&EPA
United States
Environmental Protection
Agency
Municipal Environmental Research EPA-600'2-80-01 3
Laboratory March 1 980
Cincinnati OH 45268
Research and Development
Select Topics in
Stormwater
Management
Planning for New
Residential
Developments
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironr nentaI technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. “Special” Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL PROTECTION TECH-
NOLOGY series. This series describes research performed to develop and dem-
onstrate instrumentation, equipnient, and methodology to repair or prevent en-
vironmental degradation from point and non-point sources of pollution. This work
provides the new or improved technology required for the control and treatment
of pollution sources to meet environmental quality standards.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-80-013
March 1980
SELECT TOPICS IN STORMWATER MANAGEMENT PLANNING
FOR NEW RESIDENTIAL DEVELOPMENTS
by
Robert Berwick, Michael Shapiro
Jochen Kuhner, Daniel Luecke, Janet J. Wineman
Meta Systems, Inc.
Cambridge, Massachusetts 02138
Grant No. R-805238
Project Officers
Chi-Yuan Fan and Douglas C. Ammon
Storm and Combined Sewer Section
Wastewater Research Division
Municipal Environmental Research Laboratory (Cincinnati)
Edison, New Jersey 08817
MUNICIPAL ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
-------
DISCLAIMER
This report has been reviewed by the Municipal Environmental Research
Laboratory, U.S. Environmental Protection Agency, and approved for
publication. Approval does not signify that the contents necessarily
reflect the views and policies of the U.S. Environmental Protection Agency,
nor does mention of trade names or commercial products constitute endorsement
or recommendation for use.
1 1
-------
FOREWORD
The U.S. Environmental Protection Agency was created because of increas-
ing public and governmental concern about the dangers of pollution to the
health and welfare of the merican people. Noxious air, foul water, and
spoiled land are tragic testimony to the deterioration of our natural environ-
ment. The ccmplexity of that environment and the interplay between its com-
ponents require a concentrated and integrated attack on the problem.
Research and development is that necessary first step in problem solu-
tion, and it involves defining the problem, measuring its impact, and search-
ing for solutions. The Municipal Environmental Research Laboratory develops
new and improved technology and systems for the prevention, treatment, and
management of wastewater and solid and hazardous waste pollutant discharges
from municipal and community sources, for the preservation and treatment of
public drinking water supplied and to minimize the adverse economic, social,
health, and aesthetic effects of pollution. This publication is one of the
products of that research -- a most vital communications link between the
research and the user community.
This report concerns the management of stormwater runoff from new resi-
dential developments. The authors have examined several problem areas re-
lating to the planning and implementation of non—conventional stormwater con-
trol measures in such developments, including 1) the evaluation of pollutant
accumulation and washoff data, 2) the development of production and cost
functions for stormwater management measures, 3) the formulation of stochastic
models for management planning, and 4) identification of political and insti-
tutional barriers to implementing non—conventional control measures. By
reanalyzing existing data bases, the authors show how improved statistical
methods can lead to new interpretations of the pollutant accumulation and
washoff processes. Their results suggest how elements of existing models
such as STORM and SWMM might be improved, how cost functions can be developed,
and how relatively simple stochastic models can be used to screen alternative
management strategies. The authors also demonstrate how a simulation study
can be used to evaluate alternative on-site management measures and combina-
tions of the literature on innovation in zoning, subdivision regulation, and
building codes with experience in two case studies to identify sources of
difficulty in implementing innovative control measures.
Francis T. Mayo
Director
Municipal Environmental Research
Laboratory
iii
-------
ABSTPACT
Areas of research undertaken in this study included the evaluation of
pollutant accumulation and washoff data, development of production functions
for stormwater management measures, formulation of simple stochastic models
for stormwater management, estimation of cost models for control measures,
and evaluation of institutional and political problems in implementing non-
conventional control measures.
Analysis of existing data on street surface pollutant accumulation indi-
cated that distributions are log-normal. Previous studies generally ignored
this characteristic. Accumulation is a nonlinear process, but is modeled
more appropriately by a second order relationship than the first order models
previously used. Examination of new washoff data indicated that exponential
models such as those incorporated in SWMM can fit individual storms quite welL
However, parameters are not constant across storms.
Simulation studies and two—way table analyses were used to evaluate sub-
division layout and stormwater control measures. Effects of individual mea-
sures were non—additive and interacted with site layout. Porous pavement was
the single most effective control measure of those considered, but altering
subdivision layout was an equally effective approach.
Three examples of simple stochastic analyses were developed to illustrate
their use in preliminary planning: selection of a storm event for drainage
management, design of a management system for combined sewage, and prediction
of runoff quality.
Using a simplified runoff analysis, a planning model for predicting con-
ventional drainage costs was developed and estimated from an existing data
set. The high explanatory power of this model suggests that when adequate
cost data are developed, similar models can be developed for non—conventional
management measures.
Studies on local acceptance of innovation in subdivisions design and
building codes were examined to identify possible problems in implementing
new stormwater control measures. Factors affecting the acceptance of innova-
tion included strength of the housing market, professionalism and technical
expertise in the government, and city size and geographic region. Similar
factors will influence the success of innovation in stormwater management.
The report discusses two Massachusetts case studies in detail.
This report was submitted in fulfillment of Grant Number R-805238 by
Meta Systems Inc under the sponsorship of the U.S. Environmental Protection
Agency. The project extended from July 11, 1977 to January 31, 1979; work
was completed by the end of July, 1978.
iv
-------
• 1
• 3
• 6
• 8
• 8
• 13
• 14
• 16
• 16
• 17
• 22
• 31
• 38
51
51
51
51
53
53
56
63
64
64
65
CONTENTS
Foreword
Anstract..
Figures
Tables
lthbreviations and Symbols
Acknowledgements
iii
iv
vii
x
xii
xv
1. Introduction
2. Conclusions
3. Recommendations
4. Alternatives for the Control of Stormwater Quantity
and Quality
Types of Control Measures
Subdivision Design
Selection of Control Mechanisms
5. Analysis of Loading and Washoff Data
Introduction
Exploratory Data Analysis
Analysis by Two-Way Tables
Analysis of San Jose Street Cleaning Data
Analysis of Washoff: The Envirex—DOT Data.
6. Developing Production Functions for Runoff Control
Measures
Why a Simulation Model
Design of the Model: Hydrology
The Development Patterns
The Simulation Site
Details of the Development Layouts
Analysis of Results of the Simulation Model:
Constructing Production Functions for On-Site
Control Measures
Conclusions
7. Stochastic Models
Introduction
Model #1: Storm Drainage Design
Model #2: Hydraulic Capacity of a Treatment
System
An Example
8. Estimating Costs of On-Site Control Measures . .
Previous Studies
Reanalysis of Rawis and Knapp Data
Adapting Cost Functions for Non-Conventional
Management Systems
• . . 69
• . . 72
• . 75
• . 75
78
87
v
-------
9. Institutional and Political Issues • 90
Introduction 90
Regulatory Systems Affecting Innovation in Residential
Developments 90
Planned Unit Development 91
Building Code Regulations 96
Summary of Literature Review 99
Case Studies 101
Reasons for Project Success 106
Appendix A: Analysis of Loading and Washoff Data 108
Appendix B: Developing Production Functions for Runoff Control
Measures 136
Appendix C: Listing of Computer Program 158
Appendix D: Performance of Innovative Designs for Stormwater
Management 178
Appendix E: Cost Allocation in Multipurpose Projects 181
References 190
Glossary 197
Table of U.S. Customary Standard International Conversion Constants . 203
vi
-------
FIGURES
Number Page
5-1 Raw Data: Residential Loading Rates 18
5-2 Stem-and-Leaf Display, Residential Loading Rates 18
5—3 LN (Residential Loading Rates) 19
5-4 Residential COD 20
5—5 Residuals from Two—way Fit of Land Use vs. Climatic Region . - - 28
5-6 Diagnostic Plot, Residuals from Two-way Fit, LOG (Suspended
Solids) Land Use vs. Climatic Region 28
5—7 Traffic Effects vs. Traffic Volume 29
5—8 Tropicana Street Loading Rates (Particle Size Less than 600
Microns)
5—9 Loading Rates vs. Day Since Last Rain or Swept, Tropicana Street
Area, San Jose, 1977
5-10 Accumulation Load vs. Time, “Complex” Differential Equation. . . 35
5-11 Sutherland and McCuen Empirical Curves 35
5-12 Loading vs. Days Since Last Rain 36
5-13 Cumulative Suspended Solids vs. Cumulative Flow for Harrisburg
Storms
5-14 Cumulative suspended Solids vs. Cumulative Flow for Milwaukee
Storms 39
5-15 Cumulative Suspended Solids vs. Cumulative Flow, Harrisburg
Storm #7 41
5-16 Cumulative Suspended Solids vs. Cumulative Flow, Harrisburg
Storm #2 42
5—17 Residuals from Fit, Harrisburg Storm #7 43
vii
-------
Number Page
5-18 Residuals from Fit, Harrisburg Storm #2 43
5-19 Suspended Solids vs. Rainfall Intensity, Milwaukee and
Harrisburg Storms 45
5—20 Cumulative VSS, SS Milwaukee Storm #4 46
6-1 Hypothetical Development Block 52
6-2 Bowker Woods Development Site 53
6-3 Bowker Woods Quarter—Acre Development Site (“Conventional ”) . . 54
6-4 Bowker Woods Low Density Development 54
6-5 Bowker Woods Cluster-Townhouse Development 55
6—6 Effects of Control Measures and Development Types on Runoff . 59
6—7 Effects of Control Measures and Development Types on Solids
Runoff 60
7—1 First—Flush Volumes from Colston, 1974 (Ft 3 /Sec) 74
7-2 Suspended Solids from Coiston, 1974 (Mg/L) 74
8—1 Residuals from Equation 8-21 82
A-l Raw Data: Residential Loading Rates (Lbs/Curb-Mile/Day) (TJRS
1974) 110
A-2 Steps in Constructing Stem-and-Leaf Display 110
A-3 Residential Loading Rates (Lbs/Curb—Mile/Day) (URS, 1974) . . . 112
A-4 LN(Residential Lead Loading, Micrograms/Gram) 113
A-S LN (Residential NO Loading, Micrograms/Gram) 113
A-6 Residential Cadmium Loading (Microgram/Gram) 113
A-7 LN (Commercial Orthophosphate Loading, Micrograms/Grain) 113
A-B LN (Commercial Lead Loading, Micrograms/Gram) 114
A-9 Log (Commercial Nitrate Loading, Micrograms/Gram) 114
A-b LN(Industry and Light Industry COD Loading, Micrograms/Gram). . 114
A-li LN(Industry and Light Industry Lead Loading, Micrograms/Gram) . 114
viii
-------
Number Page
A—l2 Diagnostic Plot (Suspended Solids), Land Use vs. Climatic
Region 123
A—l3 Diagnostic Plot (Suspended Solids), Land Use vs. Average Daily
Traffic Volume 124
A—14 Diagnostic Plot (Suspended Solids), Landscaping vs. Average
Daily Traffic Volume 125
A—l5 Diagnostic Plot (Suspended Solids), Landscaping vs. Climatic
Region 126
A-16 Diagnostic Plot (Suspended Solids), Land Use vs. Landscaping . . 127
A-17 Diagnostic Plot (COD), Climatic Region vs. Land Use 128
A-lB Diagnostic Plot (COD), Land Use vs. Average Daily Traffic
Volume 129
A-l9 Diagnostic Plot (Lead), Land Use vs. Climatic Region 130
A-20 Diagnostic Plot (Lead), Land Use vs. Average Daily Traffic
Volume 131
A-2l Cumulative Fe and TOC vs. Cumulative Flow Milwaukee Storm #4 . . 133
A-22 Cumulative Cl, TS vs. Cumulative Flow Milwaukee Storm #4. . . . 134
A-23 Cumulative Zinc vs. Cumulative Flow Harrisburg Storm #7 135
B-i Hypothetical Development Block 137
B-2 Bowker Woods Development Site 141
B—3 Bowker Woods Quarter-Acre Development (“Conventional”) 141
B-4 Bowker Woods Low-Density Development 142
B-5 Bowker Woods Cluster-Townhouse Development 142
B—6 Sample Output from Computer Program 148
ix
-------
TABLES
Number Page
4-1 Stormwater Control Measures
5-1 Comparison of Resilient vs. URS Results, Urban Street Loadings. . 21
5-2 Two-Way Table Analysis (Suspended Solids), Land Use vs. Climatic
Region 23
5-3 Two-Way Table Analysis (Suspended Solids), Climatic Region vs.
