2, for NOD = C2:
2 = (QIC2n+L2)(l-R2)-QC2-kNC2V-vs2(l-F2)C2As (4-10)
dt dt
StepS, for NO3-N = C3:
4-29
-------
= (QiC;n+L3)(l-R3)-QC3+kNC2V-vs3(l-F3)C3As
at
(4-11)
Step 4, for DO deficit D = Cs - DO:
V+ D=
dt dt
(4-12)
Table 4-1. Symbols and parameters for linked DO-BOD-NOD simulation, Equations 4-9 - 4-12.
Equation
Symbol
V
Ci
D
DO
cs
Qi
Q
dV/dt
Cm
i
L,
R
ki
Vsi
As
Fi
kN
k2
SOD
d
SWMM
v. 4.4h
Name
VOL
CPOL
DEFICIT
CPOL(4)
CSAT
QI
QO
DVDT
CPOL
TREMOVE
DECAY(l)
VSETL
AS
FDIS
DECAY(2)
DECAY(4)
SOD
DEAR
Units3
ft3
mg/L
mg/L
mg/L
mg/L
cfs
cfs
cfs
mg/L
cfs-mg/
L
none
I/day
ft/s
ft2
none
I/day
I/day
g/ft2-
day
ft
Definition
Volume
Concentration of constituent i
DO deficit = Cs - DO
Dissolved oxygen cone.
Saturation DO cone.
Inflow to element
Outflow from element
Change in volume of element during
one time step
Concentration of inflow
Other loads. In SWMM, these are due
to possible scour and deposition
Removal fraction for constituent i
applied to incoming load to element,
First-order BOD deoxygenation
coefficient, sometimes known as kd.
Settling velocity for constituent i
Water surface area of element
Dissolved fraction of constituent i (does
not settle)
First-order NOD deoxygenation
coefficient.
Reaeration coefficient, sometimes
known as ka.
Sediment oxygen demand
Average depth = cross sectional area /
top width = volume / surface area
Input
Para-
meter
Y
Y
Y
Y
Y
Yb
Y
Value
Rangec
0-0.9
0.1-5
0.1-2
1-5
0.006 - 1
a Units are for those required for input if an input parameter, else, for internal use in the Transport Fortran
code.
bThe reaeration coefficient may also be computed by the program as a function of velocity, depth, and
wind speed.
0 Value ranges are given only for input parameters. When no range is entered, values are too site specific
to list. Good sources for parameter estimates include Mills et al. (1985), Schnoor et al. (1987), Thomann
and Mueller (1987), Chapra (1997) and Wool et al. (2001).
4-30
-------
When the modified Streeter-Phelps option is used, CBOD, NOD, NO3-N and DO must be the first four
constituents simulated. Additional quality constituents can be identified starting with constituent 5, etc.
The total mass of a constituent consists of a dissolved and particulate form, i.e.,
Ctot = Cdis + Cpart (4-13)
where
Ctot = concentration of total mass of constituent, mg/L,
Cdls = concentration of dissolved component, mgdls/L, and
Cpart = concentration of particulate component, mgpart/L.
The dissolved fraction may be determined if the partition coefficient, Kd, is known (Chapra 1997):
r = KdCdls (4-14)
where
r = ratio of mass in particulate (adsorbed) form to mass in dissolved form, mgpart/mgdls, and
Kd = partition coefficient, L/mgdls.
The relationship (isotherm) for r between the particulate and dissolved fraction may be in a linear, power
function (Freundlich), or Monod-type (Langmuir) form (Chapra 1997) and may be determined
experimentally. The particulate concentration is related to the concentration of suspended solids, CSs, by
Cpart = r Css = Kd Cdls Css (4-16)
Hence, the total concentration is
Ctot = Cdis + Kd Cdis Css (4-17)
and the ratio of dissolved to total is
1 (4-18)
KdCss
and the particulate fraction is
. _ K,a. (4-19)
1-i-l? P
i -r jxd^ss
Hence, the dissolved and particulate fractions in Equations 4-9 - 4-12 and more generally for any
constituent can be determined if the partition coefficient is known.
Equations 4-9 - 4-12 illustrate the typical way of linking the decay of CBOD and NOD to oxygen
demand through first-order rate processes. The equations are solved in order, and the solution to earlier
equations (e.g., 4-9 for Ci and 4-10 for C2) are inserted into later equations; all of which are solved
analytically using average values for flow and volume over a time step. The solution process is described
in the version 4 documentation (Huber and Dickinson 1988) Appendix XI and documented with comment
statements in the Fortran subroutines QUAL, QUALPARM, QUALSOLN, and REAERATE.
Conceptually the solution process is likely to be simpler in SWMM5 (if implemented) through the use of
a Runge-Kutta solution for the four simultaneous equations.
4-31
-------
The equations have similar parameters, with the exception of equation 4-12 for DO deficit. No removal
fraction is assumed to be available for DO, nor is there a scour-deposition load. Both effects are
accounted for in the relation of DO to CBOD and NOD. Although nitrate is allowed to be non-
conservative, in natural waters this is usually manifested by uptake by algae. While a non-conservative
effect could be simulated with Transport, this model is unable to "close the loop" involving conversion to
organic nitrogen, etc. A complete ecological model that includes algal and phosphorus dynamics is
needed if such effects are to be simulated, such as WASP (Wool et al. 2001) or HSPF (Bicknell et al.
1997).
Additional assumptions include:
First-order rate constants are not differentiated in Transport between sewage or "bottle" rates and in-
stream rates. However, a different ki value can apply along stream segments than applies for
conversion of BODS to CBOD. The value used to convert 5-day BOD (BODS) to ultimate BOD
(CBOD) is the DECAY value entered for CBOD. The conversion is (Chapra 1997):
with units of day"1 fork]. Hence if BOD loads are provided for BODS rather than for CBOD, inflows
will be converted to CBOD by Transport using Equation 4-20. Differing in-stream values for ki and
kN may be entered for desired elements, else earlier values entered serve as the default values.
Differentiation of ki values used in the stream into stream decomposition and stream settling rates
(day"1) must be made by the user. But since settling is included directly as a removal process, this
should not be necessary.
The simplification of nitrogen dynamics into one first-order process has already been discussed. This
is the fundamental assumption of the modified Streeter-Phelps equations.
"Travel time" commonly encountered in Streeter-Phelps formulations here is volume/flow, V/Q, for
an element, each of which is treated as a CFSTR. Concentrations are averages for the entire volume
of the element, and the same as the outflow concentration.
It is important to consider the volume change term, dV/dt, especially for situations in which an
element is only draining (no inflow) or only filling (no outflow). The dV/dt term can also act to
concentrate constituents when there is evaporation. Evaporation is not currently included in the
Transport Block, but the same equations apply in the Runoff Block where the effect of evaporation
can be seen to concentrate pollutants in very shallow flows.
Settling is allowed for every constituent except DO (or DO deficit). For a constituent that is normally
dissolved, such as NO3-N, simply set vsl = 0 and/or the dissolved fraction, fj = 1.0.
Temperature corrections may be applied to the ki and kN values according to the Arrhenius equation
(Chapra, 1997):
k(20°C) = k(T°C) 0(T"20) (4-21)
where 0 is a dimensionless temperature coefficient. Typical values are 1.047 for BOD (i.e., for KI) and
1.08 for NOD (i.e., for KN) (Thomann and Mueller 1987). "Theta values" THETA1 for BOD and
THETA2 for NOD must be provided as input values to the program.
4-32
-------
4. 4. 3 Reaeration
Oxygen is replenished by flow-driven and wind-driven reaeration. Options for both are similar to those
of the WASP model (Wool et al. 2001), which uses Covar's (1976) formulation for the flow-driven
reaeration coefficient, K2, and O'Connor's (1983) formulation for wind-driven reaeration. Flow-driven
options allow a reaeration equation of the form,
where
U = average velocity in flow element, ft/s,
and Coefl, CoefZ, and Coef3 are empirical coefficients commonly employed in reaeration equations
(Rathbun 1977). The user may input his/her own values for the three coefficients or use the Covar (1976)
option for which the coefficients are set to values corresponding to three different equations depending on
the flow regime. For wind-driven reaeration, O'Connor's (1983) methods are followed, involving an
iterative solution for the drag coefficient.
4.4.4 Summary
The Modified Streeter-Phelps solution just described is applicable to the discussion of BMP simulation
since treatment (e.g., decay, settling) can occur along any flow path. Apart from the linked constituents
represented by Equations 4-9 - 4-12, the current Transport and Runoff quality routing formulations are of
the form of Equation 4-9 and permit decay, settling, and removal (with load fraction R) within any flow
element. To the extent that a BMP may be represented by a fundamental process such as decay or
settling, or by a load removal fraction, Equation 4-9 is applicable, and several variations on this basic
equation will be seen in sections that follow, particularly for ponds and wetlands. If warranted, the
equation could be adapted to one -dimensional analysis instead of the CFSTR formulation, through
inclusion of an advective term. Equation 4-9 may be solved analytically (over one time step, with
average flow parameters, e.g., Medina et al. 1981, Huber and Dickinson 1988, Appendix IX) or
numerically for the downstream concentration. Hence, BMP simulation in SWMM5 will likely include
computations involving fundamental processes as well as simple, empirical removal fractions.
4.5 INFILTRATION
Infiltration into the soil is simulated only for Runoff Block overland flow planes. With the capability to
route overland flow from one overland flow plane to another, infiltration of runoff diverted to large
surfaces such as lawns and vegetated buffers may be simulated easily (see Section 4.6). Miniature
overland flow surfaces as might be found in rain gardens, roof vegetation, and infiltration trenches may
also be simulated with this option; the key assumption is one of vertical "walls." Hence, many LID
options can be simulated in this manner, especially since overland flow planes are also subject to
evapotranspiration (ET) and possible groundwater interaction. Since overland flow is also subject to first-
order decay, constant removal fractions (based on load), and constant settling velocities, these plane
segments offer several options for simulation of infiltration BMPs. However, the lack of infiltration from
more general channel segments limits SWMM's ability to simulate swales and infiltration from porous
channels, not to mention storage devices in general. Part of the difficulty is determining the effect of
sedimentation and water depth on infiltration rates. In real systems, some measure of maintenance and
perhaps seasonality should also be considered.
Porous pavement is an important option for reduction of runoff volumes. The current SWMM can
simulate porous pavement to the extent that the infiltration through the pavement (or paving stones) can
4-33
-------
be modeled using Horton or Green-Ampt infiltration equations (see Section 12.3). Subsurface drainage
(e.g., out a highway embankment) may be simulated using the groundwater option (James et al. 2001).
But SWMM does not simulate quality processes in groundwater.
To summarize, infiltration may be simulated only for overland flow planes, not for channels. This still
permits a wide variety of LID and BMP options to be modeled, but the lack of combined infiltration and
channel flow routing is a SWMM limitation.
4.6 OTHER WET-WEATHER CONTROL OPTIONS
Other wet-weather control options include non-structural measures such as street cleaning, catch-basin
cleaning, and pollutant load reduction. Street cleaning may be simulated directly in the SWMM Runoff
Block. Catch-basin cleaning and other pollutant load reduction measures (e.g., "good housekeeping")
may be simulated only by a reduction in buildup rates or assumed EMC values.
Several hydraulic control options may be simulated in the Extran Block and to a lesser degree in the
Transport, Runoff, and S/T Blocks. Within Extran for instance, flow regulation based on stages and
timed orifice settings may be simulated, as well as complex combinations of storage, pumping, and
bypassing. The Extran capability points toward real-time control (RTC) simulation, but such capability
awaits future enhancements in SWMM5. SWMMS's ability to be stopped in the middle of a simulation
for a change in regulator settings and/or other variables is especially intriguing for RTC simulation.
Although flow rates and volumes may be tabulated along various pathways, corresponding water quality
is not yet simulated within Extran. If a constant EMC can be assigned to various pathways, after-the-fact
estimates of loads may be made.
Continuity checks can be used to reflect various removal pathways including infiltration. Iterative
application of SWMM can be used to design a facility with the desired combination of quantity and
quality control (Huber 200 Ib) - within the limits of the model to simulate such controls.
4.7 LID SIMULATION OPTIONS
Hydrologic source control is at the heart of LID, for which every effort is made to retain stormwater at or
near its source and dispose of it via infiltration and ET. For LID technologies, modeling options are
needed that allow runoff to be directed from one subcatchment to another (for areas with different slopes,
soil types or ground cover), and that allow impervious areas to be routed over pervious areas (and vice
versa), e.g., rooftop runoff to be routed over lawns. Overland flow infiltration-based controls may be
simulated in the SWMM Runoff Block by redirecting runoff from impervious areas onto pervious areas,
and from one subcatchment to another (Huber 200 la), in the manner shown in Figure 4-1. It shows
routing from the impervious sub-area of a subcatchment to the pervious sub-area of a subcatchment. The
scheme is similar for flow from pervious to impervious sub-areas, and subcatchment outflow can also be
directed to another subcatchment. Lee (2003) demonstrates the model's efficacy for analysis of
distributed stormwater quantity and quality control alternatives.
4-34
-------
Width
To channel/pipe, inlet or
another subcatchment
Figure 4-1. Conceptual routing from the impervious sub-area of a subcatchment to the pervious sub-area of a
subcatchment (Huber 2001a).
This rerouting also allows for the simulation of buffer strips or riparian zones. Inflow to the downstream
subcatchment is distributed uniformly over the downstream subcatchment in the same manner as rainfall.
This can be done because of the nonlinear reservoir flow routing method in which there is no longitudinal
variation through the subcatchment. Runoff Block overland flow planes simulate surface infiltration and
evaporation. Soil moisture accounting is possible if the subsurface flow option is used. In this case, ET
from the upper unsaturated zone and lower saturated zone may both be simulated. However, the link with
surface infiltration is indirect, affecting infiltration only if the soil becomes completely saturated.
Otherwise, infiltration capacity for the Horton or Green-Ampt method is regenerated heuristically, and
not as a function of ET. A direct link between regeneration of infiltration capacity and ET is needed.
However, the current SWMM can still simulate most LID options for hydrologic source control. This has
been shown by Lee (2003) and will be demonstrated in the discussion of SWMM simulation capabilities
for infiltration that follow, as well as in a detailed example for Portland, Oregon (Chapter 14).
As mentioned earlier in this report, overland flow planes may be used to simulate any surface with an
assumed plane surface and vertical "walls" such as rooftop vegetation, vegetated buffers, and infiltration
trenches. Redirection of flow from impervious areas to pervious (simulating the effects of downspout
disconnection, for instance) has a major impact on the predicted downstream hydrograph. Not only does
this affect the peak flow, but also the total flow volume and corresponding pollutant load.
4.8 SWMM LID MODELING NEEDS
A significant feature lacking from the Runoff and Transport Blocks is the ability to simulate infiltration
from channels. This means that swales or porous channels in which routing and cross-sectional shape
effects may be important cannot yet be easily simulated in SWMM. This capability is lacking primarily
because of lack of information on the effect of water depth on infiltration rates and because of the need
for a heuristic method to simulate the effects of sedimentation and clogging of pores and possible periodic
maintenance.
Apart from processes that truly are first-order, sedimentation in storage or flow devices is the only
4-35
-------
fundamental treatment process simulated in the model; all other fundamental treatment processes must be
mimicked through manipulation of removal equations. Biological processes, chemical and physical
processes that occur in wetland, bioswales, and riparian zones can only be simulated to the extent that
they may be characterized as defined above.
Infiltration through porous (or permeable) pavement may be performed adequately with current SWMM
algorithms for infiltration and groundwater flow (James et al. 2001). This will be explained later in the
chapter on porous pavement (Section 12.3).
Finally, sediment transport can barely be simulated in SWMM - only in the Transport Block, and only
through the use of simple scour-deposition criteria. Hurdle (2001) reviews options for improving SWMM
sediment transport routines, none of which are straightforward. However, for the model to be able to
characterize treatment based on solids settling, improvements in the overall ability of SWMM to erode,
transport, deposit, and scour sediment need to be provided.
4.9 TIME AND SPACE RESOLUTION ISSUES
The Runoff Block is very stable with regard to size of simulated subcatchments. SWMM can model very
small parcels (current research has taken catchment size down to 0.03 ac, but there really is no lower
limit); however, the time step needs to be adjusted to be smaller than the retention time (V/Q). For
instance, a time step < 1 min might be used, compared to the more typical 5-min value routinely
employed. With personal computer power this should not be an issue. Finding rainfall data on a small
enough time can be a problem, as most data come at a minimum of 15-min intervals. Fifteen-minute data
are much too coarse to simulate micro-scale hydrologic processes, but from the point of view of
assessment of the vertical water balance storage options (as opposed to peak flow rates), such data may
suffice. This is because the vertical water balance processes that are characteristic of small-scale LID
options (ET, infiltration, soil moisture routing) typically occur much more slowly than do overland flow
runoff processes. Some of these issues will be clarified through additional experience with application of
SWMM continuous simulation to miniature subcatchments (e.g., rain gardens). For a process model like
SWMM, the issue of temporal and spatial variability is largely an issue of data availability and the
amount of detail desired in the simulation run. In case that rainfall data shorter than 15-minute interval is
desired, a data disaggregation procedure is presented in the companion project report (Heaney and Lee
2006).
4.10 EXAMPLES OF SWMM BMP SIMULATION
While some additional SWMM capabilities for BMP simulation will be described while discussing the
various BMP options, example SWMM simulations reflecting LID and BMP simulation capabilities will
be provided in Chapters 14 and 15 toward the end of this report. This is done in lieu of incorporation of
examples into each BMP section since these two examples are lengthy.
4-36
-------
5 ALTERNATIVE MODELS AND APPROACHES
For each of the viable BMP/LID alternatives defined in Tables 1-1 and 1-2, modeling concepts and
mathematical formulations need to be developed that can be used in applying alternatives to the SWMM
model based on the state-of-the-engineering knowledge and information. The modeling parameters, e.g.,
soil characteristics and antecedent moisture content that reflect local site-specific conditions as well as
possible seasonal effects, should be incorporated into the formulations whenever possible, and the
modeling limits of each BMP/LID alternative that can be incorporated into SWMM based on present
knowledge should be well defined.
Several models were evaluated for their ability to simulate the quantity and quality processes in urban
BMP alternatives; a useful review of candidate models is provided by Trepel et al. (2000). For the sake
of brevity, only models found to be most applicable in urban areas were evaluated during this study; these
are listed in Table 5-1. Hence, a number of models that might be suitable for agricultural BMP analysis
are not included (Donigian et al. 1995), nor are models that do not simulate water quality. Also, models
that primarily deal with water quality in receiving streams are also not considered here, such as WASP
(Wool et al. 2001) and HSPF (Bicknell et al. 1997). The point is made later that it would be very useful
to provide an easy method to transfer hydrographs and pollutographs from SWMM to models such as
WASP and HSPF.
Models reviewed are:
DMSTA: Dynamic Model for Stormwater Treatment Areas (Walker 1995, Walker and Kadlec 2002)
MUSIC: Model for Urban Stormwater Improvement Conceptualization (Wong et al. 2002)
P8: Program for Predicting Polluting Particle Passage through Pits, Puddles, and Ponds (Walker
1990)
PREWET: Pollutant Removal Estimates for Wetlands (Dortch and Gerald 1995)
REMM: Riparian Ecosystem Management Model (Inamdar 1998a,b)
SLAMM: Source Loading and Management Model (Pitt et al. 1999b, Pitt and Voorhees 2000)
VAFSWM: Virginia Field Scale Wetland Model (Yu et al. 1998)
WETLAND: Wetland water balance and nutrient dynamics model (Lee et al. 2002)
WMM: Watershed Management Model (Wayne County, MI 1998)
Components of these models will be discussed in following sections with regards to methods for
simulating BMPs. As part of each method-related chapter, current (version 4.4h) SWMM simulation
options will also be provided.
5-37
-------
Most of the models in Table 5.1 can simulate the performance of a variety of BMPs. In some cases, a
model will be discussed primarily within the framework of just one BMP. For example, MUSIC will be
discussed primarily in Chapter 7 under the wetlands category, even though some of that discussion will
relate to other BMPs as well.
Although WMM has proven very useful for screening analyses, it is a spreadsheet model, not a process
model. That is, WMM simulates removal through the use of seasonal or annual removal coefficients,
during a static simulation. Hence, WMM procedures will not be discussed in detail in this report. On the
other hand, the model and its documentation may well be valuable for data and coefficients.
Table 5.1. Simulation models that can simulate urban BMP performance.
Model
DMSTA
MUSIC
P8
PREWET
REMM
SLAMM
VAFSWM
WETLAND
WMM
Ponds
X
X
X
X
X
Infiltration
Trenches
X
X
Grass
Swales
X
X
X
X
Extended
Detention
X
Bio-
retention
and
Wetlands
X
X
X
X
X
X
X
Dry
Wells
Filter
Strips
X
X
X
Porous
Pavement
X
Other
Devices
X
5-38
-------
PONDS
6.1 INTRODUCTION
Modeling ponds as a BMP involves storage and release of excess stormwater capture volumes based on
hydraulic controls (Urbonas and Stahre 1993). Ponds may be classified in three ways:
1. Wet retention or "wet pond," in which there is a continuous pool of water that will fluctuate up and
down during storm events.
2. Dry detention or "dry pond," which fills during a storm event and drains soon after. Dry detention
areas are often designed primarily for flood control and usually serve multiple purposes, such as
recreation.
3. Extended dry detention is essentially the same as a dry pond except that it is deliberately designed to
drain more slowly, thus providing more of a water quality benefit.
The word "retention" implies that some water retained in a pond and between storms is subject only to the
vertical water balance of ET and infiltration. The word "detention" implies that water is only detained,
and storage is temporary. Wet ponds will typically include detention storage above the permanent pool
for purposes of flood control and additional water quality benefits.
Although this chapter is entitled "ponds," the principles described apply to most types of storage devices,
that is, wet-weather control devices that provide storage, including wetlands, overland flow and flow in
swales, concrete tanks (as for combined sewer overflow control), in-system storage in pipes or channels,
and any control that may enhance sedimentation.
Many processes are responsible for the pollutant removals observed in retention and detention ponds.
Physical sedimentation is the most significant removal mechanism (Pitt and Voorhees 2000) and is
traditionally modeled based on the hydraulic overflow rate (described below, Metcalf and Eddy 2003).
However, biological and chemical processes can also contribute important pollutant reductions. The use
of aquatic plants, in a controlled manner, can provide still more pollutant removal. Wet ponds also are
suitable for enhancement with chemical and advanced physical processes. Infiltration may or may not be
an issue with ponds; often a liner or high groundwater table is required to maintain a permanent pool
(hence, infiltration is discussed in the Infiltration Trenches section of this report.)
SWMM currently models storage devices quite well, simulating settling and first-order decay directly,
with S/T removal equations for other processes. When searching for improvements to SWMM for pond
simulation one issue is biological treatment based on second-order reactions. This would be useful if
trying to model a type of activated sludge process or flocculation within a pool, although this may be a
6-39
-------
minor issue. Mineral suspensions and primary sedimentation are best characterized as discrete-particle
sedimentation and are often sufficient for characterizing water quality (Tchobanoglous and Schroeder
1985).
The models investigated have few methods for simulating biological treatment directly. Of the models
investigated, only P8 has the limited ability to simulate second-order rates of reaction directly. P8 may
also calibrate pond performance with a "particle scale removal factor" that may be used to emulate pond
processes indirectly.
Simulation of ponds as a CFSTR is a two-part process. At each time step, flow continuity is maintain, in
a variant of the lumped continuity equation,
(6-1)
where
V = pond volume,
Qj = summation of surface inflows,
Q = pond outflow to surface, e.g., through outflow structures,
A = pond surface area,
P = precipitation on pond,
ET = evapotranspiration from pond, and
G = percolation or infiltration into soil beneath the pond (could be negative).
This same equation applies to all surface BMPs, including wetlands, swales, etc. The change in volume
term, dV/dt, appears in the CFSTR kinetic equation 4-9 and is an important component of any numerical
solution. Variations of Equation 4-9 will be seen in presentations of the several pond (Chapter 6) and
wetland (Chapter 7) models that follow.
6.2 SIMULATION OF PONDS WITH P8
6.2.1 The Model
The Program for Predicting Polluting Particle Passage through Pitts, Puddles and Ponds, or P8, is used to
model generation and transport of stormwater runoff pollutants in an urban setting (Walker 1990).
Calculations are performed on continuous water-balances and mass-balances. Primary applications are
for evaluating site plans for compliance, with treatment objectives expressed in terms of removal
efficiency for TSS, and BMP design to achieve treatment objectives. Secondary (and less accurate)
predictions from this model are runoff quality, loads, violation frequencies, water quality impacts due to
proposed development and generating loads for driving receiving water quality models (Walker 1990).
6. 2. 2 Second- Order Reactions
A fundamentally different approach to simulating contaminant behavior and partitioning in devices
currently under investigation is to assign each contaminant to a separate particle class and use second-
order decay kinetics instead of first-order settling. This would reduce removal rates as concentrations
decreased. Second-order kinetics are consistent with removal mechanisms involving particle interactions
(e.g., flocculation) as opposed to discrete settling. The applicability of second-order kinetics has been
demonstrated for hydrocarbons in Nationwide Urban Runoff Program (NURP) settling column tests
(EPA. 1983, Volume II), phosphorous removal in reservoirs (Walker 1985) and TSS, phosphorous and
zinc removal in settling columns (Walker 1990). The user is required to input decay coefficients, which
6-40
-------
can make the model more flexible in modeling regional performance differences in devices at the expense
of the extra data requirements. But the limits of usefulness in using the second-order decay stem from the
lack of parameter estimates for such coefficients currently available in the model and in the literature.
In P8 the rates of reaction are used to calculate device (pond) concentrations. Each device is assumed
completely mixed for computing concentrations and outflow loads. This is essentially the same way that
SWMM currently calculates concentration in its Runoff and Transport Blocks except for the integration
of the second-order rate constant, K2 (not to be confused with the reaeration coefficient), and inclusion of
the particle scale removal factor, f
Device mass balances are calculated as follows (definition of symbols follows the four equations):
dM dVC
=W-D.M , ,
dt dt
The right hand side of Equation 6-2 is the following variant of Equation 4-9 that includes first and
second-order decay, plus settling:
D = S + f k + f K2 Cm + f vs ^ (6.3)
Assuming a constant volume, and average values over a time interval, At, the analytical solution for mass
M is:
M2 =^ + (Ml-^)e ifD>0 (6.4)
D x D
[i
=M+WAt ifD = 0 (6-5)
where
D = sum of loss terms (1/hr),
Cm = average concentration during step (mg/L),
V = average device volume during time step (ac-ft),
M = C-V = mass in device, (ac-ft-mg/L), with subscripts 1 and 2 indicating beginning and
end of time step, respectively,
At = time step length (hours),
W = total inflow load to device (ac-ft-mg/L/hr),
Q = average outflow from device, from flow balance (ac-ft/hr),
vs = particle settling velocity (ft/hr),
A = average device surface area during time step (acres),
k = first-order decay coefficient (1/hr),
K2 = second-order decay coefficient (1/hr-mg/L), and
f = device -specific particle removal scale factor.
Notice that the analytical solution is not "exact" since an average concentration, Cm, is used to evaluate
the nonlinear second-order decay term. The addition of a user-defined second-order rate equation can be
6-41
-------
used to simulate flocculation or biochemical reactions, but this is limited by a lack of defined second-
order rate constants in the literature. No default decay coefficients are provided in P8, and if desired must
be supplied by the user. Most typical K2 values available are for activated sludge processes and have
specified ranges of usefulness based on temperature and mixed liquor suspended solid concentrations
(Tchobanoglous and Schroeder 1985). These parameters are difficult to control in BMPs, and therefore
limit this modeling option's usefulness.
6.2.3 Particle Removal Scale Factor
Another P8 approach is to use the particle removal scale factor, f, which allows for the easy calibration of
an increase or decrease in pond removal efficiency. The f-values adjust the sum of removal rates for each
device, and are usually set to 1.0. The f-values can be used to account for effects of vegetation or other
factors that affect particle removal, e.g., macrophytes can increase particle removal by increasing surface
area, stabilizing bottom sediments, and through biological mechanisms. Removal efficiency curves
developed in Australian ponds with macrophytes (Phillips and Goyen 1987, Lawrence 1986, as presented
by Walker 1990) correspond to removal scale factors of 2-3 for suspended solids and 3-4 for total
phosphorous attributed to macrophyte presence in wet detention ponds. Alternatively, a removal scale
factor f < 1 can account for short circuiting or other poor hydraulic designs. See Section 6.5 for an
alternative heuristic approach leading to similar results.
6.2.4 Pollutant Removal
Pollutant removal under dynamic conditions occurs when particle settling velocities exceed the basin
overflow rate. Removing solids will also remove much of the pollutants of interest. Notable exceptions
of potential concern include dissolved forms such as nitrates, chlorides, soluble zinc, pathogens, 1,3-
dichlorobenzene, fluoranthene, and pyrene (Pitt and Voorhees 2000). The P8 model uses the traditional
hydraulic loading rate method for dynamic settling (e.g., Fair et al. 1968, Minton 2002, Metcalf and Eddy
2003) for all devices modeled (ponds, swales, bioretention facilities, filter strips), in which particles with
settling velocities greater than the hydraulic loading rate (or sometimes, the "overflow rate") are removed.
That is, particle removal occurs when
vs >q =0- (6-6)
where
vs = settling velocity,
q = overflow rate or hydraulic loading rate, flow rate/area or length/time,
Q! = inflow, and
A = surface area.
The method is used to determine removal in all devices modeled by P8 (ponds, grass swales, bioretention
facilities and filter strips). In some applications, the outflow is used instead of the inflow to determine the
overflow rate (Sections 6.3 and 6.5).
Within P8, the inflow is computed using a mean storm intensity and watershed area, so that:
A,,, R i
6-42
-------
where
q
A
Aw
Rv
i
12
= average overflow rate (ft/hr),
= pond surface area (acres),
= watershed area (acres),
= watershed runoff coefficient (runoff volume/rainfall volume),
= mean storm intensity (in/hr); ~ 0.06 in/hr is used as default in P8, and
= conversion factor from inches to feet.
An assumed particle size distribution (Table 6-1) is then used to determine the amount of settling that
occurs. The first class represents the dissolved (non-settling) fraction of water quality constituents. The
remaining classes are based on NURP settling velocity distributions. Ideally, these would be supplied by
the user with site-specific stormwater treatability data, but such data are usually sadly lacking.
Particle fractions (mg/kg), sometimes called "potency factors," are used to translate particle
concentrations of total suspended solids (TSS) into associated pollutant values. These fractions are
similar to SWMM's constituent fractions in the J4 data group of the Runoff Block and are one way to
simulate adsorption. That is, a "particle fraction" or "potency factor" or "constituent fraction" is related
to the partition coefficient, Kd, used in sorption kinetics through Equation 4-16 (Chapra 1997). These
particle fractions have been calibrated in P8 to "typical urban runoff so that the median SS EMC
corresponds to the values reported by NURP, based primarily on runoff concentrations and settling
velocity distributions (USEPA 1983, Walker 1990).
Table 6-1. Particle class default values in P8 (Walker 1990).
Class
P0%
P10%
P30%
P50%
P80%
Description
Dissolved
10thPercentile
30th Percentile
50th Percentile
80th Percentile
% of TSS
0
20
20
20
40
Settling Velocity (ft/hr)
0
0.03
0.3
1.5
15
Buildup-washoff parameters have been calibrated for both pervious and impervious areas to produce an
EMC of 100 mg/L TSS for a median site, and 300 mg/L for a 90th percentile site as per NURP. This
method is independent of stormwater volumes and ignores any variation in concentration (first-flush
effects) with large storm events and due to possible construction site runoff, which can yield much higher
TSS concentrations.
Particle compositions (mg/kg) are then used to translate particle concentrations into concentrations of
total suspended solids (TSS), total Kjeldahl nitrogen (TKN), total phosphorous, copper, lead, zinc and
hydrocarbons. These compositions have also been calibrated so that median, event-mean runoff
concentrations correspond to values reported by NURP (USEPA 1983) as listed in Table 6-2.
This calibration is based on a simulation of 1983-1987 Providence Airport rainfall. High site-to-site
variability is reflected in the 2 to 3-fold differences between the median and 90th percentile sites, and
implies considerable uncertainty in predicting actual contaminant concentrations. Calibration with local
or regional runoff data will help to reduce this uncertainty (Walker 1990).
6-43
-------
Table 6-2. Example of P8 calibrated runoff concentrations (Walker 1990).
Component
total suspended solids
total phosphorous
total Kjeldahl nitrogen
total copper
total lead
total zinc
hydrocarbons
Median, EMC, mg/L
NURP Median Site
100
0.33
1.5
0.034
.020 (a)
0.16
2.5 (b)
90th % site
300
0.7
3.3
0.093
0.05 (a)
0.5
5.0 (b)
% Dissolved
0
30
40
40
10
40
10
a - NURP lead values reduced to account for > 10-fold reduction in gasoline lead content
b- Hydrocarbons estimate from load factors reported by Hoffman (1985)
6.3 SIMULATION OF PONDS WITH SLAMM
SLAMM (Pitt and Voorhees 2000) calculates particle deposition in wet detention ponds using the same
hydraulic loading rate methodology just described for P8, although SLAMM's authors refer to it as the
"upflow velocity method" (Linsley and Franzini 1964). It is the same as hydraulic loading rate except
that Pitt and Voorhees (2000) define it as the ratio of outflow rate to surface area. This is reasonable
since pond outflow rates generally govern the time required to drain the dry pond or return a wet pond to
its permanent pool level. Hydrograph routing through the pond is first performed using the storage-
indication method (see also Section 6.5) as implemented in the RESVOR reservoir routing subroutine of
the Natural Resources Conservation Service in Technical Releases 20 and 55 (SCS 1986).
SLAMM expands on the storage-indication procedure by calculating incremental upflow velocities
(hydraulic loading rates) for each calculation interval. SLAMM automatically determines the most
efficient calculation interval. Any particle that has a settling velocity greater than this upflow velocity
will be retained in the pond. The user describes a particle size distribution for the inflowing water that
SLAMM uses to calculate the particle settling rates from Stokes' law modified for deviations from
laminar flow (e.g., Fair et al. 1968).
Stokes' equation (Fair et al. 1968), used to compute the terminal fall velocity, or settling velocity, vs (ft/s),
for a sediment particle in laminar flow is:
(6-8)
where
d
SP
v
g
= sediment diameter (ft),
= specific gravity of sediment,
= kinematic viscosity (function of water temperature), (ft2/s),
= acceleration due to gravity (ft/s2).
Limitations of Stokes' law are discussed below, in Section 6.5. SLAMM calculates the critical particle
sizes retained in each calculation interval and sums the retained particles for the complete event.
Hydraulic performance of an outfall pond is also summarized by giving the peak flow rate reduction
factor and the pond flushing ratio (ratio of incoming runoff volume to normal pond volume) for each
event. Peak flow reduction affects downstream impacts, a BMP evaluation criterion (Section 3.1).
6-44
-------
6.4 POND SIMULATION WITH MUSIC
This model simulates sedimentation using methods similar to those described for SWMM in the next
section. Since a primary application of MUSIC is to simulate the treatment obtained by storage in
wetlands, discussion of the MUSIC algorithms will be deferred to Chapter 7.
6.5 POND SIMULATION WITH SWMM
Hydraulic and water quality procedures for simulation of "storage" in SWMM are explained in
considerable detail in the documentation (Huber and Dickinson 1988, James et al. 2002a,b) as well as in
texts such as Nix (1994), Stephan Nix being the primary author of the SWMM Storage/Treatment Block.
Flow routing through storage is performed using the storage-indication (S-I, also known as the Modified
Puls) method (McCuen 1982, Huber and Dickinson 1988, Bedient and Huber 2002). The S-I method is
simply a convenient numerical scheme for solution of the combined continuity and reservoir outflow
equations for the two unknowns, storage and outflow. As shown in the detailed SWMM example of
Chapter 15, required data include surface area vs. depth (from which volume vs. depth may be computed)
and an outflow rating curve (flow vs. depth). The latter may be in the form of a table, pump curve, or
power equation (e.g., for a weir or orifice). The S-I scheme is also used in the Transport Block for flow
routing through storage, whereas storage in the Runoff Block is evaluated by the mathematical solution of
the nonlinear reservoir equation, and in the Extran Block, by solution of the Saint-Venant equations
applied to the storage device (Roesner et al. 1988). Since the S/T Block (and sometimes the Transport
Block) is the primary module used for simulation of quality in storage devices, focus will be upon its
application for ponds.
As described in Sections 4.2 and 4.3, quality transformation may be simulated in three ways:
1. first-order decay with either plug flow or complete mixing,
2. a general "all-purpose" removal equation (Equation 4-1), or
3. sedimentation theory, for use with plug flow.
Considering the third option briefly, sedimentation is obviously a function of settling velocity. While
Stokes' equation 6-8 could be used, more accurate is to compute settling velocity as a function of
turbulence level, as indicated by the particle Reynolds number, R,
R = (6-9)
v
where
vs = settling velocity (ft/s),
d = particle diameter (ft), and
u = kinematic viscosity (ft2/s).
Settling velocity is then defined by the balance between gravity and drag force,
-1) (6-10)
where
g = gravitational acceleration, e.g., 32.2 ft/s2,
CD = drag coefficient, a function of R, and
Sp = particle specific gravity.
6-45
-------
The iterative process to determine vs is explained by Fair et al. (1968) and Minton (2002) and was
implemented by Sonnen (1977) in a past version of SWMM. The current S/T Block uses Sonnen's
method (Huber and Dickinson 1988). This is in lieu of settling velocities entered directly by the user,
from a stormwater treatability analysis. Stokes' law is valid for R less than approximately 0.5, for which
CD = 24/R, and for which Equation 6-10 reduces to Equation 6-8.
Two options exist for pond (or any storage) simulation in SWMM: plug flow or complete mixing
(CFSTR, discussed in Section 4.3). Most ponds behave in an intermediate fashion between plug flow and
complete mixing due to short-circuiting, dead zones, and incomplete mixing (Thackston et al. 1987).
There are also two- and three-dimensional effects, including diffusion and dispersion that SWMM cannot
simulate. One common heuristic solution is to use Fair and Geyer's (1954) "tanks in series" (TIS)
method for analysis of imperfect sedimentation basins for water and wastewater treatment, in which the
fraction captured of particles that have settling velocity vs is:
(6-11)
NQIA
V /
where
R = 1 - Cout/Cin = fraction captured or retained in pond,
Q = pond outflow rate,
A = pond surface area, and
N = empirical measure of hydraulic efficiency, or number of CFSTRs in series.
The hydraulic efficiency factor, N, can reflect short-circuiting, for example, and ranges from 1 for "very
poor performance," to 3 for "good performance," to 5 or higher for "very good performance" (Fair et al.
1968, Driscoll 1986b, Pitt and Voorhees 2000, Minton 2002). Another interpretation of N is the number
of CFSTRs or "tanks" in series, as an approximation leading up to plug flow (Levenspiel 1972, Kadlec
and Knight 1996, Chapra 1997). That is, Equation 6-11 is the removal efficiency for N CFSTRs in series,
all with the same flow rate, and each with 1/N of the total pond area Ab. (The ratio, vsA/Q can be
expressed in several related forms, as will be discussed below in relation to Equation 6-14.) When N = 1,
the equation gives the steady-state performance of one CFSTR (solution of Equation 4-3 for dC/dt = 0).
The case N^ <» corresponds to perfect horizontal plug flow representation for a tank or pond (Chapra
1997),
lim(Eqn.6-ll) = 1-e (6-12)
N » co
The removal efficiency, R (fraction), due to quiescent settling (e.g., in the permanent pool of a wet pond),
is given by
^ - ^ 1 (6-13)
where
td = detention time, and
6-46
-------
h = pond depth.
All particles with settling velocity vs will be removed as long as the storm inter-event time is greater than
the detention time (neglecting resuspension and several other side effects). Note that Equation 6-13 is a
special case of Equation 6-11 with N = -1.
The effect of N is shown in Figure 6-1 . Several options for the dimensionless abscissa are shown that
represent equivalent design conditions, depending on the nature of the removal process being simulated,
including
(6-14)
Q/A h Q q
where
k = first-order decay coefficient (I/time),
k' = k-h = rate constant (depth/time),
q = hydraulic overflow rate (depth/time), and
V = volume.
Various relationships are implied among the parameters, including the fact that settling behaves as a first-
order removal process if k = vs/h. Kadlec and Knight (1996) define the Damkohler number, Da, as
Da=k'/q (6-15)
Thus, the various dimensionless forms shown in Equation 6-14 are variations of the Damkohler number.
Regarding Figure 6-1, so-called "very poor performance" corresponds to complete mixing in a pond, or N
= 1 CFSTR. This means that some entering particles can move "instantaneously" to the outlet, as in
short-circuiting. The other extreme is plug flow, in which particles settle while moving on one
continuous, horizontal path to the outlet, thereby maximizing their opportunity for removal. The range of
perfect mixing to plug flow corresponds to dynamic settling, that is, settling while outflow from the pond
is occurring. Quiescent settling corresponds to still water and no outflow (although this may be assumed
for discrete plugs, as indicated below). Given enough time, and neglecting turbulence and resuspension
due to wind, all particles for which vs-h/td > 1 will settle out.
SWMM S/T currently simulates the extremes of complete mixing and plug flow. For plug flow, an
enhancement is used to account for non-ideal settling conditions that were characterized by Camp (1946)
in the form of sediment trap efficiency (removal fraction) as a function of a turbulence factor. (This is the
"sedimentation theory" or third S/T option listed earlier.) The procedure was simplified by Chen (1975),
and its SWMM implementation is described in Appendix IV of the User's Manual (Huber and Dickinson
1988). In essence, removal in plugs corresponds to ideal quiescent conditions (Equation 6-13) when the
turbulence factor is low, and follows a reduced efficiency curve given by Chen (1975) for higher levels of
turbulence. However, this still assumes plug flow conditions. A possible enhancement to the simulated
removal efficiency would be to use Equation 6-1 1 to represent the reduction of efficiency as plug flow
conditions "deteriorate" to those of complete mixing, with its inherent short circuiting. The user would
need to supply the value of N or make an equivalent judgment regarding "very good" to "very poor"
performance, as defined for unit operations by Fair et al. (1968) and for ponds by Pitt and Voorhees
(2000) and Minton (2002). Additional guidance is provided in Section 7.7.
6-47
-------
o
0.0 1.0 2.0 3.0 4.0
vs/ (Q/A) = vstd/h = kV/Q = ktd = k7q
Plug flow
N=8
N=1 (CFSTR)
N=-1 (Quiesc)
Figure 6-1. Removal efficiency, R, as a function of dimensionless forms of rate of treatment or loading.
The particle scale removal factor, f, in the P8 model serves the same purpose as the hydraulic efficiency
parameter, N. Both may be used in a highly empirical fashion to account for increases or decreases in
removal efficiency, with plug flow as a starting point (Equation 6-12). For instance, values of-1 > N > -
oo represent still another set of curves on Figure 6-1 for which removal efficiency is improved to a value
between plug flow and quiescent settling (efficiencies for N = -2 are shown on the figure). Hence, use of
Equation 6-11 for plug flow in SWMM might be a very versatile way to represent performance, albeit still
empirical and of a curve-fitting nature.
6.6 EXTENDED DETENTION
In order to enhance sedimentation, extended detention basins are designed to empty their brim-full
volume in 24 to 48 hours, with no more than 50% of this volume being released during the first one-
quarter to one-third of the emptying period (Urbonas and Stahre 1993). Of course, local regulations may
modify these timing requirements, but these are typical of today's practice. Water quality processes in
extended detention (and detention) ponds obey the same principles as just discussed, including dynamic
settling. However, it is important to remember that there is no "upflow" velocity (at least not after
6-48
-------
possible spillway flow has ceased); the "overflow rate" is due to water draining horizontally or downward
toward a drain that is usually on or near the pond bottom. In essence, this is the procedure used in
SWMM S/T for plug flow with sedimentation. Water is assumed to move horizontally toward the outlet,
with pond depths gradually decreasing. Removal is always a function of detention time and may be
modified according to Camp's (1946) and Chen's (1975) adaptations for turbulence.
Of special interest is the behavior of dry detention and extended detention basins over the long term of
hydrologic events, during which a detention area will fill and empty according to the arrival of storms and
duration of inter-event dry periods. For instance, if another storm arrives before the detention area has
drained completely, the incoming flow will mix with the remaining water, and the removal analysis must
begin anew. The EPA probabilistic analysis (Driscoll 1986b, Urbonas and Stahre 1993), later enhanced
and expanded by Adams and Papa (2000), provides a statistical methodology for analyzing detention
performance by coupling rainfall and runoff statistics, e.g., storm duration and inter-event times,
determined by local meteorology, with detention drainage characteristics, determined by the hydraulic
design of the outlets. SWMM and other continuous simulation models provide the same type of analysis
by using long-term historic rainfall time series as drivers for the stormwater runoff that enters (and drains)
from detention and retention ponds. Hence, the complex interaction between storm events and filling and
drying of detention areas is analyzed empirically, through continuous simulation, with a statistical
analysis (e.g., using SWMM's Statistics Block) of the results.
6-49
-------
WETLANDS AND BIORETENTION FACILITIES
7.1 INTRODUCTION
Bioretention facilities include constructed wetlands, wetland basins, bioswales, and wetland channels.
Bioretention is a combination of processes served by other BMP types, and the processes that might
enhance SWMM's ability to model bioretention facilities are a combination of those previously listed.
Sedimentation resulting from storage is a fundamental unit process that occurs in these facilities.
Infiltration, filtration, flocculation, biochemical interaction, increased settling and decreased erosion due
to the presence of macrophytes also occur and have been discussed in conjunction with Tables 1-2 and 1-
3. A highly comprehensive analysis of the use of wetlands for stormwater treatment is provided by
Kadlec and Knight (1996).
Pollutants in stormwater can be removed by wetlands and ponds through a combination of 1)
incorporation into or attachment to sediments or biota, 2) degradation, and 3) export to the atmosphere or
groundwater (Strecker et al. 1992). The wetland hydroperiod (the seasonal pattern of water levels)
defines the rise and fall of surface and subsurface water that in some cases can lead to export of pollutants
to groundwater. In general, removal mechanisms are known to be physical, chemical and biological in
nature. Some important removal mechanisms include sedimentation, filtration, oxidation, adsorption,
volatilization, precipitation, nitrification and microbial decomposition. Removal mechanisms and which
pollutants they affect (Horner 1995) are shown in Table 7-1.
The mechanisms shown in Table 7-1 are not assumed to be independent from one another. Sedimentation
due to reduced flow velocities, adsorption, and filtration by vegetation are three of the major removal
mechanisms in many wetlands. However, it should be noted that any one of the eight mechanisms of
Table 7-1 can be dominant depending on the wetland's characteristics (e.g., as influenced by its
hydrology and hydraulics). Strecker et al. (1992) note this as a major reason why wetlands differ so
greatly in their pollutant removal efficiencies. Yu et al. (1998) also note differences in removal
efficiencies attributable to design parameters such as inlet and outlet configuration, length to width ratio,
and consequent residence times. Yu et al. (1998) reported greatest removal for sites that maximized the
length to width ratio. These factors obviously play an important role when designing a wetland or pond
system. Several options exist to an engineer who is interested in designing a wetland/pond that will
maximally remove pollutants from urban stormwater. These include increasing the hydraulic residence
time (HRT), providing an environment that encourages flow at a low level of turbulence so sedimentation
can be maximized, propagating fine, dense, and herbaceous plants, and establishing the system on a
medium-fine textured soil (Horner 1995, Kadlec and Knight 1996).
7-50
-------
Although a two or three-dimensional model could be used for simulation of surface flows through
wetlands, in all cases evaluated in this report, the lumped storage approach of Equation 6-1 has been used
for water volumes. Constituent kinetics are simulated through assumption of a CFSTR, through
variations on Equation 4-9, as for ponds.
Table 7-1 Wetlands/ponds pollutant removal mechanisms. (After Horner 1995.)
Mechanism
Physical
Sedimentation
Filtration
Chemical
Adsorption
Oxidation
Volatilization
Precipitation
Biological
Nitrification
Microbial Decomposition
Pollutants Affected
Solids, BOD, pathogens,
COD, P, N, metals
Solids, BOD, pathogens,
COD, P, N, metals
Dissolved P, metals, synthetic
organics
COD, petroleum,
hydrocarbons, synthetic
organics
Volatile petroleum
hydrocarbons and synthetic
organics
Dissolved P, metals
NH3-N
BOD, COD, petroleum
hydrocarbons, synthetic
organics
Promoted By
Low turbulence
Fine, dense herbaceous plants.
Outflow through porous media
also has an obvious filtration
effect.
High soil Al, Fe; high soil
organics, circum-neutral pH
Aerobic conditions
High temperature and air
movement
High alkalinity
Dissolved oxygen >2 mg/L,
Low toxics, temps. >5-7 °C
Circum-neutral pH
High plant surface area and
soil organics
7.2 DIFFICULTIES WITH MODELING MULTIPLE PROCESSES
The difficulty in modeling multiple processes comes from choosing the order in which each process
occurs, or trying to model processes concurrently. Some pollutants increase during some processes and
decrease during others (e.g., BOD in activated sludge vs. sedimentation). The order in which these
processes are modeled will affect the estimated effluent concentrations as well as the removal efficiency
in each process. Coupling standard removal efficiencies for multiple processes may also overestimate
pollutant removal. Many BMPs cannot remove pollutants below a certain level. Some BMPs output
consistent effluent quality that is not strictly dependant on influent concentrations (Strecker et al. 2001).
Two simple methods are described in the next sub-section, followed by descriptions of three more
comprehensive models for wetlands treatment.
7.3 BIORETENTION IN WMM AND P8
One simple method for performance simulation is to use overall device efficiency. For example, WMM
(not a process model) uses an assumed efficiency for removal that is only suitable for annual load
reduction predictions rather than for individual events. P8's use of a particle removal scale factor
7-51
-------
calibrated for simulation of the bioprocesses that occur during bio-filtration is quite simple. The presence
of macrophytes can increase the removal factor (usually f=l) to 2 or 3, as discussed in the Section 6.2.
7.4 SIMULATION OF WETLANDS WITH THE WETLAND MODEL
7.4.1 Introduction
One of the newer models available, WETLAND has been designed to model constructed wetlands and
can be adapted to model natural wetlands as well. WETLAND is a dynamic, compartmentalized
simulation model (Lee et al. 2002) and was designed as a continuous-flow, stirred-tank reactor (CFSTR),
so complete mixing is assumed to occur. Two input forms are accepted for this model. First, daily values
for hydrologic parameters and nutrients can be input, in either user-defined data or from output from a
nonpoint source model (e.g., SWMM hydrographs and pollutographs). Secondly, data can be input based
on the SCS curve number method (McCuen 1982, SCS 1986). The SCS method determines the amount
of daily runoff from the watershed, which is then multiplied by an EMC for each respective nutrient
parameter to determine nutrient inflow to the wetland system (Lee et al. 2002).
Written in Fortran 77, WETLAND models both free-water surface (FWS) and subsurface flow (SSF)
wetlands, and is designed in a modular manner that gives the user the flexibility to decide which cycles
and processes to model (Lee et al. 2002). This model has one main program that calls upon and manages
the sub-models and options that need to be simulated (Lee et al. 2002). The relationship between
WETLAND's main code and its sub-models is shown in Figure 7-1 (Lee et al. 2002).
7.4.2 WETLAND Cycles and Sub-models
Within the model, there are many wetland cycles that can be modeled. These include the hydrologic,
nitrogen, carbon, dissolved oxygen (DO), bacteria, vegetative, phosphorous and sediment cycles. In the
hydrologic sub-model, WETLAND uses a vertical water balance to account for surface storage. Treating
the wetland as a storage unit, the spatially-lumped continuity equation (Equation 6-1) used is (Lee et al.
2002) in the following form:
dV/dt = Qc + QP - Q + dVb/dt + dVp/dt + (P - PI - ET) A (7-1)
where:
dV/dt = change in surface storage (mVday),
Qc = watershed catchment runoff additions (m3/day),
Qp = additions from point sources (m3/day),
Q = daily outflow rate (mVday),
dVb/dt = change in living biomass water volume in the surface water (mVday),
dVp/dt = change in standing dead plant water volume in the surface water (m3/day),
P = daily precipitation rate (m/day),
PI = percolation/infiltration rate (m/day),
ET = evapotranspiration rate (m/day), and
A = wetland surface area (m2).
Evaporation may be computed from pan data - the primary model option. ET may also be modeled using
Thornthwaite's method (Dingman 2002). An hourly time step is used for the hydrology sub-model.
The vegetative sub-model simulates biomass growth and death rates. At the beginning of the growing
season, biomass growth rate is assumed to increase linearly for up to 20 days until a maximum rate for the
growing season is obtained (Lee et al. 2002). After 20 days, the biomass growth rate remains constant
7-52
-------
BASE
(ONE TIME)'
DELTAH
PHOSPHORUS
(S + T)3
SEDIMENT
(S + T)3
HYDROLOGY
(S + T)1
MAIN CODE
DISSOLVED
OXYGEN(S+T)2
VEGETATION
(S + T)2'1
CARBON
(S + T)2
BACTERIA
(S + T)2
NITROGEN
(S+T)2
1 Sub-mo del is called whenever WETLAND runs a simulation
2 If the NCOB cycle is simulated, then sub-mo del is called by the main code
3 If the phorphorous cycle is simulated, then sub-mo del is called by the main code
one time - Sub-mo del is called only once for the entire simulation run
T Sub-mo del is called once every time period
S + T Sub-model is called by main code once eveiy season and time period
Figure 7-1. Relationship between the MAIN CODE and respective sub-models for an entire simulation run
(Lee et al. 2002).
until the end of the growing season, when the growth rate decreases linearly to zero over a period of 10
days.
Unlike most wetland models, WETLAND explicitly accounts for the effects of biomass and microbial
dynamics in a wetland system using the dynamically-linked NCOB (nitrogen, carbon, DO, and bacteria)
cycle (Lee et al. 2002). Carbon/nitrogen ratios are accounted for in this model. WETLAND is unique in
that nitrification and denitrification are modeled using Monod kinetics, not just empirical relationships.
There are five different state variables in the carbon (denoted as "C") sub-model. They are biomass C,
standing dead C, particulate organic C (POC), dissolved organic C (DOC), and refractory C (Lee et al.
2002). The vegetative sub-model is connected to the standing dead C and the biomass C, respectively.
Biomass C is determined by multiplying a biomass C concentration by the existing biomass in the
wetland. Standing dead C is determined similarly; however, physical degradation and DOC leaching also
are taken into account. A mass balance for POC is constructed for FWS wetlands. This depends on
particulate BOD influx, microbial death, peat accumulation, and POC mineralization. The mass balance
for DOC depends on soluble BOD influx, DOC mineralization, DOC leaching, and diffusion (Lee et al.
2002).
7-53
-------
The nitrogen sub-model simulates processes such as ammonification, denitrification, immobilization of
nitrogen, and peat accumulation. Inclusion of NH3 volatilization, atmospheric deposition and nitrogen
fixation in the modeling of the overall nitrogen cycle is optional (Lee et al. 2002). Variables included in
the nitrogen sub-model are dissolved organic nitrogen (DON), particulate organic nitrogen (PON),
ammonia (NH3), ammonium (NH4+), nitrate (NO3~), immobilized N and refractory N. DON influent may
enter a wetland from point sources, direct catchment runoff, atmospheric deposition and percolation. The
physical degradation of decaying plant mass can add to the accumulation of DON. PON modeling is
similar to DON, except that PON is also assumed to accumulate in peat as refractory N (Lee et al. 2002).
Nitrogen fixation is also an available option for modeling as a way to increase DON and is modeled with
a zero-order equation. NH4+ enters a wetland from catchment runoff, seepage, atmospheric deposition
and point sources. Immobilized N is the sum of DON and PON immobilization, NO3" uptake and NH4+
uptake. Increases in nitrate in wetland waters come from influent and nitrification, whereas decreases
come from plant uptake and denitrification.
In the dissolved oxygen (DO) sub-model, oxygen is assumed to be added to the wetland by point sources,
incoming streamflow, precipitation, reaeration from the atmosphere, and biomass flux. Dissolved oxygen
is the only state variable in the DO sub-model. Oxygen is assumed to be passed from vegetation to the
wetland bottom at a constant rate during the growing season.
The bacteria sub-model describes the microbial interactions within the wetland and includes all of the
microbial activity of the model. Both autotrophic and heterotrophic bacteria are modeled in WETLAND.
pH is not modeled because wetlands are known to drive pH towards neutrality (pH = 7). The growth rate
of bacteria is modeled using Monod kinetics (Chapra 1997). Heterotrophic bacteria are modeled using
Monod kinetics where growth is dependent on TOC.
Sedimentation is modeled in the sediment sub-model of WETLAND. There are five different sediment
classifications in this sub-model. They are inflow, outflow, deposition, resuspension and decomposition
(Lee et al. 2002). Sediment inflow depends on the option chosen, while outflow is a function of the
resuspension, settling velocity, and total amount suspended in the water. A first-order user-defined rate
equation is used to model decomposition of sediment.
The remaining sub-model is the phosphorous sub-model and is based on the assumption that all of the
suspended sediment particles provide surface area to which phosphorous can be attached and
consequently settled, resuspended, or transformed (Lee et al. 2002). Phosphorous coming into the
wetland is modeled as a direct input, or by sorption using the Freundlich or linear isotherms (Chapra
1997). Additions of phosphorous from biomass decomposition and mineralization can also be modeled.
Input phosphorous concentrations are used to determine the particulate phosphorous concentrations when
modeling with Freundlich or linear isotherms. Mineralization is modeled with first-order equations for
each particle class (Lee et al. 2002). Settling and resuspension of particulate phosphorous is related to the
quantity of sediment particles. The amount of phosphorous from physical degradation is directly related
to plant biomass.
7.4.3 WETLAND Output
Time series of concentrations in the effluent (and within the well-mixed wetland) are provided for: NFL,,
NO3, Org-N, DON, PON, DO, BODS, TSS, DP (dissolved P), and TP (total P). Hydrologic state
variables include water depth (implying a water volume) and the outflow hydrograph. In comparisons
with measurements, all predicted chemical constituents except DO had a significant correlation with
measured data in the example presented by Lee et al. (2002). Wetland effectiveness as a BMP is
computed as the reduction of effluent loads relative to influent loads, i.e., percent removal based on loads.
The strength of the model is its linked Monod kinetics for the chemical state variables. In this respect it is
7-54
-------
similar to EPA WASP model capabilities (Wool et al. 2001). Weaknesses include the need for requisite
kinetic parameters (similar to the need of most models) and the well-mixed assumption, which does not
allow the study of hydraulic short-circuiting and shape effects.
7.5 SIMULATION OF WETLANDS WITH VAFSWM
7.5.1 Introduction
The Virginia Field Scale Wetland Model (Yu et al. 1998) was developed with the help of extensive
monitoring of constructed wetlands and detention ponds in eastern Virginia. The model is available
through the Virginia Transportation Research Council in Charlottesville and was developed to fill a gap in
analytical tools perceived on the basis of the Virginia studies.
7.5.2 VAFSWM Components
The water balance is essentially that of Equation 6-1. Constituents are simulated in three forms: 1) water
column suspended solids, 2) water column constituent, for which the particulate form is a fraction of the
SS concentration, and 3) sediment/substrate concentration. The substrate consists of sediment and near-
surface root zone for the aquatic vegetation. The substrate water volume must account for the porosity of
this zone. The model is simpler than WETLAND inasmuch as nutrient kinetics are not linked; each
constituent is simulated in each of the three forms in the manner of a CFSTR using a variant of Equation
4-9. Suspended solids simulation includes only settling as a removal mechanism. Water column
constituent simulation includes the following removal mechanisms:
Settling of the particulate fraction (Equation 4-19)
First-order decay of the dissolved fraction (Equation 4-18) by adsorption to plants and plant uptake
Adsorption to substrate
Substrate concentrations are affected by:
Settling from the water column
Settling within the substrate area (using a different settling velocity)
Adsorption and plant uptake from the water column
Dissolved and particulate fractions are based on a partition coefficient, Ka (Equations 4-18 and 4-19).
The three CFSTR equations are solved simultaneously using a fourth-order Runge-Kutta (RK4)
integration scheme. Insufficient data were available for the authors to fully verify the model, but TSS and
TP simulations provided removal efficiencies (efficiency ratio, Section 3-4) in the range values observed
at the test site.
7.5.3 Implications for S WMM Improvements
The principal contribution of the VAFSWM formulation is explicit inclusion of settling, adsorption, and
first-order kinetics for the water column and substrate in a simpler form (i.e., one that does not involved
the linked nutrient dynamics of WETLAND). Although there are approximately 15 required input
parameters (for simulation of just one constituent in addition to TSS), some guidance is available for
parameter estimates, and the overall formulation is amenable to inclusion in SWMM, since the
fundamental processes are already included in Equation 4-9 (currently implemented in SWMM 4.4h).
The main additional effort would be to include a representation of substrate concentrations, with possible
complications in linkages to subsurface flow pathways.
7-55
-------
7.6 SIMULATION OF WETLANDS WITH PREWET
7.6.1 Introduction
Another program used to model wetlands is the Pollutant Removal Estimates for Wetlands (PREWET)
model, developed at the Waterways Experiment Station of the Army Corps of Engineers. Online help is
available to answer some questions about the model, and a brief description is available from the website
http://www.wes.army.mil/el/elmodels/index.htmltfwqmodels. PREWET uses equations and logic that are
programmed in C++, and a commercially available graphical user interface (GUI) library, "Zinc," which
is also written in C++, is used (Dortch and Gerald 1995).
PREWET contains algorithms to model a wetland's ability to remove contaminants such as TSS, BOD,
total nitrogen, total phosphorous, total coliform bacteria and other contaminants. PREWET does not
model microbial growth and decay or seasonal/annual processes like vegetation growth and decay.
Therefore, this model cannot be used to assess seasonal/annual effects. In fact, because it is a steady-state
model, its usefulness is limited primarily to help in parameter estimation, as will be seen. Hydrological
parameters are input into the model based on knowledge about the wetland.
7.6.2 PREWET Removal Mechanisms
The main assumption made in PRETWET is that the modeled wetland is at steady-state. This means
flows and concentrations of pollutants are constant over time. Obviously wetlands are rarely at steady-
state. However, average values or long-term values are the goal of PREWET, for which the steady-state
assumption is valid. There are two conditions for which this model works. It assumes the wetland is
either a CFSTR, or plug-flow reactor (PFR). The mass balance equation PREWET uses for a CFSTR is
essentially the same as Equations 4-3 and 6-2 under steady-state conditions,
d(VC)/dt = 0=W-QC-kVC (7-2)
where:
V = volume of the wetland (volume)
C = pollutant concentration leaving the wetland (mass/volume)
t = time
W = loading of pollutant entering wetland (mass/time)
Q = flow rate (volume/time)
k = first-order biological degradation rate of pollutant (I/time)
Equation 7-2 can be solved for the steady-state concentration (Chapra 1997),
C= W/Q (7-3)
1+kV/Q
The ratio V/Q is recognized as the hydraulic residence time. Sedimentation is modeled in PREWET by
total suspended solids removal (settling velocity) based on a balance among settling, resuspension, and
sediment layer burial (Thomann and Mueller 1987). BOD removal occurs through settling of the
particulate fraction of BOD from the water column to the sediments, and adsorption to benthic biota
(Dortch and Gerald, 1995). These removal processes are combined into a first-order rate constant, kr.