Traffic Density 24
5-4 Two-Way Fit, Log (Suspended Solids), Land Use vs. Traffic Volume. 26
5-5 Two-Way Fit, Log (Suspended Solids), Landscaping VS. Traffic
Density 27
5-6 Summary of Two-Way Table Results (Analysis by Log or LN (Median
Constituent)) 30
5-7 Fraction of Pollutant Associated With Each Particle Size Range. . 31
5-8 DOT Catchment Description 38
5-9 Regression Fits for Harrisburg Storms 40
5-10 Regression Analyses, Cumulative Suspended Solids vs. Cumulative
Flow
5-11 Non-Linear Models for Various Pollution Parameters Milwaukee A,
Storm #4. . 46
5-12 predicted and Observed Solids Loads for SWMM/STORM and Fitted
Exponential Models 48
5—13 Results of Exponential Model for Predicting Suspended Solids
Washed Of f over an Interval 49
6-1 Peak Flow (cfs) Simulation Model 57
6-2 Total Flow from Simulation Model 57
x
-------
Number Page
6-3 Peak Solids from Simulation Model 58
6—4 Total Solids from Simulation Model 58
6-5 Two-Way Analyses of Peak Flow Effects 61
8-1 Data Set Summary, Rawls and Knapp 80
8—2 Comparison of Drainage Area Characteristics 88
9—i Ordinance Design Standards for PUD.s 94
A-l Log-Transformed Loading Data 115
A-2 Transformed Loading Data Minus Row Median 115
A-3 Transformed Loading Data with Row and Column Medians 3ubtracted . 116
A-4 Log(Median Suspended Solids Loading, Lbs/Curb-Mile/Day) 116
A-5 Two-Way Fit, Log (Suspended Solids), Landscaping vs. Climatic
Region 117
A-6 Two-Way Fit, Log (Suspended Solids), Land Use vs. Landscaping. . 118
A-7 Two-Way Fit, LN(COD), Climatic Region vs. Land Use 119
A-B Two-Way Fit, LN (COD), Land Use vs. Average Daily Traffic Volume 120
A-9 Two-Way Fit, LN (Lead), Land Use vs. Climatic Region 121
A-b Two-Way Fit, LN (Lead), Land Use vs. Average Daily Traffic
Volume 122
B-i Model Parameters Changed by Control Measures 139
B-2 Model Parameters for Conventional Development 143
B-3 Model Parameters for Low-Density Development 145
B—4 Model Parameters for Townhouse Development 146
B—5 Transport System Parameters (Conventional Development) 147
E-l Comparison of Methodologies to Measure Water Quality Benefits. . 188
xi
-------
LIST OF ABBREVIATIONS AND SYMBOLS
ABBREVIATIONS
acre—ft —— acre—foot
ADT —- average daily traffic flow
BOD 5 —- five-day biological oxygen demand
cfs —— cubic feet per second
COD 2 3 -- chemical oxygen demand
ft(ft , ft ) —— feet (square feet, cubic feet)
ft 3 /sec — — cubic feet per second
gal/SY —- gallons per square yard
ha —— hectare
in/hr -- inches per hour
kg -- kilogram
km -- kilometer
lb -- pound
m —— meter
m 3 /sec —— cubic meters per second
mg/2 —- milligram per liter
rni n -- minute
R 2 —— coefficient of determination
se —— standard error
TS -- total solids
SYMBOLS
A —— drainage area, in acres
a —— empirical parameter (section 5)
—- multiplier on dry weather flow (D), aD signifies design
capacity (section 7)
—— empirical parameter (section 8, defined on page 77)
Ad —— developed acres in drainage area
ai —— constant in runoff coefficient calculation, evaluated
separately for each level of impervious area
AT —— total drainage area, in acres
a 1 ,a 2 ,a 3 —— empirical parameters in general flow formula
a —— empirical parameter
b -- empirical parameter
bi ,b 2 —— empirical parameters in pipe cost equation
—- empirical parameter
C —— runoff coefficient
c —— empirical parameter
Cp —- cost of providing storage basin
CD -- damage function associated with stormwater quality and quantity
xii
-------
C c l
Cj (
Cl -
Cp
Cr
Cs
CT
ci ,c2
C 3
D
D(i)
DB
DE
Ds,Dw
d 1 ,d 2 ,d 3 ,
d ,d 5 ,d 6
E
e 1 ,e 2 ,e 3 ,ei 4
f(i)
F(i)
F 1 ( ),F 2 (
Fe
flow
r
I
I
115
i*
K
k
k 1
K(i)
L
A
M
MS
n
(t)
-- cumulative cadmium loading
—— present value of costs of control measure i
-- cumulative chloride loading
—— pipe cost in dollars per foot (1963 dollars)
—— cumulative chromium loading
—— cost of collection and storage system
—— total cost of storm runoff, in 1963 dollars
—— empirical parameters in system length equation
—- empirical parameter relating A/Ad ratio to total storm cost
—- dry flow (section 7)
—— pipe diameter in inches (section 8)
-- damage caused by inundation of property
—— smallest pipe diameter, in inches
—- largest pipe diameter, in inches
-- locational dummy v riables in total cost equation, for
southern and western regions of the U.S., respectively
—— empirical parameters in storm runoff total cost equation
-— pipe costs
—— empirical parameters in cost per foot equation
—- density function of rainfall intensity
-- Frequency function of rainfall intensity
—— production functions relating to control levels and rainfall
to water quality and quantity, respectively
-— cumulative iron loading 3
—— flow accumulated up to and including period t, in ft
—— gamma function
—— empirical parameter
-- design rainfall intensity, in inches per hour
—— control measure (section 4)
—- rainfall intensity, in inches per hour (section 7)
-- 15 minute storm intensity associated with design storm
—— design rainfall intensity
—— location specific empirical parameter in rainfall intensity
equation
—— multiple of the increase in a large number of random variables
(section 5, p. 20)
—— empirical constant (section 5, p. 40)
-— decay constant (section 5, p. 47)
—— empirical constant (section 7)
—— accumulation constant
—— erosion constant
—— storm sewer cost
-- solids washoff pounds
—- maximum potential solids washoff
—- total length of drain, in feet
—— an empirical parameter, used with r, such that r/A = mean and
= variance of storm intensity in Gamma distribution
—— number of inlets and manholes
—— mean squared flow
—— Manning’s roughness coefficient
xiii
-------
N02 -— cumulative nitrite loading
N03 -— cumulative nitrate loading
O-P0 4 -- cumulative orthophosphate loading
P(t) -- probability of storm event
Pb -- cumulative lead loading
P0 4 -- cumulative phcsphate loading
Pr (D+a)>aD -— probability that D+a is greater than aD
Q -- storm runoff (section 7)
-— design storm outlet capacity (section 8)
Q(t) -- storm water quantity for rainfall event t
q(t) —— storm water quality standard for rainfall event t
r —— an empirical parameter, used with A such that r/ = mean, and
nA 2 = variance of storm intensity (section 7, p. 70)
—— runoff in inches per hour (section 7, p. 71)
r 1 —- runoff inches/time in period 1
r 2 -- runoff inches/time in period 2
S —— average slope, in percent
ss —— cumulative suspended solids loading
sst -- suspended solids, in lbs., accumulated up to period t
T —— storm frequency, in years
—— design storm frequency, in years
-— index of rainfall event (some combination of duration,
intensity, etc.) (section 4)
—— time (section 5, p. 32)
t —- time, in minutes (section 5, p. 49)
TOC —— cumulative total organic carbon loading
TKN -- cumulative total Kjeldahl nitrogen loading
t 1 —— elapsed time from storm start, time 1 (hours)
t 2 -- elapsed time from storm start, time 2 (hours)
U -- pipe utilization factor
U 0 -- maximum pipe utilization factor
V —— runoff volume
VSS —— total volatile suspended solids loading
x —— empirical parameter (section 8)
xi,x 2 —— empirical parameters
Z -- total cost of storm runoff
z —— random variable
Zi —— set of random variables
xiv
-------
ACKNOWLEDGNENTS
BSC Engineering and in particular Robert Daylor and John Thomas provided
valuable information on stormwater control measures and development costs.
They also contributed the case studies in Section 9. Alfred Leonard, our
research assistant, performed a substantial amount of the statistical analy-
sis on the accumulation and washoff data. Chi—Yuan Fan and Douglas ? imnon,
Project Officers and Richard Field, Chief, Storm and Combined Sewer Section,
U.S. Environmental Protection Agency, Municipal Environmental Research Labo-
ratory, provided helpful guidance.
We would also like to thank Byron Lord of the Department of Transporta-
tion, and Nic Kobriger of Envirex, Inc., for permission to use the Envirex-
DOT data analyzed in Section 5, and Robert Pitt of Woodward-Clyde Consultants
for the use of the San Jose street cleaning data.
Our special thanks to Dianne C. Wood, who organized the preparation and
production of the final report. Her careful supervision improved both the
appearance and content of the document.
xv
-------
SECTION 1
INTRODUCTION
According to U.S. Census Bureau estimates, the population of the United
States is expected to grow from 203 million in 1970 to 264 million in 2000.
If the average household size remains constant, 18 million additional dwelling
units will be needed by the year 2000 -- a 30 percent increase. Replacement
of attrition in the existing housing stock may require roughly an equal number
of new units. In other words, a substantial proportion of the nation’s hous-
ing stock in the year 2000 will have been constructed over the next 20 years.
Thus the manner in which these new units are designed, constructed, and main-
tained will be an important factor in the contribution of the residential
sector to the nation’s environmental quality. To the extent that proper plan-
ning today can mitigate possible adverse consequences of such new development,
environmental quality can be preserved and enhanced. This report deals with
one aspect of the environmental consequences of new development: the impact
of stormwater runoff and the management of that impact.
Unfortunately, it is not possible today to develop a manual or handbook
that will routinely and accurately predict the runoff impacts from various
types of new development and prescribe cost—effective mitigation measures.
The state of knowledge in several crucial areas, although improving rapidly,
is not yet well—enough developed. This study has had a much more modest
goal: to develop methodologies and approaches for dealing with several
selected aspects of the planning problem.
1. The development of in roved methods for estimating pollutant
accumulation and washoff from street surfaces
The accumulation and washoff functions currently used in stormwater
quality computations are crucial elements in such computations, yet both the
basic formulations and the techniques used to estimate parameters for these
formulations can be questioned. Using a variety of statistical techniques,
this study has reexamined much of the existing data in these areas and has
begun an exploratory analysis of newly acquired data.
2. The development of production functions for describing the
effectiveness of stormwater control measures and site layout
patterns
Virtually no actual data exist from which such performance measures can
be estimated. Our approach was to simulate the effects of various control
measures and layout patterns on runoff from a small subdivision. The results
of these simulations were then used to explore methods for summarizing the
impacts of various alternatives.
1
-------
3. The formulation of stochastic models for stormwater management
Since stormwater runoff events are inherently stochastic, relatively
simple probabilistic models may be formulated for many planning purposes to
capture the essence of the decision problem. Several such models were formu-
lated in this study and are discussed in this report.
4. The estimation of cost models for control measures
In order to develop general guidelines for stormwater management in new
developments, planners recjuire cost models which can be employed at a pre-
liminary planning level in the absence of detailed engineering design data.
This study investigated how such models can be developed and estimated, using
as a specific example the costs of conventional stormwater drainage systems.
5. The evaluation of institutional and political problems in
implementing non—conventional control measures
Two case studies of non—conventional approaches were examined. Both are
small subdivisions in eastern Massachusetts and incorporate porous pavement
in their design. The report describes the problems encountered by the devel-
oper in gaining approval for these designs and places them in context by dis-
cus sing the general problems of innovation in residential development.
A closely related issue is that of allocating the costs of stormwater manage-
ment among the affected parties. Drawing upon the general economic literature,
this report discusses some of the basic principles for evaluating various
allocation mechanisms.
2
-------
SECTION 2
CONCLUSIONS
ANALYSIS OF ACCUMULATION AND WASHOFF DATA
• Our reanalysis of available pollutant accumulation data for street
surfaces indicates that loadings follow log-normal distributions.
This is true both of solids and specific pollutants. Previous studies
have ignored the nature of the frequency distribution, and as a con-
sequence their reported sunimary statistics are often misleading.
• Models for explaining the solids and pollutant accumulation as a func-
tion of variables describing land use, traffic, and geography can be
developed using the technique of two—way tables. Our results suggest
that average daily traffic is a consistently important variable.
Solids loadings tend to decrease with increasing traffic levels.
These results suggest that scouring is an important mechanism affect-
ing accumulation. In the case of lead, loadings first increased then
decreased with increasing traffic volume, suggesting that both deposi-
tion and scour are important factors. Other variables were less con-
sistently related to accumulation. In general the unexplained varia-
tion in these models was high —- on the order of the effects of the
variables themselves. Thus with the present data base, detailed
predictive models are not feasible. Outside of a possible correction
for traffic volume, there is no reason to use other than median (or
some typical value) accumulation values.
• An examination of the available data on accumulation over time indi-
cates that the process is nonlinear; accumulation is initially rapid
and then tapers off. The process is not well—modeled by a simple
first order process, as has been commonly assumed. A second order
accumulation model appears to be more appropriate. The data also
suggest that since the bulk of the material accumulates within the
first day, street sweeping would have to be very frequent to be
effective as a water quality management practice.
• Analysis of washoff data indicates that accumulated pollutant loads
from individual storms can be modeled as a function of accumulated
washoff using log linear formulations or exponential models similar
in structure to those in STORM and SWMM. However, the parameters vary
widely across both watersheds and storms within the same watershed.
3
-------
CONSTRUCTING PRODUCTION FUNCTIONS FOR ON-SITE CONTROL MEASURES
Given the lack of actual data on on—site stormwater controls, simulation
studies can be used to compare the relative effectiveness of sets of con-
trols. However, the results of these studies should not be viewed as being
representative of the real performance of control measures.
• Because of interactions alTonç control measures and among controls and
development types, analysis of variance or two—way table methods are
a prerequisite for a proper analysis of the effects of control mea-
sures and development layouts.
• For the set of three residential developments considered, the
simulations indicate that porous pavement is the most effective single
control measure in reducing peak and total flow and peak and total
solids washoff. Flows and solids are about 25-35 percent of the base
case (no controls). Additional controls (swales, roof drain discon-
nection), account for an additional ten percent. Control measure
effects are not additive; use of additional controls beyond two shows
diminishing return (zero—one percent further reduction). Vegetative
cover is an exception, reducing solids washoff even when used with
two or three other controls already in place.
• Cluster development lowers flow or solids 12 to 15 percent compared to
conventional developments with the same number of total housing units.
• The interaction between on—site controls and developments is
multiplicative. Therefore, by carefully planning the proper combina-
tions of control measures and development layouts, substantial reduc-
tions in stormwater runoff may be effected. This is a key finding
for the planning of new residential developments.
COST MODELS
• Relatively simple models based on a physical model of runoff control
can be estimated to explain the costs of stormwater management
facilities. Such models explained 60 to 90 percent of the variation
in costs in an available data set depending upon the extent of detailed
hydrologic information assumed.
STOCHASTIC MODELS
• For preliminary planning and regional studies, relatively simple
stochastic models can prove to be useful adjuncts or alternatives to
simulation approaches. Examples illustrated in this report are select-
ing planning events for drainage facilities, evaluating design alterna-
tives for combined sewer systems, and computing the expected loading
from runoff events.
4
-------
INSTITUTIONAL AND POLITICAL ISSUES
• Factors that influence the acceptance of innovations in residential
development include the strength of the local housing market, the
degree of professionalism among city officials and agency personnel,
the size of the city, and the region of the country. Major difficul—
ties in the case of planned unit development (PUD) ordinances and new
building codes arise when special interest groups feel threatened by
new innovations. It is anticipated that these patterns will assert
themselves in the case of stormwater management innovations. In two
cases a developer (BSC Engineering) was able to convince Massachusetts
towns to accept innovative measures after meeting considerable initial
resistance. A crucial element in gaining acceptance for the plans was
the developer’s ability to convince town officials that the full risks
of the designs would fall on the developments. In both cases the
innovative measures required considerably more time for approval than
would conventional designs.
5
-------
SECTION 3
RECOMMENDATIONS
ACCUMULATION AND WASHOFF DATA
• Runoff quality predictions.are sensitive to street loading. Our
analysis indicates that there is considerable unexplained variation
in the existing data. This may be due either to the inherent vari-
ability in the processes involved or inadequate models. More data
and continuing analysis using resistant statistics are needed to
resolve this question.