Total coliform bacteria are also modeled using a first order decay rate. This is due to death, settling, and
predation of the bacteria (Dortch and Gerald, 1995). First-order decay rates, such as kB for bacteria, are
7-56
-------
adjusted for temperature using the customary Arrhenius formulation, Equation 4-21, with a recommended
value for 9 of 1.07 for bacteria.
In PREWET, the only phosphorus component modeled is total phosphorous (TP). The model only
considers the natural, long-term removal mechanism of sediment burial (Dortch and Gerald 1995). TP
retention decreases with time in wetlands due to the sediments becoming saturated with phosphorous.
After a period of time, the sediment reaches saturation equilibrium, reducing the rate of phosphorous
uptake. A net first-order removal rate is obtained from the coupled water column processes of settling,
resuspension, burial, and diffusion from the bed.
Total nitrogen (TN) is the only nitrogen variable considered with this model, and denitrification is the
only TN process modeled. A nitrate balance cannot be used to compute the loss of TN because nitrate
may also be gained through nitrification (Dortch and Gerald 1995). Instead, TN is estimated through loss
of nitrate via denitrification. The steady-state relationship between TN and NO3-N is used to help
determine the TN loss rate from better knowledge of the denitrification rate, by
dTN/dt = 0 = -kTN TN = -kdn NO3 (7-4)
where:
kTN = first-order removal rate for TN,
kdn = denitrification rate,
NO3 = concentration of NO3-N
Equation 7-4 is then used to estimate the TN first-order removal rate, kTN, on the basis of better-known
values for k^ and representative TN and NO3-N concentrations. Finally, wetlands removal efficiency, R
(%), as a BMP is evaluated on the basis of steady-state load reduction,
R (%) = 100 x (W - QC)/W (7-5)
7.6.3 PREWET Usefulness for Urban BMP Evaluation
Steady-state hydrology and water quality are of little usefulness in the urban stormwater setting.
However, the value of PREWET is in its array of parameters (beyond those discussed above) related to
sorption, settling, phosphorus cycling, and decay and other degradation processes. As a model it could
also be used to check order-of-magnitude removal efficiencies on an annual basis to those computed
using a continuous simulation model such as SWMM.
7.7 SIMULATION OF WETLANDS WITH DMSTA
7.7.1 Introduction
The Dynamic Model for Stormwater Treatment Areas (DMSTA) simulates daily water and mass balances
in a user-defined series of wetland treatment cells, each with specified morphometry, hydraulics, and
phosphorous cycling parameters (Walker and Kadlec 2002). An in-depth description of this model can be
viewed at http://wwwalker.net/dmsta/index2.htm. However, the web site does not fully explain all the
processes included in the model, and there appears to be no more detailed explanation available short of
obtaining the model code. DMSTA was designed primarily to model total phosphorous concentrations in
Everglades stormwater treatment areas (STAs) near Lake Okeechobee, Florida that receive agricultural
runoff and releases. The goal of the STA project is to achieve TP outflow concentrations of 50 ug/L or
less by 2007. The DMSTA model will help predict outflow phosphorous concentrations through the
means of sedimentation, filtration, and adsorption. Within this model, up to six stormwater treatment
areas (STAs) can be linked together at one time. STAs are areas capable of treating stormwater that
7-57
-------
contain vegetation, and are often wetlands, though not always. These STAs can be linked in either
parallel or series systems to show compartmentalization and management to promote specific vegetation
types (Walker and Kadlec 2002). Furthermore, each STA is broken down individually as a CFSTR and
can be modeled as such. This model has been coded in Visual Basic for Applications, and the user
interface is a Microsoft Excel workbook (Walker and Kadlec 2002).
7.7.2 DMSTA Model Features
The primary purpose of this model is to predict phosphorous cycling in STAs. Several of the input
parameters required by the model are given below (Walker and Kadlec 2002):
Linkage of treatment cells (up to 6 cells in series and/or parallel)
Morphometry (length, width, area and cell configuration)
Number of stirred tanks in series for each treatment cell
Daily time series (for calibration runs only)
Inflow and outflow volume
Inflow and outflow TP concentration
Mean depth
Rainfall
Evapotranspiration
Descriptive data
Seepage rates
Community description
Phosphorous storage (metadata: macrophytes, periphyton, and soil)
Other factors considered by the model include (Walker and Kadlec 2002):
Temporal variations in inflow volume, load, rainfall, and evapotranspiration
Hydraulic compartments (cells, flow distribution levees)
Residence time distribution
Water level regulation
Compartmentalization of biological communities
Dry-out frequency and supplemental water needs
Bypass frequency, quantity and quality
Inflow pulse modulation by upstream storage reservoir
Seepage collection and management
The model is operated on a daily time step and has been run for periods of up to 31 years. Required input
time series include daily values for inflow, inflow phosphorus concentration, rainfall, and ET. Seepage
and outflows are computed through a model specific to the Everglades STAs being simulated. Spatial
variation within a cell (i.e., within an STA) can be approximated by breaking the cell into a series of
CFSTRs.
7.7.3 DMSTA Phosphorous Cycling Model
TP computations within DMSTA are similar to those within WETLAND and are essentially a dynamic
version of PREWET. That is, phosphorus parameters are very similar to those required by PREWET, but
unlike PREWET, DMSTA is a dynamic model, not a steady-state model. However, DMSTA output has
been compared to steady-state simulations of the same areas (Walker 1995).
DMSTA considers storage of phosphorus in biomass, and nonlinear relationships (typically second-order
kinetics) are used to simulate the exchange of phosphorus between the water column and the biomass.
Biomass itself is essentially wetland vegetation (emergent macrophytes, submerged aquatic vegetation,
and periphyton). It appears that DMSTA does not include growth models for the three types of vegetation
7-58
-------
and instead represents their biomass only by a long-term means. Many coefficients have been obtained
for the Everglades region through calibration using several monitored systems.
7.7.4 DMSTA Usefulness for Urban BMP Evaluation
This model is useful in similar ways to PREWET, that is, for evaluation of coefficients that might be
included in a similar algorithm in SWMM. WETLAND, PREWET and DMSTA all include complex,
nonlinear kinetics for phosphorus cycling in the manner of WASP (Wool et al. 2001). Inclusion of such
terms in SWMM might be warranted for large wetland BMPs with standing water throughout the year.
But such complexity is probably not warranted for simpler devices such as bioswales.
7.8 SIMULATION OF WETLANDS AND OTHER BMPS WITH MUSIC
7.8.1 Introduction
The Model for Urban Stormwater Improvement Conceptualization (MUSIC) was developed by the
Cooperative Research Centre for Catchment Hydrology (CRCCH) in Melbourne, Australia. MUSIC is
capable of continuous simulation, with time steps ranging from 6 minutes to 24 hours. The model was
designed to operate over a range of temporal and spatial scales, suitable for catchment areas from 0.01
km2 to over 100 km2 (Wong et al. 2002). It is important to note that MUSIC was developed as a decision
support system and is not a detailed design tool (Wong et al. 2002). The model is intended to be a tool
used in conjunction with other techniques to evaluate differing strategies for treating urban stormwater.
MUSIC is capable of modeling wetlands, ponds, infiltration strips, buffer strips, swales, sedimentation
basins, and gross pollutant traps. For the purpose of this report, only the wetland and pond portion of the
model will be reviewed. Output from the model includes time-series graphs of flows, pollutant loads or
concentration, statistical summaries, and cumulative probability plots (Wong et al. 2002). MUSIC does
not contain the necessary complex algorithms for runoff routing, catchment contaminant build-up and
wash-off processes, and does not enable the detailed sizing of structural stormwater quantity and/or
quality facilities. Additional general and specific information about MUSIC is available at the web site:
http://www.toolkit.net.au/products/music/index.htm.
7.8.2 MUSIC Algorithms
The algorithms used in MUSIC are based on the recognized routine characteristics of known stormwater
quality improvement measures. The rainfall-runoff algorithm used in MUSIC was developed by Chiew et
al. (1997). This is a water-balance equation similar to those used in models such as WETLAND and
PREWET. MUSIC routes runoff using the Muskingum-Cunge equation (Bedient and Huber 2002,
Dingman 2002).
TSS, TP, and TN concentrations are generated using a stochastic process involving cross correlation
between TSS and TP and serial correlation of water quality time series (Wong et al. 2002). Pollutants that
enter a stormwater treatment area (e.g., wetland or pond) in MUSIC are treated in CFSTRs. The option
exists to treat the wetland/pond as either a series of CFSTRs or as a plug flow reactor (PFR), as discussed
in Section 6.5. Plug flow implies a very small amount or no short-circuiting of flow.
Wong et al. (2001) present the background for the estimate of the parameter, N, the number of CFSTRs in
series (Equation 6-9). Hydraulic inefficiency, i.e., the tendency to deviate from the ideal of plug flow and
not to realize the full potential residence time of a wetland or pond can be due to at least two primary
factors (Thackston et al. 1987):
Dispersion effect caused by unsteady flow rates, wind, entrance and outlet effects, shear stresses at
the sides and bottom, etc.
7-59
-------
Volume effect caused by "dead zones" in which velocities toward the outlet are considerably less than
average and in which recirculation currents exist. Dead zones are not part of the volume through
which water flows; hence, the effective volume is less than the total volume, V, and the effective
residence time is less than the theoretical residence time, V/Q (Thackston et al. 1987).
These effects can be quantified through tracer tests described by several authors, including Fair et al.
(1968), Levenspiel (1972), Thackston et al. (1987), and Kadlec and Knight (1996). Some of these studies
attempt to derive an overall indicator of "hydraulic efficiency" such that a number near 1 .0 indicates good
hydraulic efficiency (close to plug flow) and a number near zero indicates poor hydraulic efficiency
(close to complete mixing). Recall that the number of CFSTRs (or "tanks in series," TIS), N = 1
(Equation 6-1 1) for complete mixing and oo for plug flow. Thus a natural efficiency indicator for mixing
is
emix = 1 - 1/N, typically: 0 < emix < 1 (7-6)
The value for N can be estimated from tracer data from moments of the residence time distribution
(RTD). There is a continuous analog to N discrete tanks in series, for which the analytical solution for the
unit impulse response is a gamma distribution (Kadlec and Knight 1996, Eqn. 9-108),
N ( tV"1 -N-
f(t) = -HN- e * (7-7)
r(N)l ij
where
T = mean of tracer distribution, not necessarily = V/Q for real data,
F(N) = gamma function = (N-l)! if N is an integer (N is not required to be an integer).
The function is plotted in Figure 7-2. The concept is that each "tank" contributes a fractional residence
time T/N. The mean of the distribution is at t = T, or equivalently, t/ T = 1 . It is clear that as N increases,
the distribution becomes more peaked, approaching a PFR. The special case N=l corresponds to the
exponential distribution for complete mixing in one tank,
f (t) = e-t/T (7-8)
If the mode (temporal location of peak) is tp, then from the moment relationships (Kadlec and Knight
1996),
= 1 or ,_1 = IL = 8 ,7-9)
T. T mix v J
NT. T
N T
where Equation 7-7 has been included to indicate one measure of mixing efficiency. Equation 7-10
provides one way to evaluate N from moments of the tracer distribution (Kadlec and Knight 1996, Eqn. 9-
1 1 1), although Kadlec and Knight offer Levenspiel's (1972) assessment that another moment relationship
is preferable for flat observed distributions where the mode may be difficult to identify,
= ^- = CV2 (7-10)
N T2
where
7-60
-------
o = variance of tracer distribution, and
CV = coefficient of variation of tracer distribution.
0
0.5 1 1.5 2
Dimensionless residence time, t/tau
2.5
1
10
Figure 7-2. Theoretical residence time distribution (Equation 7-7) for unit impulse to N tanks in series.
Values of N are given in the legend.
Persson et al. (1999) discuss difficulties of obtaining moments from real tracer data and provide
alternatives to Equations 7-9 and 7-10 based on percentiles of the distribution. However, these authors
end up using Equation 7-9 to evaluate mixing efficiency on the basis of simulated flows (using the MIKE-
21 model, http://www.dhisoftware.com/mike21/) for various geometric layouts of wetlands (or ponds).
The authors simulated 13 hypothetical ponds (Figure 7-3) and generated an output tracer distribution from
a "spike" (unit impulse) input. Moments of the simulated tracer distribution were analyzed to compute
emix from Equation 7-9. The effective volume ratio, evoi, was computed in the manner of Thackston et al.
(1987) for a through-flow Q,
Evol Veffective/Vtotal ~
(7-11)
where
Veffectiv
Vtotai
T
td
= effective volume through which passes the flow,
= total volume of wetland or pond, not all of which is encountered by the flow,
= mean detention time or first moment of the tracer distribution, = Veffective/Q, and
= nominal or theoretical detention time = Vtotai/Q.
7-61
-------
Figure 7-3. Pond shapes simulated by Persson et al. (1999). Pond G has baffles, ponds O and P have islands,
and pond Q has a sill.
Again, the measured mean detention time, T, is usually less than the nominal detention time, td, due to
dead zones. Hence, the volume efficiency is usually a number between 0 and 1 (Thackston et al. 1987).
Finally, Persson et al. (1999) arrive at an overall "hydraulic efficiency," X, as the product of the mixing
and volume efficiencies,
^ = Gmix ' Gvol (7-12)
Values ofk for the 13 shapes in Figure 7-3 are given in Table 7-2, ranked in order of decreasing hydraulic
efficiency. Highest efficiencies are for ponds with a distributed inflow (pond E), baffles (pond G), and
very elongated flow or high length to width ratio (pond J).
Table 7-2. Numerical results of Persson et al. (1999) for pond shapes of Figure 7-3. The qualitative rating of
hydraulic efficiency is by Persson et al.
Pond
J
G
E
P
Q
I
K
A
B
O
D
H
C
Svol
1.00
1.00
0.89
0.96
0.93
1.00
0.78
0.74
0.79
0.73
0.34
0.44
0.46
£.
mix
0.90
0.76
0.85
0.64
0.64
0.41
0.46
0.41
0.33
0.35
0.52
0.25
0.23
>,
0.90
0.76
0.76
0.61
0.60
0.41
0.36
0.30
0.26
0.26
0.18
0.11
0.11
N«l/(l-^)
10.0
4.2
4.1
2.6
2.5
.7
.6
.4
.4
.3
.2
.1
.1
Qualitative
Rating
Good
Satisfactory
Poor
7-62
-------
From an analysis of the data of Persson et al. (1999), Wong et al. (2001, 2002) recommend for MUSIC an
approximation forN, the number of CFSTRs or tanks in series,
N«1-1A, (7-13)
Values of N from Equation 7-13 are included in Table 7-2. In effect, this attributes all imperfect mixing
just to the dispersion effect and none to the volume effect. Thus, Table 7-2 and Figure 7-3 provide a
qualitative estimate forN for use in Equation 6-11. The application of the numerical results of Persson et
al. (1999) to real wetlands is also discussed by Wong and Breen (2002).
The pollutants described by MUSIC are modeled using a first-order kinetic model. This is Kadlec and
Knight's (1996, Eqn. 9-103) k'-C* model and is expressed similarly as:
(Cout-C*)/(Cm-C*) = e-kVq (7-14)
where
C* = background concentration (mg/L),
Cm = input concentration (mg/L),
Cout = output concentration (mg/L),
k' = rate constant (m/y), and
q = hydraulic loading or overflow rate (m/y).
The argument of the exponential in Equation 7-15 is the Damkohler number, discussed in relation to
Equation 6-14.
Equation 7-14 was adapted from another, earlier Australian model, the Universal Stormwater Treatment
Model (USTM), and is used to simulate pollutants as they pass from one CFSTRto another (Wong et al.
2001, 2002). This equation is computed separately for each time step at each CFSTR. The main
difference between this equation and ordinary first-order decay modeling is the inclusion of C*, the
equilibrium or background concentration. This means that effluent concentrations will not be reduced
below C*. It may be noted from Equation 6-14 that
e-k'/q = g-ktd (?.15)
That is, using depth, h, as a linking variable,
k' kh
q h/t
= ktd (7-16)
where
k = first-order decay coefficient = k'/h (I/time),
h = average depth, and
td = nominal detention time = V/Q.
Some recommended k' and C* values are given in Table 7-3, based on limited model calibration for TSS,
TP, and TN in urban areas near Melbourne (Wong et al. 2002). If depths were used to compute k' values,
they are not reported. In fact, one advantage of the formulation using rate constant k' (depth/time) is that
it avoids having to specify an average depth for odd natural configurations.
7-63
-------
Table 7-3 Calibrated k' and C* values from MUSIC based on limited model simulations.
Treatment
Measures
Sedimentation
Basins
Ponds
Vegetated Swales
Wetlands
k' (m/yr)
TSS
15,000
1,000
15,000
5,000
TP
12,000
500
12,000
2,800
TN
1,000
50
1,000
500
C* (mg/L)
TSS
30
12
30
6
TP
0.18
0.13
0.18
0.09
TN
1.7
1.3
1.7
1.3
Refinement of the parameters for the k'-C* model to suit local conditions (particle size distributions in
particular) and treatment measure design specifications, is currently being undertaken (Wong et al. 2002).
It is expected that the parameter C* will vary with discharge and the influence of chemical and biological
processes during the inter-event period. Derived k values for ordinary CFSTR modeling that result from
using the TSS and k' values in Table 7-3 are shown in Table 7-4 for three representative depths. But it
will be seen in Chapter 15 that these values are probably too high for simulation of "decay" of TSS in
most ponds.
Table 7-4. First-order decay values converted from k' values for TSS in Table 7-3 for assumed depths.
Treatment Measures
Sedimentation Basis
Ponds
Vegetated Swales
Wetlands
Depth = 1 ft
TSS k, I/day
134.8
8.99
134.8
44.9
Depth = 3 ft
TSS k, I/day
44.9
3.0
44.9
15.0
Depth = 5 ft
TSS k, I/day
27.0
1.80
27.0
8.90
The same k'-C* model is used for ponds, wetlands (Wong and Breen 2002), grass swales (Fletcher et al.
2002), and gravel filters (Wong et al. 2002). It has proven adaptable to fitting of many observed BMP
performance data in Australia (Wong and Breen 2002).
7.8.3 MUSIC Evaluation
MUSIC has been calibrated and tested for its ability to predict pollutant removals by various different
companies and consultants. The equations used in MUSIC are fairly simplistic and are first order. It
appears to predict the removal of TSS and TN well, and TP moderately well. The uniqueness of this
model is its ability to link together different treatment options such as a wetland coupled with a swale, or
a wetland coupled with a pond, through a convenient GUI. This type of simulation (known as a
"treatment train") yields higher pollutant removal efficiencies than from just one BMP alone. Presently,
SWMM is capable of simulating treatment devices in series in the Storage/Treatment Block, or in Runoff
and Transport channels and storages. However, sufficient information is not stated within the MUSIC
training manual or other references reviewed as to whether or not correct simulation of the treatment train
occurs. Correct simulation depends upon recognition that the upstream device removes the easiest
materials (e.g., heavy solids). Downstream devices are left to remove fine particles and dissolved
constituents. Therefore, downstream removal efficiencies will progressively lessen, and it is not clear
from the documentation whether or not this is observed in MUSIC.
Once again, another model recognizes the efficacy of multiple CFSTRs or the tanks in series approach
(Equation 6-9). The use of tracer data to identify the number, N, of the series of CFSTR would be helpful
for parameter estimation, should such an approach be implemented in SWMM.
The way MUSIC models first-order decay (Equation 7-14) may be worthy of implementation into
SWMM. Specifically, the inclusion of the C* term (the background concentration) might be included in
7-64
-------
SWMM first-order decay modeling. However, C* is a constant for the simulation. A better alternative
might be a distribution of effluent concentrations (Strecker et al. 1991).
7.9 SWMM SIMULATION OF WETLANDS AND BIORETENTION DEVICES
As described in Chapter 1, SWMM simulates wetlands in the same manner as it simulates storage or
ponds (Chapter 4 and Section 6.5). Hydraulic efficiency options range from plug flow in the S/T Block,
to complete mixing (CFSTR) in S/T or in any Runoff or Transport Block channel or storage device.
Similarly, bioretention devices may be also be simulated in the manner of storage (Minton 2002) as
discussed in Section 6.5. First-order decay and settling are readily simulated, and the "universal removal
equation" (Equation 4-3) provides the option for curve-fitting of observed removal performance. What is
missing is interaction among state variables, as exemplified by the WETLAND model, and performed
more simply with the VAFSWM model. However, the complex nutrient dynamics involved in a model
like WETLAND might be much too sophisticated for the typical analysis and design employed by
stormwater engineers. The EPA WASP model (Wool et al. 2001) provides an alternative should such
complex dynamics need to be considered. Hence, the most useful recommendation regarding SWMM is
probably to enhance the linkage between land-surface runoff models, such as SWMM, and receiving
water quality models, such as WASP or HSPF.
Interfacing of different models is not necessarily an easy task; Lin and Medina (2003) present one
example of linking (through appropriate interface files) three USGS models: a stream transient model
(DAFLOW), a groundwater flow model (MODFLOW), and a solute transport model (MOC3D plus a
one-dimensional stream solute transport model. Even for models from the same agency, time-step and
other considerations made the interfacing difficult. Interpolation in time series in order to match time
steps between models can be a particularly difficult problem. There is an opportunity for leadership with
the development of SWMM5: an interface file "standard" could and should be developed to facilitate
exchange of time series data between models that are primarily watershed runoff models, such as
SWMM, with models that are primarily receiving water quality models, such as WASP.
7-65
-------
8 INFILTRATION TRENCHES
8.1 INTRODUCTION
Modeling infiltration trenches involves storage and release of stormwater capture volumes to
groundwater. Infiltration is the major process and is simulated in several models.
8.2 SIMULATION OF INFILTRATION TRENCHES WITH SLAMM
8.2.1 Introduction
Infiltration devices are one of the control practices evaluated by SLAMM (Pitt and Voorhees 2000).
Volume reduction from infiltration is dependant on the study area, runoff rates, infiltration rates and
physical trench parameters. One approach is to take infiltration rates from SCS data, altered to reflect
infiltration during micro-storms (storms with depths less than about 0.10 in.), and to adjust volumetric
runoff coefficients. Similarly, the SWMM Horton or Green-Ampt methods could be used. But
infiltration in trenches, swales, and channels may also depend upon the depth of water in the device, depth
to shallow groundwater, clogging, etc. and thus require additional simulation efforts. The SLAMM
procedure essentially ignores these complications, and is discussed below.
8.2.2 SLAMM Calculation Procedures for Infiltration Devices
Infiltration devices are assumed to affect water volume, but not pollutant concentrations. As the water
volume is reduced, the pollutant yield (load) is obviously decreased. SLAMM calculates the runoff
volume reductions for each source area (served by an infiltration device) for each individual rain event in
the study period. Figure 8-1 shows the spatial breakdown for the SLAMM model.
Runoff volume reduction fraction is assumed to be:
fractional volume reduction = - L (8-1)
loJUJ
where
Qp = the percolation volume rate of the device (cfs),
Qr = the runoff rate to the device (cfs),
As = the area draining to the device (acres), and
At = the total study area (acres).
8-66
-------
It can be seen that the fractional volume reduction is the product of the fraction of runoff infiltrated times
the fraction of area served.
runoff to device (Qr)
area not draining to swale
drains to swale (As)
Figure 8-1. Schematic of SLAMM model area breakdown.
The ratio Qp/Qr used in this equation can never be greater than 1.0, because the device cannot infiltrate
more water than is delivered into the device. The percolation volume rate, Qp, is the capacity of the
infiltration device to infiltrate runoff, expressed as cfs. Pitt and Voorhees (2000) assume (no other basis
given) that each side wall of a vertical trench infiltrates 1/3 of the rate along the trench bottom. For a
vertical-walled (rectangular) trench of depth h, width w, and length L, and infiltration rate (percolation
rate) f (depth/time), the volume rate of percolation is thus:
2 2h
QD = Lwf + -Lhf = Lwf (1 + )
3 3w
(8-2)
This yields the version cited by Pitt and Voorhees (2000) in which percolation area = Lw,
0.67 1
1H (percolation rate) (percolation area)
width to depth ratio J
(8-3)
No specifications are given for trench design (trapezoidal or vertical side walls are not specified) in
SLAMM.
Much of the effort within SLAMM is involved in generation of runoff, including the use of runoff
coefficients obtained from extensive analysis by Pitt (1987) based on the evaluation of data obtained from
NURP (EPA 1983), the EPA's Urban Rainfall-Runoff-Quality Data Base (Huber et al. 1982,), and from
the Humber River portion of the Toronto Area Watershed Management Study (Pitt and McLean 1986).
Since runoff generation is already provided by SWMM and not the main thrust of this presentation about
BMPs, runoff generation by SLAMM is not included herein.
However, for purposes of presenting a brief SLAMM example, volumetric runoff coefficients, Rv, are
defined conventionally by
8-67
-------
runoff volume =RV (area draining to device) (rain depth), (8-4)
and Pitt (1987) presents empirical data relating the duration of runoff (hours) to the duration of rainfall as
Runoff duration = 0.90 + 0.98 (rain duration, in hours) (8-5)
An example of use of this procedure follows:
Percolation rate = 3 in./hr
Total rain = 1.7 in.
Rain duration = 6 hours
Volumetric runoff coefficient = 0.35
Area served by infiltration trench =1.3 acres
Total area in study = 5.6 acres
Trench bottom area (percolation area) = 5500 ft2
Trench width/depth ratio = 2
Therefore:
runoff volume = 0.35 (1.7 in.)(1.3 acres) = 0.774 ac-in.
runoff duration = 0.90 + 0.98(6 hours) = 6.78 hours
Qr = 0.774/6.78 = 0.114 ac-in./hr= 0.115 ftVsec.
Qp = [1 + 0.67/2] (3 in./hr) (5500 ft2) (ft/12 in) (hr/3600 sec) = 0.510 ftVsec.
Therefore Qp/Qr = 0.51/0.114 = 4.43, which is greater than 1.0, so 1.0 must be used in Equation 8-1.
In this example the infiltration trench is oversized for this event since all of the runoff from the service
area is infiltrated. This means that Pitt's effectiveness criterion is simply the ratio of area served to the
total area. The study area volume reduction performance is therefore: 1.3 acres/5.6 acres = 0.23 (23 % of
the runoff and pollutant load are infiltrated).
8.2.3 Infiltration in Disturbed Urban Soils
Disturbed urban soils do not behave as indicated by typically used models. More rain infiltrates through
pavement surfaces and less rain infiltrates through soils than typically assumed (Pitt et al. 1999a, Pitt and
Voorhees 2000). Double-ring infiltrometer test results from urban soils in Oconomowoc, WI (Table 8-1)
indicated highly variable infiltration rates for soils that were generally sandy (NRCS A/B hydrologic
group soils).
Many infiltration rates actually increased with time during these tests. In about one third of the cases, the
observed infiltration rates remained very close to zero, even for these sandy soils. Areas that experienced
substantial disturbances or traffic (such as school playing fields) had the lowest infiltration rates, typically
even lower than concrete or asphalt (Pitt and Voorhees 2000). These values indicate the large variability
in infiltration rates that may occur in areas having supposedly similar soils.
In an attempt to explain much of the variation shown in the Wisconsin tests, Pitt and his students
conducted tests of infiltration through disturbed urban soils in the Birmingham, AL area (Pitt and
Voorhees 2000). Eight categories of soils were tested, with about 15 to 20 individual tests conducted in
each of eight categories (comprising a full factorial experiment). Numerous replicates were needed in
each category because of the expected high variation in infiltration rates. The eight categories in Table 8-
2 were tested. These tests resulted in the default infiltration parameters distributed with SLAMM (Table
8-3).
8-68
-------
Table 8-1. Ranked double ring infiltration test results and observed urban soil infiltration rates from
Oconomowoc, WI (Pitt and Voorhees 2000).
Initial Rate (in/hr)
25
22
14
5.8
5.7
4.7
4.1
3.1
2.6
0.3
0.3
0.2
<0.1
<0.1
<0.1
<0.1
Final Rate (after
2 hours) (in/hr)
15
17
79
9.4
9.4
3.6
6.8
3.3
2.5
0.1
1.7
<0.1
0.6
<0.1
<0.1
<0.1
Total Observed
Rate Range (in/hr)
11 to 25
17 to 24
4 9.4 to 17
0.2 to 9.4
5.1 to 9.6
3.1 to 6.3
2.9 to 6.8
2.4 to 3. 8
1.6 to 2.6
<0.1to0.3
0.3 to 3.2
<0.1to0.2
<0.1to0.6
alKO.l
alKO.l
alKO.l
Tab
e 8-2 Categories tested for infiltration rates (Pitt and Voorhees 2000).
Category
1
2
3
4
5
6
7
8
Soil Texture
Sand
Sand
Sand
Sand
Clay
Clay
Clay
Clay
Compaction
Compact
Compact
Non-compact
Non-compact
Compact
Compact
Non-compact
Non-compact
Moisture
Saturated
Dry
Saturated
Dry
Saturated
Dry
Saturated
Dry
Table 8-3 Percolation rates for different soil texture and moisture used in SLAMM (Pitt and Voorhees 2000).
Soil Description
Non-compact sandy soils
Compact sandy soils
Non-compact and dry clayey soils
All other clayey soils
Number of
tests
29
39
18
60
Average
Infiltration rate
(in/hr)
17
2.7
8.8
0.69
CV
0.43
1.8
1.1
2.1
8-69
-------
The CV, or coefficient of variation, is the ratio of the standard deviation for a variable to the mean value
of the variable. This is used by Pitt and Voorhees (2000) to measure the imprecision in survey estimates
introduced by sampling. A coefficient of variation of 1% would indicate that an estimate could vary
slightly due to sampling error, while a coefficient of variation of 50% means that the estimate is very
imprecise.
Although the CV values shown in Table 8-3 for the infiltration tests are generally high, Pitt and Voorhees
(2000) claim that they are much less than if compaction was ignored. The high variation within each of
the four main categories makes it difficult to identify legitimate patterns, implying that average
infiltration rates within each event may be most suitable for predictive purposes. Other infiltration rates
for clayey and sandy soils can be taken from Figures 8-2 and 8-3.
8.2.4 SLAMM Procedures for SWMM
Infiltration trench procedures used in SLAMM appear to have little value for SWMM except for the very
positive aspect of providing infiltration rate data (previous section). Runoff generation is already
performed in SWMM using a dynamic procedure not involving runoff coefficients or regression. The
latter work well in SLAMM and may be additionally useful in spreadsheets, but do not appear to usefully
enhance the current SWMM Runoff Block procedures. Infiltration itself is also dynamic (Horton or
Green-Ampt), and a work-around procedure is available to simulate trenches, explained in the following
section.
Figure 8-2. 3-D plots showing interactions affecting infiltration rates in sandy soils (Pitt and Voorhees 2000).
8-70
-------
10
Figure 8-3. 3-D plots showing interactions affecting infiltration rates in clayey soils (Pitt and Voorhees 2000).
8.3 SIMULATION OF INFILTRATION TRENCHES IN SWMM
In spite of the fact that the current SWMM flow routing procedures do not allow infiltration from
channels, the ability to route flow from one overland flow plane onto another (Section 4.7) allows runoff
block subcatchments to serve as vertical-walled infiltration trenches. The procedure in SWMM (current
or SWMM5) is as follows:
1. Simulate subcatchment runoff by usual procedures and route it downstream to the infiltration trench
subcatchment.