• Further consideration should be given for incorporation in STORM and
SWMM of a second order model accumulation relationship such as the one
developed in this report.
PRODUCTION FUNCTIONS FOR ON-SITE CONTROL MEASURES
• In planning on—site management measures, the subdivision layout
should be considered carefully, since such measures can be at least
as effective as control technologies and can influence the effective-
ness of such technologies.
• While simulation studies can be used to explore techniques for fitting
production functions and to suggest the relative effectiveness of
alternative measures, there is a lack of data for verifying such
studies. A high priority should be given to obtaining data on the
effectiveness of control measures. Porous pavement in particular
looks very attractive and should receive accelerated study.
STOCHASTIC MOL LS
•Stochastic models can be used effectively in preliminary planning
studies. However, this type of analysis has been under—utilized in
the past because planners and designers lack familiarity with the
techniques involved. EPA planning materials and manuals should
incorporate models similar to those discussed in this report and dis-
cuss the nature and limitations of their use.
DST FUNCTIONS
• In order to develop general cost models for non—conventional control
measures, a set of synthetic cost data should be developed that
6
-------
represents the range of control measures, development types, and site
conditions likely to be encountered in practice.
INSTITUTIONAL P ND POLITICAL ISSUES
• Some type of legislative action may be necessary to minimize resis-
tance to innovative control measures at the local level. Such action
might range from educational programs and n del regulations to a
program of encouraging states to enact preemptive legislation (as was
done by Operation Breakthrough). 208 agencies would be appropriate
vehicles for educational programs.
7
-------
SECTION 4
ALTERNATIVES FOR THE CONTROL OF STORMWATER QUANTITY AND QUALITY
TYPES OF CONTROL MEASURES
Once it is recognized that the objective of stormwater management is not
solely to remove runoff from the site as quickly as possible, a large variety
of alternative approaches become available to planners and engineers. These
measures may alter the runoff peak, the total quantity of runoff, and/or the
runoff quality and may also provide additional secondary benefits to the
developer and residents of a subdivision. Of course, each measure is also
associated with a vector of primary and secondary costs. Thus the task of the
engineer is generally perceived as that of selecting a measure or combination
of measures to meet a set of objectives (described in terms of runoff quantity
and quality and possibly other secondary benefits) while minimizing the costs.
More generally, the objective might be specified as maximizing net benefits.
In either case a necessary first step in the analysis must be an enumeration
of the options available and their general characteristics. In this section
we present such a description, and, as an introduction to the remainder of
the report, discuss some of the current limitations to carrying forward the
complete analysis.
Table 4 -1 presents a list of control measures, organized by principal
control mechanism and described in terms of a variety of positive impacts
(benefits) and negative impacts (costs). Considered in these categories are
not only monetary costs and the primary impacts upon water quality, but also
secondary benefits and Costs coninonly or necessarily associated with these
measures. Three types of primary benefits are included. Measures that
reduce the peak or total quantity of flow enhance the ability of downstream
works to treat the runoff, while measures that remove pollutants (chiefly
suspended solids) directly enhance water quality.
The allocation of costs among the relevant interests is not a si nple
matter. For a discussion of different methodologies, see Appendix E.
The major control mechanisms are as follows:
Storage/Detention . Measures that provide a means to capture and store
runoff for some period of time.
Overland Flow Modification . Measures that influence the rate of runoff
by altering the velocity or direction of overland flow.
Infiltration . Measures that increase the ability of the ground to accept
storn ater by infiltration.
8
-------
TABLE 4-1. STORMWATER CONTROL MEASURES
Bone f_its
Primarj Secondar — — — — Costs
4) —l
44) 0 4-4 I 4)
4) )C - .4 -‘4 - -I U “-. I 4)
0’ 4444 4.4 —I U I .G 044 •.4
U ‘0 -‘-4 ‘-4 4) 4)4
OeO.c I )) 44 14 4 4)4) 0 ’—I 4-4 . )48J1JU.. -4
. 4)4-4 4.4 4) .ic 4c (444 w ‘oc cc>
( 44J t4O 0 0 44 U 044’4 04)
04)0-14 44 ‘4 4dW .E 4 44(4 Bi(fl ’4C Q’d
‘-40 4.40 - . 4 444)0444)
-4 -14.40-4 . 4 . 0’UW 4.’ ZB
4 . 4Q4J 04.4 ‘0 944140 4) UU )Z1 . 0>,’ 1 4 ’ 1 4 ’ o4- ’ ,OA.4-4.C
o 4)444-14’ 4) :rd - )0 .0 4444 0 . .4-144. ) ,’000W - ’ 4W 44
1444 J ’ - 4 x 4.4 4)44.044.044 4)q4 ’-44) BH44OQ . 0
m 0e. . u oog , . 0 e c o.uu . 4)44 U - 4(44 )O 4 b ’0O4J4. )
, W44OO ’ -4 , 4444 0 4)4.444 14 ’-40 O
9 (. ) .44 O 0 0.V)Z0( 4
Stor ge
Rooftop
X
)(
X
X
x
X
X
Parking Lot
X
X
X
X
Wet Detention Basins
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Dry Detention Basins
underground
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
— —
House Lot
X
X
X
X
X
X
X
X
X
X
Small Dikes, Traps,
etc.
X
X
X
X
X
Fountains
X
X
X
X
X
X
Infiltration
Trenches
X
X
X
X
)
X
Pits
X
X
X
X
X
X
Porous Pavement
Roads
Driveways
Sidewalks
Parking
x
x
x
x
x
x
x
x
x
x
. .
x
i
x
x
x
x
X
x
x
A.
—
—
X
X
L-
X’
—
Gravel
Roads
Driveways
X
X
X
X
X
X
X
X
X
ç
*
X
X - Possible or likely impact.
-------
TABLE 4.-i. (CONTINUED)
Pri mary
Ccrnnrlaru
Costs
Cn 4 m;
0 )5 0
0) t4 1 .4
b’ I .i 111 41
Iw II.0 0
OOO.O 0)
.0 044-i 4.)
00050 0
0 1fl0-4 8.1
• . 4 . 4G.-I 0i
4)04-’ 01-’
U u84
. 0 541—4
54100)
0)5000—i
0 . E-I0C1 i .
0
0
—4
4)
ITS
S
84
U
0)
1 1 .4
.4
0
1 .-441
0) • . 40
) W
50) E
C) 0
I8.440
4041
0.0405
00
845
-------
H
H
TABLE 4-1. (CONTINUED)
Benefits
Primary — Seconda — — — — Costs — — — —
r’ - i 5
55 0
0 O’4.’
()
omo: m
.5 014 .1 J
CUSsO 0
0 50 .- I U
O’-40i
JOJ 04)
C) Q .-4U- .-40
a)1 .’—lo
‘oa)04ca)o
5 5 1 4)00 5—I
0. E- ’QC a).
C
0
--i
4. )
14
14)
U
U
a)
5
‘ i -
- -l
0
.-1J
a • c
J Q )
a)w .5
;4.i- -1U
CU4J5
o.cee
OUPS
USO’S
U
. 4
4.)
U
.5
U)
U
e
l
4-i
a m
Ca)
a )s
)4
we
4J
. 1ia)
O
• I S
U ‘. I—i a)
-
a) 55 04.J0’.- l .4 ) 4’U.-l
U U5 .-5u)nja) U oc cc>
• . i Ca)
QI54C4. I . U ,dm Oo’a)Oa)EwiJi..
.i5)..4,. a)uJOQ )U
USZBaJ• . 4Q 0 >,’-I-I ’C . 1J
C OO . .4 . -4• .44J0 50W - . -4U U
0’QIO UU .U) 9 . 1SUO—44.4U ’UO4a
-IX - ICUOOSSa)OUUOa)55. ..45...4
Soil Stabilization
X
X
(Continued)
Chemical Tackifiers
Hydromulching
X
X
x
x
—
x
x
Management
StreetCleaning
x
—
x
x
SewerFlushing
X
X
—
Sewer Flow Control
x
x
x
——
Cleaning of Catch
Basins
x
x
Filtration
Drain Filters
X
X
X
StraworHayBales
X
X
X
——
——
——
Sources: Poertner (1974); Soil Conservation Service (1975); Engineering Science (1975);
Hittinan Associates (1976); IJLI, ASCE. and NABS (1975); Fairfax County, Va. (1974).
-------
Soil Stabilization . Measures that reduce soil erosion by binding the
soil particles and/or by dissipating the energy of falling rain drops.
Filtration . Measures that strain suspended sediment particles from
stormwater flow.
Management . Measures that maintain or operate existing infrastructure
to alter the timing of storinwater flows and reduce their pollution
potential.
In Table 4—1 measures have been classified according to their principal
mechanisms as listed above, but it should be understood that certain measures
might involve more than one mechanism. For example, while grasses and mulch
provide direct stabilization of the soil, they also alter the pattern of over-
land flow by providing higher flow resistance than does bare soil. If prop-
erly maintained, certain types of storage basins will greatly enhance
infiltration. Thus the categories presented are not necessarily mutually
exclusive, but are intended to highlight the principal mechanisms of control.
Table 4-1 provides an indication of which costs and benefits have been
attributed to specific measures. No attempt is made to quantify these
attributes, either in relative or absolute terms. Methods for quantifying
certain of the attributes, particularly the primary benefits and costs, will
be discussed in later sections of the report. Other impacts can be assessed
only qualitatively. Further, some of the adverse impacts indicated in the
table can be alleviated with proper management. For example, the water level
in a wet basin can be controlled to eliminate mosquito breeding. Table 4-1
is intended to be a framework for later analyses and the selection of combi-
nations of measures. The paragraphs below discuss by mechanism the major
areas of benefits and costs that have been identified.
Stormwater storage measures reduce stormwater peaks by dampening extreme
storm flows and improve the quality of the water by providing a period of time
for the larger suspended particles to settle out. Generally, storage measures
do not reduce the total quantity of flow, although depending on the facility
type there may be some reduction through evaporation or infiltration. The
secondary benefits of storage facilities depend upon the scale of the facili-
ties and their management. The larger wet storage basins have the greatest
potential for beneficial impacts, providing a variety of opportunities for
recreational uses, a possible habitat for aquatic organisms, and aesthetic
enhancement of a site. All storage measures provide some degree of flood
protection as do all other measures that reduce peak flow. However, stormwater
quality improvement and flood protection are not completely complementary.
For example, if the first flush of pollutants is significant, it might be
desirable from a quality standpoint to retain the early flows long enough to
achieve maximUm removal by settling. Flood control objectives, on the other
hand, might dictate discharging the early flows as rapidly as possible in
anticipation of later peaks. There are some opportunities for water reuse
using storage measures, although additional treatment may be necessary
depending upon the nature of the reuse.
Most storage measures have significant capital costs and operating and
maintenance expenses associated with cleaning sediment and debris from the
12
-------
facilities and adjacent areas. Wet basins need particularly good maintenance
in order to obtain the full range of potential benefits. Also there may be
health, safety and insect problems associated with standing bodies of water.
Storage facilities that comprise a secondary use of some facility, such as
parking lots, backyards or roadways, tend to be the least expensive in terms
of direct capital costs, but may lower the value of the facility for its
principal use.
Infiltration measures reduce both the peak and total flow and in the
process provide a means for enhanced groundwater recharge. Because of the
contamination of urban stormwater by toxic organic and inorganic materials
(Colston, 1974) there exists a potential groundwater contamination problem
which has not been fully assessed.
Overland flow modification measures are generally small in scale and must
be implemented extensively to have a significant impact on a development.
Some of these measures serve chiefly to enhance water quality by reducing the
erosion potential of overland flow, while others act primarily to alter the
timing, and, to some extent, the quantity of runoff. Swales provide all
three primary benefits and in addition may be preferred on aesthetic grounds
to traditional curbs, since they help preserve the natural character of lower
density development.
Drain filters and bales are used mainly to remove sediment from construc-
tion site runoff. They also provide some attenuation of runoff peaks, but
are employed as temporary measures only. cleaning after each storm is neces-
sary, particularly for drain filters.
Most soil stabilization measures enhance infiltration and alter runoff
patterns. Grass and shrubs may be employed either as temporary or permanent
stabilization measures, while mulches and tackifiers may be employed to
stabilize and protect the soil until a vegetative cover can develop and mature.
These measures increase the capital costs of site development, but can be
cost—effective when compared with downstream removal of deposited sediments
(Engineering Science, Inc., 1973).
Management methods have been directed primarily at quality improvement
and generally have high operating costs. Sewer flow management systems employ
the storage capacity of sewers to attenuate the storm peak flows and thus
enhance the capacity of treatment plants to remove pollutants. Aesthetic
improvement is listed in Table 4—1 as a secondary impact of street cleaning.
Of course, street cleaning as practiced today is primarily intended for its
aesthetic effects and probably has little impact on stormwater quality (URS,
1974)
SUBDIVISION DESIGN
The distinction between new and existing residential developments is a
factor in stormwater management decisions for two reasons. First, existing
developments impose structural limitations on the types of options that can
be considered. There may not be enough room for detention basins, for example.
In theory these limitations do not apply to new subdivisions, although in
13
-------
practice the options available to the designer depend upon the point in the
subdivision process at which stormwater management is considered. If little
or no thought is given to this factor until after all the important layout
and structural decisions have been made, then for all practical purposes the
designer may be as limited as in an existing development.
A second opportunity that is available only for new developments is the
management of stori ater through the basic layout of the subdivision itself.
For example, by clustering, dwelling units (that is, using smaller lot sizes
in return for preserving common open space), the length of road network and
hence the impervious area can be greatly reduced. An example from the Costs
of Sprawl (Real Estate Research Corporation, 1974) illustrates the type of
savings involved. For 1000 single family homes in a conventional subdivision
with one—third acre (1,344 square meters) lots, 60,000 feet (18,293 meters) of
streets would be required, covering 37 acres (149,233 square meters) of land.
If the same homes were clustered on one-fifth acre (807 square meter) lots,
44,750 feet (13,643 meters) of street would be required, covering 31 acres
(125,033 square meters) -- 16 percent less than the conventional layout. With
more intensive clustering (e.g., the use of townhouses) even greater savings
are possible. If the clustered layout has the same gross density as the
conventional subdivision, the reduction in street surface would be translated
into open space and thus a reduction in overall impervious area. In general,
this would lead to reduced runoff from the site and make possible the preser-
vation of natural drainage and wetland areas that provide natural treatment
for the runoff.
In clustered subdivisions, however, activities are more concentrated;
accumulation of street pollutant per curb mile could be greater than in con-
ventional layouts. These considerations may attenuate the advantages of
clustering to some extent, but there are no data upon which to base an assess-
ment of these possibilities. In general, it is still likely that layout
patterns which minimize impervious area are beneficial to stormwater quality
by keeping gross population density constant. Without a quantitative measure
of the effectiveness of clustering, it is not possible to compare the storm-
water impacts of a conventional subdivision against a clustered design with
a higher gross density.
SELECTION OF CONTI JL MECHANISI4S
If the secondary costs and benefits of stormwater control are ignored for
the moment, the problem of selecting the scale or activity level for the
various storinwater control measures can be stated as a problem of minimizing
the expected costs of runoff events:
N
minimize: E C (q(t)IQ(t)) p(t) + I C.(x.) (4—1)
t i=l 1
subject to: q(t) = F 1 (X 1 , ... XN)1 (42)
14
-------
Q(t) = F 2 (X 1 , ... X )l (4—3)
where q(t) = stormwater quality standard for rainfall event t,
Q(t) stormwater quantity for rainfall event t,
X. = level of control measure i, i = 1, ... N (N = total num-
ber of measures),
CD = damage function associated with stormwater quality and
quantity; this cost may reflect the costs of conveyance
and treatir nt to some prescribed level of quality, or
more generally, the damages associated with water quality
deterioration,
t = index of rainfall event, which may be appropriately
defined by some combination of duration, intensity,
antecedent conditions, etc.,
c.( ) = present value of costs of control measure i,
F 1 ( ),F 2 ( ) = production functions relating control levels and rainfall
to water quality and quantity, respectively, for storm
event t, and
p(t) = probability of storm event t.
While equations (4-1) to (4—3) are conceptually straightforward, it is
difficult to plan stormwater control measures based directly on this approach
for the simple reason that the necessary technological and economic linkages
are not yet well-established. This statement is particularly true for new
developments, but applies to existing areas as well. In addition to these
technical considerations, planners recognize that a variety of political and
institutional factors impinge upon the environmental decision—making process
and that a solution to the design problem which does not take these factors
into account could be unworkable in practice. This is true for all environ-
mental control measures, but stormwater management poses some unique problems
because of the degree to which local governments, as opposed to state or
federal agencies, control the design process through zoning ordinances, sub-
division codes, and building standards.
In moving from a simple enumeration of options such as Table 4—1 to a
selection of appropriate measures for a specific location or region, a variety
of obstacles must be overcome. In succeeding sections of this report we
examine in detail selected problems that appear to pose special difficulties
in the planning of stormwater management for new subdivisions.
15
-------
SECTION 5
ANALYSIS OF LOADING AND WASHOFF DATA
INTRODUCTION
Two basic strategies have been adopted in gathering data for use in quan-
tifying urban stormwater pollutant loadings (see Appendix A). One strategy
has focused on measuring the accumulation and composition of dust and dirt
on street surfaces. The other has focused on “end of pipe” or waterway
measurements of flow and concentration during storm periods. While data from
both strategies have been combined and analyzed (URS, 1974), it is unfor-
tunate that to our knowledge only a few studies (Pitt, Woodward-Clyde Consul-
tants, 1977; Envirex-Department of Transportation, 1977) are being performed
in which both types of measurements are employed simultaneously in the same
study area. Such data would provide a basis for development of a mass bal-
ance for quantitative comparisons of the two measurement strategies.
It is evident that differences between the two strategies and general
problems associated with measurement and data reduction have considerably
hindered attempts to develop statistical models for use in predicting non—
point source loadings or in calibrating theoretical runoff models.
As discussed in Appendix A, one of the major studies which attempted
analysis and model development was done by UPS. It utilized univariate
statistical techniques in its intensive analysis. But a basic problem in
applying univariate analytical strategy is that it could easily have led to
false causal inferences, considering the multicollinearity in the data. The
authors noted that a multivariate analysis (e.g., analysis of covariance)
would have been more desirable, but they considered the data insufficient.
Apparently, other analytical strategies, such as multiple regression analysis
using dummy variables to represent various classes (e.g., Northwest climatic
region or industrial land use), or a robust two—way table approach, were also
rejected by the authors for similar reasons. Thus, in general, most in-
vestigations of urban pollutant loadings from nonpoint sources are considered
to have yielded disappointing results to date. For example, Singh concluded
at the 1977 ASCE meeting
The main objective of this study was to determine the rates at which
solids accumulate on street surfaces.. .Initial efforts in fitting
simple conceptual models and use of regression analysis were unsuc-
cessful due to extreme scatter in data...
One can argue that the error lies in looking at “extreme scatter in data” as
a nuisance rather than the interesting property of loading rates. Why is
16
-------
there such scatter? For that matter, is there really such scatter? There is
confusion in the literature; for example, Singh (1977) as well as Colston
(1974), Whipple (1976), and Haimner (1976) find no correlation between elapsed
days of accumulation and pollutant concentrations. Sutherland and McCuen
(1975) claim the reverse. In part these contradictions are due to sloppiness
in the definition of “concentrations” and the failure to consider the col-
linear effects between independent variables; but there is also real uncer-
tainty in the physical systems being modeled.
Considering the current lack of a thorough understanding of the accumu-
lation/washoff processes, any kind of straightforward regression technique is
doomed to failure. Therefore methods designed to discover something about
loading rates should be based on the following premises:
1) Exploratory , rather than predictive data analysis; and
2) Stochastic , rather than deterministic model building.
The exploratory process partially relieves us from the burden of having a
specific explanatory model of the data at hand; the stochastic premise is
similar -- it throws everything we cannot specifically describe into a lumped
box of noise. Actually, the two premises are not distinct, since we use
either one to guide the other.
EXPLORATORY DATA ANALYSIS
The commentary on urban loading rates (e.g., Sartor and Boyd, 1972;
Hammer, 1976; URS, 1974) invariably mentions their extreme range. As a typi-
cal example, the URS data for residential loading rates (lbs/curb-mile-day)
shows a range of 8-770 (2.25-217.09 kg/curb-km-day), with a standard devia-
tion of 195 (54.98 kg/curb-km-day). For most categories of land use, climate,
and traffic patterns, the coefficient of variation (standard deviation/mean)
is greater than 1, indicating large variability. This variability in the
loading rates was examined by using the URS data, which is a summary of all
major studies done before 1974 (all raw data normalized to lbs/curb-mile-day).
Figures 5-1 and 5-2 show the residential suspended solids loadings of
the u study in raw form and organized into a stem-and-leaf display. 1
In the spirit of exploratory analysis, we make two preliminary remarks
about the completed stem-and-leaf display of the residential loading rates.
shown in the figures:
1) The data are skewed; this immediately makes questionable the
use of the standard deviation as a confidence—interval esti-
mate (as was done by Singh, 1977), or the use of regression
1
The basic purpose of such displays is to arrange raw data into roughly
numerical order -— like a histogram —— so that one can easily answer ques-
tions such as: what is the largest value? the smallest? what does. the distri-
bution of values look like? Unlike a histogram, however, the display retains
some of the identity of the original data by using the actual digits of the
6ata values to construct the display. This makes it easier to see which data
value is located where in the histogram (See Appendix A for details).
17
-------
FIGURE 5-1. RAW DATA: RESIDENTIAL LOADING RATES
URS LBS/CURB-MILE-DAY
400 600 390 210 170 019 032
121 148 081 062 121 135 148
019 020 096 153 060 022
032 035 024 033 041 028
070 092 2700 690 260 860
220 372 659 418 70 85 24
77 238 18 34 103 93 40
770 950 205 950 100 67 93
33 11 8 3 295 31 165
13 69 17 27 18 6 8
39 45 22 12
Number of observations = 71
FIGURE 5-2. STEM-AND-LEAF DISPLAY, RESIDENTIAL LOADING RATESt
(LBS/CURB-MILE-DAY)
loading rate x 0.1
27** 0 Milwaukee
9 55
8 6
7 Number of Observations = 71
95
6 0 Resultstt: upper hinge: 205(57.80
kg/curb-kin-day)
median: 70(19.74
4 01 kg/curb-kin-day)
97 lower hinge: 27(7.61)
69 spread: 178(46.52)
2 1230
756
1 2423400
8696797879696
0** 131223323422134210031121003421
tURS, 1974.
ttmedian: the data value half-way in from either end; half the
data lie above or below this data point.
hinge: upper (lower)--half-way from the high(low) data extreme
to the median, i.e., three—quarters (or one—quarter) of the data
are below this point.
spread: difference between upper and lower hinges, i.e., con-
tains one—half of the data.
18
-------
analysis without data transformation. Specifically, the shape
is suggestive of a log-normal distribution (see below).
2) If we calculate the median of the data, it is 70 (19.74). URS
reported a mean of 149 (42.01), revealing the skew of the data.
(As a comparison, the geometric mean —- essentially a log trans-
formation of the data -- gives a value of 84 (23.68)). Reporting
only this arithmetic mean plus a standard deviation of 195
(54.98) is misleading; it says that about 67 percent of the data
was between 149 and 195 (42.01 and 54.98). If fact, the true mean
is even larger, since URS threw out the highest and lowest values
in each group of data. Using the full data, the mean is 201.3
and the standard deviation is 379.7 (56.75 and 107.05).
We can see that the mean does a very poor job of summarizing the
c’ nter of the data. We can also calculate the upper and lower
“quarters” of the data to get an idea of how it is spread out,
rather than use the standard deviation. If we calculate these
hinges (see Figure 5—3), one—half of all data falls within this
range. In general, this kind of analysis by medians is much less
sensitive (i.e., the analysis is robust) to extreme fluctuations
in data.
Most importantly, however, the data look log-normally distributed. If
we take logarithms of the loading rates, the resulting display (Figure 5-3)
seems much more well—behaved. First, it is symmetric (mean 4.32 (3.05),
close to median 4.24 (2.98)) and even somewhat “normal-looking;” second, the
data do not appear so “wild” except for the prominently high value of 7.9
(6.63) from Milwaukee. Now we can use more confidently the mean of 4.32
(3.05) and standard deviation of 1.397 (0.13) as summary figures for solids
FIGURE 5-3. LN (RESIDENTIAL LOADING RATES)T
(LBS/CURB-MILE-DAY)
7* 9
7
6* 56788
6 034
5* 56999
5 0113334
4* 55556677999
4 012222334
3* 556678
3 00112344444 (ln scale)
2* 5888999 Results: upper hinge: 5.32(4.05)
2 0034 median: 4.25(2.98)
1* 7 lower hinge: 3.29(2.02)
1 0 spread: 2.03(0.76)
mean: 4.32(3.05)
standard
deviation: 1.397(0.13)
tURS, 1974.
19
-------
loading rates. What is more, the log-normal distribution suggests an under-
lying model of the accumulation process. The standard explanation is this:
suppose the increase in loading, i x, is some multiple, k, of the increase in
a large number of other variables (z), about most of which we don’t know,
i.e., E x kz x; and suppose the z. ’s are random variables. Then the sum of
the z’s is th sum of these change in x:
r 1 jxdx 1 x
z=L z. —=—log—
1 kx 1 x k
The random variable z, being the sum of many random variables, is, by the
central limit theorem, normally distributed; and the loading x., log-normally
distributed. Therefore, if the observed loading is the multiplicative effect
of many factors, we will see the kind of pattern displayed in Figure 5-3.
One question that should be asked at this point is whether this effect
is found with other constituents. A problem is that often there are too few
measurements for materials such as heavy metals to reveal a distinct pattern
to their distribution. When there is a sufficient sample size, the log-nor-
mal distribution seems to appear. An excellent example is COD (Figure 5—4).
Mean COD loading is 90936 micrograms/gram/day, but this value is greater
than almost three—quarters of the data (the upper hinge is 91700) and far
larger than the median of 46000 (the UPS mean is 82000, differing because of
the removal of extreme values by UPS). After a log—normal transformation
the mean is 10.97; the median, 10.74. The transformation does a good job of
making the data symmetric.
FIGURE 5-4. RESIDENTIAL COD MICROGRAMS/GRIU4 (UPS, 1974)
COD x .001
5 1
4**
3 1
2** 9700
.2
18
16
.8
14
12* upper hinge = 91700
10 median = 46000
11 lower hinge = 30000
8*
2 mean = 90935.9
6 5118
308 standard
4* 560 deviation 105749.
24244082
2 9441064
88
0*
20
-------
To summarize, detailed examination of the distribution of loadings for
the constituents COD, Pb, cadmium, organic nitrogen, orthophosphate (O-PO 4 ),
and nitrate reveals generally.a log-normal pattern. (See figures in Appendix A
for representative stem-and-leaf displays). Exceptions are usually found
where the sample size is small (less than 20); this is to be expected from
statistical considerations. In most cases log—transformations do a good job
of making the data more symmetrical. We can conclude from this that the re-
ported mean values and standard deviations are not good values to use for
summarizing loading rates. use of log-transforms, along with medians, is
suggested. Table 5-1 gives a summary of the findings.
Residential
Suspended
Solids (ibs/day)
Residential
COD
Residential
0-Pa
Residential
Cd
Residential
Pb
Residential
NO 3
Residential
Organic N
Commercial
COD
Commercial
O-P0 4
Commercial
Pb
log—normal
log transform is approp-
riate
looks somewhat uniform,
not normal after log tran9—
form(it is symmetric)
log transform is approp-
riate
log transform is approp-
riate
in transform is approp-
riate (lstrayvaiue =
Lawrence, Mass.)
skewed opposite from all
others(a few extreme
values)
not normal—looking after
log transform (small num-
ber of cases)
more uniform after log
transform
log transform is approp-
riate
TABLE 5-1. COMPARISON OF RESILIENT VS. UPS RESULTS,
URBAN STREET LOADINGS
(MICROGRAMS/GRAM UNLESS OTHERWISE NOTED)
Material
Median
Value
UPS
I Reported
Mean Mean*
Number
of Cases
Comments
on
Distribution
201
90935.9
3 .05
1779
70
46000
635
2.7
1100
544
1950
77000
14.88
3312
149 71
82000 39
936 42
3.0 50
1484.66 53
550 20
1880 26
269000 16
22.50 15
3400 17
1
*UPS mean different because extreme values are not included.