2. Simulate an infiltration trench as a 100% pervious subcatchment of width w and length L (trench
dimensions). The depth is implicitly infinite since there is no maximum subcatchment water depth for
overland flow planes. However, depression storage could be set equal to the trench depth, thus ensuring
no horizontal outflow for water depths less than or equal to the depression storage depth. But the modeler
would have to ensure that the trench could accept all inflow (that is, not flood), unless there was provision
to accept such "overflow" as legitimate flow to an auxiliary drain.
3. Infiltration may be simulated by Horton or Green-Ampt; if a constant rate is desired, it is easier to
manipulate the Horton equation (maximum infiltration rate = minimum infiltration rate). Note that water
depth will have no effect on infiltration within the SWMM model formulation. The infiltration rate might
be adjusted higher to reflect the fact that there will be some infiltration out through the side walls that the
model cannot simulate. Alternatively, a larger planar area than the actual length and width could be used,
but both methods are judgmental.
4. A combination of low slope, high Manning's n, and/or very small conceptual width should be provided
to eliminate horizontal outflow out of the trench - unless such outflow actually occurs, into a drain, say,
when the water level is above the depression storage. If this work-around procedure does produce water
depths higher than the trench depths, results should be carefully checked.
8-71
-------
Drainage from the infiltration trench subcatchment can be directed to a groundwater component if further
tracking is desired, an advantage. Note that water in the trench will also be subject to evaporation and
rainfall on the trench itself. There is no ready way to simulate reduction of infiltration capacity due to
sedimentation apart from running the model with different parameter sets.
A better formulation would consist of a channel (with the option of a weir, to prevent outflow for water
levels below the weir level = trench depth). This would permit other cross sections besides vertical
walled, such as trapezoidal. But the model needs to be modified to allow infiltration from channels,
including the possibility that infiltration might increase with water depth, and some consideration of
clogging over time. Infiltration methods, including SWMM's Green-Ampt procedure, which might be
suitable for this purpose, are reviewed by Williams et al. (1998).
8.4 TRANSITION TO SIMULATION OF RAIN GARDENS AND GREEN ROOFS
In the same manner that SWMM may currently be used to simulate infiltration trenches, the model can be
used to simulate rain gardens and green roofs. Here, the conceptualization is more accurate than for
infiltration trenches because the drainage is confined between vertical walls (i.e., the walls of the
vegetated plots, as in a planter). The vegetated area may simply be a source subcatchment or a
subcatchment to which flow is directed from an upstream source, such as an impervious portion of the
roof. It is important to include the groundwater modeling option in order to obtain a vertical water
balance and to provide for vertical drainage through subsurface geo-fabrics, screens, or other soil
structures.
Another, more indirect simulation of green roofs can be performed by modeling each soil layer as an S/T
unit. The upper unit drains to a lower unit on the basis of a prescribed rating curve. The advantage of this
conceptualization is that quality parameters may be tracked through the units. The disadvantage is that
the vertical water balance must be simulated indirectly, as in a prescribed time series of ET and/or
outflow hydrograph.
8-72
-------
GRASS SWALES AND FILTER STRIPS
9.1 INTRODUCTION
Grass swale drainages can be used in place of concrete curb and gutter drainages in most land uses,
except strip commercial, manufacturing industrial, and high-density residential areas (Pitt and Voorhees
2000). Grass swales reduce urban runoff problems by a combination of mechanisms. Infiltration of the
runoff and associated pollutants is probably the most important process of removal in grass swales.
Filtering of particulate pollutants in grassed waterways may also occur, but the flows are usually too large
(and deep) to permit effective filtering by grass (Pitt and Voorhees 2000). However, Minton (2002)
points out that settling is a primary unit process as water flows through swales, and in fact, as long as
there is storage (water depth) along the swale, solids removal may be simulated in the same way as for
ponds; see Section 6.5. This will be explored further at the end of this section.
Filter strips differ from swales in the sense of not necessarily consisting of a channel, rather just an
overland flow path. A large swale may be conceptualized as having its upper banks consist of filter strips,
whereas the active channel is the swale. Performance data are sometimes differentiated on this basis.
Groundwater contamination concerns are frequently raised whenever stormwater infiltration is proposed.
Pitt et al. (1996) reported that groundwater contamination is not a major concern for most stormwater if
using surface spreading (such as occurs in grass swales). Pitt et al. (1999b) also reported on the
accumulation of stormwater pollutants in the surface soils of swales, minimizing groundwater
contamination problems.
9.2 SIMULATION OF GRASS SWALES WITH P8
P8 uses the same methods to describe settling and decay in swales as in ponds. Particle and pollutant
removal are also calculated similarly, although runoff velocities in swales are calculated with Manning's
equation. An added process modeled in swales is infiltration, and the associated filtration. For pervious
areas, infiltration is calculated with the SCS runoff curve number technique, and thus does not include the
additional complexities of depth and sedimentation mentioned earlier. Infiltration rates used in P8 are
listed in Table 9-1.
Filtration efficiencies for all infiltration particle fractions are assumed to be 100%, and 90% is assumed
for the dissolved fraction to account for the adsorption, precipitation and other reactions between
dissolved contaminants and the soil matrix. There is no method to calculate resuspension of particulates
9-73
-------
if critical velocities for incipient motion are reached. Therefore, surface water outflows from grass swales
have reduced pollutant mass due to the reduction in water volume from infiltration.
This method for tracking groundwater may be useful if modeling BMP interactions with a shallow
groundwater system, although a 90% removal rate of all dissolved pollutants seems optimistic. This
should strongly be a function of the soil type and the pollutant in question. For instance, nitrate does not
sorb strongly, but heavy metals might. This would only be important to modelers wishing to track
pollutants in groundwater, which is not available in SWMM (but is in HSPF).
Table 9-1 Infiltration rates used in P8.
References:
Soil Textures
sand
loamy sand
sandy loam
silt loam
loam
silt loam
sandy clay loam
clay loam
silty clay loam
sandy clay
silty clay
clay loam
Sources: a- Rawls
Musgrave(1955)
(a)
Infiltration
(b)
rate (in/hr)
(a)
Infiltration rate
(c)
(in/hr)
SCS soil group
4.64
1.18
0.43
0.26
0.13
0.06
0.01
0.04
0.03
0.02
0.01
etal. (1983)
8.27 A
2.41 B
1.82 C
0.27 D
0.52
0.27
0.17
0.09
0.06
0.05
0.01
0.02
values for saturated hydraulic
0.43
0.26
0.19
0.03
conductivity. b-Shaver
.30-.45
.15-.30
.05-.15
.00-.05
(1986) c-
9.3 GRASS SWALE PERFORMANCE CALCULATIONS IN SLAMM
SLAMM calculates the performance of grass swales in a similar manner as other infiltration devices, by
assuming (Qp/Qr) (As/At) as indicative of swale infiltration (refer to Equation 8-1).
The water percolation rate in the swale is calculated by:
Qp = (dynamic percolation rate) (percolation area)
(9-1)
where
percolation area = swale length times the swale width, and
percolation rate in the swale is for dynamic flow conditions and has been found to be about one-half of
the typically measured static infiltration rate in some Florida locations (Wanielista et al. 1983).
This procedure is generally independent of swale routing; it assumes that the water is in the swale long
enough to be infiltrated. "Long" swales serving "small" service areas encourage infiltration. Grass
swales include infiltration as a function of flow distance for different slopes and infiltration rates and can
therefore be used to estimate needed flow length in swales (Pitt 1985, 1987). Obviously, swale design
9-74
-------
(like all other controls) must be carefully done to encourage performance. As an example, these
procedures would not be appropriate for steep swale gradients. The ratio of area served by swales to total
area therefore needs to be reduced if steep swales are present, or if the swales are "short."
An example of the calculations for swale performance follows:
Total contributing flow volume = 1140 ft3
Rain duration = 5.5 hours
Dynamic percolation rate in swale = 3.5 in./hr (1/2 of measured static infiltration rate)
Swale density = 350 ft/acre
Wetted swale width = 5 ft
Area draining to swales =1.5 acres
Study area =3.3 acres
Therefore the runoff duration (Equation 8-5) = 0.90 + 0.98 (5.5 hours) = 6.29 hours, and
Qr = 1140 ft3/6.29 hrs = 181 ft3/hr = 0.050 cfs
Qp = (3.5 in./hr)(350 ft/acre)(1.5 acre)(5 ft)(hr/3600 sec)(ft/12 in.) = 0.21 cfs
Therefore Qp/Qr = 0.213/0.05 = 4.26, which is greater than 1.0 and the swale is larger than necessary for
this rain (total infiltration). The study area runoff reduction is therefore 1.5 acres/3.3 acres = 0.46 (46
percent reduction in flows and pollutant yields due to the swales).
Once again, SLAMM procedures for swales appear to offer little to be added to SWMM, except for good
data. A review of output from SLAMM also suggests that SWMM might be improved through additional
tables of effectiveness measures, such as volume reduction, etc.
9.4 SIMULATION OF VEGETATED FILTER STRIPS WITH REMM
The processes that occur in filter strips are sedimentation, filtration, infiltration and biochemical
interactions. Discussions of these processes in the previous sections are applicable to this BMP. A model
that thoroughly investigates the biochemical processes in filter strips is the Riparian Ecosystem
Management Model (REMM), developed by the USDA in partnership with the Southeast Watershed
Research Laboratory at Tifton, GA. It was developed for natural resource agencies and researchers as a
tool that can help quantify the water quality benefits of riparian buffers in response to changes in upland
agricultural use (Inamdar et al. 1998a,b; Inamdar et al. 1999a,b; USDA 1999; Lowrance et al. 2000). The
following discussion is based on portions of all of these six references. Additional discussion of this
model is in the Appendix. REMM simulates: (a) the movement of surface and subsurface water; (b)
sediment transport and deposition; (c) transport, sequestration (capture) and cycling of nutrients; and (d)
vegetative growth.
The strengths of the REMM model are its ability to deal with subsurface fate and transport of nutrients.
REMM can be applied to:
Quantify nitrogen and phosphorus trapping in riparian buffer zones and to determine buffer width for
a given set of riparian conditions and upland loadings.
Determine buffer effectiveness under increased loads.
Evaluate influence of vegetation type on buffer effectiveness.
Determine impacts of harvesting on buffer effectiveness.
Investigate long-term fate of nutrients in riparian zones, sequestration in vegetation, or loss to
atmosphere (denitrification in case of N).
Investigate N / P saturation in riparian buffers.
9-75
-------
The REMM model is the only model reviewed that integrates subsurface flow (three layers) and
groundwater interaction when simulating buffers. REMM also closely models the N, P, and C cycles in
the buffer that is broken into three zones: 1) closest to stream, 2) middle, and 3) farthest from stream.
Although determining site-specific nutrient cycling parameters requires extensive field data to simulate
accurately, the most useful information from the REMM model may be the default rate constants used in
these calculations.
The model operates on a daily time step and requires daily loadings from upland areas and daily
meteorological data. Output includes daily time series of surface and subsurface flows and water quality.
Comparison of outflow and influent loads yields BMP effectiveness. The model can be used to study the
effectiveness of buffer strips of various lengths (in the direction of water flow), for instance. If
subsurface outflow to an adjacent stream is important, REMM can also provide those fluxes.
Because the writers were unable to obtain feedback from the USDA at Tifton, current support for this
model is minimal and algorithms used for modeling calculations are unavailable. As documentation for
REMM becomes available the model should be revisited. But overall, the nutrient dynamics algorithms
appear to be beyond what would be reasonably expected of SWMM, in the manner of the WETLAND
model. A more general description of this model is presented in the Appendix.
9.5 SIMULATION OF GRASS SWALES WITH SWMM
The "obvious" way to simulate grass swales in SWMM is to model them as channels that infiltrate.
Unfortunately, SWMM version 4.4h cannot infiltrate directly from open (or closed) channels. Work-
arounds for infiltration include the option for entering a negative hydrograph upstream of the channel, to
simulate outflows, but this is not very satisfactory. Alternatively, a swale could be modeled as a
rectangular channel by simulating it as a subcatchment, as discussed in Section 8.3 for infiltration
trenches. Again, since real swales are usually trapezoidal and since infiltration might depend upon water
depth, this is less than satisfactory, but perhaps the best current option since pollutant routing across
downstream subcatchments does reflect first-order decay, a constant settling velocity, and/or constant
removal fraction.
Still another option within the current SWMM is to simulate swales with the S/T Block, as a storage
device. Infiltration may be simulated either as 1) monthly evaporation, or 2) residuals outflow. The latter
could be adjusted to provide infiltration as a function of depth through a rating curve. Minton (2002)
indicates that swales are essentially shallow clarifiers, and sedimentation occurs by dynamic settling,
which is therefore a function of hydraulic loading rate, as discussed in Section 6.5. This is supported for
TSS removal by data from Brisbane by Fletcher et al. (2002), for which the k'-C* model discussed in
Section 7.7 provides a good fit. S/T simulation is probably the most accurate, if least intuitive method
available in the current SWMM for simulation of pollutant removal in swales. It also has the advantage
of providing options for simulating removal of soluble pollutants, should any such removal be noted in
swale BMPs.
It is discouraging that the literature review for this and related projects has not uncovered simpler
functional relationships to represent removal in buffer strips or along overland flow. Overland flow and
riparian removal effectiveness intuitively would depend upon the length of the overland flow pathway.
An example is shown in Figure 9-1 (Huber et al. 2000) and is similar in principle to pollutant reduction in
swales observed by Fletcher et al. (2002). As one would intuitively expect, removal efficiency does
increase with length of overland flow, but there is no unifying result from this set of experiments, at any
rate. Overland flow removal effectiveness data tend to be found in the agricultural literature, although the
data of Barrett et al. (1997) are from a highway median. A problem in analysis of such data is the lack of
consistent reporting of the physical conditions of the filter strips, such as slope, soil type, vegetation,
9-76
-------
antecedent conditions, etc. In addition, agricultural experiments often utilize animal wastes as loading
materials and thus may not be reasonable approximations of stormwater characteristics for urban runoff.
Additional review of the literature may yield better data in an effort in support of SWMM algorithm
improvement.
100
90 4-
80
70 +
60
50
(5 40
2 30
20
10
0
o
it
LU
0)
0
Dry AMC
Wet AMC
- Barrett etal., 1997
Magetteetal., 1989
468
Filter Length (m)
10
12
Figure 9-1. TSS removal effectiveness of vegetated filter strips, based on total mass of suspended solids
entering and leaving the strip (Huber et al. 2000). ("AMC" refers to antecedent moisture condition.)
9-77
-------
10 DRY WELLS
10.1 INTRODUCTION
Dry wells are usually holes several feet deep, that are filled with porous material and that fill with water,
which then infiltrates. The P8 method for modeling storage and infiltration would adequately describe
storage and removal in dry wells that capture a given storm water runoff volume and infiltrate it to the
ground. In P8, infiltration is simply subtracted from the device flow according to the device dimensions
and the SCS infiltration rates. However, in reality, infiltration of moderately deep water in a dry well is a
three-dimensional problem of unconfined flow from a partially penetrating well and can involve very
complex analytical techniques (Freeze and Cherry 1979). Under suitable soil conditions, however, it may
be possible to design dry wells that infiltrate the entire design inflow over the duration of a storm or
somewhat longer, and thus may be modeled simply by assuming all the water will infiltrate, without
becoming bogged down in groundwater modeling efforts. In any event, SCS infiltration rates are very
unlikely to be appropriate for infiltration from the bottom and sides of a dry well.
10.2 SIMULATION OF DRY WELLS WITH SWMM
In SWMM, dry wells can be modeled as a pipe or inlet with a specified inflow capacity (in the manner of
a combined sewer overflow regulator). This assumes that the well has the capacity to accept the diverted
flow from the "regulator." Another option is to follow exactly the same procedure as outlined in Section
8.3 for infiltration trenches. This would provide one way in which flooding of the well could be
simulated if infiltration capacity were exceeded. But for the case of deep dry wells with a small surface
area, the one-dimensional model of infiltration is even more inappropriate.
Still another means of simulation is as a storage device with constant or head-driven outflow. This could
be done using 1) a Transport Block storage element, 2) a S/T Block storage unit, or 3) an Extran Block
orifice or rating curve. (Rating curves in the current Extran must be mimicked using a pump Q vs. h
curve, which works well but is not intuitive.) It is unlikely that the quality of the water in the dry well
would need to be simulated, but if it were, Extran could not be used. SWMM offers the option for
continuous simulation in all cases, with which to characterize the performance and operation of the well.
10-78
-------
11 CISTERNS
11.1 INTRODUCTION
Cisterns are usually a barrel or other tank placed beneath a downspout, with outflow controlled by a
valve. The difference between a cistern and, say, a dry well, is that the cistern has a fixed capacity, and
excess or bypassed runoff must be accounted for. In addition, the emptying of cisterns is more complex
from a modeling viewpoint since they are usually drained very deliberately, for use for irrigation, for
example. Hence, information on the timing of releases is required.
In a modified version of SLAMM (Pitt and Voorhees 2002), it is possible to designate only a fraction of
flow to treatment areas. As an example, a fraction of the roof runoff and driveway runoff can be directed
to a cistern for storage for later use during dry weather for on-site irrigation, toilet flushing (gray water),
boiler feed water, etc. On-site water treatment might be required to improve quality for some uses, such
as gray water. In the rain barrel/cistern "outlet/discharge" option in SLAMM, monthly water uses are
entered so the model can track water use and re-filling of the tanks during storms. Hence, a storage
accounting method is needed for the cistern storage units including supply and use schedule.
11.2 SIMULATION OF CISTERNS WITHIN SWMM
A cistern may be simulated in the manner of any storage device, but since cisterns usually collect
relatively clean runoff from roofs, quality simulation may not be necessary. Outflows and bypasses may
be directed downstream in the catchment as well. The principal need in SWMM is the ability to input a
water use schedule, as described above for SLAMM.
11-79
-------
12 POROUS PAVEMENT
12.1 INTRODUCTION
Porous or permeable pavement is a "hard" surface that can support a certain amount of activity, while still
allowing water to pass through. Porous pavement is generally used in areas of low traffic, such as service
roads, storage areas, and parking lots. Several different types of porous pavement exist (Pitt and
Voorhees 2000). Open mixes of asphalt have a much higher porosity than regular asphalt, and concrete
grids can have open holes up to several inches wide, possibly containing sand, gravel or planted with
grass. This kind of surface is often marketed as "pavers," i.e., concrete paving stones designed to allow
easy passage of water between joints. Porous pavements can be effectively used in areas having soils
with adequate percolation characteristics. The percolation rates of the soils underlying the porous
pavement installation only need to exceed the rain intensity directly. In most cases, several inches of
storage is available in the pavement base to absorb short periods of very high rain intensities. Diniz
(1980) states that the entire area contributing to the porous pavement can be removed from the surface
hydrologic regime if all runoff infiltrates. Porous pavement can be designed to eliminate all of the runoff
from paved areas, and recent tests have found few problems with porous pavement in areas having severe
winters (Pitt and Voorhees 2000). Work at the University of Guelph in Ontario (Thompson and James
1995) has shown that porous pavement systems can also be effective filters to remove particulate
pollutants from runoff.
Experiments in Bordeaux and Paris, France have shown that porous pavements were very efficient in
reducing the pollutant loads discharged into the receiving water (Balades et al. 1995a,b). These French
studies have shown that the pollutant removal efficiencies for suspended solids can be between 50 and
70%, between 54 and 89% for COD, and between 78 and 93% for lead. These reductions were associated
with the high amount of infiltration of water, and associated pollutants, through the pavements, away
from the surface drainage. These experiments confirm results from previous studies in other countries
(Hogland et al. 1987, Pratt et al. 1989, Pratt et al. 1995).
The primary objective of using porous pavements is to mimic natural flow and infiltration conditions as
closely as possible. It is therefore very important to pay attention to the following aspects to reduce
groundwater contamination potential (Pitt et al. 1996):
Depth to groundwater
Groundwater uses
Risks due to industrial activities in the catchment
Use and traffic levels on the porous pavement
12-80
-------
Use of de-icing salts on the street
12.2 SLAMM CALCULATION PROCEDURES FOR POROUS PAVEMENTS
SLAMM (Pitt and Voorhees 2000) uses a calculation procedure for porous pavement performance that is
almost identical to the general infiltration device procedure of Chapters 8 and 9. However, porous
pavements are only assumed to treat the paved area, with no additional flows from upland areas
discharging to the pavement.
Therefore:
f
fractional volume reduction =|
i
^
A,
(12-1)
where
f = the percolation (infiltration) rate of the porous pavement: the pavement base, or the
soil, whichever is less (depth/time),
i = the rain intensity = total rain/rain duration,
Ap = the paved area, and
At = the total study area.
Again, the ratio f/i of Equation 12-1 must be less than or equal to 1.0.
An example follows:
Percolation rate = 3 in./hr
Total rain = 1.7 in.
Rain duration = 6 hrs
Porous pavement area = 0.7 acres
Total study area = 5.3 acres
Therefore i = 1.7 in./6 hrs = 0.283 in./hr
The ratio of f/i therefore = 3/0.283 = 10.6 which indicates an over-design for this rain, requiring the use of
1.0 in the performance equation.
The volume reduction is therefore 0.7 acres/5.3 acres = 0.13 (13% reduction in flow and pollutant yield).
SLAMM documentation does not cite a source for porous pavement percolation rates. The example used
in the SLAMM inputs a percolation rate of 3 in./hr. Use of this method to determine porous pavement
performance requires the user to have a percolation value, which may be equal to the underlying substrate
percolation rate.
12.3 SIMULATION OF POROUS PAVEMENT WITH SWMM
When using SWMM, porous pavement may be treated as a pervious surface and either the Horton or
Green-Ampt infiltration equations employed. Because porous pavement installations often have
provision for subsurface drainage, laterally away from the paved area, the SWMM subsurface flow
routines may be used for this purpose. James et al. (2001) demonstrate how SWMM can be used
effectively in this manner. These authors discuss the key parameter choices necessary to simulate surface
and subsurface runoff from the current Runoff Block hydrologic routines and include an extensive
discussion of parameter selection. For ease of use, an interface for parameter selection has been included
in the PCSWMM graphical user interface (www.chi.on.ca). The reader is referred to James et al. (2001)
for details. One limitation is that the SWMM groundwater routine does not perform routing of infiltrated
quality constituents. Instead, the quality of effluent groundwater is input as a constant concentration in
12-81
-------
the current SWMM. If linked surface-groundwater quality routing must be performed, HSPF is one
model that does this (Bicknell et al. 1997).
12-82
-------
13 HYDRODYNAMIC DEVICES
13.1 INTRODUCTION
Hydrodynamic devices range from oil-water separators, which are essentially flotation devices but may be
simulated on the basis of distinct pathways for one portion of the flow vs. another portion of the flow, to
much more complex and often proprietary devices, such as a swirl concentrator, StormCeptor,
Vortechs, etc. (Minton 2002 provides a description of several such devices). These latter devices often
rely on a vortex or similar secondary flow pattern to separate heavier grit from a cleaner overflow.
Modeling these devices by process would be difficult due to the individual variation between devices.
When trying to describe performance of this widely varying group, a black box method may be the most
realistic choice that identifies volume treated vs. volume bypassed and manufacturer specifications. This
approach neglects maintenance, malfunctioning, or poorly sized devices, and would probably reflect
maximum treatment rates. The EPA/ASCE BMP Database is currently collecting performance
information on such devices. In the future this may be a good source for treatment efficiencies, although
currently there are few listed.
13.2 SIMULATION OF HYDRODYNAMIC DEVICES WITH P8
Particle (and associated pollutants) removal from hydrodynamic devices can be modeled in P8 with a
particle scale removal factor. Hydrodynamic devices only work for particulates (no dissolved nutrient
removal, no decay). Loading rate, particle size, and maximum treatment rates are all of concern when
modeling these devices.
Just as this parameter was adjusted in the Pond Section, the Particle Removal Scale Factor allows for the
calibration of an increase or decrease in device removal efficiency. This adjusts the removal rates for
each device, and is usually set to 1.0. Other values can be used to account for effects of filters or other
factors that affect particle removal and can be calibrated to the manufacturer's specifications (essentially
outflow concentration = removal factor times inflow concentration). However, the possible usefulness of
the particle scale removal factor in SWMM has already been discussed (Chapter 7). It may be preferable
to use the tanks in series model (N CFSTRs, Equation 6-11) as a more conventional empirical tool for
unit process design.
13.3 SIMULATION OF HYDRODYNMIC DEVICES WITH SWMM
The SWMM S/T Block is currently able to use performance data (i.e., removal equations) to simulate
these kinds of devices. The Extran Block is often able to simulate the complex hydraulics of such
13-83
-------
devices, but without corresponding water quality computations. Nonetheless, flow treated vs. flow
bypassed could likely be computed with Extran.
Although the data are seldom presented in terms of overflow rate, because residence time in a secondary
flow device increases as volume and surface area increase, so must sedimentation. A brief review of
vortex separation by Minton (2002) indicates that performance increases as the diameter of the vortex
chamber increases. Hence, removal based on overflow rate (Equation 6-6) will likely work, if
performance data are available.
13-84
-------
14 CASE STUDY: LID SIMULATION IN PORTLAND
14.1 OBJECTIVES
To aid in identifying strengths and weaknesses of SWMM's ability to simulate LID and BMP options,
application to real basins is most useful. Various locations for LID simulation were discussed in Chapter
2, including locations in Portland, Oregon, for which monitoring data are available for both catchments
and BMPs, although the latter are more limited, as will be seen in Chapter 15. Advantages of the
Portland location were discussed in Chapter 2, including proximity to Oregon State University (OSU),
good cooperation and information by the operative agency, the Bureau of Environmental Services (BBS),
and most of all, an extensive data collection and archival effort (http://www.cleanrivers-pdx.org/).
This chapter discusses a Portland study area used for LID simulation in detail, and the application of
SWMM for this purpose. A different Portland area is described in Chapter 15 for a BMP simulation
example. In this chapter, the Runoff and Extran Blocks are used to represent a portion of the Sullivan
area combined sewer system down to the parcel (individual lot) level using actual rainfall-runoff data
monitored by the BES during fall 1998 - spring 1999. The ability of SWMM to simulate LID scenarios
is demonstrated. In the next chapter, the Runoff and Transport Blocks are used to perform quantity and
quality simulation of a detention pond serving a small catchment in the Lexington Hills area of southeast
Portland, including water quality simulation. LID and BMP simulation capabilities and limitations are
noted on the basis of these simulations.
14.2 PORTLAND COMBINED SEWER STUDY AREA
The Portland, Oregon Sullivan combined sewer basin is a 1,700-ac area located on both sides of the
Banfield Freeway (Interstate 84), from about NE 25th to NE 55th Avenue in northeast Portland. The land
use is primarily single-family residential with localized zones of commercial properties. The overall
basin imperviousness is about 46%. Detailed information on the Sullivan area and neighboring Stark and
Holladay areas is provided by Carollo Engineers (1999). Additional information is provided by Adderley
and Mandilag (2000a, 2000b) and Hoffman and Crawford (2000). Additional information on the study
area and modeling is given by Huber and Cannon (2002). A portion of the Sullivan Basin is shown in
Figure 14-1.
The City of Portland's BES has been modeling this area since the late 1990s. The area has been targeted
for the City's Downspout Disconnection Program, and the program has very detailed information on
percentages of roof area that are currently disconnected, including a complete description of each parcel.
Hence, optimization for LID can be studied using this monitored, real system.
14-85
-------
Initial efforts to use BBS data resulted in huge amounts of data in formats not readily accessible to
researchers at OSU. BBS has modeled much of the City's combined sewers and uses Maplnfo for their
GIS information. With the help of BBS personnel, OSU staff were able to define a usable study area for
modeling and convert GIS information from Maplnfo into the Arc View format usable at OSU. The BBS
performed much of the basic data preparation described subsequently, and the OSU study described
below is a subset of the larger Sullivan area evaluated by the BBS.
1 HE BRAZEH ST
'
S ii -fil i1
: ^LWUDU
it I" I" !' f" o'l'
I Sknt
- : u^iC'^-.:.^
.' IH ^^H ,^^ii__
-ji 11
.
,
,
MY ;:
- I
,iL"JJ.,4.
ST^C
s^si*S₯r«
madSj
V'^iioiT [
^
«f=-" -v
j^F ^
--|M ,^
% - =3=
J%%^' "s=aT*s**
\-
^
L'lPW
'\
Bureau of Environmental Services
Sullivan, Stark, Holladay Ba&ins Sewer
,.) Relief and Reconstruction Predesign
Probfem Characterization Report
SW Sullivan Sub-Basin
Future Conditions
Basement and Non-Basement
Flooding Risk/Characterization
uiiU Assoclaied Finn*
Figure 21
Figure 14-1. Sullivan area (Carollo Engineers, 1999). The sub-area used for this study is just south of Grant Park.
Color code relates to basement flooding risk. Red: very high risk; yellow: high risk; green: medium risk; gray: some
risk; white: low risk.
The selected study area is between 33rd Ave. and 37th Ave. (west and east boundaries) and between the
Banfield Freeway (1-84) on the south and Grant Park on the north (Figures 14-2 and 14-2). A six-block
area of 115 single-family residential lots (also referred to as parcels ortaxlots) totaling 16.9 acres with
35% impervious area was modeled. Average lot size is just over 0.1 acres. An aerial photo of the area is
shown in Figures 14-3, which emphasizes the density of housing and small lot size.
This area has been monitored extensively because of basement flooding in the area. There is a flow
monitor in a 16-in. pipe draining the study area and a tipping bucket rain gage a few blocks away (Figure
14-4). Monitoring data are available from November 98 - May 99 at 5-minute intervals. Evaluation of
14-86
-------
rating curves indicated flows measured at depths greater than 2.3 inches, corresponding to 2 cfs, were
more reliable than lower flows; hence most reliance was placed on these higher flows. The Arc View
layer in Figure 14-4 also shows the roof density of the neighborhood and the main sewer laterals.
Figure 14-2. Combined sewer study area, south of Grant Park (Carollo Engineers 1999). See color codes in
caption to Figure 14-1.
Figure 14-3. Aerial photo of combined sewer study area.
14-87
-------
14.3 DATA PREPARATION METHODS
14.3.1 Required Parameters
The SWMM Runoff Block converts rainfall into runoff using a nonlinear reservoir technique (Huber and
Dickinson 1988). Each subcatchment is characterized by the following parameters:
Area
Imperviousness
Width
Slope
Depression storage (pervious and impervious subareas)
Manning's roughness (pervious and impervious subareas)
Three Green-Ampt (G-A) infiltration parameters
These parameters were developed for both an aggregated and disaggregated schematization of the 16.9-
acre basin, as described in the following sections. However, the same values for depression storage,
roughness, and infiltration parameters were used for all subcatchments, as indicated in Table 14-1. These
values are based on BBS simulations and data. All model runs shown are uncalibrated! Although this
report's authors relied on BBS estimates for baseflow, there was no attempt to improve upon the
parameter estimates described below to obtain better fits, mostly because the initial comparisons were
very good. In fact, all comparisons of simulated and monitored flows are generally good, due in part to
the relatively high imperviousness of the overall basin.
Table 14-1. Constant model parameters.
Parameter
Depression storage, in.
Manning roughness
G-A suction, in.
G-A hydraulic conductivity, in./hr
G-A initial moisture deficit
Impervious Area
0.03
0.013
n/a
n/a
n/a
Pervious Area
0.25
0.25
2.56
1.1
0.08
The Runoff Block was used only for surface runoff simulation; the sewer network was simulated in the
Extran Block. Hence, while most of the discussion that follows deals with the Runoff Block, simulated
vs. monitored flows rely upon Extran Block output at the monitor location. Extran block input for all
pipes in the system is shown in Table 14-2.