21
-------
TABLE 5-1 (CONTINUED)
Material
Median
Value
Mean
Reported
Mean*
j Number
L0f Cases
on
Distribution
Commercial 1259 158 14 symmetrical after in trans-
NO 3 form
All Indus- 73000 88000 U symmetrical, but not normal-
trial COD looking
All Indus-
1400 1250 11 very scattered
trial O-P0 4
All Indus— log transform not quite
1100 1160 24 normal—looking (square
trial Pb
root transform better)
very scattered distribu-
All Indus-
trial NO 410 246 7 tion (small number of
3 cases)
* p g mean different because extreme values are not included.
ANALYSIS BY TWO-WAY TABLES
The model described above assumes that the processes that account for
the observed loading on streets are numerous and otherwise unanalyzable, that
is, a black box. To continue the exploratory analysis, we specify a set of
factors that we presume contribute to the final loading level. Other resear-
chers have attempted to specify such categories as land use or perhaps traf-
fic density as intuitively plausible factors; we would expect industrial and
commercial land to have patterns of accumulation that differ from low—density
residential. Nevertheless, it may well be that other factors, including a
generally high “residual” variability, may mask these effects. This was
found to be the case in the URS study: high variability within categories --
as measured via t-tests —— hampered the attempt to distinguish differences
in accumulation among land use categories. As we have seen, this is due in
part to the use of raw data; log transformed values do not behave quite so
badly. Another approach is to use medians instead of a least squares analy-
sis because medians are insensitive to extreme fluctuations in the data. To
perform the analysis, we disaggregate the data into a model of the form:
LOADING = common value + land use effect
(over all cate— + traffic effect
gories) + climate effect
+ residual
22
-------
where the residual is a normally distributed source of error. We are in
effect making the categories into “dummy variables.” 2
Appendix A describes all the steps to be performed in applying this tech-
nique. The results are presented in Tables 5-2 to 5-5. The numbers left in
the center cells are that part of the loading not accounted for by climate,
land use, or overall value —— the residuals.
TABLE 5-2. TWO-WAY TABLE ANALYSIS, LOG MEDIAN SUSPENDED SOLIDS LOADING,
LAND USE VS. CLIMATIC REGION, LBS/CURB-MILE/DAY (URS, 1974)
Original Data Region Row
Land Use Northeast Southeast Northwest Southwest Median
residential 2.13 1.70 1.59 1.43 1.65
commercial 2.01 1.61 1.20 1.61 1.61
light
industry 2.13 2.02 1.77 1.93 1.98
industry 2.71 1.64 1.59 1.87 1.76
Common Value+
The Two—Way Fit: Loading = Land Use (row) Effect + Region (column)
Effect + Residual
Row
Effect
residential 0.01 —0.01 0.13 —0.25 —0.08
commercial —0.01 0.01 —0.15 0.04 —0.18
light
industry —0.29 0.02 0.02 —0.04 0.22
industry 0.43 -0.22 —0.02 0.04 0.07
Column Effect 0.44 0.03 —0.23 —0.01 1.76
(common value)
2
The classic statistical approach to fitting a linear regression model to
observed data is to use a “least squares” criterion for the fit. If we dis-
played the data in categories (as in Tables A-i and A-2) we could follow this
technique by subtracting out the row and column means for each category, and
by arriving at a formula such as
fit = grand mean + row effect + column effect + residual,
it can be formally shown that the linear decomposition in a model of this
kind is equivalent to a least squares analysis. However, least squares is
ineffective with highly variable data because, as_the least squares derivation
shows, points farther away from the grand mean (X)contribute more to the place-
ment of the least squares line than points closer to the mean (the exact fac—
tor is f(X —X)/ x, where is the particular observation and X the mean).
But why should “untypical” points contribute more to the direction of the
line? Shouldn’t it rather be the “typical” points that determine the line?
In the face of these arguments it has been suggested that the medians be sub-
tracted out instead of the means, thus providing an inherently “typical”
value, not one influenced by extreme values. This gives us Tables 5-2 and
5—3.
23
-------
TABLE 5-3. TWO-WAY TABLE ANALYSIS, LOG MEDIAN SUSPENDED SOLIDS LOADING
(LBS/CURB-MILE/DAY) CLIMATIC REGION VERSUS TRAFFIC DENSITY
Original Data Traffic Density (ADT)
Climate <500 500—5000 5000—15000 >15000/day
Northeast 2.55 2.08 2.16 2.32
Southeast 1.76 1.83 1.66 1.23
Southwest 1.15 1.72 1.41 1.55
Northwest absent 1.34 1.62 1.20
Row
The Two-Way Fit
effect
Northeast 0.12 0.27 —0.12 0.34 0.75
Southeast —0.03 0.12 0.02 —0.11 0.11
Southwest —0.47 0.12 —0.12 0.32 0
Northwest absent —0.23 0.12 0 —0.03
Column Effect 0.12 0.04 —0.03 —0.33 1.56
(common
value)
The “common value” is an overall measure of the load. The “effects”
are what they say they are —— in the case of land use and region, the ef-
fect of being in a particular land use or climatic category. For example,
in Table 5-2 being in the Northeast adds 0.440 to the common value of 1.76;
being residential adds -0.08 (the URS”overall” value is 156 (43.98) —- too
high compared to the model above (anti-log 1.76 = 57.5)). Notice that this
model is linear in a log scale, hence actually multiplicative.
The basic results are these:
land use ersus climatic region; suspended solids loading (Table 5-2):
• Land use effects
— Commercial areas are somewhat lower in loading rates; light
industrial is somewhat higher. (There are few observations
for the “light industry” category, so this latter result
must be viewed with some caution).
• Climatic region effects
- In general, climate has larger effects on loadings than does
land use.
— The Northeast has a substantial positive effect; the North-
west, a negative one. The other regions have almost neglig-
ible effect. One explanation is that the aged infrastruc-
ture in the Northeast contributes to increased loadings,
whereas the long rainy periods (or different vegetation?)
in the Northwest prevent large build—ups; however, these
are just speculations, particularly since the relationship
between loading and days since last rain is problematic.
24
-------
• Residuals
— There are large residuals for residential Southwest (—0.25),
Northeast light industrial and industrial (-0.29, 0.43), and
Southeast industrial (—0.22) —— as large as any land use or
climate effects. These all suggest details unaccounted for
by the gross categories of land use and climatic region,
particularly for the Northeast. Are these local disturbances?
• The simple model does a better job than the URS classification
method: the residential Northeast mean given by URS method is
291 (82.04) , whereas the reported value is 197 (55.54) . That
is an error of 48 percent. The two-way table method predicts
a median value of 132 (37.22) compared to an actual value of
135 (38.06) —- an error of two percent. Of course, in an area
with large residuals, the Northeast, neither method can be
expected to do well.
• Examination of residuals in a stem-and-leaf display (Figure 5—5)
indicates that the median is 0, as it should be. The residuals
look about normal with some high and low values as noted. A diag-
nostic plot of residuals versus (row effectxcolumneffect)/common
value (Figure 5-6) is used to reveal any systematic trend in the
residuals that would indicate an interaction between land use and
climatic region. (The additive, linear form of the table assumes
there is no such interaction; i.e., the effects of land use are
independent of the effects of climatic region. The diagnostic
plot is designed to test for the interaction) In this case
there is no clear pattern to the residuals. However, we note
the large residuals for industrial and light industrial north-
east areas.
Traffic density versus climate, land use, and landscaping beyond
the sidewalk, respectively, for solids loading (Figures 5-3, 5-4,
5—5)
• Traffic effect
In general, in all three tables increasing traffic from the
lowest to the highest category lowers the accumulation up to
.3 to .5 log units. This effect is more notable for the
traffic versus region fit. Weassumethat the result is due
to scouring caused by larger volumes of traffic. (Figure
5-7 plots this effect versus traffic volume) To test this
directly, least squares regressions of average daily traffic
volume versus accumulation were calculated; however, in no
case was the R 2 value greater than 0.15. It appears that
the information is only sufficient to do analysis by cate-
gories of traffic volume, as in two-way tables.
• Climate, land use, and landscaping effects
For climatic regions the Northeast has a sizeable positive
effect apparently due to commercial areas being substantially
lower (street sweeping); landscaping matters very little.
25
-------
TABLE 5—4. TWO-WAY FIT, LOG (AVERAGE DAILY MEDIAN SUSPENDED SOLIDS)
(LBs/CuRB-MILE/DAY) LAND USE VS. TRAFFIC VOLUME (ADT)
(URS, 1974)
Original Data Traffic Volume
Land Use ess than gr ater row
500 500—5000 5000-15000 than 15000 median
Residential
Commercia),
All Industrial
2.18 2.08 1.73 1.98
absent 1.49 1.46 1.40
2.44 2.09 1.94 1.26
2.030
1.460
2.015
The Fit Traffic Volume
Land Use less than greater row
500 500—5000 5000-15000 than 15000 median
Residential
Commercial
industrial
—0.097 0.057 —0.143 0.079
absent —0.032 0.088 0.000
0.096 0.000 0.000 0.678
0
-0.501
0.067
column effect 0.321 0.067 —0.083 —0.055 1.956
(common
value)
Residuals x .100
2
5789
O** 000
— 39
—1 4
—2
4 upper hinge = 0.06
n = 11 median = 0.00
lower hinge = 0.06
-6
.7
*
26
-------
TABLE 5-5. TWO-WAY FIT, LOG (MEDIAN SUSPENDED SOLIDS LOADING,
LBS/CURB-MILE/DAY) LANDSCAPING VS. TRAFFIC DENSITY (ADT)
(URS, 1974)
Original Data Landscaping
Traffic Volume landscaped row
.grass trees buildings none median
less than 500
500—5000
5000—15000
‘ 15000
2.217 absent 1.900 absent
1.863 1.505 2.170 1.968
1.613 1.544 1.663 1.778
1.230 1.982 0.062 1.301
2.059
1.916
1.638
1.266
The Fit Landscaping
Traffic Volume landscaped row
grass trees buildings none median
less than 500
500—5000
5000—15000
> 15000
0.143 absent —0.093 absent
—0.008 —0.259 0.380 0.000
—0.038 0.000 0.093 0.030
0.009 0.868 —0.538 —0.017
+0.313
0.110
—0.110
—0.540
column effect 0.015 0.092 0.066 —0.017 1.746
(common
value)
Residuals x 0.001
8** 6
6
4* *
8
2
+0 00930
—0 9031
_2** 5 n = 14 upper hinge = 0.093
4 median = 0.000
3 lower hinge 0.038
_6**
27
-------
FIGURE 5-5. RESIDUALS FROM TWO-WAY FIT OF LAND USE VS. CLIMATIC REGION
Residuals x 100
43
3
2
13
+0 004224
—0 0042
—l 5
—2 592
—3
—4
median = 0.0
FIGURE 5-6. DIAGNOSTIC PLOT, RESIDUALS FROM TWO-WAY FIT
LOG (SUSPENDED SOLIDS, LBS/CURB-MILE/DAY) LAND USE VS. CLIMATIC REGION
X
4 5—36
3-
X
XX
2 xx x
-. X
K X
x
0
U,
X
-2 K
x
K
-3
-4
I I I J
-30 - - 0 C 20
COIo RIS0N LLE x 100
28
-------
- traffic effects from 2—way
0 - table vs. climate
+ traffic effects from 2—way
table vs. land use
TRAFFIC VOLUME, thousonds/day
FIGURE 5-7. TRAFFIC EFFECTS VS. TRAFFIC VOLUME
• Residuals
— There are large residuals remaining even after traffic and
climate, land use, or landscaping effects are subtracted
out. This indicates substantial “unexplained” variation
remaining in some categories, though in some cases (e.g., traf-
fic versus land use) this is likely because of the small nuxn—
ber of sample values in a category. Residual plots are
qiven following the tables which show generally random scat-
ter, indicating proper linear disaggregation of the row
and column effects.
Instead of an in—depth discussion of other two—way tables, a summary of
the results is given in Table 5—6. The reader can refer to the tables and
diagnostic plots in Appendix A for details.
Briefly, we can see that:
• the northeast and southwest
tive or negative effects--
Generally the Southwest has
regions fluctuate —— either with posi-
more than any other climatic region.
moderately higher loadings.
traffic effects from 2—way
- table vs. landscaping
+
— a_a aS —
0
‘C
I-
0
liJ
LL
IL
ILl
0
Li
La
it:
F-
-3
-4
-5
6
16 18 20 22
24 26
29
-------
TABLE 5-6. SUMMARY OF TWO-WAY TABLE RESULTS
(ANALYSIS BY LOG OR LN (MEDIAN CONSTITUENT))
Material
Effect
Table
#
Effects: Comments
Residuals:
Comments
Diag.
Plot
Solids
LS*
Climatic
Region
A-5..
“trees” much lower
NE* higher
interaction be—
tween climate/
LS beyond side-
walk
A-15
COD
Land Use
LS
A—6
commercial lower
LS little effect
interaction
A-l6
Climate
Land Use
A-7
industrial lower
residential-corn-
mercial same
SW higher
some large resi—
duals
A-17
Lead
Traffic
Land Use
A-B
traffic effect
clear
some inter-
action
A-lB
Lead
Land Use
Climate
A-9
industrial lower
NE* lower, SW
higher
interaction
A-l9
Lead
Land Use
Traffic
A-b
scouring at high
volumes; deposit
at other volumes
commercial much
higher
no interaction
A-20
*LS =
Landscaping;
NE = Northeast; SW = Southwest
• residential land uses do not contribute much to loadings, but
industrial and commercial uses do have some sizeable effects.
Lead and COD are higher in commercial areas; lower in indust-
rial areas. Solids are lower in commercial areas; higher in
industrial areas.
— It is interesting that the lead and COD land use effects are
different in sign from the solids land use effects. This
suggests that solids deposition differs perhaps significantly
from those processes that specifically accumulate COD or lead.
The implication is that models that rely on solids accumula-
estimates alone cannot be expected to do a reasonable job of
estimating particular pollutants like BOD or metals that do
not deposit or erode as solids do.
30
-------
• lead deposition by traffic follows an intuitively satisfying
pattern: deposit at low-medium volumes; scouring at highest
volumes.
• There are interactions between land use, climate, and land-
scaping for the majority of the tables. In the case of
landscaping this interaction is possibly the result of as-
sociation between vegetation types and climate types —— for
example, more trees in the Northeast.
ANALYSIS OF SAN JOSE STREET CLEANING DATA
One of the most recent studies of street surface accumulation is that done
in San Jose, California by Woodward-Clyde Consultants in 1977. Importantly,
this report contains a series of before—cleaning/after—cleaning samples done
in the same locations under carefully controlled conditions. The number of
samples taken was based on an initial estimate of.the variability of the
loading to establish a uniform confidence estimate for the figures obtained.
The reporting is particularly complete, giving particle size distributions
(in percentage of total and absolute amount) for each sample. This is useful,
since some models (for example, that of Sutherland and McCuen, 1976) posit
washoff varying with particle size. Further, the proportion of particular
contaminants (e.g., BOD) seem to vary with particle size (See Table 5-7).
Validation of the Accumulation Model
The San Jose data can be used to test the validity of the accumulation
model presented above. There we stated that solids accumulation could be
fitted as:
load = overall load + row effect + column effect + residual
where “row effect” and “column effect” correspond to our knowledge of, for
example, what land use and climate category the chosen site is in, and “resi-
dual” is a randomly distributed remaining error.
TABLE 5-7. FRACTION OF POLLUTANT ASSOCIATED WITH EACH PARTICLE SIZE RANGE
(% BY WEIGHT)*
MICRONS
Total Solids
>2000
840—2000
246—840
104—246
43—104
>43
24.4
7.6
24.6
27.8
9.7
5J9
Volatile Solids
11.0
17.4
12.0
16.1
17.9
25.6
BOD 5
COD
7.4
2.4
20.1
4.5
15.7
13.0
15.2
12.4
17.3
45.0
24.3
22.7
Kjeldahl Nitrogen
Nitrates
9.9
8.6
11.6
6.5
20.0
7.9
20.2
16.7
79.6
26.4
18.7
31.9
Phosphates
0
0.9
6.9
6.4
29.6
56.2
*Source: Sartor, J.D., and G.E. Boyd, Water Pollution Aspects of Street
Surface Contaminants , Report prepared for U.S. EPA, EPA-R2-72-081, Washing-
ton, D.C., 1972.
31
-------
If we take a fixed street location in the San Jose test area (“Tropicana”
Street), we can presumably fix the row and column effects; we further control
for possible effects of accumulation time by looking only at a single day
following a street cleaning. Therefore, the remaining observed variation in
accumulation should be due solely to what we have termed “residual error.”
If we plot these residual loading rates in a stem—and-leaf (Figure 5-8 for
particles less than 600 microns), we see that there is a large remaining
variation (median = 44 lbs (12.3 Kg) spread = 57 lbs (15.9 Kg)). Even with
presumably all effects of location and time of accumulation taken into ac-
count, the remaining spread of loads is quite large. In particular, there
are several occasions when measured load on the street is lower the day after
a sweep (negative loading rates in the figure). This surprising result could
be caused by wind or traffic scouring; it is quite large in some instances.
Further, the distribution is roughly of the shape expected by our model.
If we look now at the larger particles in a so—called box—and-whiskers
plot (see Figure 5-9, particles greater than 600 microns), we find generally
lower loading rates (median 19 (15.32)) and less variation. In the figure
the upper and lower “tails” represent the maximum and minimum data values.
Between them is a rectangular box whose top is the upper hinge and whose
bottom is the lower hinge of the data. The line inside the box is the
median value. One can see that the display is designed to give a quick
visual idea of the spread of the data, since about 50 percent of the data
values lie inside the box. If the data are not symmetric, the median is like-
wise not centered in its box. We suspect the cause to be that larger parti-
cles are less susceptible to variation caused by wind or traffic, hence have
a “tighter” distribution. If we study the size distribution of particles,
we see that cleaning removes a disproportionate share of large particles in
part by grinding the larger particles into smaller ones.
Clearly the inability of the sweepers to pick up smaller particles also
plays a role in these effects. There are several other cases in the San Jose
data where the day following a cleaning shows an increased rate of large
particle deposition relative to small sizes; this makes sense if one looks
at a simple rate equation by particle class -— there is more room for increase
in the large particle class if the sweeper has differentially removed mate-
rial from that class.
What of the effects of days of accumulation before a street cleaning?
Figure 5-9 compares the distribution of loadings after one day and after four
or more days of accumulation. The figure tells us that the substantial varia-
tion in loadings found in the smaller particle size range for less than one
day accumulation is also present in the four— or more—day loadings (median
seven days). Likewise, the smaller variability in the loading of larger par-
ticles is found even after four days of accumulation. Comparing the median
load after one and after seven days, we see that for particles less than 600
microns, after less than one day loads have reached 68 percent of their seven-
day value; the corresponding figure for larger particles is 59 percent. For
the generally used exponential accumulation model, L = L 0 (l_e t), the first
result implies a “k” value of 1.14, which means that 90 percent of the ulti-
mate load is attained after only about two days. The conclusion is an
32
-------
important one for certain control measures such as street sweeping; for exam-
ple, in order to reduce the average material available for washoff to 50 per-
cent of its maximum value, one would have to sweep at intervals of under two
days.
FIGURE 5-8. TROPICANA STREET LOADING RATES, LBS/CURB-MILE-DAY
(PARTICLE SIZE LESS THAN 600 MICRONS)
(219,204) unit = lbs 15 2
15 14
14 9 13
13 12 9
12 11
11 10 5
10 9 6
9 8 635
84 77
7 8 6 05
6 00 5 911
5 4333 4 2
4 No. Observations = 23 3 1 No. Observa-
3 169 2 tions = 17
2 14
1 8 60 = upper hinge = 96 0
0 24303 44 = median = 65 —i
3 = lower hinge = 42 -2
—2 _3 5
—3 3 —4
less than 1 day greater than 4 days
5 2 accumulation accumulation
FIGURE 5-9. LOADING RATES VS. DAY SINCE LAST RAIN OR SWEPT, TROPICANA
STREET AREA, SAN JOSE, 1977
200 -
150..
c i )
•r1 100 —
5O - F __ H __
10_ti II I
less than greater than less than qreater than
600 miciuns 600 microns 600 microns 600 microns
< 1 day > 4 days (7 days median)
33
-------
A second important conclusion to draw from Figure 5—9 concerns the large
variability of loadings. Any control measures must consider that after only
one day in the Tropicana Street location, one-half of the small particles
loadings were between 60 and 3 lbs/curb-mile (16.80 and 0.84 kg/curb-kin);
after 7 days the spread is 54 (15.12). Therefore, the efficacy of any con-
trol program is subject to wirxd fluctuations; it appears that perhaps due to
wind and traffic, street cleaning is inherently an uncertain kind of control,
one that is not robust. What this means is that in order to achieve a vir-
tually certain level of ‘ollutant reduction on the street (for example, 90
percent sure), it would be necessary to raise the median cleaning level to
possibly even a daily regime.
Mathematical Modeling of the Accumulation Data--
The basic loading rate equation is:
=k 1 -k 2 L (5-1)
where k = accumulation constant
1
= erosion constant
L = loading
t = time
If k 1 and k 2 are generalized, we obtain:
dL/dt = f(t) - f(L) + random component (5-2)
Suppose the deposition is non-linear, say a/(l+bt) 2 . so that possible deposi-
tion diminishes with time. There is no analytic solution to this differen-
tial equation (there is an integral equation solution), but a series solution
expanded to two terms gives
L = —a/(l+bt) + ac/b 2 (ln(l+bt) exp — (ct))
+ c(1+bt) exp — (ct)
+ c 2 (l+bt) 2 exp — (ct)
+ (a/b+c+c 2 /4). (5-3)
The load approaches its asymptotic value quickly for a selection of a,
b, and c, corresponding to the “standard” differential equation (5-1). Fig-
ure 5-10 shows a graph with some typical a, b, c values -- the asymptotic
limits are just fictional. It can be quickly seen that this more general
equation can cover the simpler case, equation 5—1, as well as the empirically
derived curves of Sutherland and McCuen (1976), Figure 5-11:
load (industrial) = 1388(1 — exp(—.19t))
load (commercial) = 500(1 — exp(—.535t))
load (residential)= 1089t(1+l.3t)).
34
-------
FIGURE 5-10. ACCUMULATION LOr D VS. TIME; “COMPLEX” DIFFERENTIAL EQUATION
V.
C
0
.0
0
I
c
0
-J
FIGURE 5-11.
SUTHERLAND AND McCUEN EMPIRICAL CURVES
TOTAL
SOLIDS
LBS.
PER
CURB-MILE
4 5 6 7 8 9 10
DAYS OF ACCUMULATION
35
-------
Further, as is shown in the analysis of the San Jose street cleaning data,
the fast rise time to a saturation level after two to three days is generally
confirmed by data from the “Tropicana” and “Keyes” Streets test sites.
The Confusion About Estimated Days of Accumulation (EDA)--
Given that the simplest differential equation for loading is of the form
load = function (time since last washoff), it is natural that investigators
have tried to confirm this empirically. There have been contradictory re-
sults. Coiston, Hammer, URS, and Whipple, among others, found no relation-
ship between EDA -- number of days since last washoff -- and the loading.
On the other hand, Sutherland and McCuen (1976) reported the opposite.
Least squares regressions of solids and lead versus “days since last
rain” using the URS data gave extremely small R 2 values (less than 0.1). It
is concluded that not much of the observed variation in loading is due to the
accumulation as measured in days. See also Figure 5—12 (this pattern corres-
ponds closely to the form found by Whipple, 1977).
In the case of the URS data the negative finding is not surprising,
since most samples of different EDA’s are from different cities: Baltimore,
with nine data points, all 26 days after the last washoff; Atlanta, all2 days,
and so on. Within one city there are few different EDA’s; thus to expect
the EDA to account for loading differences between cities is to believe all
the data to be from one “city,” but this is not the case. Further, the San
Jose data show a rise time in one to three days to a loading “saturation.”
To capture this fluctuation, sampling would have to be done at intervals much
shorter than a single day; that is, hours.
FIGURE 5-12. LOADING VS. DAYS SINCE LAST RAIN (uRS, 1974)
3000
U i
-J
2000
-1000 -
2
S
I I
1 2 x 5 i 3
I . ’- I 1
0 20 40 60
DAYS SINCE LAST RAIN
36
-------
Conclusions
Implications for Modeling Accumulation--
Observations Modeling Implication
1) fast rise time to saturated 1) use non-linear accumulation
loading equation
2) deposition and pollutant corn- 2) disaggregate deposition depend-
position varies with particle ing on severity of variation
size
3) a. large variability in error 3) incorporate variance estimate
term of statistical esti— into effectiveness of control
mate of load; measures like street cleaning
b. production of small par-
ticles by street cleaning
4) conflicting reported effects 4) smaller sampling intervals (less
of “days of accumulation” than 1—2 days) required only if
necessary to capture the effect
of accumulation time
37
-------
? NALYSIS OF WASHOFF: THE ENVIREX-DOT DATA
To study the changes in pollutant concentrations at different stages of
storm runoff flow, it is necessary to find data that: 1) measure different
pollutant parameters over the evolution of storm; and 2) measure storm run-
off independently of stream flow and sanitary sewer flow. Such datahave been
collected in a recent study by the Envirex Corporation for the Department of
Transportation, and that portion of the data released in time for evaluation
in this study (i.e., 12 rainfall events in 3 different catchments) is the
basis of this section of our report. Table 5—8 displays some relevant chara-
teristicsof these catchments.
TABLE 5-8. DOT CATCHMENT DESCRIPTION
Catchment
No.
Rainfall
Events
Runoff Area
ADT*
Description
of Area
Harrisburg Catchinent
7
18.5 acres
24000
paved/unpaved
Milwaukee Catchinent A
(Highway 45)
4
106 acres
85000
31% imper-
vious
Milwaukee Catchment B
(Highway 794)
1
2.1 acres
52000
completely
impervious
*Average Daily Traffic
Volume.
The DOT data include observations of precipitation, flow, suspended
solids, total organic carbon, and pH for each storm. Most storms have obser-
vations of other parameters, including Cd, Cu, Cr; Fe, Pb, C1, NO 2 , NO 3 ,
TKN, P0 4 , BOD 5 , COD, and oil and grease.
Regression analysis was used to search for equations which appear to
explain the level of pollutants as a function of runoff flow volume. The
analysis focused on the suspended solids parameter in recognition of its
general physical properties and the relatively complete information avail-
able, but other parameters were examined as well. Analysis was undertaken
for individual storms, for aggregates of storms in each catchment, and for an
aggregate of data over all catcbments.
In the analysis of individual storms the relationship of suspended
solids to flow was studied by comparing cumulative suspended solids to cumu-
lative flow. Figures 5-13 and 5-14 show this relationship for each storm.
Simple observation of these plots suggest that 1) a first flush effect is
found in each storm, and 2) cumulative suspended solids decay exponentially
relative to cumulative flow. (See Appendix A for a discussion of the tech-
nique used to develop data for these studies.)
38
-------
FIGURE 5-13.
U,
U)
-j
0
C ,)
LL 3
z
1
0
U)
U)
tii
>
I-
-3
D
0
CUMULATIVE SUSPENDED SOLIDS VS. CUMULATIVE FLOW
FOR HARRISBURG STORMS
FIGURE 5-14. CUMULATIVE SUSPENDED SOLIDS VS. CUMULATIVE FLOW FOR
MILWAUKEE STORMS (INCLUi)ING THE HARRISBURG STORMS)
CUMULATIVE FLOW (Ft 3 )
Storm 8
Storm 6
St
CUMULATIVE FLOW (Ft 3 )
10,.
8,
Storm
(4ncomplsl, doso)
7
U,
0
‘I)
0
-3
0
(1)
0
w
0
z
U)
D
U)
w
>
I -.
—I
0
Storm 4
Storm S
39
-------
Cumulative Plots of the Harrisburg Storms
Figure 5-13 suggests the existence of an envelope or a characteristic
shape of suspended solids-flow curves whereby storms of a threshold intensity
will rise a Long a common slope before diverging toward their individual hori-
zontal asymptotes. The Milwaukee catchment A plot (Figure 5-14) also shows
this, although two of these storms, events 5 and 7,are missing data in the
latter portions of the flow and consequently do not show the decay effect
observed in the other storms. Figure 5—14 also serves to contrast a charac-
teristicHarrisburg runoff pattern (Harrisburg Storm #8) to the Milwaukee
storms, indicating the degree of variation in the suspended solids—flow
relationship that may be due to differences in catchments. This last obser-
vation is partially explained by the difference in runoff area of the two
catchments (the Harrisburg site is 18.5 acres (7..40 ha), while the Milwaukee
catchinent A is 106 acres (42.9 ha)).
Some summary results of the regression analysis are shown in Table 5-9.
Complete fitted equations are only shown for Harrisburg storms 7 and 2, but
R 2 values for the different regressions are shown for each storm.