14.3.2 Directly Connected, or DCIA Subcatchments
14.3.2.1 Introduction
Directly connected impervious area (DCIA) consists of the impervious area of each parcel that is directly
connected to the sewer through laterals. DCIA subcatchments are delineated for each pipe with service
laterals, i.e., for every parcel with a sanitary sewer connection. They contain only the impervious area of
a parcel and are therefore 100% impervious (except for the disconnection program described below). It is
assumed that the pervious areas of the parcels (and impervious areas not directly connected to the sewer
system) drain to the street and are therefore included in the Surface Water (SW) Subcatchments.
Although this is generally the case for single-family parcels it may not always be the case for commercial
areas - but there are no commercial areas in the small study area.
14-88
-------
Table 14-2. Extran Block conduit input data. Conduits are identified in Figure 14-6.
Conduit
No.
62
61
57
58
56
60
55
54
51
49
50
52
Upstream
Junction
62
61
57
58
56
60
55
54
51
49
50
52
Downstream
Junction
64
62
62
57
57
56
56
55
55
51
51
50
Diameter,
ft
1.33
0.67
1.17
0.67
1.17
0.67
1.00
0.67
1.00
0.67
0.67
0.67
Length, ft
130
433
109
526
108
427
133
477
129
429
97
580
ZPl*,ft
0
0
0
0
0
0
0
0
0
0
0
0
ZP2*, ft
0
0.31
0
0.31
0
0.64
0
0
0
0.3
0
0.3
Manning
n
0.013
0.013
0.013
0.013
0.013
0.013
0.013
0.013
0.013
0.013
0.013
0.013
*ZP = vertical displacement of pipe invert above junction invert. ZP1 = upstream end, ZP2 =
downstream end.
14.3.2.2
Delineation
Two primary models will be described below. In one model ("disaggregated"), every individual house
parcel (lot) is considered individually, and separated into directly connected impervious area (DCIA) and
the remainder (pervious plus non-DCIA imperviousness). Although the computation of DCIA is
reasonably precise, through a combination of aerial photos and GIS analysis, it is complicated by
Portland's downspout disconnection program (described below), which applies to the study area. The
second basic model is one in which DCIA and the remaining surface area is aggregated into 14 bigger
subcatchments, to test the effect of aggregation.
Parcels draining to multiple sewers (i.e., along a low ridge, such that front and back yards drain in
different directions) were divided into smaller areas, each with its own (sewer) lateral pointer in the GIS.
Runoff from the DCIA Subcatchments is typically inserted into the model at the upstream manhole of
each major sewer lateral.
14.3.2.3
Impervious Area
The impervious roof and parking areas for each parcel (see Figure 14-4) were obtained from
photogrammetric maps of the City. The model was based on actual impervious areas only; no impervious
area assumptions were made based on land use. However, the City of Portland has an active downspout
disconnection program, with incentives for homeowners of $53 per disconnected downspout
(http://www.cleanrivers-pdx.org/get_involved/downspout_disconnection.htm). Hence, some of the roofs
and parking lots (none of the latter in this study area) are disconnected from the service laterals and flow
to vegetated areas or drywells. Areas with drywells are termed "sumped" areas in text below. DCIA
imperviousness was then modified, as follows, based on information provided by the BBS that applies to
the whole Sullivan area, not just this small study area:
14-89
-------
Figure 14-4. Arc View map of study area showing individual house parcels and rooftop imperviousness.
Aggregated subcatchments used by BBS are shown in heavy black. The sewer network is shown in heavy blue. The
monitoring station on NE 35th Ave. is shown with a star and the raingage with an asterisk (*). The top (north) east-
west street is Tillamook, the next is Hancock, and the east-west street above the monitoring station is Schuyler. The
north-south street on the east is 37th Ave. The north-south street on the west, not drawn, would be 33rd Ave.
Dimensions of the figure border are approximately 1050 ft (320 m) wide by 590 ft (180 m) high.
1) Twenty percent (20%) of single-family roofs are disconnected from the combined sewer in unsumped
areas. This disconnection is assumed to be 70% effective (i.e., some of the disconnected water flows
over the curb and into the sewer through a street inlet). The effective disconnection rate accounting
for both of these factors is 14%, with the surface water subcatchment receiving the remaining 6%, as
discussed in the next section.
2) Thirty percent (30%) of single-family roofs are disconnected in sumped areas, and this disconnection
is 100% effective since the sumps (drywells) are assumed to have capacity to accept all the roof
runoff.
3) Twenty percent (20%) of commercial roofs and parking lots are disconnected to a drywell in sumped
areas, but this is irrelevant to this small study area.
Based on disconnection survey data, the impervious area for the DCIA Subcatchments was
computed by the BBS as outlined below. The following equations use the existing impervious
areas (measured from aerial photos using the GIS) and assumed disconnection rates to calculate
the impervious area, subcatchment area, and impervious percentage of each parcel that is directly
connected to the sewer (assuming no commercial or parking areas in the catchment). That is,
each DCIA area is scaled back uniformly for each lot over the study area.
14-90
-------
In an unsumped area:
Impervious Area = 0.86 x Area Single Family Roofs
In a sumped area:
Impervious Area
= 0.70 x Area Single Family Roofs
(14-1)
(14-2)
The assumptions are incorporated into the Runoff Block subcatchment data supplied by BBS and
incorporated into the SWMM runs described subsequently. The remaining area of each parcel may
include some imperviousness and is described below under "Surface Water Subcatchments" (SW
Subcatchments). DCIA runoff enters the sewer system in the model at the upstream end of each main
lateral.
14.3.3 Surface Water Subcatchments
14.3.3.1 Introduction
Surface Water Subcatchments (SW Subcatchments) are delineated for each of six sub-basins. Street and
sidewalk pavement and pervious areas from the parcels are included in the SW Subcatchments. All
surface water runoff is collected by sub-basin and delivered to the six corresponding nodes. When this is
broken down to the parcel level, each parcel has a SW Subcatchment associated with it. However, they
do not include the impervious areas (DCIA) of each parcel that are connected to the sewer through service
laterals (i.e., sewer pipes connected to individual homes). These areas are already included in the DCIA
Subcatchments, as discussed in the previous section. Delineation of Subcatchments and determination of
impervious area, width, and slope with the GIS are discussed below.
14.3.3.2
Delineation
The SW Subcatchments were delineated by the BBS with the aid of a digital terrain model (DTM). The
DTM was created from contour maps of the project area using Vertical Mapper, a third party application
for Maplnfo. Slope and aspect grids were created from the DTM. The aspect grids show the direction
that each grid drains expressed as degrees from north. A vector representation of the aspect was created
for each grid to assist with delineation of the SW Subcatchments; an example is shown in Figure 14-5.
Figure 14-5. Example slope and aspect grids from the digital terrain model.
14-91
-------
14.3.3.3 Impervious Area
SW Subcatchments with infiltration sumps that were installed prior to the date being modeled were given
a data flag so that they would not be included in the SWMM Runoff model. This assumes that 100% of
the runoff is taken by the sump and removed from the combined sewer system. Otherwise, parcel
imperviousness was computed for SW Subcatchments as follows:
Unsumped area:
Parcel Impervious Area = Inefficient Portion of Disconnected Single Family House
= 0.14 x Area Single Family Roof (14-3)
Sumped area:
Parcel Impervious Area =0 (14-4)
For both areas:
Parcel Pervious Area = Area of Parcel - Impervious Area (14-5)
Imperviousness in sumped areas is just the street pavement since the sumps are assumed to be
100% effective. The pervious area includes sidewalks and driveways that drain onto adjacent
lawns. "Parcel Impervious Area" above is treated like DCIA in the model.
Aggregated and disaggregated subcatchment models were run, as described below. For the
aggregated simulation,
Impervious Area = Parcel Impervious Area + Street Pavement Area (14-6)
14.4 THE MODELS
14.4.1 Two Model Types
Two models have been created for this area:
1. Aggregated subcatchment model (A-Model): 11 DCIA Subcatchments plus three combination (pervious
plus impervious) Subcatchments = 14 total Subcatchments. The main purpose of these runs was to
compare OSU's efforts with prior BBS efforts, and to compare aggregated vs. disaggregated simulation
results.
2. Disaggregated subcatchment model (I-Model): 115 house parcels (containing DCIA and pervious area)
plus three street Subcatchments =118 total Subcatchments). The purpose of these runs was to determine
the added value of a highly discretized and detailed subcatchment schematization and to be able to
separate impervious from pervious area more accurately.
Both models contain three sub-basins (moving north to south) that correspond to the three east-west cross
streets (north to south: Tillamook, Hancock and Schuyler, Figure 14-4). Each sub-basin has a north-south
sewer main line and two east-west service laterals on either side of the main line (Figure 14-6). Pipe sizes
range from 6 to 16 inches throughout the study (Table 14-2). The street and sidewalk components for the
aggregated and disaggregated models are shown in Figure 14-7.
14-92
-------
49
GO
56
61
50
52
55
V
57
Figure 14-6. Pipe segment ID for the study area, as simulated in Extran. The aggregated subcatchment model
(A-model) includes the DCIA in parcels in the six areas draining to laterals 49, 52, 54, 58, 60 and 61. Five smaller
DCIA areas (not shown) drain to north-south trunk sewer segments 51, 55, 56, 57 and 62. For the aggregated
subcatchment model, streets, sidewalks and all pervious areas are lumped into three surface subcatchments. The
street and sidewalk components of these surface subcatchments are shown in Figure 14-7.
Figure 14-7. Aggregated areas for three street subcatchments include roads, sidewalks and grass strips. The
three areas shown are the actual areas used for the disaggregated modeling and conceptual areas for the aggregated
modeling, since the aggregated modeling adds all pervious area from house parcels into the three surface
subcatchments. Subcatchment 28349051 drains to pipe 51 (Figure 14-6), etc. (The first four digits of the
subcatchment IDs are not included in the model data.)
Model A aggregates all DCIA for each of the six main laterals (49, 52, 54, 58, 60 and 61 in Figure 14-6)
into one DCIA Subcatchment, plus just one SW Subcatchment for the entire sub-basin. SWMM input
14-93
-------
data for Model A are heavily based on earlier BBS SWMM runs. The SW Subcatchment aggregates all
surface water in the sub-basin (over both pervious and impervious areas) and concentrates it into the
sewer mainline for the sub-basin (see example in Figure 14-8). The three SW Subcatchment for the I-
Model are shown in Figure 14-7. Model I takes the catchment down to the parcel level, as shown in
Figure 14-9. SWMM input data for Model I were prepared entirely by OSU personnel.
Monrtor2_piirtshp_^.
Raingage_poinUhf
EH
; Build83a_region.s
cn
yj PiPes_polyline.shp
A/
_| Nodes2_point.shp
Surfac
EH
Figure 14-8. The highlighted sub-basin 28349056 (referred to as 56) has four DCIA Subcatchments: one for
each pipe segment, main line (1) and lateral (2), with DCIA, and one SW Subcatchment for all the remaining area.
,^
*
_| Trees_eliipse.shp
_| Trees_region.shp
1 1
_| Build33a_region.sh
_| Untitled polyline.sh
A/
^f Parcefe region. shp
CD
^\ Nodes2_point.shp
*
_| Nodes2_point.shp
*
^ Surface subcatchrr
"^ CD
<
,
N
J^~
/
209
209
209'
2084
>083
n
i i
b_124 E178
5 P159 C19
212
212
212C
2119
m
215'
215C
2147
2144
,
?174
2173
2171
219;
2191
2190
2189
2188
^222
!22(
221E
225
-n^
/
2312
228'
228;
2311
2309
2215
2211
208
2250
2249
?248
2278
2275
2308
7307,
>270
268
_3p_6
305
2323
r
^ ^V
235"
2352
2353
238(
9377
2346
'347
!376
1
]
241 :
241 ;
741 C
2406
?402
-^' L
iNr
B
338
!
243
243"
j
243E
'434
7433
2432
r
245
245C
2448
2447
2446
i
247;
2471
746S
2464
>461
i
250!
2504
2500
?496
252'
252;
'520
2518
>516
[
254J
'547
7542
J
!257
3 2608
/(
2577
2597
'
>566
2594
S
/
4
-*
Figure 14-9. Individual parcels for the disaggregated model.
14-94
-------
Although it may appear that subcatchment 2268 should drain south (to the bottom of Figure 14-9), in fact,
the surface drainage is north across subcatchments 2306 and 2270, as indicated in Figure 14-10.
Similarly, it appears in Figure 14-9 that the DCIA of subcatchments 2268 and 2305 drains to the sewer
line below the monitor, but according to the BBS this is not true and is only an artifact of the Maplnfo
schematic.
The SWMM Runoff Block simulates a subcatchment as having pervious and impervious (DCIA)
subareas. Alternatively, each subcatchment could be split into two, for each surface type. The two
methods were compared for the I-Model runs, that is, 115 parcel subcatchments with both pervious and
impervious subareas vs. 115 pervious subcatchments plus 115 impervious subcatchments totaling the
same area. (Because some individual parcel subcatchments were already 100% impervious, the total
number of subcatchments for the latter option was actually 216 + 3 = 219 instead of the expected 115 +
115 + 3 = 233.) Since results were identical, the 115 combined land surface subcatchments were used for
most of the simulations. The parcel sub-area distinction is shown in Figure 14-11. Runoff from both the
impervious and pervious subareas is routed to the upstream end of the main sewer lateral for the street.
Remaining area in the sub-basin not associated with an individual plot (streets, sidewalks, grass strips,
etc.) is aggregated as one of three additional SW Subcatchment (Figures 14-7 and 14-8) and routed to the
top of the sewer lateral serving the sub-basin.
All DCIA that is within the sub-basin but directly connected to a lateral outside of the study area was
omitted in the I Models (Figure 14-12), which had the effect of reducing the total area from 16.9 acres in
the A Models to 16.4 acres in the I-Models. Surface runoff from the portion of these plots draining into
the modeled sub-basins is included in the aggregated SW Subcatchment (Figure 14-13).
V R*tai**t?,p«Mt
-------
fill
Figure 14-11. Dark areas (blue) are DCIA and are surrounded by lighter (lavender) pervious areas. Both
subareas are represented for each parcel in the disaggregated modeling. For the aggregated modeling, the pervious
(lighter) areas are combined and added to the appropriate one of five street subcatchments and the DCIA (darker)
areas are combined to form six DCIA subcatchments.
Menttor2_pomtshp_±
Raingage2 point.sf
cn
Build83a_region.sh
\ Hodes2_point,shp
Nodes2_point.shp
urface_£ubcatchrr
Figure 14-12. The dark (blue) DCIA connected outside the sub-basin is not included in the model as the
runoff effects are not noticed at the monitor. The DCIA connected to service laterals in the highlighted sub-basin
is all modeled within individual DCIA Subcatchments (I-Models) or summed for the appropriate aggregated
subcatchment (A-Models).
14-96
-------
Figure 14-13. Parcel land surface definitions. The pervious (light, pink) parcel area adjacent to the highlighted
sub-basin (yellow, at bottom) is included in the aggregated SW Subcatchment for the sub-basin in the A-Model.
Pervious area (within the watershed, Figure 14-11) is included within each of the 115 individual parcels for the I-
Model.
14.4.2 Width and Slope
The Runoff Block requires a modeling parameter -width to indicate the shape or flow path of the
subcatchment
Width = Parcel Area/Flow Length
where, for individual parcels in the I-Models, Flow Length was assumed to be:
Flow Length (ft) = 25 ft of overland flow +
subcatchment house connection length/6
(14-7)
(14-8)
This equation for width generally attempts to discount the channelized flow in the pipe and make the
DCIA Subcatchments concentrate faster then the SW Subcatchments. In the I-Model runs, most widths
were rounded to 50 ft in order to facilitate sensitivity analysis. The width parameter does not have a great
influence on model results for these relatively small Subcatchments, since the time of concentration of the
subcatchment is much less then the duration of most storms. Hence, equilibrium (peak) flow is reached
for each hyetograph interval regardless of the width.
For aggregated A-Model Subcatchments, the length of overland flow was assumed to be 150 ft of flow
from the back of a lot to the street plus the distance from the farthest pavement point in the subcatchment
to the surface water inlet. The distance from the farthest pavement point was determined with the GIS by
using a routine to create points at 5-ft intervals along a street centerline map and querying for the farthest
point within the subcatchment. Again, Equation 14-7 was used to compute the widths for the aggregated
Subcatchments.
14-97
-------
The slope of all DCIA Subcatchments is assumed to be 6%, a combination of single-family roof slopes
and lawns. The slope of each SW Subcatchment was calculated from the slope grid (Figure 14-5). The
average slope of all of the 40-ft grids within the subcatchment was used as the subcatchment slope.
14.4.3 A-Model Input Data
Because there are only 14 subcatchments for the aggregated model, Runoff Block subcatchment data are
shown for this simulation in Table 14-3. This includes the three SW Subcatchments (labeled with 9000-
numbers) and the 11 DCIA subcatchments (labeled with 8000-numbers). The numbering scheme
corresponds to the last two numbers of the sewer segments shown in Figure 14-6. For instance, SW
Subcatchment 9056 flows to pipe 56 and drains the entire middle, yellow-highlighted sub-basin shown in
Figures 14-7 and 14-8. DCIA Subcatchment 8060 drains to pipe 60, etc. All DCIAs drain to the top end
of the lateral. It is assumed that the pervious areas of the parcels drain to the street and are therefore
included in the surface water catchments. The remainder of each of the three main sub-basins is
aggregated into one SW Subcatchment representing all pervious areas and the impervious area not
directly connected to the sewer system (but including the streets, since this was the scheme used by BBS).
These drain to the sewer mainline for the sub-basin. All slope, hydraulic connectivity, infiltration, and
other parameter variables (Table 14-1) are taken from BBS data.
Table 14-3. Subcatchment input data for aggregated models (A-Models). Other parameters are the same for all
subcatchments and listed in Table 14-1.
Subcatchment
ID
9051
9056
9062
8049
8051
8052
8054
8055
8056
8057
8058
8060
8061
8062
Flow to
pipe:
51
56
62
49
51
52
54
55
56
57
58
60
61
62
Width,
ft
248
276
274
190
18
181
226
21
25
36
221
215
245
116
Area,
ac
4.38
4.7
4.6
0.42
0.02
0.51
0.54
0.02
0.02
0.04
0.57
0.47
0.55
0.12
Imperviousness,
%
16.8
18.1
18.2
100
100
100
100
100
100
100
100
100
100
100
Slope
0.077
0.072
0.08
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
0.06
Since the basin is a combined sewer area, baseflow (dry-weather flow) needs to be considered. BBS data
indicate an average of about 0.032 cfs at the monitor. The BBS provided distributed baseflow estimates
for each sub-basin, and these were added to the upstream end of each lateral in the Extran simulation.
However, the 0.032 cfs baseflow is almost impossible to discern during storm events.
14.5 MODELING RESULTS
Uncalibrated SWMM output, including baseflows, for January 13-18, 1999 for the aggregated and
disaggregated model representations is shown in Figure 14-14. This time period corresponds to a high 5-
14-98
-------
day rainfall total of 2.42 in. (61 mm) and was used because BBS focused strongly on this time period
during their modeling efforts. There is good correlation with monitor flow data, but the comparison is
difficult to make visually when the five days are compressed into just one plot. Hence, the event of
January 17 is considered in more detail in Figure 14-15, wherein it may be seen that there is practically no
difference between the aggregated and disaggregated model simulations. This is good news for
continuous modelers, for whom an aggregated model representation will require much less computer
time.
1.6
1.4
1.2
1.0
Monitor Flow
Model A with base flow
included
01/13/99 01/14/99 01/15/99 01/16/99
Date
01/17/99
01/18/99
Figure 14-14. Five-day comparison of simulated and measured flows at the monitoring site. A visual
comparison is difficult because of the crowded scale.
7:12:00 8:24:00 9:36:00 10:48:00 12:00:00 13:12:00 14:24:00
Time
Monitor Flow
-A-Model Flow (aggregated model)
l-Model Flow (disaggregated model)
Figure 14-15. Simulated and measured flows for seven-hour period on January 17,1999. The aggregated and
disaggregated models show very little difference.
14-99
-------
Generally, the model vs. monitor comparison is good with regard to shape, but the rising limb of the
modeled hydrographs is somewhat high, and the modeled recession limb is somewhat low. This might be
better simulated by detaining a little more water on the land surface prior to letting it run off. Recall that
no calibration or adjustment has been attempted. Extran continuity is very good for all runs, on the order
of 0.4%.
Another method for comparison is total runoff volume; these values are shown in Table 14-4. Baseflow
is necessary in order to obtain close to the monitored runoff volume. Runoff volumes are comparable for
the aggregated and disaggregated models, as would be expected from the inspection of the hydrographs.
This confirms the utility of aggregated simulations for long-term simulation.
Table 14-4. Model total flow comparison for five-day event, Jan 13-Jan 18,1999. In bottom row, Ks is the
saturated hydraulic conductivity. (After Huber and Cannon 2002.)
Rainfall for event
Monitored flow
Baseflow volume, @ 0.032
cfs
Model A (aggregated
catchments)
No baseflow
With baseflow included
Model I (disaggregated,
separate subcatchment for
each parcel)
With baseflow included
Model I-LID (re-routing
impervious runoff over
pervious area)
With baseflow included
Same as above but with Ks =
I/ 10 previous value
Total flow comparison for 5-
day event, Jan 13-Jan 18, 1999
(cubic ft)
148,987
58,130
13,824
47,244
60,459
60,492
26,829
27,762
(in.)
2.42
0.94
0.22
0.77
0.98
1.02
0.45
0.47
7-hour interval, 7:00 am to 2:00
pm, January 17, 1999
(cubic ft)
37,555
14082
806
12,492
13,365
13,416
4,494
4,872
(in.)
0.61
0.23
0.01
0.20
0.22
0.23
0.076
0.082
14.6 LID SIMULATION
The essence of LID is to retain as much water on site as possible at the parcel level (Prince George's
County 2000, Wright et al. 2000). One essential technique is to minimize direct connections to the
drainage system by routing runoff from roofs, driveways, etc. over pervious areas to promote infiltration.
The SWMM Runoff Block has been adapted for this purpose by Huber (200 la) as discussed in Section
4.7. As indicated in Figure 4-1, overland flow can be rerouted internally within subcatchment subareas
(i.e., from pervious to impervious and vice versa) and also may be routed from one subcatchment onto
another. The simulation shown below simply routes the impervious area runoff from each of 115 parcels
over the pervious area within the same parcels, sort of a massive, hypothetical LID effort for the
neighborhood.
14-100
-------
The I-LID simulation uses the I-Model as the base and reroutes 4.0 ac of impervious area runoff from
rooftops and driveways over 7.9 ac of pervious areas of each parcel. Street surface runoff (and sidewalk
and grass strip runoff from the three street subcatchments is unaffected. A drastic reduction in the runoff
hydrograph (Figure 14-16) and runoff volume (Table 14-4) is produced by the LID option. The 5-day
runoff volume is reduced by 56% and the 7-hr runoff volume by 67% (Table 14-4). This is to be
expected for this hypothetical simulation for which the saturated pervious area hydraulic conductivity of
1.1 in./hr will accept any intensity of runoff associated with typical western Oregon rainfall, including the
additional non-DCIA runoff from the roofs, etc.
o
7:12:00
8:24:00 9:36:00 10:48:00 12:00:00 13:12:00 14:24:00
Time
Monitor Flow
l-model, with baseflow I-LID Model
Figure 14-16. Comparison between Model I and Model I-LID simulations for a seven-hour interval, January
17,1999.
Soil is often compacted in urban developments, with a lower hydraulic conductivity than for pre-
development conditions (Pitt et al. 1999a). An additional run was made with a value of saturated
hydraulic conductivity, Ks, equal to one-tenth of the value used for the rest of the modeling. That is, the
new value of Ks = 0.11 in./hr compared to the previous value of Ks = 1.1 in./hr. The much lower
hydraulic conductivity resulted in only slightly higher runoff volumes (Table 14-4) for the 5-day and 7-hr
duration events. (The hydrograph is also a little higher but very close to the I-LID hydrograph shown in
Figure 14-16 and has not been plotted.) This reflects generally low rainfall intensities in western Oregon
(typically less than 0.11 in./hr, but of long duration). Thus, an even lower infiltration rate would be
necessary to reduce the effectiveness of the LID option simulated. This is to say that this LID option is
likely to be even more effective in climatological regions with rainfall (and runoff) intensities that are
characteristically low in magnitude.
This LID simulation is not completely hypothetical. Disconnected flows due to the downspout
disconnection program are directed to dry wells. Although this program mainly affects rooftop runoff, its
impact could be significant if implemented over a large area.
Infiltration is assumed to remove quality constituents as well, either through advection of dissolved
constituents into the soil, or by sedimentation of particulates as water enters the soil. Hence, infiltration is
14-101
-------
effective in cleaning up surface water. There may be concern, however, about the impact of infiltration
on groundwater (Pitt et al. 1996).
14.7 SUMMARY AND CONCLUSIONS
Version 4.4h of SWMM has been applied to a 16.9-ac combined sewer catchment in the Sullivan Basin in
northeast Portland, Oregon. With the aid of prior modeling work performed by the Portland Bureau of
Environmental Services, the SWMM Runoff and Extran Block simulation of the area compares well with
monitoring data for a 118-subcatchment disaggregated simulation (I-Model) of every house parcel (tax
lot). The simulations are uncalibrated and could easily be improved through additional effort, including a
better definition of baseflow in the combined sewers. In addition, the disaggregated simulation of every
individual house parcel compares well with an aggregated simulation that uses only 14 subcatchments to
simulate all the parcels, pervious areas and street surfaces. This suggests that long-term modeling (i.e.,
continuous simulation, if performed) can conveniently be done with a less detailed model representation.
When a typical LID option of routing non-directly connected impervious area runoff over pervious areas
is simulated, the hypothetical SWMM Runoff Block output indicates an expected reduction in discharge.
Although no hypothetical quality simulation was performed for this study area, quantity reductions by
infiltration induce corresponding quality reductions (Huber 2001a,b).
To summarize the key points presented in this chapter (Huber and Cannon 2002):
SWMM may be used to simulate LID options that involve routing of runoff from non-directly
connected impervious area (non-DCIA) over pervious areas, e.g., roof and driveway runoff over lawn
surfaces.
SWMM may also be used to direct surface runoff from one subcatchment over another.
Simulation of an aggregated model representation (14 subcatchments) performed about as well as a
disaggregated model representation (118 subcatchments) for the Portland, Oregon test area.
A hypothetical LID option that infiltrates all roof and driveway runoff in the dense Portland study
area is predicted to result in over a 50% reduction in runoff volumes and peak flows in the combined
sewer system. Although a reduction would be expected, the modeling allows a better quantification
of the potential results.
14-102
-------
15 CASE STUDY: BMP SIMULATION IN PORTLAND
15.1 OBJECTIVES
A detailed example of a hypothetical LID simulation in a real, well-monitored Portland, Oregon
catchment was presented in Chapter 14. A similar, if somewhat less detailed example is presented in this
chapter for simulation of a detention pond, representing a very typical BMP (Chapter 6). Once again,
excellent cooperation was received from Portland BBS personnel, including extra help with interpretation
of data and graphics. Additional details of the simulation are provided by Stouder (2003).
15.2 LEXINGTON HILLS AREA BACKGROUND
The modeled area was the Lexington Hills BMP area, which is an area with a constructed pond located in
the Johnson Creek drainage in southeastern Portland (Liptan 2001). The actual site is just to the west of
SE 162nd Avenue, approximately %-mile south of SE Foster Road (Figure 15-1). While there are two
ponds on the site and one planned, data was collected only for the northernmost pond (Pond 3),
constructed in 1996, which is located at SE 162nd Avenue and SE Flavel Drive. Aerial photographs of
the pond area are shown in Figures 15-2 and 15-3.
Although during the wet season there can be a small pool below the orifice outlet, Pond 3 is intended to
act as extended detention and is designed to fill for a storm size of 0.83 in. of rainfall in a 24-hr period. It
receives runoff from a 26.57-ac residential neighborhood with 8.75 ac of (Figure 15-4) impervious area.
The pond has approximately 3,000 ft3 of dead storage plus another 8,000 ft3 of active storage (Figure 15-
5) above the outlet orifice. Influent stormwater enters via a 24-in. concrete pipe, and outflow is
discharged through a 6-in. PVC pipe fitted with a 2.5-inch reducer (orifice), designed to empty the live
storage (extended detention storage) in about 24 hrs. There is also an overflow weir. Both the orifice and
the weir then discharge into a stilling well. After the stilling well, water is routed approximately 75 ft
down an open channel to Kelly Creek, a tributary to Johnson Creek.
In 2000 and 2001, pollutant samples were taken by the Portland BES at the inlet and outlet of the pond
(Figure 15-6) during seven different storm events (Liptan 2001) in order to test the pond's effectiveness at
pollutant removal. Storm events 1,4,5,6 and 7 were the focus of the SWMM modeling. Events 2 and 3
were discarded due to minimal rainfall. All storms except event 7 were selected by the BES so that 24-
hour antecedent rainfall was minimized in order to get results for a "first flush" of pollutants. The
sampling dates are shown in Table 15-1.
15-103
-------
Table 15-1. Storm events used in SWMM simulations.
Event
1
4
5
6
7
Date
4/13/00
10/9-10/00
3/1/01
5/14/01
5/15-16/01
Event Rain*, in.
0.73
0.81
0.49
0.56
0.30
Pond inflow, ft3
8,275
8.276
7,538
10,688
7,898
Pond outflow, ft3
8,075
6,815
6,217
13,364
8,917
*Holgate rain gage.
Vicinity Map
Lexington Hills
BMP Monitoring
ENVIRONMENTAL SERVICES
crrv OP PORTLAND
Figure 15-1. Location map for Lexington Hills BMP site (Liptan 2001).
15-104
-------
"'ft^'
TVC. vl>
»-V £A*:*s»*
»: "'; HI;;^\Y1ip<\
' ' ' T*p ' :' 'VV^i// j
*rw/
f ^fe%/: :
'«^t.'.;; ' rfc^LrfSl
Figure 15-2. Lexington Hills pond vicinity photo (Liptan 2001).
Figure 15-3. Lexington Hills pond site photo (Liptan 2001). Pond 3 is to left of intersection.
15-105
-------
.; v) -: :fio
~L
UtF'AHTMtHl 01
PUBLIC Ulll IIIIS
Si I62N0 AVE. SOUW OF FOSf&? HO.
HAWTHORNE RIDGE - PHASE I
STORM DRAIN LINE KEY MAP
SD10
Figure 15-4. Lexington Hills BMP site in relation to catchment (Liptan 2001). Pond 3 is at upper right corner of figure, to the left of the upper-right
intersection.
15-106
-------
Hl.l-."!»(.
~ t~ 1C (Iff N \ ::-^S '5
Figure 15-5. Details of Lexington Hills extended detention Pond 3 (Liptan 2001).
15-107
-------
Outlet
(Stilling Well)
City of Portland
Environmental Services
Site Map
Lexington Hills BMP Monitoring
Figure
3-2
Figure 15-6. Location of influent (Site 1) and effluent (Site 2) monitoring (Liptan 2001).
15-108
-------
Flow-weighted composites were prepared from sample bottles in order to determine event mean
concentrations (EMCs) at the influent and effluent of the pond. Flow measurements were conducted at
the pond entrance and in the stilling well of the pond exit (thus measuring the sum of the orifice and weir
flows). Details of the sampling methods are provided by Liptan (2001).