The two Harrisburg storms were chosen to represent a range of different types
of storm that can be described by the same regression models ( 4 t 7 had a single
peak flow, while #2 had two periods of high flow). While the curves repre-
senting the storms differ greatly (see Figures 5-15 and 5-16), the regres-
sion coefficients and residuals are different - - both storms exhibit high
R 2 values for several of the fitted models. In other words, while there is
unexplained variation between storms, the within-storm variation is explained
rather well by bivariate regression models.
Generally, the most successful model was of the form suspended solids
= Lo (l—exp (flow*k)), where suspended solids (lbs) and flow (ft 3 ) are cumu-
lative, Lo is a constant representing the potential total amount of suspended
solids that could be washed off, and k is an ex irica1ly derived constant.
This so-called “exponential” model is related to the functional form used by
the SWMM computer program (see discussion below). Coefficients for the
model were found using non-linear regression techniques, but a variation
of the “method of moments,” developed by Moore, Thomas, and Snow (1950) for
analyzing BOD data can be en!ployed as a manual tool for deriving the model
coefficients.
TABLE 5-9. REGRESSION FITS FOR HARRISBURG STORMS
Harrisburg
Storm No. 7
ss
= —55.89 + 40.49 * log 10 (flow)
R 2
=
.889
ss
= —15.22 + 6.68 * flow 3 Z6 )
R 2
=
.989
ss
= 111.00 * (1. —exp(—flow * .000563))
R 2
.995
Note:
ss
is cumulative
suspended solids and flow is
cumulative flow.
40
-------
TABLE 5-9 (CONTINUED)
Harrisburg Storm No. 2
ss = —205.16 + 69.1 * log 10 (flow) R 2 .865
ss = —15.22 + 6.68 * f1ow( 9 0 9 ) R 2 .968
ss = 163.19 * (1. —exp(flow * .000059)) R 2 .964
All Storms -- R 2 values
Hi
H2
H3
H5
H6
H7
H8
MA5
MA6 I MA7
MA8
MB1
Linear—log model
Non—linear model
Exponential model
.980
.983
.976
.865
.968
.964
.982
.988
.995
.562
.902
.910
.912
.987
.993
.884
.989
.995
.973
.997
.961
.788
.977
.995
.770Th i
.997 .995
.994 .991
.760
.999
.997
.988
.994
.994
Note: ss is cumulative suspended solids and flow is cumulative flow.
ss = pounds
flow= cubic feet
FIGURE 5-15. CUMULATIVE SUSPENDED SOLIDS VS. CUMULATIVE FLOW
HARRISBURG STORM #7
100
d Z/: —7-’
U)
a 80 7/ /‘
w 7/ /
o ‘/1/
U i
cL 60 / 1/,’
(I ) / P,’
D 1 1/
If
40 l — Observed
o Linear- log model
v P4rs -hneov model
20 7
______ o ExponentIal model
2000 4000 6000 8000 K).000
FLOW (ftx)
41
-------
FIGURE 5-16. CUMULATIVE SUSPENDED SOLIDS VS. CUMULATIVE FLOW
HARRISBURG STORM #2
(A
.0
o 80
/ .__ 4 _g , 0
60
w 40 - ,o — — — z Observed
o Lmear - log model
- v Non-linear model
20 —-—-- o Eeponentiol model
C)
. 4 . ___ O j I ii
2000 4000 6000 8000 K),000 12,000
CUMULATIVE FLCW (ft 3 )
Examination of Residuals
Even though the R 2 values for the exponential model are quite good, it
is worthwhile to examine where the fitted values differ from those observed.
Figures 5—17 and 5—18 display the residuals remaining in the Harrisburg
Storms #7 and 4 2 after an exponential fit. Both fits show observable trends
in the residuals, indicating that the exponential model consistently over—
predicts suspended solids in the early portions of storms. Harrisburg
Storm #2, in particular, shows a sinusoidal pattern of residuals: the model
does well in predicting the middle portion of the storm, but not the begin-
ning or end (Figures 5—15 and 5—16 also tell the same story). There are
several conclusions that might be drawn:
1) Even though the R 2 values (the overall fits) are quite good,
this is misleading. If one is interested in predicting first
flush effects or total final solids, then the actual fit is
not quite so good. The implications for computer modeling of
storm events is discussed below.
2) The unexplained variation in the early sections of storms
suggests a more complicated relationship between solids
washed off and flow than that presented here.
3) The analysis of residuals is recognized as limited by the
data. All of the storms had more observations in the later
(higher cumulative flow) portion because measurements were
divided into equal time, rather than flow time increments.
42
-------
FIGURE 5-17. RESIDUALS FROM FIT, HARRISBURG STORM #7
6
x
4-
x
x
2 X
X
I I
I I I
20 40 60 80 100 120
FITTED VALUE
-2 FOR EXPONENTIAL MODEL
-4
-6
x
FIGURE 5-18. RESIDUALS FROM FIT, HARRISBURG STORM #2
10 x
8
x
6
4
32
x
C I — I I I I I
10 20 30 40 50 60 70 80 90
W FFTTED \ALUE FOR EXPONENTIAL
MODEL
K
43
-------
When the data were aggregated within a catchinent the level of explained
variation was less exceptional, but still rather high, as shown in Table 5-10.
The simple cumulative flow term serves as a proxy for more complex effects in
each storm, but it is much less able to do so for the aggregated data. Ex-
ploratory correlations were made between suspended solids and total flow,
average flow, peak flow, and mean squared flow (see Table 5—10). An examina-
tion of average suspended solids concentration versus peak precipitation (see
Figure 5-19) showed some degree of correlation and visibly different patterns
for the different catchments. By including a mean square flow intensity term
in the exponential models for aggregate data regressions, the level of unex-
plained variation was reduced: the R 2 value for the Harrisburg model in-
creased from .933 to .972 while the R 2 value for the Milwaukee model increased
from . 65 to .968.
TABLE 5-10. REGRESSION ANALYSES
CUMULATIVE SUSPENDED SOLIDS VS. CUMULATIVE FLOW
Aggregate Models
Harrisburg
(63)* ss = 9.61 + .0118 * flow R 2 = .900
ss = —534.826 + 195.79 * log 10 (flow) .515
ss = —32.60 + .202 * flow .919
ss = 1157.23 * (1 —exp(—flow * 1.554 E—5)) .933
ss = 984.316* (1 _exp(_flow* 1.23 E—5)) +8.50* .972
Milwaukee A
(48)* ss = 900.169 + .033 * flow .914
ss = —8131.91 + 2646.84 * log 10 (flow) .739
ss = —355.347 + 6.440 * flow .976
ss = 11066.5* (l_exp(_flow*6.086E_6)) .965
ss = 10507.8* (1_exp(_flow*6.29E_6)) +l.826*MS .968
All Storms
(l24)* ss = 76.996 + .0366 * flow .895
ss = —5677.6 + 1833.94 * log 10 (flow) .508
ss = —247.37 + .531 * flow .912
ss = 16819.8 * (1 -exp(—flow * 2.939 E—6)) .915
ss = 15179.8 * (l_exp(_flow*3.l3E_6)).I 2.944*MS .919
ss = cumulative suspended solids in pounds
flow= cumulative flow in cubic feet
— . ( cubic feet
MS — mean squared flow in ( mm J averaged over storm
*N ther of observations.
Aggregating the data from all these catchinents introduced another ele-
ment of variation, which was expected after observing the distinctly differ-
ent solids—flow relationships for each catchinent (Figures 5—13 and 5-18).
The regression results are shown in Table 5—10, and while the relatively high
R 2 suggests a good fit, the substantially larger flows in Milwaukee catch—
ment A have allowed that data to dominate the regression.
44
-------
FIGURE 5-19. SUSPENDED SOLIDS VS. RAINFALL INTENSITY, MILWAUKEE AND
HARRISBURG STORMS
z
Q
a:
I—.O6
z
I
I __
/
2:
Ui
a-
(I)
/
(I)
Lu
a:
Lu
>
3 4 5 6 7 B
PEAK PREC P TATION (rn/hr)
Other Parameters
Similar ana1y es were performed for ol ier pollutant parameters, though
less extensive data limited aggregation or comparisons between storms. Fig-
ure 5-20 shows selected results for Milwaukee Event Number 4 and shows that
all forms of suspended solids measured have constant or decreasing concentra-
tion with increasing flow (See Appendix A, Figures A-21 and A-22 for the re-
maining plots.) Milwaukee Storm #4 showed a similarity of non-linear regres-
sion coefficients as suspended solids, Fee Cu, Cr, Pb, and Zn all fitted ex-
ponents between .43 and .53 (see Table 5—11). (A representative plot of
zinc versus flow for Harrisburg Storm #7 is given in Appendix A, Figure A-23.)
The similarity of the decay coefficients for different constituents suggests
that all of these heavy metals behave alike in their washoff and deposition
characteristics. If so, it may be possible with more extensive data to de-
velop a single equation (perhaps based on particle size (see Pisano, 1977))
to account for observed amounts of all these metals. In contrast, total
solids and chlorides show increasing concentrations with increasing flow; in
this particular storm this is probably the result of snowmelt.
—— 0 UILWAUKE
X I$ARRISBURGH
x
x
x
45
-------
FIGURE 5-20. CUMULATIVE VSS, SS MILWAUKEE STORM #4
5000 - 0 VSS
ss
U)
U)
/*
4000 -
U)
U)
3000 -
L ii
>
2000 —
1000
I I
20.000 40,000 00,000 00,000 100,000 120,000 140,000
CUMULATIVE EL( (ft 3 )
TABLE 5-11. NON-LINEAR MODELS
MILWAUKEE
FOR
A,
VARIOUS POLLUTION PARAMETERS
STORM #4
SS = —413.19 + 16.494 * flow 9 5 R 2 = .990
= —34.58 + 1.355 * f1ow 503 = .989
Cu = —.0919 + .00525 * f1ow 85 = .996
ca = —8.937 E—4 + 1.591 E—5 * flow 82 ’ = .998
Cr = —.0271 + .00183 * f1ow 699 = .993
Fe = —11.786 + .8180 * f1ow 5 7 = .990
Zn —.2956 + .01394 * flow ” 5319 = .997
Pb = —1.088 + .0932 * f1ow 331 = .993
NO 2 = .0276 + 1.395 E—5 * f1ow 9315 = 999
TKN = .0624 + 2.613 E—5 * f1ow 1 ”° 98 = .990
= 1.261 E—4 + 4.781 E—4 * flow” 7229 = .997
NO 3 .111 + 3.798 E—7 * flow 1.3738 = .998
vss = —63.632 + 2.457 * flow’ 5377 = .987
—- E-n is equivalent to lO’
-- Dependent variables in pounds
-- Flow in ft 3
46
-------
Conclusions: Suspended Solids Fits
1) Cumulative flow accounts for a high proportion -- about 90
percent —— of the variation in suspended solids loading.
2) However, the pattern of consistent over- and under—predic-
tion with increasing flow suggests remaining variation in
solids levels that must be explained by a more complex
model.
IMPLICATIONS FOR STORMWATER MODELING
All of the attempts at regression analysis above (and related work by
Espey and Huston Associates, 1976) are in a sense simple formulations of
proxy models of suspended solids washoff. While it may be objected that
these functional forms predicting solids are not complex, they are like those
used in the state—of—the—art computer models STORM and SWMM. It is there-
fore relevant to verify these functions and to examine the assumptions they
embody.
STORM and SWMM both employ an exponential decay washoff function for
impervious areas: change in L = Lo (1 -exp(kr 1 t 1 )) - Lo(l -exp(-kr 2 t 2 ))
(time 1 to time 2)
where Lo = ultimate loading (potential maximum)
runoff inches/hour, time period 1
= runoff inches/hour, time period 2
= elapsed time from storm start, time 1 (hours or fractions of hours)
t 2 = elapsed time from storm start, time 2
k = decay constant (= default values 4.6 in SWMM, 2 in STORM)
L = washoff
To test this model, data from two Harrisburg storms (#6 and #8) were
converted to the proper units. Then ultimate load values (Lo) were taken
as known in advance. For storm #6 Lo = 610 lbs; for #8, 778 lbs. Notice
that in general these values would not be known with certainty. Using
STORM (k = 2) and SWMM (k = 4.6) equations, predicted loadings were calcu-
lated for the storms. (To convert from ft 3 to inches, flow was apportioned
for STORM by that area of the test basin that was impervious and for SWMM
by the total area.) The results are given in Table 5—12. While both STORM
and SWMM do a good job of predicting the ultimate loading (Lo), this is not
surprising, since the Lo value was assumed given. In fact, for an exponen-
tial equation with a “k” value of 2.0 or more the asymptotic load will be
attained quite rapidly; as can be seen in the table, past a time of 0.75
hours there is little further washoff.
47
-------
TABLE 5-12.
PREDICTED AND OBSERVED SOLIDS LOADS FOR S NM/STORM
AND FITTED EXPONENTIAL MODELS
time, hrs
cumulative
solids
observed SWMM value
STORM value
fitted
exponential
change in solids = 1233.08 (exp
(—1.66861 E—5)t x flow)
R 2 = .981
change in solids = 628.9 18(exp
(—5.527 E—5)t x flow)
R 2 = .389
Note: solids in pounds; flow in cubic feet.
In Harrisburg Storm #6, the SWMM formulation does almost as well as the
fitted exponential model in predicting cumulative solids, but STORM over-
predicts solids. Notice that this was a brief, intense storm which might be
expected to produce a classical exponential washoff.
Harrisburg Storm #6
0
0
0
.33
15.54
9.7
16.0
10.94
.58
271.48
294.4
398.9
276.8
.83
551.15
536.7
589.9
536.7
1.08
591.15
585.7
606.6
601.49
1.33
596.91
601.7
610.0
617.67
2.08
602.81
610.0
610.0
637.83
2.83
607.18
610.0
610.0
654.39
4.08
609.34
610.0
610.0
671.8
7.58
609.96
610.0
610.0
684.58
fitted exponential model:
Harrisburg Storm #8
0
0
0
.25
6.59
3.85
6.19
9.35
.50
132.3
34.9
55.45
41.49
.75
363.8
278.9
397.29
223.62
1.0
419.8
532.3
656.14
361.50
1.25
574.2
751.4
773.81
542.10
1.5
688.4
777.0
777.0
601.67
1.75
714.8
777.0
777.0
610.27
2.0
772.4
777.0
777.0
624.50
fitted exponential model:
48
-------
Further, SWMM and STORM first underpredict (in the time period 1/2 hour
to .75 hours elapsed time), then overpredict solids for Storm #8. The reason
is that this storm exhibits non-uniform changes in slope (see Figure 5-13).