15.3 SWMM MODELING
15.3.1 Background Information
Fortunately for this project, the BBS had performed prior SWMM modeling of the overall Johnson Creek
watershed. A map of Johnson Creek subcatchments is shown in Figure 15-7, with a more localized view
shown in Figure 15-8. BBS SWMM data were used as the starting point for simulations described herein.
PCSWMM (http://www.computationalhydraulics.com/) was used as the graphical user interface (GUI) to
run SWMM version 4.4h.
The City of Portland has an extensive, cooperative network of tipping bucket rain gages (with the U.S.
Geological Survey, USGS) that provide rainfall at 5-min. intervals (http://oregon.usgs.gov/non-usgs/bes/).
The SWMM Rain Block provides the ability to process such long-term data for input to the model. Rain
values were supplied from the Holgate rain gage ("gage 21"), located approximately 2 miles to the
northwest of the site (a general map, from the web site, is shown in Figure 15-9). Continuous 5-min.
rainfall data for the period from January 1, 2000 to December 31, 2000 were imported into the Rain
Block. Rain Block output was then linked to the Runoff Block. At first, actual rainfall values used for
the measured data were averaged from the Holgate and Pleasant Valley rain gages (Figure 15-9).
However, after reviewing the rainfall data, it was determined that no significant changes in simulated
rainfall occurred that warranted the averaging of the two rain gages for SWMM modeling; hence, just the
Holgate data were used for all runs.
Next, a Runoff Block input file was generated using the subcatchment data provided by the BBS.
Interestingly, the BBS used an imperviousness percentage of 14, compared to the ratio 8.75/26.57 = 33%
given in the BBS report (Liptan 2001). The 14% value was used in the simulations because it gave much
more reasonable results than the 33% value (see discussion of volumes below). Another slight
discrepancy is in catchment area: 26.98 ac in the BBS SWMM data vs. 26.57 ac in the data report. The
former was used in these simulations.
The wet time step simulated was 15 minutes, while the dry time step was set to 24 hrs. Dates of
simulation were set to correspond to the storm events in which sampling took place by the BBS. The
SWMM 4.4h Runoff Block input file is provided in Table 15-2.
15-109
-------
Figure 15-7. Regional subcatchments used in BES modeling of Johnson Creek Watershed. Lexington Hills is at subcatchment JC35 in center of the map.
Johnson Creek runs in red, from right (east) to left (west).
15-110
-------
RIVO
V022 i ra|V021
Figure 15-8. Localized view of Johnson Creek Watershed subcatchments. Johnson Creek is shown in red and flows west (right to left) in upper portion of
map. The Lexington Hills subcatchment is number JC35 in center of map.
15-111
-------
Pleasant Valley School-'
Figure 15-9. General map of BES raingage network for Johnson Creek near Lexington Hills (Lexington Hills is between the Holgate and Pleasant
Valley School gages). The large river on the left (west) is the Willamette, draining to the north.
15-112
-------
Table 15-2. Runoff Block input for Lexington Hills Pond 3 simulation (Stouder 2003).
* File control and file linkages not needed with PCSWMM
*SW 1 0 9
*MM 9 11 12 13 14 15 16 17 18 19
$ANUM
$RUNOFF
* Title Lines
Al 'Lexington Hills'
A2 'BMP Pond 3, April 13, 2000'
*
*
Bl
*
B2
*
B3
*
B4
*
Run Co
METRIC
0
IPRN1
0
WET
900
PCTZER
0
ntrol
I SNOW
0
IPRN2
1
WETDRY
900
REGEN
0
NRGAG
1
IPRN3
0
DRY
86400
INFILM KWALTY I VAP NHR NMN
11000
IRPNGW
LUNIT LONG
3 365
NDAY MONTH
13 4
IYRSTR IVCHAN
2000 0
* ROPT
Dl 1
^Rainfall for November 26 2000 through November 29 2000
*Entered from PCSWMM Meteorological module.
*
* Evaporation Data
^r
* VAP(l) VAP(2) VAP(3) VAP(4) VAP(5) VAP(6) VAP(7) VAP(8) VAP(9) VAP(IO) VAP(ll) VAP(12;
* Use default of 0.1 in./day
*
* NAMEG NGTO NP GWIDTH GLEN G3 GS1 GS2 G6 DFULL GDEPTH
*
* Conduits/Channels
*
* None used in this simulation.
*
* Subcatchments
15-113
-------
0.25
0.03
0.1
9.0 0.4
Table 15-2 (Continued. Please note wrap-around on HI - H3 and J2 lines.)
* JK NAMEW NGTO WIDTH WAREA IMPERV WSLOPE IMPER_N PERV_N WSTORE1 WSTORE2 SUCT HYDCON
IMD
HI 1 'KC00372#2' 'KC00372' 160 26.98 14 0.06
0.00005
* NMSUB NGWGW ISFPF ISFGF BELEV GRELEV STG
H2 'KC00372#2' 'KC00372' 0 0 0.0 8.5
* Al Bl A2 B2 A3 POR WP
H3 0.0014 1. 0. 1.0 0.0 .35 0.1
* HCO PCO GET DP DET
H40.1 5. .5 0 1.0
0.15
BC
TW
7 9
FC HKSAT TH1
0.15 0.15 0.15
* Water Quality
*
JJ
*
Jl
*
J2
*
RCOEF
J3
0
J3
0
J3
0
*J3
5
J3
0
*
J4
*
J5
*
LI
IMUL
0
NQS JLAND
4 1
LNAME METHOD
'Residential' 0
PNAME PUN IT
IROS
0
JACGUT
1
NDIM
IRQ SAD
0
DDLIM
1
KALC
DRYDAY CBVOL
5.00 3.00
DDPOW DDFACT
.OE04 1.0
KWASH KACGUT
DRYBSN
5.00
CLFREQ
1.0 7
LINKUP
RAINIT
0
AVSWP
.0
QFACT1
REFFDD KLNBGN KLNEND
000
DSLCL
1.0 0
QFACT2 QFACT3 QFACT4 QFACT5 WASHPO
CBFACT CONCRN REFF
'TSS' 'mg/L'
0 125
'Nitrate' 'mg/L
0 0.23
'BOD' 'mg/L'
0 15
'TotPh' 'ug/L'
1.0 0
TotZn 'ug/L'
0 65.8
KTO KFROM
2 1
TSS Nitrate
125 0.23
NAMEW KL
KC00372#2 1
0
0
0
0
0
0
1
0
1
0
Fl
0.02
BOD
15
BASINS
0
1
1
1
1
1
TZn
65.8
GQLEN
0
0 1
0 1
0 1
0 1
0 1
NDIM P(TSS)
0 0
0
0
0
0
0
P(Nit)
0
0
0
0
5
0
P(Cu)
0
0 0 0 0 2.0
0 0 0 0 2.0
0000 2.0
1 0.1 0 0 0.8
0000 2.0
P(Pb) P(Zn)
0 0
* Print Control
15-114
-------
Table 15-2 (concluded)
*
Ml
*
M2
*
M3
*
*M4
*
NPRNT
1
NDET
1
IPRNT
KC00372
MDEEP
0
INTERV
8
STARTP1 STOPPR1
20000412 20000413
KDEEP
0
*ENDPROGRAM
15-115
-------
The Transport Block was used to simulate the pond since it has the ability to simulate both storage and
first-order decay within the storage. Inflow and loads from Runoff occurred at an upstream "manhole" in
Transport. The pond was then simulated as a storage unit, draining to another "manhole" for tracking.
Bathymetry measurements were obtained from the construction drawings provided by the BBS, which
showed detailed contour lines (Figure 15-5). These were used to develop stage-area-volume-outflow data
that Transport uses for storage-indication flow routing. However, additional effort was needed to develop
the outflow rating curve, as described below.
Effluent from the pond exits through a 2.5-in. orifice, which is designed to drain to the orifice level in
about 24 hrs. When there is more water than can exit through the orifice, water spills over a broad-
crested weir located above the orifice (Figure 15-5). A rating curve was developed by summing the
theoretical orifice and weir equations,
1.5
(15-1)
305.0 MSL,
Q = Cd Ao [2g(h-h0)]1/2 + Cw Lw (h-hw)
where
Q = outflow (ft3/s),
Cd = orifice discharge coefficient, assumed ~0.9,
A0 = area of orifice = 0.0341 ft2 for a diameter of 2.5 in.,
g = gravitational acceleration = 32.2 ft/s2,
h = elevation of water in pond (ft), above pond bottom at elevation
h0 = elevation of orifice centerline = 306.0 ft,
Cw = broad crested weir coefficient (ft0 5/s), assumed -3.3,
Lw = length of weir = 4.0 ft,
hw = weir crest elevation = 307.7 ft.
Surface areas were planimetered and interpolated from the Pond 3 geometric plans, Figure 15-5, yielding
the information in Table 15-3 for input into the Transport Block storage element. It should be noted that
the outflow rating curve in the stilling well used by BBS for monitoring is useful for hydrograph
verification but not for modeling, since the stilling well is downstream of the orifice and weir used in the
model (Figure 15-5).
Table 15-3. Rating curve development for Pond 3 at Lexington Hills.
Depth, h,
ft
0.00
0.50
1.00
1.25
1.50
2.00
2.70
3.00
3.25
Area,
ft2
2124
3152
4180
4350
4844
5510
6350
6850
7200
Volume,
ft3
0
1319
3152
4218
5367
7956
12107
14087
15843
Q orifice,
cfs
0
0
0
0.12
0.17
0.25
0.32
0.35
0.37
Q weir,
cfs
0
0
0
0
0
0
0
2.17
5.38
Q total,
cfs
0
0
0
0.12
0.17
0.25
0.32
2.52
5.75
Orifice sill
Weir sill
15.3.2 Initial Modeling
Inflows from the five storm events modeled were compared with actual flow volumes recorded during the
sampling events. This is illustrated by comparing recorded and simulated inflows to rainfall, as shown in
Figure 15-10.
15-116
-------
12000 -i
10000
1
o
6000
2000
0.1 0.2 0.3 0.4 0.5 0.6
Rainfall (in.)
0.7
0.8
0.9
* Recorded data
ISWMM simulation
Figure 15-10. Comparison of recorded and simulated flows vs. rainfall for the five simulated events.
It is clear from Figure 15-10 that recorded inflows to the site vary only slightly with rainfall. This leads
to questions about the confidence and reliability of the measured data. SWMM simulations indicate a
steady increase of inflow with increased rainfall, which is what is expected. Because rainfall amounts
were exactly the same for the recorded and SWMM data, it was decided that only the closest two events
would be used for further modeling. These were event 1 (April 13, 2000) and event 4 (October 9-10,
2000).
15.3.3 Quantity Modeling Results
Quantity modeling results are summarized in Table 15-4, in which it may be seen that pond influent
(catchment runoff) and pond effluent measured and simulated volumes agree well. For the two
simulations of the pond, estimates of initial volume were made on the basis of the delay in the starting
time of the outflow hydrographs.
Table 15-4. SWMM simulated inflows and outflows versus measured data for Pond 3 at the Lexington Hills
BMP site.
Storm Date
April 13, 2000
October 9- 10, 2000
Initial Vol.
Estimate, ft3
2692
2629
Measured
Inflow, ft3
8275
8276
Simulated
Inflow, ft3
8058
8208
Measured
Outflow, ft3
8075
6815
Simulated
Outflow, ft3
7749
7339
15-117
-------
SWMM predicts slightly higher outflow for the October 2000 storm than was measured. However, the
actual measured outflow value was assumed to be low by the BBS sampling crew due to clogging of the
orifice from debris such as leaves, needles, etc. (Liptan 2001).
15.3.4 Quality Simulations
Modeled constituents were TSS, BOD, NO3-N, and total zinc, from a much longer list of samples for
which EMC values were available from BBS monitoring (Liptan 2001). No attempt was made to
calibrate Runoff Block nonpoint source water quality values. Instead, Runoff Block concentrations were
set equal to measured BBS influent concentrations (pond influent = subcatchment effluent) by setting
Runoff Block rainfall and groundwater concentrations to the measured influent EMC values. This was
done to avoid a time-consuming calibration process when the point of the study was to simulate the BMP,
not the watershed. Results are summarized in Table 15-5.
Table 15-5. Measured and simulated quality results for Lexington Hills Pond 3. Concentrations are EMCs.
Loads (Ib) are computed as product of EMC and measured or simulated volumes (Table 15-4). Pond influent is same
as catchment effluent.
TSS,
mg/L
TSS,
Ib
BOD,
mg/L
BOD,
Ib
N03-N,
mg/L
N03-N,
Ib
Tot. Zn,
ug/L
Tot. Zn,
Ib
Event 1, April 13, 2000
Influent, measured
Influent, simulated
Effluent, measured
Effluent, simulated
125
125
35
98.5*
64.5
64.4
17.6
47.6*
15
15
8
15
7.7
7.5
4.0
7.3
0.23
0.23
0.28
0.23
0.119
0.118
0.141
0.111
65.8
65.8
28.9
55.3*
0.034
0.034
0.0146
0.027*
Event 4, October 9-10, 2000
Influent, measured
Influent, simulated
Effluent, measured
Effluent, simulated
222
222
114
138*
115
114
48.5
63.5*
5
5
6
5
2.6
2.6
2.6
2.3
0.25
0.25
0.36
0.25
0.128
0.129
0.153
0.115
52.4
52.4
38.2
35.0*
0.0271
0.0268
0.0162
0.0160*
*See text regarding "decay" used for simulated TSS and total zinc in pond effluent.
Pond effluent EMCs for TSS and total zinc are less than influent EMCs for both storms. The pond would
not be expected to remove much, if any, nitrate; in fact, measured effluent nitrate values are higher than
influent values. The BES (Liptan 2001) offers no explanation for higher effluent nitrate values, but some
could be hypothesized, such as nitrification in standing water during interevent times. The quality of any
water standing below the orifice at the start of an event was not measured, so there is no way to determine
if the initial pool added mass to the effluent. Influent BOD values are higher than for the effluent for
event 1 and about the same for event 2. Although some BOD decay might be expected, the relatively
short detention time would not yield much change. Thus, for the simulations, both BOD and NO3-N were
treated as conservative constituents, and it may be seen that inflow concentration is the same as outflow
concentration in Table 15-4.
The reduction in TSS and total zinc is expected by sedimentation, of the solids themselves and of the
adsorbed zinc. The only decay in the Transport Block is by a first-order process (not including the
relatively untested July 2001 modifications to Transport that include a settling velocity). Hence, an
equivalent first-order decay coefficient was created from a settling velocity by (see also Equation 6-14),
= vs/h
(15-2)
15-118
-------
where
k
vs
h
= decay rate (s"1),
= settling velocity (ft/s), and
= average depth (ft).
The average depth of outflow was determined to be about 1.2 ft from review of the Transport output files.
A particle size distribution is provided by BBS for the October 9-10, 2000 and later events (not for the
April 13, 2000 event). These data are presented in Table 15-6, from which a weighted average diameter
can be computed of 25.2 jam, using range midpoints as representative diameters. From this a typical
diameter of 25 jam was used in Stokes' law (Equation 6-6).
Table 15-6. Particle size distribution for event of October 9-10,2000. Counts are in units of particles per 100
mL.
Size range,
(im
Total particles
5-15
15-25
25-50
50-100
>100
Range mid-
point, urn
10
20
37.5
75
Influent
Number
75100
32400
4500
36900
1300
n/d*
Effluent
Number
29000
26800
1600
500
100
n/d*
*n/d = not detected
For a measured temperature of 15°C, kinematic viscosity, u = 0.0115 cm2/s (Chapra 1997). Using g = 981
cm2/s and a diameter of 25 jam = 0.0025 cm, Equation 6-8 gives a settling velocity of 0.0444 cm/s =
0.00146 ft/s = 126 ft/day. Dividing by an average depth of 1.2 ft gives a first-order decay coefficient of
105 day"1. It may easily be recognized even without any simulation that any constituent with such a large
coefficient will be completely removed in just an hour or so, e.g., the half-life,
t,/2 = In2 / k = 0.0066 day = 0.16 hr
(15-3)
In fact, this value of 105 day"1 is in same order of magnitude as values derived in Table 7-4. But not all of
the Pond 3 TSS and total zinc is removed.
Another approach would be to use the SWMM S/T Block with a particle size distribution, but the particle
counts would have to be converted to concentrations. Logistical constraints prevented this approach (this
work was done too late in the project period). In the interest of expediency, overall removal percentages
(Liptan 2001) were used to compute approximate first-order decay coefficients for Pond 3. For the seven
monitored events, average TSS removal was 57%, average total zinc removal was 45%, average BOD
removal was 6%, and average nitrate removal was -11%. BOD and nitrate were treated as conservative,
especially since the model has no physically realistic mechanism for increasing the concentration of a
pollutant, other than starting with a higher concentration in the permanent pool (discussed earlier).
Assuming TSS and total zinc reduction occurs in t = 1 day, an equivalent first-order decay coefficient can
be computed from
C/C0 = 1 - removal fraction = e"kt
yielding a TSS decay coefficient of 0.84 day"1 and a total zinc decay coefficient of 0.60 day"
(15-4)
15-119
-------
The effluent TSS values predicted by SWMM are above the measured effluent EMCs for both events,
although closer for the October event. Simulated total zinc EMCs are higher for the April event and
slightly lower for the October event. Refinements could be made, but it would simply amount to a curve-
fitting exercise. Instead, the kind of results made possible by use of the somewhat maligned removal
fraction (Section 4.3) can be imagined. Constituents treated as conservative reflect this fact through
equality of influent and effluent EMCs in Table 15-5.
15.4 SUMMARY
It is difficult to obtain good data sets for testing of BMP simulation, including monitoring of influent and
effluent flows and concentrations. The Lexington Hills Pond 3 monitoring conducted by the BES comes
close, inasmuch as influent and effluent flows were monitored along with composite quality sampling.
However, the flow monitoring itself is somewhat suspicious on the basis of the lack of a relationship of
inflow to rainfall (Figure 15-10). A water quality sample from any standing water in the pond would
have been useful, as well as noting the depth of the pond at the start of a storm event. Liptan (2001) notes
a few other monitoring problems, such as issues of orifice clogging and rating curve generation.
Nonetheless, the data set has served a valuable function in evaluating SWMM pond simulation
capabilities.
Some conclusions follow:
Using SWMM Runoff Block catchment data supplied by the BES, seemingly reasonable simulations
of runoff hydrographs were obtained. However, the influent monitoring appears to give similar total
volumes for all events, which is unexplained. Quality simulations were conducted for the two events
for which simulated and measured runoff volumes most closely agreed (Figure 15-10).
SWMM simulation of the pond hydraulics worked well even while using an outflow rating curve
based on theoretical equations for the weir and orifice outlets. This is based on the comparison of
simulated and measured total outflow volumes (Table 15-4).
Approximation of TSS and total zinc losses in a SWMM Transport Block storage unit through an
equivalent first-order decay coefficient is crude and serves only to demonstrate that the measured
outflow EMCs could be replicated by the model. Better would be to attempt to simulate
sedimentation processes in the S/T Block using the limited particle size distribution information
provided, but time did not permit this approach.
15-120
-------
16 RECOMMENDATIONS FOR SWMM BMP MODELING
IMPROVEMENTS
16.1 INTRODUCTION
Evaluating the methods for model BMPs by process and type have much overlap. A breakdown of
applicable methods for modeling each BMP is given in Table 16-1, followed by recommendations by
type. While nearly all BMPs may be described using removal coefficients or with the effluent probability
method (EPM), these methods do not model processes, i.e., fundamental unit processes of environmental
engineering. Neither would they predict overloaded or undersized BMPs, or describe failing BMPs. This
makes these methods less useful for sizing and design of BMPs. Nonetheless, if the EPM is the best
representation of BMP performance for an otherwise hopelessly complicated set of processes, it would be
a useful option to include in SWMM. This would be implemented by enhancing the statistical analysis
capability, currently in the SWMM Statistics (Stat) Block, to provide comparative lognormal plots of
influent and effluent EMCs. In addition, an option for specifying a lognormal (or other distribution) for
influent and/or effluent EMCs is needed.
A somewhat general CFSTR formulation of fundamental source-sink processes that can be used to
represent BMP impacts was provided in Section 4.4, with regard to Equation 4-9, which is repeated here:
V+ C = (QICm+L )(1-R )-QC -kC V-vs(l-F)C As (16-1)
dt dt
where
C = constituent concentration,
V = volume of conveyance/storage object,
t = time,
Qi = inflow to object,
Cm = inflow concentration,
L = other loadings, e.g., from sediment or precipitation,
R = removal fraction due to BMP,
Q = outflow from object,
k = first-order decay rate,
vs = settling velocity,
F = fraction of constituent in soluble (non-settleable) form, and
A, = surface area.
16-121
-------
The source-sink terms of this equation can be readily adapted to a one-dimensional advective-dispersion
equation, if desired. Of course, the equation is solved immediately after the flow routing equation for the
corresponding conveyance/storage object. Flow routing accounts for all inflows and outflows, including
"vertical" processes of precipitation, ET, and infiltration. For CFSTR routing, these are reflected in the
dV/dt term. If Equation 16-1 were used for a particular settling velocity range, it would be well-suited for
treatment-train simulation as well, since particles with higher settling velocities would settle sooner,
upstream. The settling velocity formulation could easily be enhanced to something other than a constant
vs as well, as in the current SWMM S/T Block.
16.2 PONDS
SWMM currently does a good job of modeling storage devices based on hydraulic controls, including
settling and first-order decay. When searching for improvements to SWMM for pond simulation, one
remaining (minor) issue is biological treatment based on second-order reactions. The use of second-order
rate reactions used in P8 is novel to SWMM, but much less useful without a default set of reaction
constants. The effectiveness of macrophytes on pollutant removal might be simulated with the use of a
removal factor similar to the particle scale removal factor, f, in P8 (Section 6.2.3).
Although sedimentation in the SWMM S/T Block can be simulated with reasonable sophistication at the
moment, input is not phrased in common environmental engineering terms such as overflow rate. Use of
terminology that is more consistent with the profession might be helpful to the model user.
Environmental engineering texts and references often refer to the "tanks in series" model, i.e., a series of
N CFSTRs, Equation 6-11, as a useful transition between one extreme of plug flow to the other extreme
of complete mixing in a storage device (or chemical "reactor"). This could readily be implemented as an
option for treatment in a SWMM5 "object" that includes storage, such as a channel, pipe, or storage unit.
However, it will still require user judgment as to whether mixing is "good," "very good," etc., thus
retaining the empiricism that seems difficult to avoid (see discussion of Equation 6-11). On the other
hand, if tracer data are available, quantitative inferences about the degree of mixing can be made - see
discussion in Section 7.8.2 regarding the MUSIC pond algorithms. The MUSIC implementation of
Equation 6-11 provides qualitative guidance as to the number of CFSTRs that would be a good start for
any SWMM implementation.
SWMM is well equipped to deal with detention and extended detention storage from the standpoint of
timing issues. The relationship of storage availability and draw-down to local meteorology and
catchment conditions may be analyzed quantitatively via continuous simulation. The principal
enhancements for SWMM would be more useful statistics, e.g., analyze statistics of the time series of
volume (or depth), in addition to inflows and outflows from storage. This only requires the ability to
create additional time series of model state variables.
The hydrology of most extended detention ponds can be modeled in the same way as for wet ponds.
However, a missing hydrologic component for both extended detention and for ponds in SWMM is
infiltration and evaporation (at better than monthly averages). The vertical water balance may only be
simulated crudely, through the input of "negative rainfall."
16-122
-------
Table 16-1. Clarification of method applicability to modeling BMPs.
Bioretention
Simulation Practices
Effluent probability method
Removal coefficient
Second-order decay
Particle scale removal factor
Tanks (CFSTRs) in series
Fractional volume reduction*
Substrate modeling**
Ecological and nutrient cycling modeling
Ponds
X
X
X
X
X
X
X
Facilities
X
X
X
X
X
X
Infiltration Grassed Dry Porous Hydrodynamic
Trenches Swales Wells Cisterns Pavement
X
X X
X
X
X XXX X
Devices
X
X
*As implemented in SLAMM, e.g., Equation 8-1.
**As implemented in VAFSWM.
16-123
-------
16.3 WETLANDS AND BIORETENTION FACILITIES
The particle scale removal factor in P8 (usually set to 1) is applied as a calibration parameter for devices.
This can be increased to represent better treatment and removal in bio-filtration facilities. Increased
removal by macrophytes is documented in the P8 manual wherein the removal factor is set to 2 or 3 based
on improved removal rates. But modeling multiple processes can lead to conflicts and difficulties with
sequence of treatment. Moreover, the particle scale removal factor is akin to heuristic removal
efficiencies and is difficult to represent as a fundamental unit process.
A more conventional alternative within the unit process literature is the tanks-in-series model (Equation
6-11), discussed just above and in Section 6.5 with regard to ponds and Section 7.8 with regard to
wetlands. Parameter N, the number of CFSTRs in series, may be estimated somewhat quantitatively from
tracer studies (Kadlec and Knight 1996, Persson et al. 1999), or qualitatively on the basis of an estimate
of short-circuiting and dead zones (Fair et al. 1968, Pitt and Voorhees 2000, Wong et al. 2001, Minton
2002). Kadlec and Knight's (1996) k'-C* modification of exponential decay used in MUSIC (Section
7.8) is a useful means by which to provide a lower bound (C*) from a CFSTR with first-order decay. The
lower bound could also be in the form of a frequency distribution of effluent EMCs.
Certainly wetlands, bioswales, and similar storage areas where there is strong interaction between water
quality and biological processes within the area (e.g., vegetation, sediment) are one of the most difficult
challenges for BMP simulation. Three models that perform this kind of simulation (WETLAND,
PREWET, DMSTA) include complex kinetic formulations and several state variables to represent
nutrient cycling. Whether or not this effort is needed or can be supported for SWMM is to be determined.
The VAFSWM simplifies the process by simulating only a "substrate" (in addition to the constituent in
the water column) consisting of water in sediment and vegetation, and without the linked nutrient
kinetics. This concept would be easier to implement in SWMM. A useful enhancement regardless of
whether physical and biological process options are updated to better represent wetlands is the ability to
link SWMM time series output to other models. For instance, wetland quality processes may be
simulated for the most part with the EPA WASP model (Wool et al. 2001), with HSPF being another
option (Bicknell et al. 1997). A standard for protocol interfacing is required to facilitate such linkages.
That is, SWMM simply may not be the best choice for a "universal treatment model" when it comes to
simulation of natural receiving water processes.
16.4 INFILTRATION TRENCHES
One of the biggest needs for SWMM with regard to infiltration trenches and swales (below) is the ability
to simulate infiltration from channels as well as for overland flow planes. All conveyance and storage
options within the model need to include a full vertical water balance that includes precipitation, ET, and
infiltration. Within this framework, SWMM conveyance and storage elements should be able to simulate
quality processes characteristic of infiltration devices, bio-swales, etc. Notwithstanding the need to be
able to infiltrate from channels, infiltration trenches can be simulated with the current SWMM as small
overland flow planes, with the ability to receive runoff from an upstream overland flow plane.
Depression storage is used to define the depth of the trench, as described in Section 8.3.
Other models perform this function using highly empirical, though often useful, methods. For instance,
the methods described for SLAMM produce a fractional volume reduction based on adjusted watershed
runoff coefficients that reflect smaller storms, and adjusted percolation rates that account for soil moisture
and compaction. Unfortunately, infiltration rates themselves must be known a priori; however,
SLAMM's reference base is helpful in this regard, e.g., for disturbed soils.
16-124
-------
16.5 GRASS SWALES
Regarding grass swales, filter strips, and similar BMPs that feature flow through vegetation, the same
needs for a vertical water balance in conveyance and storage devices just discussed apply here. However,
to the extent that flow through swales behaves like a sedimentation device for which performance is a
function of hydraulic overflow rate (Minton 2002), swales can be simulated in the current SWMM S/T
Block. But a ready means of simulating infiltration from S/T storage is still needed. The simplified
infiltration procedures in SLAMM are applicable to grass swales (see infiltration trenches section), but
these procedures do not seem to offer many SWMM enhancement opportunities. P8's assumption of
adsorption of some dissolved constituents while traveling through swales is an implementation possibility
for SWMM, easily simulated through the S/T removal equation.
REMM closely models the N, P, and C cycles in three surface buffer zones (linked to three subsurface
layers), which may be applicable to SWMM, although this will require extensive field data to simulate.
The most useful information from the REMM model when good documentation becomes available may
be the default rate constants used in calculations.
The REMM techniques are certainly applicable in vegetated filter strips, but the REMM methods are
complex in the same way as the nutrient cycling simulations are complex in the wetlands models.
Moreover, REMM includes a linked surface-subsurface hydrologic model for the riparian zone. This
quasi-2D model (three horizontal by three vertical zones) is beyond the capability of SWMM at the
moment and likely not worth the effort of implementation given SWMM's common stormwater design
applications. Nonetheless, similarly to the wetlands models, there is good opportunity for parameter
estimation help from REMM publications, and such a model could be used to calibrate simpler techniques
that might be added to SWMM.
What the literature review of this study has not yet uncovered - which is not to say that they are not
available - are functional relationships between performance and design parameters for filter strips,
although this kind of information does exist for swales (Fletcher et al. 2002). For instance, one expects
removal to be proportional to flow path length, type of vegetation, type of soil, etc. (Huber et al. 2000).
While SWMM can now simulate load removal associated with infiltration, concentrations are not changed
through sedimentation or biological processes in the Runoff Block overland flow routines, although they
may be changed using a constant settling velocity, first-order decay rate, or removal fraction. When
empirical or theoretical relationships are found with causative parameters - e.g., length, vegetation, soil,
slope, and maintenance - they should probably be implemented in SWMM.
16.6 DRY WELLS
Modeling dry wells can involve complex analytical techniques for both quantity and quality for
simulation of the groundwater flow regime. Under suitable soil circumstances, though, it may be possible
to infiltrate the entire design inflow over the approximate storm duration. SWMM is currently capable of
accepting flows into a device that represents a dry well but not capable of simulating the complex flow
net that results as water enters the soil. This latter effort is probably not necessary to simulate dry wells
as BMPs as long as their capacity can be provided as a function of time or other simple relationship (e.g.,
a rating curve). If current SWMM infiltration routines are used (from subcatchments), a linkage to
groundwater routines can be made to follow the pathway of water further, should that be desired.
16.7 CISTERNS
For cisterns what is needed in SWMM and similar models is an overall water use simulation capability
that includes all sources (e.g., cisterns, city water supply) and an irrigation schedule (as in SLAMM) in
order to dispose of runoff collected in cisterns. Water stored in cisterns can also be used for gray water
16-125
-------
supply in buildings and boiler feed water, both of which might require some on-site treatment. The
irrigation schedule might in turn be linked to a soil moisture accounting model, leading to greater
complexity. Otherwise, SWMM is capable now of simulating the storage associated with cisterns -just
not the timed release!
16.8 POROUS PAVEMENT
SLAMM models porous pavement with infiltration rates and rainfall events manipulated to focus on the
micro-storm effect. SWMM currently can simulate porous pavement by using its Horton or Green-Ampt
infiltration equations, and the resulting infiltrated water can be tracked through use of the groundwater
routines, except for water quality. James et al. (2001) demonstrate how the current version of SWMM
can be used to simulate roadways and parking lots with porous pavement or permeable pavers.
16.9 HYDRODYNAMIC DEVICES
Because of the variability between devices in this category, the effects may best be simulated with
removal equations based on the manufacturer's specifications, or with data from the EPA/ASCE BMP
Database. On the other hand, removal as a function of overflow rate may work just as well. Checks need
to be established to determine if devices are overloaded or for maximum treatment values. Extran is
capable of simulating the hydraulics of some common devices through proper use of orifice settings in
particular, but water quality is not tracked.
16.10 LID AND OTHER RELATED NEEDS
Almost all BMP design is linked to a description of the influent stormwater. A characterization of solids
is especially important, i.e., treatability data on settling velocities and/or particle size and specific gravity
distributions. The SWMM S/T Block requires this information to perform sedimentation computations,
but this is entered only for the S/T Block and as a constant for the inflowing stormwater or combined
sewage. Upstream blocks should supply this time series - a very complicated process since it relates to
erosion, scour, deposition, and sediment transport, all of which are poorly represented in SWMM and in
most any alternative model. Nevertheless, typical particle size distributions are available from several
sources (e.g., Minton 2002) that can be provided as a default, in lieu of site-specific data. Within the rest
of SWMM, particle sizes can be tracked by treating a size range as a separate constituent subject to a
constant settling velocity. This is not too bad an assumption in the current Runoff and Transport Block
routines for which settling in essence can be simulated as a function of residence time, V/Q. Even
without improved scour and deposition routines, simply linking the constituents grouped by particle size
(or settling velocity) range to the S/T Block would be a huge enhancement to the overall SWMM
simulation capability. If SWMM5 can perform this linkage, then it will immediately provide an improved
BMP simulation capability.
Several of the models discussed in this report include heuristic parameters (e.g., efficiency ratio, particle
scale removal factor). While these parameters do not represent fundamental unit processes and are
difficult to evaluate a priori, if they gain favor in the profession, SWMM might include them as options
for simulation of BMPs. The N tanks-in-series model (Equation 6-11) would be relatively easy to
implement as a treatment option within any device that incorporates storage, including any conveyance
element.
Regarding the suitability of LID simulation, the ability to route flows from one overland flow plane to
another is useful (e.g., see Section 4.7) but insufficient. Infiltration and ET simulation are also required
for open channels and for ponds and storage devices. The vertical water balance is critical in infiltration
devices such as swales as well as in wetlands. SWMM currently has only limited capabilities in these
areas. Another useful LID simulation procedure would be to provide a way in which stored water could
16-126
-------
be distributed seasonally, as from a cistern. An irrigation schedule would be useful here, and coupling
with other components of the urban water cycle (e.g., water supply, wastewater removal) should be
provided.
SWMM does not simulate water quality in the subsurface zone, i.e., in the Runoff Block groundwater
routines. Examples of models with this capability include HSPF and REMM. This would require a major
renovation of model capabilities, but it would be useful for simulatinf all water budget components in an
area subject to infiltration. However, alternative models, e.g., HSPF and REMM, already provide this
capability.
Another useful addition to SWMM would be the ability to prescribe a lognormal (or other) frequency
distribution to represent EMCs in surface runoff and/or effluent quality from BMPs, e.g., as characterized
by the effluent probability method. The modeling challenge would be in part to ensure conservation of
constituent mass if EMCs and hence loads are inserted into the effluent (or nonpoint source runoff) from a
prescribed frequency distribution.
Finally, SWMM is unlikely ever to perform all the functions that a user might need. Particularly for
complex nutrient dynamics and subsurface water quality, alternative models such as WASP, HSPF,
REMM or WETLAND should be considered. In order to interface one model with another, standards for
time series transfer are needed. Such standards would facilitate many multi-media modeling needs within
the environmental community in general and Environmental Protection Agency in particular.
16.11 FINAL SUMMARY OF SWMM BMP SIMULATION NEEDS
For the convenience of the reader, proposed BMP/LID simulation enhancements are summarized in Table
16-2. A summary commentary is provided in Chapter 17. The priority assigned in the table is purely the
opinion of the authors of this report. In their opinion, probably the highest priority enhancement should
be item 10 of the table: inclusion of the vertical water balance (precipitation, ET, infiltration) for all flow
objects, especially channels. This is urgently needed for proper simulation of any porous conveyance or
storage, such as bioswales and wetlands.
16-127
-------
Table 16-2. Summary of proposed SWMM BMP/LID
order of discussion in Chapter 16. Priority: H = high, M
simulation enhancements. These are listed generally in
medium, L = low.
1
2
3
4
5
6
7
8
9
10
Simulation proposal
Emphasize fundamental unit processes.
Provide for automated evaluation of inflow
and outflow EMCs by effluent probability
method.
Provide for lognormal and other frequency
distributions of watershed runoff EMCs and
BMP effluent EMCs.
Provide a generalized source-sink
formulation of the type given in Equation
16-1 for conveyance and storage objects.
Second-order decay is useful for some
constituents.
Particle removal scale factors as used in P8
can help with calibration.
The "tanks in series" model provides a
useful bridge between simulation of
conveyance and storage objects as well-
mixed (worst performance) and plug flow
(best performance).
A lower bound on exponential decay
through a first-order process in a CFSTR
can be provided with the k'-C* model.
Upgrade SWMM statistical analysis
options.
The "vertical" water budget must be
available for all flow objects, including
overland flow on subcatchments,
conveyance channels, and storage objects.
This implies inclusion of precipitation, ET,
and infiltration for all flow objects.
Comment
Most flexible regarding "real world"
settings.
Will facilitate comparisons with published
evaluations using this method, often
recommended for BMP performance
evaluation.
BMP influent and effluent EMCs are often
characterized by a lognormal distribution
and performance may sometimes best be
characterized simply by the effluent
distribution, rather than, for example, a
removal fraction.
This formulation provides for first-order
decay, settling, and generalized "removal."
A significant problem with implementation
is lack of data.
These scale factors amount to a linear
increase in settling and first-order decay
terms. This effect can also be provided for
in other ways.
Adapting this formulation, often used in
environmental engineering practice and
implemented in the MUSIC model, might
be one of the most useful enhancements to
SWMM.
The lower bound, C*, could be specified in
a numerical solution. The lower bound
could also be in the form of a frequency
distribution.
Related to item 1, graphical, regression,
and other statistical evaluation options
would make a model such as SWMM5
even more powerful as an analysis tool.
Implementation of this vertical water
budget is essential for simulation of
conveyances such as swales, bioswales,
and wetlands. Load reduction associated
with infiltration, for instance, cannot be
ignored.
Priority
M
M
M
H
L
L
H
M
H
H
16-128
-------
Table 16-2. (Continued)
11
A ready method with which to
interface SWMM to other models is
urgently needed, including a
standard format for time series files.
This will facilitate linkage of SWMM to models better
suited to certain kinds of receiving water and subsurface
water analysis, such as WASP and WETLAND. SWMM
cannot "do everything," and the ability to link SWMM's
watershed processes with the receiving water simulation
strengths of other models will facilitate use of the right
model for the right task.
H
12
Settling velocity ranges should be
tracked through the runoff-
transport-treatment pathways.
This will facilitate proper simulation of "treatment
trains."
M
13
Related to item 12, even though
stormwater treatability data are
relatively uncommon, SWMM
should be adapted to use such
information when available.
The model must be able to track particle ranges based on
settling velocities (or size - specific gravity ranges)
throughout the whole model: sources, transport, and
treatment.
H
14
Improved mechanisms for
simulation of sediment, including
scour and deposition, will facilitate
item 13.
This is one of the most difficult set of physical processes
to simulate in the context of urban hydrology.
M
15
Although dry wells can be
simulated as an outflow or storage
within SWMM, a link to
groundwater routines would provide
additional continuity.
Water leaving the surface through dry wells could be
tracked similarly to infiltration from subcatchments.
16
Storage of water in cisterns can be
simulated with the addition of a
water distribution (water use)
algorithm or input table.
Stored water can be used for irrigation, gray water, boiler
feed water, etc. A schedule is needed for such
distributions.
17
Data and parameters from other
models should be adapted for use in
SWMM to the extent possible.
Other models reviewed in this report have different ways
of conceptualizing processes such as infiltration and
porous pavement, but parameter values from such models
may be very useful in SWMM simulations. The effect of
disturbed soils on infiltration parameters is particularly
important.
M
18
Additional information is needed on
hydrodynamic devices to determine
how well they can be simulated
using fundamental processes.
Simpler removal fractions might be
used in the meantime.
Proprietary devices are especially difficult to evaluate
simply on the basis of manufacturers' information.
M
19
Simulation of subsurface water
quality would be useful for
continuity of constituent loads as
well as flow.
But this can be very difficult, with the need to account
for sorption, etc. Linkage to groundwater quality models
is another option.
16-129
-------
17 CONCLUSIONS
The EPA Storm Water Management Model was developed in 1968-71 to simulate combined sewer
systems, including evaluation of management strategies. The user community immediately, and logically,
applied the model to separate stormwater systems as well as combined sewer systems and eventually into
the entire spectrum of urban and non-urban watershed response. All of this was facilitated by EPA
support for continuous enhancement and development of the model in the intervening years. The
Storage/Treatment Block within the original model included algorithms that allowed it to mimic removal
for a few specified pollutants and for an array of CSO controls available at that time; these algorithms
were generalized and enhanced to permit broader simulation of series-parallel arrangements of S/T units
beginning with SWMM III (Nix et al. 1978). In 2004, municipalities and their consultants are
increasingly under pressure to implement BMPs for control of both stormwater quantity and quality. As
new information and data are collected about the performance of BMPs, SWMM and similar models must
be updated further to reflect this new technology. LID (hydrologic source control) offers yet another
opportunity for stormwater management through distributed and localized control options, with emphasis
upon infiltration, evapotranspiration, and water reuse. This report evaluates a number of alternatives for
simulation of conventional BMP and LID options, with emphasis upon incorporation of fundamental unit
processes into the algorithms wherever possible. Several ways in which SWMM can successfully
accomplish these simulation goals are summarized below, along with several more areas in which
improvements are needed.
When modeling BMPs, the ability of SWMM to redirect flow from impervious areas to pervious
(simulating the effects of downspout disconnection) greatly enhances the model's ability to simulate LID.
This has been shown in the literature (Lee 2003) and is demonstrated in an extensive case study for
Portland (Chapter 14). However, for the model to be able to characterize treatment based on solids
settling, improvements in the overall ability of SWMM to erode, transport, deposit, and scour sediment
(i.e., to incorporate scarce treatability data) need to be provided, as well as the ability to simulate
infiltration in channels and ponds. The vertical water balance must be computed for all conveyances and
storages, not just for overland flow planes.
The Runoff Block is very stable with regard to size of subcatchments that are simulated. SWMM can
model very small parcels (and consequently BMPs and LID facilities on an individual lot scale or
smaller). An immediate consequence of simulation of areas on the order of a fraction of an acre is the
need for a smaller Runoff Block (or SWMM5 subcatchment object) time step (e.g., < 1 minute),
compared to the more typical 5-minute value routinely employed. For a process model like SWMM, the
issue of temporal and spatial variability is largely an issue of data availability and the amount of detail
17-130
-------
desired in the simulation run. For long-term (continuous) simulations, aggregated subcatchment
schematizations can work just about as well as highly discretized schematizations, as demonstrated in the
Portland case study presented in Chapter 14. However, this advantage becomes less relevant as computer
speeds continue to increase.
Regarding BMP simulation, the S/T Block has the most capabilities for fundamental unit process
simulation but is unsophisticated hydraulically and hydrologically. Insertion of fundamental processes
into any conveyance or storage "object" in SWMM5 should make the model much more useful,
especially if quality routing can be performed when dynamic routing (as in the current Extran) is being
used for pipes and channels. The current Transport Block provides simpler but somewhat flexible
mechanisms for simulation of storage BMPs, as demonstrated for a Portland detention pond in Chapter
15. The Transport Block also permits simulation of decay, settling, and removal within any conveyance
or storage element, as described in Section 4.4. The general form of source-sink terms used in this
formulation (Equation 4-9) is suitable for SWMM5 conveyance and storage objects.
When considering the options for modeling BMP performance that can be integrated into SWMM, the
most applicable improvements would be infiltration from swales and trenches, storage and reuse in
cisterns, storage and infiltration in dry wells, and inclusion of infiltration from porous pavements
(although current SWMM infiltration and groundwater routines can be made to simulate porous pavement
satisfactorily - Section 12.3). The SLAMM documentation has an extensive review of case studies and
many default parameters, which may be useful when modeling these BMPs. P8's use of the particle scale
removal factor may be a simple method for calibrating BMPs to actual performance data. The tanks-in-
series model from environmental engineering practice appears to supply a similar qualitative parameter in
N, the number of CFSTRs. REMM has the potential for useful algorithms and default parameters when
modeling nutrient cycling in buffers. However, models such as REMM, WETLAND and DMSTA likely
contain more complex nutrient cycling descriptions than need to be supported in SWMM and for which
parameter estimates would be even more difficult than they are now. In this case, linkage to
"downstream" receiving water models or groundwater models should be considered. Standards for
transfer of time series from one model to another will facilitate such linkages and are urgently needed.
Modeling multiple BMPs in a "treatment train" (several BMPs in series) is difficult unless solids removal
is represented through fundamental sedimentation processes - and separate solids "ranges" (i.e.,
characterized by settling velocity or else particle size and specific gravity) are carried from one part of the
model to another, e.g., from the current Runoff to the Transport to the S/T Block. Difficulties with using
removal rates in series can easily lead to over prediction of removal (if the first BMP removes 90% TSS,
it is unlikely that the second BMP in series will remove 90% of the remaining suspended solids). Often,
instead, the output may be relatively constant compared to influent concentrations. Some BMPs may
increase effluent concentration (such as BOD or nitrate in biological systems) if influent concentrations
are low. None of these were considered in the models surveyed in this report. Nor would the current
SWMM be able to simulate this effect in other than "work-arounds." One option to handle this effect
would be to input a frequency distribution of effluent EMCs.
Another issue is the malfunctioning or maintenance issues associated with BMPs. Reduced removal
efficiencies are likely without regular maintenance, and effects of clogging or other reduction in proper
performance should be included in any modeling procedure.
The United States Environmental Protection Agency can be proud of the current state of stormwater
modeling using SWMM. Of the models surveyed in this study, SWMM has the most extensive and
versatile capabilities for simulation of BMPs. Implementation of SWMM5, the "next generation" version
of SWMM, should enhance the model's overall status for use by practitioners in stormwater and wet-
weather flow management.
17-131
-------
APPENDIX: USING REMM TO PREDICT RIPARIAN BUFFER
PERFORMANCE
RIPARIAN ECOSYSTEM MANAGEMENT MODEL (REMM)
This appendix is intended to supplement the information on the Riparian Ecosystem Management Model
(REMM) provided in Chapter 9. Source material is taken primarily from USDA-ARS (1999) and
Inamdaretal. (1998a,b).
REMM has been developed for natural resource agencies and researchers as a tool that can help quantify
the water quality benefits of riparian buffers. REMM simulates: (a) the movement of surface and
subsurface water; (b) sediment transport and deposition; (c) transport, sequestration, and cycling of
nutrients; and (d) vegetative growth.
When looking at the processes REMM is capable of modeling and applying those processes to all
applicable BMPs, the strengths of the REMM model are the ability to deal with movement of subsurface
water and fate and transport of nutrients, neither of which are currently available in SWMM. The
applicable wet-weather control (WWC) and program suitability are listed in Table A-l.
REMM can be applied to:
Quantify nitrogen and phosphorus trapping in riparian buffer zones and determine buffer width for a
given set of riparian conditions and upland loadings.
Determine buffer effectiveness under increased loads.
Evaluate influence of vegetation type on buffer effectiveness.
Determine impacts of harvesting on buffer effectiveness.
Investigate long term fate of nutrients in riparian zones, sequestration in vegetation, or loss to
atmosphere (denitrification in case of N) investigate N / P saturation in riparian buffers.
A-132
-------
Table A-l. REMM wet-weather controls program suitability.
WWC Option
Source Controls
Overland Flow, Swales,
Infiltration, Porous
Pavement
Major Benefits
Major Drawbacks
REMM
Modeled as reduced input from adjacent lands.
Useful for grass swales. Infiltration simulated with a modified Green-Ampt
equation, vertical unsaturated conductivity with Campbell's equation. Simulates
surface and subsurface water, sediment transport and deposition, transport,
sequestration and cycling of nutrients. Capable of modeling hydrology budget
with losses to seepage, transpiration and interception. Porous pavement is not
modeled.
Detailed analysis on nutrient cycling in buffer strips including grass filter strips.
Can determine effects of vegetation type on buffer effectiveness.
While REMM does model complex nutrient cycling for grass and forested
buffers, it is most applicable to rural areas.
BUFFER HYDROLOGY
The riparian system is characterized in the model as consisting of three zones parallel to the stream and
representing increasing levels of management away from the stream. These zones include a narrow,
undisturbed forest area adjacent to the stream for protecting the stream bank and aquatic environment, an
area with managed woody vegetation for sequestering sediment and nutrients from upland runoff, and a
grass strip for dispersal of incoming upland surface runoff and sediment and nutrient deposition.
The soil is characterized in three layers through which vertical and lateral movement of water and
associated dissolved nutrients are simulated. Water movement and storage are characterized by processes
of interception, evapotranspiration, infiltration, vertical drainage, surface runoff, subsurface lateral flow,
upward flux from the water table in response to evapotranspiration losses, and return flow. These
processes are simulated for each of the three zones. The storage and movement of water between the
zones is based on a combination of mass balance and rate controlled approaches.
Each of these processes is simulated on a daily basis and described briefly in the following paragraphs.
For a more complete description of the processes and the equations used, the reader is referred to Inamdar
etal. (1999a).
Canopy interception is an exponential function of the canopy storage capacity and the amount of daily
rainfall and is simulated using a modified form of the Thomas and Beasley (1986) equation. Potential
rates of leaf evaporation and transpiration are both computed using a modified form of the Penman
Monteith equation (Running and Coughlan 1988). Unsaturated soil hydraulic conductivity is described
by Campbell's (1974) equation. Soil evaporation is computed in two stages (Gardner and Hillel 1962).
Infiltration in the model is simulated using a modified form of the Green-Ampt equation (Stone et al.
1994). Surface runoff entering the riparian area is routed downslope using a simplified procedure based
on the depth of runoff and flow velocity.
Evapotranspiration flux is determined using the Darcy Buckingham equation as described by Skaggs
(1978). Actual transpiration loss is limited by the availability of moisture in the soil and competition
among the roots of the various plant types present. As the soil dries, water extraction depends on both the
root distribution and the rate at which water can move to the roots. The maximum rate of water uptake
A-133
-------
from a layer is limited by its soil hydraulic conductivity. Vertical unsaturated conductivity is simulated as
a function of the soil moisture content using Campbell's equation (Campbell 1974). This allows any
excess demand that is not realized by a layer to be transferred to the layer below.
Subsurface lateral movement is assumed to occur when a water table builds up over the restricting soil
layer. The lateral movement of the water is simulated using Darcy's equation. In the model, saturated
lateral soil conductivity is assumed the same as vertical saturated conductivity. Down-slope subsurface
flow between the component zones is driven by the gradient of the water table. The potential hydraulic
gradient that determines the subsurface movement from zone 1 to the stream is assumed equal to the
smaller of the surface slope of zone 1 and the gradient from the water table elevation from the mid point
of zone 1 to the stream thalweg. Stream thalweg is a user-defined input.
Vegetation and associated litter material provide physical barriers to water and sediment transport over
the ground surface. Deposition of organic matter by plants provides a substrate supporting important
biological transformations of chemicals in the soil. Plants also sequester nutrients such as nitrogen and
phosphorus that contribute to water pollution. The zone immediately adjacent to the stream helps to
protect the stream bank and aquatic habitat.
The litter layer is important as the locus for the mixing of surface water with the soil surface. This mixing
process results in an equilibrium of dissolved and adsorbed chemical concentrations, which determines
amounts of chemicals that are subsequently leached, deposited on the ground surface, or carried along in
surface runoff. Concentrations of dissolved and adsorbed chemicals are recalculated as water moves
through each of the other soil layers.
Climate parameters required are the following: rainfall amount and duration, solar radiation, maximum
and minimum air temperatures, dew point temperatures. If actual measured data are not available the
model uses a subroutine, CLIGEN, to generate climate data. The model operates on a daily time step.
EROSION AND SEDIMENT
Erosion and sediment is calculated separately for each of the three riparian zones. The Universal Soil
Loss Equation (USLE) is used to predict erosion for each storm event (Wischmeier and Smith 1978).
Parameterization of the USLE for forested area is accomplished using guidelines presented by Dissmeyer
and Foster (1984). The method used for sediment routing uses equations developed by Foster et al.
(1981) and Lane (1982) and is as applied in the AGNPS model (Young et al. 1989). The effective
transport capacity is computed using a modification of the Bagnold stream power equations (Bagnold
1966). A detailed presentation is made in Young et al. (1989).
NUTRIENT DYNAMICS
REMM is also capable of modeling nutrient dynamics. Simulation of the carbon dynamics is based on
the Century Model (Parton et al. 1984, Inamdar et al. 1999b). Stoichiometric relationships are assumed
among C, N, and P in organic matter. N and P are released and immobilized in proportion to
transformations of C. The decomposition rates of the organic matter pools area calculated according to
first-order rate equations modified by temperature, moisture and C:N and C:P ratios (Inamdar et al.
1999b).
Nitrification is calculated with a first-order rate equation modified by temperature, moisture and pH. Rate
coefficients are determined following the approach of Reuss and Innis (1977) and Godwin and Jones
(1991) based upon a Michaelis-Menten (Monod) function.
A-134
-------
Phosphorous is simulated using the EIPC model (Jones et al. 1984) with two pools of inorganic P
unavailable to plant uptake, and a labile form that may be dissolved or adsorbed according to a
partitioning coefficient (Williams et al. 1984).
REMM rate constants may be a good source of K values for process modeling. Sources for upland data
may be from site monitoring or from the use of an upland model such as GLEAMS (Knisel et al. 1993).
Output from REMM includes predicted sediment yields, depth to groundwater, and predicted C-N-P
distribution throughout the system in many forms. BMP effectiveness may be evaluated on the basis of
comparison of incoming and outgoing loads.
RECOMMENDATIONS
The REMM model is only model reviewed that integrates subsurface flow and groundwater interaction
when simulating buffers. REMM also closely models the N, P, and C cycles in the three buffer zones,
although this requires extensive field data to simulate. The most useful information from the REMM
model may be the default rate constants used in these calculations. Its overall structure is likely beyond
what is required in SWMM.
A-135
-------
REFERENCES
Adams, B.J. and Papa, F. (2000). Urban Stormwater Management Planning with Analytical Probabilistic
Models, John Wiley and Sons, NY.
Adderley, V.C. and Mandilag, A.M. (2000a). "Detailed Design Modeling for large Scale CSO Facilities -
Facing our Fears and Simplifying Assumptions," Bureau of Environmental Services, Portland, OR.
Adderley, V.C. and A.M. Mandilag, A.M. (2000b). "Decisions," Water Environment and Technology, pp.
35-44, December.
American Society of Civil Engineers (1996). Hydrology Handbook, Second Edition, ASCE Manuals and
Reports on Engineering Practice No. 28, Reston, VA.
Altier, L.S., Lowrance, RR, Williams, R.G., Inamdar, S.P., Bosch, D.D., Sheridan, J.M., Hubbard , K.K.
and Thomas, D.L. (Undated). Riparian Ecosystem Management Model (REMM): Documentation. USDA
Conservation Research Report, Tifton, GA.
Bagnold, R.A. (1966). An Approach to the Sediment Transport Problem from General Physics. Prof.
Paper 422-J. U.S. Geological Survey. Reston, VA.
Balades, J.D., Bourgogne, P. and Gourioux, (1995a). "A Reservoir Structure Pavement for the Urban
Storm Drainage of a Supermarket Area of 120.000 m2 in Bordeaux." (in French) International
Conference NOVATECH95, Lyon, pp. 329- 336.
Balades, J.D., Legret, M. and H. Madiec, H. (1995b). "Porous Pavements as Tools to Control the
Pollutant Loads on Urban Streets." (in French) International Conference NOVATECH 95, Lyon, pp. 345-
352.
Barrett, M.E., Walsh, P.M., Malina, J.F. and Charbeneau, Jr. R.J. (1997). Performance of Vegetative
Controls for Treating Highway Runoff, University of Texas, Department of Civil Engineering, Austin,
TX.
Bedient, P.B. and Huber, W.C. (2002). Hydrology andFloodplain Analysis, Prentice-Hall, Upper Saddle
River, NJ, Third Edition.
R-136
-------
Benjamin, J.R. and Cornell, C.A. (1970). Probability, Statistics, and Decision for Civil Engineers,
McGraw-Hill Book Co., NY.
Benson, M.A. (1965). "Spurious Correlation in Hydraulics and Hydrology," Proc. American Society of
Civil Engineers, Journal of the Hydraulics Division, 91(HY4):35-42.
Bicknell, B.R., Imhoff, J.C., Kittle, J.L., Jr., Donigian, A.S., Jr. and Johanson, R.C. (1997). Hydrologic
Simulation Program - Fortran (HSPF): User's Manual for Release 11, USEPA, Office of Research and
Development, Athens, GA. (Available from: http://water.usgs.gov/software/hspf.html)
Bosch, D.D., Sheridan, J.M. and Lowrance, R.R. (1996). "Hydraulic Gradients and Flow Rates of a
Shallow Coastal Plain Aquifer in a Forested Riparian Buffer." Trans, of the ASAE, 39(3): 865-871.
Brown, A. (2003). Development of a BMP Evaluation Methodology for Highway Applications. Master of
Science Project Report, Dept. of Civil, Construction, and Environmental Engineering, Oregon State
University, Corvallis.
Brown, W. and Schueler, T. (1997). National Pollutant Removal Performance Database for Stormwater
BMPs. Center for Watershed Protection. Chesapeake Research Consortium, Ellicott City, MD.
Burke, C.B. (1979). The Comparative Performance of Several Methods for Storm Sewer Design. M.S.
thesis, School of Civil Engineering, Purdue University, West Lafayette, IN.
Burke, C.B. and Burke, T.T. (1994). HERPICC Stormwater Drainage Manual. Highway Extension and
Research Project for Indiana Counties and Cities (HERPICC), Purdue Research Foundation, West
Lafayette, IN.
Burke, C.B. and Gray, D.D. (1979). A Comparative Application of Several Methods for the Design of
Storm Sewers. Purdue University Water Resources Research Center Technical Report No. 118, West
Lafayette, IN.
Burton, G.A., Jr. and Pitt, R.E. (2001). Stormwater Effects Handbook: A Toolbox for Watershed
Managers, Scientists, and Engineers. Lewis Publishers, Boca Raton, FL.
CalTrans (2003). Stormwater BMP Handbook-New Development and Redevelopment, Appendix B:
General Applicability of Effluent Probability Method. January. Available at
http://www.cabmphandbooks.com
Camp, T.R. (1946). "Sedimentation and the Design of Settling Tanks." Trans. Am. Soc. of Civil Eng.,
Ill, paper 2285,895-958.
Camp Dresser & McKee Inc. and Larry Walker Associates. (1993). California Storm Water Best
Management Practice Handbooks, California State Water Resources Council Board, Alameda CA,
March.
Campbell, G.S. (1974). "A Simple Method for Determining Unsaturated Conductivity from Moisture
Retention Data," Soil Science, 117:311-314.
Cannon, L. (2002). Urban BMPs and their Modeling Formulations, Master of Science Project Report,
Dept. of Civil, Construction, and Environmental Engineering, Oregon State University, Corvallis, August.
R-137
-------
Carollo Engineers, (1999). Sullivan, Stark, and Holladay Basins, Sewer Relief and Reconstruction
Predesign. Draft Problem Characterization Report, Report to City of Portland, Bureau of Environmental
Services, Portland, OR, October.
Carollo Engineers and CH2M Hill (2000). Sanitary Sewer System Condition Assessment for Sanitary
Sewer Overflow Elimination Program. Vallejo Sanitation and Flood Control District, Vallejo, CA.
Chapra, S.C. (1997). Surface Water-Quality Modeling, McGraw-Hill, NY.
Chen, C. (1975). "Design of Sediment Retention Basins." In: Proc. National Symposium on Urban
Hydrology and Sediment Control. University of Kentucky, Lexington, 28-31 July 1975. Parsons,
Brinkerhoff, Quade & Douglas, NY. pp. 285-298.
Chiew, F.H.S., Mudgway, L.B., Duncan H.P. and McMahon, T.A. (1997). Urban Stormwater Pollution,
Industry Report 97/5, Cooperative Research Centre for Catchment Hydrology, Melbourne, Australia,
http://www.catchment.crc.ore.au/index.shtml (1997).
Covar, A.P. (1976). "Selecting the Proper Reaeration Coefficient for Use in Water Quality Models,"
Proceedings of the Conference on Environmental Modeling and Simulation, W.R. Ott, ed., Cincinnati,
OH, EPA-600/9-76-016, Environmental Protection Agency, Washington, DC, July, pp. 340-343.
Debo, T.N. and Reese, A.J. (2003). Municipal Storm Water Management, Lewis Publishers, Boca Raton,
FL, Second Edition.
Dingman, S.L. (2002). Physical Hydrology. Prentice-Hall, Upper Saddle River, NJ, Second Edition.
Diniz, E.V. (1980). Porous Pavement; Phase I, Design and Operational Criteria, EPA-600/2-80-135.
Environmental Protection Agency, Cincinnati, OH, August.
Dissmeyer, G.R. and Foster, G.R. (1984). A Guide for Predicting Sheet and Rill Erosion on Forested
Land. United States Department of Agriculture, Forest Service Southern Region, Technical Publication
R8-TP 6.
Donigian, A.S., Jr., Huber, W.C. and Barnwell, T.O. Jr. (1995). "Modeling of Nonpoint Source Water
Quality in Urban and Non-urban Areas," Chapter 7 in Nonpoint Pollution and Urban Stormwater
Management, V. Novotny, ed., Technomic Publishing Co., Inc., Lancaster, PA,, pp. 293-345.
Dooge, J.C.I. (1973). Linear Theory ofHydrologic Systems, Technical Bulletin No. 1468, Agricultural
Research Service, U.S. Department of Agriculture, Washington, D.C.
Dortch, M.S. and Gerald, J.A. (1985). Screening-Level Model for Estimating Pollutant Removal by
Wetlands. Wetlands Research Program Technical Report WRP-CP-9. U.S. Army Corps of Engineers,
Waterways Experiment Station, Vicksburg, MS.
Driscoll, E.D. (1986a) "Lognormality of Point and Non-Point Source Pollutant Concentrations," in Proc.
Stormwater and Water Quality Users Group Meeting, T.O. Barnwell and W.C. Huber, eds., Orlando, FL,
EPA/600/9-86/023 (NTIS PB87-117438/AS), Environmental Protection Agency, Athens, GA, September.
Driscoll, E.D. (1986b). Methodology for Analysis of Detention Basins for Control of Urban Runoff
Quality, EPA-440/5-97-001, Environmental Protection Agency, Office of Water, Washington, DC.
R-138
-------
Fair, M.F. and Geyer, J.C. (1954). Water Supply and Waste-water Disposal, John Wiley and Sons, NY.
Fair, M.F., Geyer, J.C. and D.A. Okun, D.A. (1986). Water and Wastewater Engineering, John Wiley and
Sons, NY.
Fan, C-Y., Field, R., Lai, F., Heaney, J.P., Sample, D.J. and Wright, L.T. (2000). "Costs of Urban
Stormwater Control." Proc. ASCE 's 2000 Joint Conference on Water Resources Engineering,
Minneapolis, American Society of Civil Engineers, Reston, VA, CD-ROM.
Field, R., Heaney, J.P. and Pitt, R.E., eds. (2000). Innovative Urban Wet-Weather Flow Management
Systems, Technomic Publishing Co., Lancaster, PA. (See also Web site:
http://www.epa.gov/ednnrmrl/publish/book/main.htm for Adobe Acrobat copy.)
Field, R. and Sullivan, D., eds. (2003). Wet-Weather Flow in the Urban Watershed. Lewis Publishers,
Boca Raton, FL.
Field, R., Sullivan, D. and Tafuri, A.N. eds. (2004). Management of Combined Sewer Overflows. Lewis
Publishers, Boca Raton, FL.
Fletcher, T.D., Peljo, L., Fielding, J., Wong, T.H.F. and Weber, T. (2002). "The Performance of
Vegetated Swales for Urban Stormwater Pollution Control," In Global Solutions for Urban Drainage,
Proc. Ninth International Conference on Urban Drainage, E.W. Strecker and W.C. Huber, eds., Portland,
OR, American Society of Civil Engineers, Reston, VA, CD-ROM, September (2002).
Foster, G.R. (1982). "Modeling the Erosion Process," In: C.T. Haan, H.P. Johnson, and D.L. Brakensiek
(ed.). Hydrologic Modeling of Small Watersheds, ASAE Monograph Number 5, pp. 297-380.
Freeze, R. A. and Cherry, J.A. (1979). Groundwater. Prentice Hall, Inc, Englewood Cliffs, NJ. (1979).
Gardner, W.R. and Hillel, D.I. (1962). "The Relation of External Evaporative Conditions to the Drying of
Soils," J. Geophysical Research, 67:4319-4325.
GeoSyntec Consultants, Urban Drainage and Flood Control District, and Urban Water Resources
Research Council of ASCE (2002). Urban Stormwater BMP Performance Monitoring, A Guidance
Manual for Meeting the National Stormwater BMP Database Requirements. EPA-821-B-02-021 (CD-
ROM), Environmental Protection Agency, Washington, DC, April, or at:
http://www.bmpdatabase.org/docs/Urban%20Stormwater%20BMP%20Performance%20Monitoring.pdf.
Godwin, D.C. and Jones, C.A. (1991). "Nitrogen Dynamics in Soil-plant Systems." In: J. Hanks and J.T.
Ritchie (ed.) Modeling Plant and Soil Systems. Agronomy Monograph no. 31. ASA-CSSA-SSSA,
Madison, WI.
Heaney, J.P., Huber, W.C., Medina, M.A., Jr., Murphy, M.P., Nix, S.J. and Hasan, S.M. (1977).
Nationwide Evaluation of Combined Sewer Overflows and Urban Stormwater Discharges, Volume II:
Cost Assessment and Impacts. EP2-600/2-77-064b. NTIS PB-266005, Environmental Protection Agency.
Cincinnati, OH. March.
Heaney, J.P. and Lee, J.G. (2006). Optimization of Urban Sewer Systems during Wet Weather Periods,
Final Report to Environmental Protection Agency, Contract No. 68-C-01-020, Dept. of Civil,
Environmental, and Architectural Engineering, University of Colorado, Boulder.
R-139
-------
Heaney, J.P., Sample, D.J. and Wright, L.T. (1999a). Costs of Urban Stormwater Systems. Final Report
to USEPA, National Risk Management Research Laboratory, Cincinnati, OH.
Heaney, J.P., Sample, D.J. and Wright, L.T. (1999b). Geographical Information Systems, Decision
Support Systems, and Urban Stormwater Management. Final Report to USEPA, National Risk
Management Research Laboratory, Cincinnati, OH.
Heaney, J.P., Sample, D.J. and Wright, L.T. (1999c). "Integration of GIS and Optimization for Evaluating
Urban Storm Water Quality Options," Proc. ASCE Conference on Water Resources Planning and
Management, Tempe, AZ, American Society of Civil Engineers, Reston, VA, June.
Heaney, J.P., Sample, D.J. and Wright, L.T. (1999d). "Robust Optimization of Urban Stormwater
Systems," In Proc. 8th International Conference on Urban Storm Drainage Joliffe, I. and J.E. Ball, Eds.,
Sydney, Australia, Institution of Engineers, Canberra, Volume 4, pp. 1896-1903, August.
Heaney, J.P., Wright, L.T. and Sample, D.J. (1999e). Innovative Methods for Optimization of Urban
Stormwater Systems. Final Report to US EPA, National Risk Management Research Laboratory,
Cincinnati, OH.
Helsel, D.R. and Hirsch, R.M. (1992). Statistical Methods in Water Resources, Elsevier, Studies in
Environmental Science 49, NY.
Hoffman, E.J. (1985). "Urban Runoff Pollutant Inputs to Narragansett Bay: Comparison to Point
Sources." EPA-440/5-85-001, Environmental Protection Agency, pp. 159-164.
Hoffman, J. and Crawford, D. (2000). "Using Comprehensive Mapping and Database Management to
Improve Urban Flooding Modeling," Bureau of Environmental Services, City of Portland, OR.
Hogland, W., Niemczynowicz, J. and Wahlman, T. (1987). "The Unit Superstructure during the
Construction Period." The Science of the Total Environment, 59:411-424.
Horner, R. (1995). Constructed Wetlands for Urban Runoff Water Quality Control. Dept. of Civil
Engineering, University of Washington, Seattle, WA.
Huber, W.C., (1996). "BMP Simulation Using the EPA Storm Water Management Model (SWMM).
"Proc. Seventh International Conference on Urban Storage Drainage, University of Hannover,
Hannover, Germany, September, Vol. Ill, pp. 1629-1633.
Huber, W.C. (200 la). "New Options for Overland Flow Routing in SWMM," Urban Drainage Modeling,
R.W. Brashear and C. Maksimovic, eds., Proc. of the Specialty Symposium of the World Water and
Environmental Resources Conference, ASCE, Environmental and Water Resources Institute, Orlando,
FL, May, pp. 22-29.
Huber, W.C. (200Ib). "Wet-weather Treatment Process Simulation Using SWMM," .Proc. Third
International Conference on Watershed Management. National Taiwan University, Taipei, Taiwan, pp.
253-264.
Huber, W.C., Borst, M. and Koustas, R.M. (2000). "SWMM Modeling in the New Century," Proc. Joint
Conference on Water Resources Engineering and Planning and Management, Minneapolis, MN,
American Society of Civil Engineers, Reston, VA, CD-ROM, July.
R-140
-------
Huber, W.C. and Cannon, L. (2002). "Modeling Non-Directly Connected Impervious Areas in Dense
Neighborhoods," In Global Solutions for Urban Drainage, Proc. Ninth International Conference on
Urban Drainage, E.W. Strecker and W.C. Huber, eds., Portland, OR, American Society of Civil
Engineers, Reston, VA, CD-ROM, September.
Huber, W.C. and Dickinson, R.E. (1988). Storm Water Management Model Version 4, User's Manual,
EPA/600/3-88/00la (NTIS PB88-236641/AS), Environmental Protection Agency, Athens, GA.
Huber, W.C., Driscoll, E.D. and Ink, P.L. (1987). "Use of Monitored Water Quality Frequency
Characteristics for SWMM Quality Calibration," in Urban Storm Water Quality, Planning and
Management, W. Gujer and V. Krejci, eds., Proc. Fourth International Conference on Urban Storm
Drainage, Lausanne, Switzerland, September, pp. 208-213.
Huber, W.C., Heaney, J.P., Aggidis, D.A., Dickinson, R.E., Smolenyak, K.J. and Wallace, R.W. (1982).
Urban Rainfall-Runoff-Quality Data Base, EPA-600/2-81-238 (NTIS PB82-221094), Environmental
Protection Agency, Cincinnati, OH, July.
Hurdle, J. (2001). Improving the Modeling of Sediment Transport in Sewers in SWMM: A Feasibility
Study, Master of Science Project Report, Department of Civil, Construction, and Environmental
Engineering, Oregon State University, Corvallis.
Inamdar, S.P., Altier, L.S., Lowrance, R.R., Williams, R.G. and Hubbard, R. (1998a). "The Riparian
Ecosystem Management Model: Nutrient Dynamics." Proceedings of the First Federal Interagency
Hydrologic Modeling Conference, Las Vegas, NV, April, pp. 1.73-1.80.
Inamdar, S.P., Sheridan, J.M., Williams, R.G., Bosch, D.D., Lowrance, R.R., Altier L.S. and Thomas,
D.L. (1998b). "The Riparian Ecosystem Management Model: Evaluation of the Hydrology Component."
Proceedings of the First Federal Interagency Hydrologic Modeling Conference, Las Vegas, NV, April,
pp. 7.17-7.24.
Inamdar, S.P., Sheridan, J.M., Williams, R.G., Bosch, D.D., Lowrance, R.R., Altier, L.S. and Thomas,
D.L. (1999a). "The Riparian Ecosystem Management Model (REMM): I. Testing of the Hydrologic
Component for a Coastal Plain Riparian System." Transactions ASAE, 42(6): 1679-1689.
Inamdar, S.P., Sheridan, J.M., Williams, R.G., Bosch, D.D., Lowrance, R.R., Altier, L.S. and Thomas,
D.L. (1999b). "The Riparian Ecosystem Management Model (REMM): II. Testing of the Water Quality
and Nutrient Cycling Component for a Coastal Plain Riparian System." Transactions ASAE, 42(6): 1691-
1707.
James, W, Huber, W.C., Pitt, R.E., Dickinson, R.E. and James, W.R.C. (2002a). Hydrology, A Guide to
the rain, Temperature, Runoff, Combine and Statistics Modules of the USEPA SWMM4, Water Systems
Models [1], Computational Hydraulics International, Guelph, ON, Version 2.4, October.
James, W, Huber, W.C., Pitt, R.E., Dickinson, R.E., Roesner, L.A., Aldrich, J.A. and James, W.R.C.
(2002b). Hydraulics, User's Guide to SWMM4 Transport, Extran, and Storage Modules and to
PCSWMM4, Water Systems Models [2], Computational Hydraulics International, Guelph, ON, Version
2.4, October.
James, W.R.C., James, W. and van Langsdorff, H. (2001). "Stormwater Management Model for
Environmental Design of Permeable Pavement," Chapter 26 in Models and Applications to Urban Water
R-141
-------
Systems, Proc. Conference on Stormwater and Urban Water Systems Modeling, W. James, ed.,
Computational Hydraulics International, Guelph, Ontario, pp. 423-445.
Jones, C.A., Cole, C.B., Sharpley, A.N. and Williams, J.R. (1984). "A Simplified Soil and Plant
Phosphorous Model: I. Documentation." J. Soil Science Society of America, 48:800-805.
Kadlec, R.H. and Knight, R.L. (1996). Treatment Wetlands, Lewis Publishers, Boca Raton, FL.
Keblin, M., M. Barrett, J.F. Malina, and R.J. Charbeneau, (1997). The Effectiveness of Permanent
Highway Runoff Controls: Sedimentation/Filtration Systems, Center for Research in Water Resources,
Bureau of Engineering Research, University of Texas at Austin.
Knisel, W.G., Leonard, R.A. and Davis, P.M. (1993). "Nutrient Component Documentation," In: W.G.
Knisel (ed.) GLEAMS (Groundwater Loading Effects on Agricultural Management Systems) Model
Documentation, Version 2.10, UGA-CPESBAED Publication #5, University of Georgia, Tifton, GA.
Lane, L.J., (1982). "Development of a Procedure to Estimate Runoff and Sediment Transport in
Ephemeral Streams," In Recent Developments in the Explanation and Prediction of Erosion and Sediment
Yield. Pub. No. 137. International Association for Hydrological Science, Wallingford, England, pp. 275-
282.
Lawrence, I., (1986). "Source and Fate of Urban Runoff Constituents and Their Management," Proc. of
12th Symposium on Stormwater Quality in Urban Areas, Water Research Foundation of Australia,
University of Woollongong, New South Wales, Australia, July.
Lee, E.R., Mostaghimi, S. and Wynn, T.M. (2002). "A Model to Enhance Wetland Design and Optimize
Nonpoint Source Pollution Control," Journal of American Water Resources Association, 38(1): 17-32,
February.
Lee, J.G., (2003). Process Analysis and Optimization of Distributed Urban Stormwater Management
Strategies, Ph.D. Thesis, Department of Civil, Environmental and Architectural Engineering, University
of Colorado, Boulder.
Levenspiel, O., (1972). Chemical Reaction Engineering, John Wiley and Sons, NY, Second Edition.
Lin, Y.C. and Medina, M.A., Jr. (2003). "Incorporating Transient Storage in Conjunctive Stream-Aquifer
Modeling." Advances in Water Resources, 26(9): 1001-1019.
Lindsay, W.L. (1979). Chemical Equilibria in Soils. John Wiley and Sons. NY.
Linsley, R.K. and Franzini, J.B. (1964). Water Resources Engineering, McGraw-Hill Book Co., NY.
Liptan, T. (2001). Storm Summary Report, Lexington Hills BMP Monitoring, Bureau of Environmental
Services, City of Portland, OR.
Lowrance, R.R., Altier, L.S., Williams, R.G., Inamdar, S.P., Sheridan, J.M., Bosch, D.D., Hubbard, RK.
and Thomas, D.L. (2000). "REMM: The Riparian Ecosystem Management Model." J. Soil and Water
Conservation. 55(l):27-34.
Magette,W.L., Brinsfield, R.B., Palmer, R.E. and Wood, J.D. (1989). "Nutrient and Sediment Removal
by Vegetated Filter Strips," Transactions of the ASAE, Volume 32, No. 2, pp. 663-667.
R-142
-------
Malmquist, P. (1978). "Atmospheric Fallout and Street Cleaning - Effects on Urban Stream Water and
Snow," Prog. Wat Tech., 10(5/6): 495-505. Pergamon Press, Great Britain, September.
McCuen, R.H. (1982). A Guide to Hydrologic Analysis Using SCSMethods, Prentice-Hall Inc.,
Englewood Cliffs, NJ.
Medina, M.A., Huber, W.C. and Heaney, J.P. (1981). "Modeling Stormwater Storage/Treatment
Transients: Theory," Journal of Environmental Engineering Division, Proc. ASCE, 107(EE4): 781-797.
Metcalf and Eddy, Inc. (2003). Wastewater Engineering. Fourth Edition, McGraw-Hill, NY.
Mills, W.B., Porcella, D.B., Ungs, M.J., Gherini, S.A., Summers, K.V., Mok, L., Rupp, G.L., Bowie, G.L.
and Haith, D.A. (1985). Water Quality Assessment: A Screening Procedure for Toxic and Conventional
Pollutants in Surface and Ground Water (Revised 1985), Parts 1 and 2, EPA/600/6-85/002a (NTIS PB86-
122496) and EPA/600/6-85/002b (NTIS PB86-122504), Environmental Protection Agency, Athens, GA,
September.
Minton G.R. (2002). Stormwater Treatment; Biological, Chemical and Engineering Principles, Resource
Planning Associates, Seattle, WA.
Musgrave, G.W. (1955). "How Much of the Rain Enters the Soil?" In: Water, U.S. Dep. Ag. Yearbook.
Washington, DC, pp. 155-159.
Nix, S.J. (1994). Urban Stormwater Modeling and Simulation, Lewis Publishers, Boca Raton, FL.
Nix, S.J., Heaney, J.P. and Huber, W.C. (1978). "Revised SWMM Storage/Treatment Block,".Proc.,
Storm Water Management Model Users Group Meeting, Annapolis, Maryland, H.C. Torno, ed.,
EPA-600/9-79-003 (NTIS PB-290742/6BE), Environmental Protection Agency, Washington, DC,
November, pp. 1-27.
O'Connor, D.J. (1983). "Wind Effects on Gas-Liquid Transfer Coefficients," J. Environmental
Engineering, 109(3):731-752.
Parton, W.J., Stewart, J.W.B. and Cole, C.V. (1988). "Dynamics of C, N, P and S in Grassland Soils: A
Model," Biogeochemistry 5:109-131.
Persson, J., Sommes, N.L.G. and Wong, T.H.F. (1999). "Hydraulics Efficiency of Constructed Wetlands
and Ponds" Water Science and Technology, 40(3):291-300.
Phillips, B.C. and Goyen, A.G. (1987). "Guidelines for Improving Urban Stormwater Quality Using
Control Structures," Proc. Seminar on Urban Runoff Water Quality, Institution of Engineers Australia,
A.W.W.A. Sydney, July 12.
Pitt, R.E. (1986). "The Incorporation of Urban Runoff Controls in the Wisconsin Priority Watershed
Program." In: Advanced Topics in Urban Runoff Research, (Edited by B. Urbonas and L.A. Roesner).
Engineering Foundation and ASCE, NY. pp. 290-313.
Pitt, R.E. (1987). Small Storm Flow and P articulate Washoff Contributions to Outfall Discharges. Ph.D.
dissertation, Department of Civil and Environmental Engineering, the University of Wisconsin - Madison,
November.
R-143
-------
Pitt, R.E., Clark, S., Farmer, K. and Field, R. (1996). Groundwater Contamination from Stormwater. Ann
Arbor Press, Inc. Chelsea, MI.
Pitt, R.E., Lantrip, J., Harrison, R, Henry, C.L. and Xue, D. (1999a). Infiltration Through Disturbed
Urban Soils and Compost-Amended Soil Effects on Runoff Quality and Quantity, EPA/600/R-00/016,
Environmental Protection Agency, Cincinnati, OH.
Pitt, R.E. and McLean, J. (1986). Toronto Area Watershed Management Strategy Study - Number River
Pilot Watershed Project. Ontario Ministry of the Environment, Toronto, Ontario, June.
Pitt, RE., Lilburn, M., Nix, S.J., Durrans, S.R., Burian, Voorhees, J. and Martinson, J. (1999b). The
Source Loading and Management Model (SLAMM) Guidance Manual for Integrated Wet Weather Flow
(WWF) Collection and Treatment Systems for Newly Urbanized Areas (New WWF Systems). U.S.
Environmental Protection Agency, Edison, NJ, December.
Pitt, R.E. and Voorhees, J. (2000). The Source Loading and Management Model (SLAMM), A Water
Quality Management Planning Model for Urban Stormwater Runoff. University of Alabama, Department
of Civil and Environmental Engineering, Tuscaloosa, AL.
Pratt, C.J., Mantle, J.D.G. and Schoffield, P.A. (1989). "Urban Stormwater Reduction and Quality
Improvement through the Use of Permeable Pavements." Urban Discharges and Receiving Water Quality
Impacts. Pergamon Press, Oxford, 1989, pp. 123-132.
Pratt, C.J., Mantle, J.D.G. and Schoffield, P.A. (1995). "UK Research into the Performance of Permeable
Pavement, Reservoir Structures in Controlling Stormwater Discharge Quantity and Quality,"
International Conference NOVATECH 95, Lyon, pp. 337-344.
Prince George's County (2000). Low-Impact Development Design Manual. Dept. of Environmental
Resources, Prince George's County, Laurel, MD.
Rathbun, R.E. (1977). "Reaeration Coefficients of Streams - State-of-the-Art," J. Hydraulics Division,
Proc. ASCE, 103 (HY4): 409-424.
Rawls, W.J., Brakensiek, D.L. and Miller, N. (1983). "Green- Ampt Infiltration Parameters from Soils
Data," Journal of Hydraulic Engineering, 109(1):62-70.
Reuss, J.O. and Innis, G.S. (1977). "A Grassland Nitrogen Flow Simulation Model." Ecology. 58:379-
388.
Roesner, L.A., Aldrich, J.A. and Dickinson, R.E. (1988). Stormwater Management Model User's Manual,
Version 4: Extran Addendum. EPA/600/3-88/11 Ob (NTIS PB88-236658), Environmental Protection
Agency, Athens, GA.
Running, S.W. and Coughlan, J.C. (1988). "A General Model of Forest Ecosystem Processes for Regional
Applications I. Hydrologic Balance, Canopy Gas Exchange and Primary Production Processes." Ecol.
Model. 42:125-154.
Sample, D.J., Heaney, J.P., Wright, L.T. and Koustas, R.M. (2001). "Geographical Information Systems,
Decision Support Systems, and Urban Stormwater Management" Journal of Water Resources Planning
and Management, 127(3): 155-161.
R-144
-------
Schnoor, J.L., Sato, C., McKechnie, D. and Sahoo, D. (1987). Processes, Coefficients, andModels for
Simulating Toxic Organics and Heavy Metals in Surface Waters, EPA/600/3-87/015, Environmental
Protection Agency, Athens, GA, June.
Schueler, T.R. (1987). Controlling Urban Runoff: A Practical Manual for Planning and Designing Urban
BMPs, Metro Washington DC Council of Governments, July (1987).
Shaheen, D.G., Contributions of Urban Roadway Usage to Water Pollution. 600/2-75-004. U.S.
Environmental Protection Agency, Washington, DC, April.
Shaver, H.E. (1986). "Infiltration as a Stormwater Management Component," Urban Runoff Quality
Impact and Enhancement Technology, B. Urbonas and L. Roesner, eds., American Society of Civil
Engineers, NY.
Singh, V.P., ed. (1995). Computer Models of Watershed Hydrology, Water Resources Publications,
Highlands Ranch, CO.
Skaggs, RW. (1978). A Water Management Model for Shallow Water Table Soils. WRRI, Report no.
134, University of North Carolina, Durham.
Soil Conservations Service (SCS, now NRCS, Natural Resources Conservation Service) (1986). Urban
Hydrology for Small Watersheds. U.S. Dept. of Agriculture Tech. Release No. 55 (revised), June.
Available at: http://www.wcc.nrcs.usda.gov/hydro/hydro-tools-models-tr55.html.
Sonnen, M.B. (1977). "Subroutine for Settling Velocity of Spheres," Journal of the Hydraulics Division,
ASCE, 103 (HY9): 1097-1101.
Stone, J.J., Hawkins, R.H. and Shirley, E.D. (1994). "An Approximate Form of the Green Ampt
Infiltration Equation," J. Irrigation and Drainage, 120(1): 128-137.
Stouder, M. (2003). Simulation Methods for Wetland/Pond BMPs, Master of Science Project Report,
Dept. of Civil, Construction, and Environmental Engineering, Oregon State University, Corvallis, June.
Strecker, E.W. (2001). "Low-Impact Development (LID): Is It Really Low or Just Lower?" In Linking
Stormwater BMP Designs and Performance to Receiving Water Impact Mitigation, Proc. Engineering
Foundation Conference, B.R. Urbonas, ed., American Society of Civil Engineers, Reston, VA, pp. 210-
222.
Strecker, E.W., Kersnar, J.M., Driscoll, E.D. and Horner, R.R. (1992). The Use of Wetlands
for Controlling Storm Water Pollution. The Terrene Institute, Alexandria, VA.
Strecker, E.W., Quigley, M.M., Urbonas, B.R, Jones, J.E. and Clary, J.K. (2001). "Determining Urban
Storm Water BMP Effectiveness," Journal of Water Resources Planning and Management, 127(3): 144-
149.
Tchobanoglous, G., and Schroeder, E.D. (1985). Water Quality, Addison-Wesley Publishing Co. Inc.
Englewood Cliffs, NJ.
Terstriep, M.L. and Stall, J.B. (1974). The Illinois Urban Drainage Area Simulator, ILLUDAS, Bulletin
58, Illinois State Water Survey, Urbana, IL.
R-145
-------
Thackston, E.L., Shields, F.D., Jr. and Schroeder, P.R. (1987). "Residence Time Distributions of Shallow
Basins," Journal of Environmental Engineering, 113(6): 1319-1332, December.
Thomann, R.V. and Mueller, J.A. (1987). Principles of Surface Water Quality Modeling and Control,
Harper and Row, NY.
Thomas, D.L., and Beasley, D.B. (1986). "A Physically-based Forest Hydrology Model I: Development
and Sensitivity of Components." Trans. oftheASAE, 29:962-972.
Thompson, M.K. and James, W. (1995). "Provision of Parking Lot Pavements for Surface Water
Pollution Control Studies." Modern Methods for Modeling the Management of Stormwater Impacts.
Edited by W. James. Computational Hydraulics International, Guelph, Ontario, pp. 381 - 398.
Trepel, M., Dall'O', M., Dal Cin, L., de Wit, M., Opitz, S., Palmeri, L., Persson, J., Pieterse, N.M.,
Timmermann, T., Bendoricchio, G., Kluge, W. and J0rgensen, S.E. (2000). "Models for Wetland
Planning, Design and Management." EcoSys 8:93-137. Available at: http://www.ecology.uni-
kiel.de/~michael/wet/mod.pdf
Urbonas, B.Rand Stahre, P. (1993). Stormwater: Best Management Practices and Detention for Water
Quality, Drainage, and CSO Management, Prentice Hall, Englewood Cliffs, NJ.
USD A-ARS-Southeast Watershed Research Laboratory (1999). Riparian Ecosystem Management Model,
U.S. Department of Agriculture, Agricultural Research Service, Tifton, GA,
http://www.cpes.peachnet.edu/remmwww/remm/remmoldwww/default.htm
U.S. Environmental Protection Agency (EPA) (1983). Final Report for the Nationwide Urban Runoff
Program. Water Planning Division, Washington, DC, December.
U.S. Environmental Protection Agency, National Risk Management Research Laboratory (EPA
NRMRL). (1996). Wet-Weather Flow Research Plan, Edison, NJ,
http://www.epa.gov/ednnrmrl/repository/wwfplan/wwf_plan.pdf
Van Buren, M.A., Watt, W.E. and Marsalek, J. (1997). "Application of the Log-Normal and Normal
Distribution to Stormwater Quality Parameters," Water Research, 31(1):95-104, (1997).
Vogel, R.M. (1986). "The Probability Plot Correlation Coefficient Test for the Normal, Lognormal, and
Gumbel Distributional Hypotheses," Water Resources Research, 22(4):587-590, April.
Walker, W.W., (1985). "Empirical Methods for Predicting Eutrophication in Impoundments Report 3.
USAE Waterways Experiment Station, Vicksburg, MS, March.
Walker, W.W. (1990). P8 Urban Catchment model Program Documentation Version 1.1, Prepared for
IEP, Inc. and Narragansett Bay Project, October, (http://wwwalker.net/p8/)
Walker, W.W. (1995). "Design Basis for Everglades Stormwater Treatment Areas." Water Resources
Bulletin, 31(4):671-685, August.
Walker, W.W. and Kadlec, R.H. (2002). Dynamic Model for Stormwater Treatment Areas, Sponsored by
U.S. Department of the Interior, (http://wwwalker.net/dmsta/).
R-146
-------
Wanielista, M.P., Kersten, R. and Eaglin, R. (1983). Water Quantity and Quality Control. Ann Arbor
Press, Ann Arbor, MI.
Water Environment Federation and American Society of Civil Engineers (1998). Urban Runoff Quality
Management, WEF Manual of Practice No. 23, ASCE Manual and Report on Engineering Practice No.
87, Water Environmental Federation, Alexandria, VA.
Wayne County, Michigan (1998). User's Manual: Watershed Management Model Version 4.0, Rouge
River National Wet Weather Demonstration Project, Technical Memorandum RPO-NPS-TM27.01,
Detroit, MI, September.
Williams, J.R., Jones, C.A. and Dyke, P.T. (1984). "A Modeling Approach to Determining the
Relationship Between Erosion and Soil Productivity," Trans. ASAE, 27:129-144.
Williams, J.R., Ouyang, Y., Chen, J-S. and Ravi, V. (1998). Estimation of Infiltration Rate in the Vadose
Zone, Volume II: Application of Selected Mathematical Models, EPA/600/R-97/128b, Environmental
Protection Agency, National Risk Management Research Laboratory, Ada, OK. Available at:
http://www.epa.gov/ada/csmos/ninflmod.html.
Wischmeier, W.H. and Smith, D.D. (1978). Predicting Rainfall Erosion LossesA Guide to
Conservation Planning. USDA Agriculture Handbook No. 537, Washington, DC.
Wong, T.H.F. and Breen, P.P. (2002). "Recent Advances in Australian Practice on the Use of Constructed
Wetlands for Stormwater Treatment," In Global Solutions for Urban Drainage, Proc. Ninth International
Conference on Urban Drainage, E.W. Strecker and W.C. Huber, eds., Portland, OR, American Society of
Civil Engineers, Reston, VA, CD-ROM, September.
Wong, T.H.F., Duncan, H.P., Fletcher, T.D. and Jenkins, G.A. (2001). "A Unified Approach to Modelling
Urban Stormwater Treatment," Proc. 2nd South Pacific Stormwater Conference, Auckland, New Zealand,
pp. 319-327, June.
Wong, T.H.F., Fletcher, T.D., Duncan, J.P., Coleman, J. and Jenkins, G.A. (2002). "A Model for Urban
Stormwater Improvement Conceptualisation," In Global Solutions for Urban Drainage, Proc. Ninth
International Conference on Urban Drainage, E.W. Strecker and W.C. Huber, eds., Portland, OR,
American Society of Civil Engineers, Reston, VA, CD-ROM, September.
Wool, T.A., Ambrose, R.B., Martin, J.L. and Comer, E.A. (2001). Water Quality Analysis Simulation
Program (WASP), Version 6.0 Draft User's Manual. Environmental Protection Agency, Environmental
Research Laboratory, Athens, GA.
Wright, L.T., Dent, S., Mosley, C., Kadota, P. and Djebbar, Y. (2001a). "Comparing Rainfall Dependent
Inflow and Infiltration Simulation Methods," Chapter 16 in Models and Applications to Urban Water
Systems, Proc. of User's Group Conference, W. James ed., Monograph 9, Computational Hydraulics
International, Guelph, Ontario, pp. 235-258.
Wright, L.T, Heaney, J.P., Dent, S. and Mosley, C. (2001b). "Optimization of upstream and downstream
controls for sanitary sewer overflows," In Urban Drainage Modeling, Proc. of the Specialty Symposium
of the World Water and Environmental Resources Congress, R.W. Brashear and C. Maksimovic, eds.,
Orlando, FL, American Society of Civil Engineers, Reston, VA, pp. 156-164.
R-147
-------
Wright, L.T., Heaney, J.P. and Weinstein, N. (2000). "Micro-scale Modeling and Monitoring of Low
Impact Development Management Strategies for Wet weather Flow Control." ASCE 2000 Joint
Conference on Water Resources Engineering and Water Resources Planning and Management-Building
Partnerships, Minneapolis, MN, American Society of Civil Engineers, Reston, VA, CD-ROM, 11 pp.
Yang, C.T. (1996). Sediment Transport Theory and Practice, McGraw-Hill Series in Water Resources
and Environmental Engineering, NY.
Young, R.A., Onstad, D.D., Bosch, D.D. and Anderson, W.P. (1989). "AGNPS, Agricultural Non-Point-
Source Pollution Model: A Watershed Analysis Tool." J. Soil and Water Conservation. 44:168-173.
Yu, S.L., Fitch, G.M. and Earles, T.A. (1998). Constructed Wetlands for Stormwater Management,
VTRC 98-R35, Virginia Transportation Research Council, Charlottesville, VA. Available at:
http://www.virginiadot.org/vtrc/main/online reports/pdf/99-rl4.pdf
Websites
US EPA Center for Exposure Assessment Modeling, HSPF, May 2002
http://www.epa.gov/ceampubl/hspf.htm,.
EPA/ASCE BMP Database
http: //www .bmpdatabase .org/
Riparian Ecosystem Management Model REMM:
http://www.cpes.peachnet.edu/remmwww/remm/remmoldwww/default.htm
New Jersey Standard for Bioretention Systems
http://www.state.nj.us/dep/watershedmgt/DOCS/BMP_DOCS/chapter5.PDF
Characterization of models from the USEPA, USDA / ARS / NRCS, Federal Highway Administration,
Universities and others, model characterization.pdf, May, 2002.
http: //www .glc. org/proj ects/sediment/allweb3 2 .pdf
Minnesota BMP manual
http://www.pca.state.mn.us/water/pubs/swm-appendices.pdf
The most current version of SWMM (version 4.4h) can be downloaded from the Oregon State University
website: http://www.ccee.orst.edu/swmm.
SWMM5 is available from the EPA web site:
http://www.epa.gov/ednnrmrl/models/swmm/index.htm
R-148
------- |