Even if we fit the data directly - - obtaining an “individualized” k- value
for each storm, (see Table 5-13 and last column of Table 5-12) -- there is
much variation, particularly in early periods of the storm, that cannot be
accounted for. This is significant because, given this data, it is the best
fit any exponential-SWMM—type model could hope to have. If we further gene-
ralize the equation and use a fixed k over all storms, as is done in the
SWMM or STORM models, then we will do worse.
TABLE 5-13. RESULTS OF EXPONENTIAL MODEL FOR PREDICTING SUSPENDED
SOLIDS WASHED OFF OVER AN INTERVAL
x2 flow(t—l) x2 flow(t)
Model: ss xl(e -e
where ss is suspended solids in pounds, accumulated up to period t
flow(t) is flow accumulated up to and including period t in ft 3
t is time in minutes
Catchment Event Xl X2 R 2 Remarks
Harrisburg 1 26.21 8.35 E-4 .773 very few (5) data pts.
2 206.795 4.994 E—5 .643 double storm
3 18.045 7.36 E—4 .948
5 22.694 1.77 E—4 .138 time intervals are very
uneven (scattered
showers)
6 1253.92 1.683 E—5 .981
7 99.913 5.49 E—4 .942
8 628.92 5.527 E—5 .389 double storm
Milwaukee A 4 6228.1 1.42 E—5 .917
5 12939.5 4.46 E—6 .884
6 1503.6 6.743 E—6 .894
7 13662.5 5.83 E—6 .904
Milwaukee B 1 99.496 1.33 E4 .92
What conclusions for computer modeling can we draw?
• Use of individualized “k” values
First, if data are available, the variation of “k” values from
catchment—to-catchrflent would seem to require individualized fits.
49
-------
• Different behavior of heavy metals.
This result is further strengthened by the different behavior of
heavy metals from suspended solids. Use of “standard” k values
must be considered carefully; metals do not behave as suspended
solids do.
• Bad predictions of early storm loads.
The exponential model by itself cannot do a good job of pre-
dicting early variations in solids during some storms. Harris-
burg storm #8 shows sections of increasing, then decreasing
second derivative suggestive of a second-order differential
equation unlike those currently used.
• Alternative functional forms.
It is possible that some form of a Michaelis—Menton equation
or a second—order form as discussed with the loading equation
earlier in this section may do a better job of fitting this
observed behavior. Because these latter models have two or
more parameters rather than the single “k” value in SWMM,
they could perhaps be better fit to observed data. This re-
mains to be investigated.
• Importance of good estimation of Lo (maximum potential washoff).
If one is using SWMM or S RM to predict ultimate loads, then
accurate estimation of the Lo term is essential because the
models quickly attain that limiting value. As we have seen,
accurate prediction of Lo is itself contingent upon careful
use of transformed loading values. Given that one cannot use
the exponential model for very accurate estimation of early
storm loadings, we find it reasonable to place relatively
more effort into obtaining good values for Lo so that total
storm loads might be predicted. If the Lo value is only
guessed at, the use of the exponential equation must be viewed
as questionable, for then both of its parameters are possibly
in error.
• Need for more early storm measurements.
Part of the poor results for fits in the early sections of the
storms is almost certainly due to the scarcity of observed
data from these time periods.
50
-------
SECTION 6
DEVELOPING PRODUCTION FUNCTIONS FOR RUNOFF CONTROL MEASUBES
WHY A SIMULATION MODEL
There is a lack of data on the effects of different on-site runoff
control measures. For example, there is only one place known to us where
monitoring of water quality runoff of porous pavement has been done: the
Woodlands, Houston, Texas (Espey, Huston and Associates, l976).1 Since we
cannot use test data, one way to obtain some insight into the relative effects
of control measures is to simulate a series of experiments utilizing different
controls. Given a synthetically modeled site, we do not make accurate pre-
dictions of the runoff, but only comparisons among different options. By
running the model under a simulated set of randomized trials, we can create
synthetic experimental data for each set of on—site options. Just as if we
had actually performed such experiments, we can then analyze the effects of
different options in a two-way table.
DESIGN OF ThE MODEL: HYDROLOGY
Let us look at a schematic representation of the development site (Figure
6-1). We want a procedure to find the hydrograph and pollutographs at a
point A, before runoff enters a possible storage/treatment section. An
observer at point A sees the effects of site controls only by changes in flows
and concentrations at this single point: the development effects are aggre-
gated into a one—measurement location.
A simulation that purports to develop hydrographs must have a space
coordinate and a time coordinate to measure the effects of distance and
changes in the state of nature, i.e., the time it takes runoff to reach ‘A’
and the increments of time when rain occurs. In our case, we use the simplest
possible coordinates: “blocks of land,” i, and “rain intervals,” j. Both
blocks of land—space— and rain intervals-time— are scaled to be compatible
with the selected site: if the time of concentration for the block in ques-
tion is on the order of a few minutes, the rain interval is also of the same
order of magnitude. The details of the model are described in Appendix B;
a complete listing of the computer program appears in Appendix C.
THE DEVELOP NT PATTERNS
The layout of the development site can have important consequences for
the hydrologic regime of an area. The model is constructed to easily test
1 An interesting case study of a porous pavement site appears in Appendix D.
51
-------
FIGURE 6-1. HYPOThETICMJ DEVELOPMENT BLOCK
BLOCK I
IMPERVIOUS PERVIOUS
OBSERVATION
POINT
TREATMENT
alternative development plans: one simply constructs a new set of cells with
different areas allocated to the pervious or impervious sub—sections. As the
number of housing units increases, for example, the area devoted to “imper-
vious” and “road” grows as well. On the other hand, a “townhouse” type
development can put the same number of units onto a smaller land area, with a
smaller change in cover. Other model parameters are changed to reflect the
different distribution of cell areas.
In this study, three layouts were chosen for simulation:
1. a “conventional” development, with quarter—acre lots;
2. a “low—density” development, one acre lots;
3. a “cluster—townhouse” development, with four units per townhouse,
and clusters of three—five townhouses per cell.
These different development patterns were selected so as to incorporate a
broad range of alternatives, from those that develop almost all the area of
site (alternative 1, conventional) to those that leave most of the area
untouched (alternative 2). In addition, alternative 3 (townhouse), allows us
to test whether one can put the same number of total units onto an area as in
conventional developments, but with less impact on the hydrology.
The following situations were simulated with the three development
alternatives.
1. I o soil regimes: highly pervious, highly impervious
2. Six control measures:
a. none
b. porous pavement
c. swales
d. roof—gutter disconnection
e. on—lot storage
f. vegetative cover
52
-------
THE SIMULATION SITE
Bowker Woods, a residential development in Norwell, Massachusetts, was
chosen as the simulation site. As one of two areas developed by the sub-
contractor, BSC Engineering, hydrologic information and detailed maps of soil
type and topography were readily available. The total area covered is about
26.4 acres (10.75 ha). Figure 6-2 shows a schematic drawing of the area with
U.S. Soil Conservation soil classification as shown: “A” is highly permeable;
“B “ less permeable; “C “ impermeable; “D “ highly impermeable (sCS, 1975).
For each of the three site layouts seven or eight simulation model cells
were formed. Cells were drawn roughly the same size in each layout. A
schematic drawing of each layout is given in Figures 6-3 to 6-5 (see Appendix
B for the details)
FIGURE 6-2. BOWKER WOODS DEVELOPMENT SITE
DETAILS OF THE DEVELOPMENT LAYOUTS
Development Number One: Conventional Design
Figure 6-3 shows a conventional layout (quarter—acre lots). A total of
68 units are placed, and simulation cells are as marked on the figure.
2 The unnumbered cell in the southeast corner of Figures 6-3 to 6-5 is
primarily conservation land and is assumed to be pervious and not draining
directly into an impervious area. In the simulation the land in this cell is
allocated between the two adjacent cells.
I — 50O’— --i
53
-------
FIGURE 6-3 - BOWKER WOODS QUARTER-ACRE DEVELOPMENT (“CONVENTIONAL”)
FIGURE 6-4. BOWKER WOODS LOW-DENSITY DEVELOPMENT
1 500’
I- -.500’ —1
54
-------
FIGURE 6-5.
BOWXER WOODS CLUSTER-TOWNHOUSE DEVELOPMENT
Parameter values for each of the cells and their sub—sections are given in
Appendix 13, Table B-2. Within each cell the sub-sections are defined as:
1) pervious, draining impervious area (receives runoff from sub—section
4); lawns, open field, etc.;
2) pervious, not draining impervious area;
3) impervious draining to transport system (roof drains, if connected);
4) impervious draining to pervious (driveway, rooftop if not connected
to transport system);
5) impervious main road
Development Number Two: Low Density Layout
There are only 16 units placed on the large lots, roughly corresponding
to the density proposed in the actual plan by BSC Engineering (ten units).
This means that the percent area that is converted to impervious is much
smaller for this type of development than for the conventional development
(road and house surface area is smaller). Figure 6-4 gives the sketch of the
layout. In Appendix B, Table B-3 gives the parameter information for the low—
density (one+acre) layout.
Development Number Three: “Cluster” Layout
Each cluster is made up of groups of townhouses, four units/townhouse.
There are 17 clusters, for a total of 68 units, so we are constructing the
same total number of units as the conventional layout. Further modifications
include the cul-de—sac roads that lead into each cluster; this adds to the
— 500’ -I
55
-------
surface area taken up by roads. Table B-4 (Appendix B) presents information
for the “cluster” development. A sketch of the layout is given in Figure 6-5.
ANALYSIS OF RESULTS OF THE SIMULATION MODEL: CONSTRUCTING PRODUCTION
FUNCTIONS FOR ON-SITE CONTI )L MEASURES
The effects of on-site controls. . Table 6-1 presents a summary of the
mean peak flow from ten randomized trials drawn from storms of a given volume
and rainfall distribution. The rows represent the three developments and the
columns list the control measures. The value within a particular cell in a
row and column is the mean peak flow over the storms for that development
(row) and control measure (column); in this way the table gives a comparison
among all possible combinations of development types and control measures.
Tables 6-2 to 6—4 give corresponding results for total flow, peak solids
and total solids.
Returning to Table 6—1 we can analyze this table as if it were a real
experiment with a set of treatments (control measures) and a set of experi-
mental “plots” (development types). In order to separate the effects of
different developments and controls, either classical analysis of variance
or two—way table methods could be used. Further, we then can see whether the
effects of certain control combinations are additive —- that is, is the
change in flow from porous pavement control to porous pavement and vegetative
cover control the same as adding the changes from porous pavement and vege-
tative cover separately?
Figures 6-6 and 6-7 provide a better way to compare the options; they
give ratio changes of effects, comparing the conventional/no control cases.
Without any additional analysis it is easy to see that the low density
development has greatly reduced flow and solids (baseline conventional
development). This is not surprising, since a smaller portion of the site
is impervious in this development. Further, one notes that porous pavement
has a stronger effect in reducing flow and solids than any other single
control measure. Nonetheless, it is crucial to use a two—way analysis to
analyze the results because it separates the total change in flow into a
nodel where the effects are additive.
flow = development types effect + control measure effect + commen
value + residual
We cannot merely assume, however, that the effects of different options and
developments are additive. In fact, they are not: there is aluvDst a perfect
factorial interaction between development type and control option. That is,
the effect of having porpous pavement and low-density housing is
multiplicative. This multiplicative interaction can also be fit into the
two-way tables. The medel then becomes:
flow = development type effect + control measure effect
+ interaction (development x control)
+ conmon value
+ residual
The results of the two-way analysis appear in Table 6-5. The remaining
residuals, as displayed below the two—way table, show no clear pattern, so
that we may assume that the medel is additive as given. However, there are
56
-------
TABLE 6-1. PEAK FLOW (CFS) SIMULATION MODEL
1” Storm Volume
Mean of 10 Simulations per cell
control measure
.135
.090
.136
.133
.131
.123
.129
.088
.088
.081
.092
.068
.061
.063
.070
.063
068
.181
.10
.179
.174
.173
.163
.167
.09
.092
.100
.104
.100
.082
.081
.084
.081
083
Key N no controls
PP porous pavement
V = vegetative cover
R roof—drain disconnection
ST = on—lot storage
SW = swales
I = impervious soils
TABLE 6-2. TOTAL FLOW FROM SIMULATION MODEL (inches)
eyt N
Pp
V
S
ST
S W
I
1” Stors Vol é
Mean of 10 Simulations per Cell
control r.easure
— on controls
— porous pavensat
= v.getative cov.r
roof—drain disconnection
on—lot storage
— swalss
— i .rvious soils
devlopeent type
V
N
PP
R
S7
low d.nsity
clustsr
conventional
.022 .01
Pp
V
R
SW
.023
I
Pp
V
I
PP
V
pp
V
SW
Pp
SW
pp
ST
pp
R
SW
ST
V
pp
S
pp
R
SW
PP
S
SW
ST
.024 .022 .020 .021 .016 .014 .016 .016 .016 .015 .015 .020 .015 .015
Asvslnr .nt tvoe
N
Pp
V
S
ST
SW
low—density
cluster
conventional
Pp
V
I
I
PP
V
PP
V
R
PP
V
SW
.09 .05 .091 .091 .091 .091 .113 .058
pP PP
SW ST
PP
S
PP
S
SW
PP
S
SW
ST
Pp
S
SW
Sr
V
.058 .058
.058
.72
.46
.719
.729
.726
.730
. l
.463
.541
.462
.465
.270
.272
.276
.269
.278
.24r
.76
.431
.723
.807
.774
.791
.956
.411
.546
.393
.432
.456
.462
.463
.457
.465
.461
.058 .058 .058 .058 .058 .05
57
-------
TABLE 6-3. PEAK SOLIDS FROM SIMULATION MODEL (ibs)
1 Storm Vol e
p’ an of 10 Sinulations
control measure
Xays I - no controls
PP — porous pavement
V - veqstative cover
R • roof-drain disconnection
ST — on-lot storage
SW — swab.
I • isçsrvious soil.
TABLE 6-4. TOTAL SOLIDS FROM SIMULATION MODEL (ibs)
1 Storm Vo1t e
Mean of 10 Siaulations per Cell
control measure
eyz N — no trols
PP — porous pavement
V . vegetative cover
S roof-drain disconnection
ST — on-jot StOrage
SW — swaiss
I — i srvjous soil.
Ii
dpve1ec ent tvue
PP
V
R
ST SW
PP
V
7
lou—density
cluster
conventional
I
PP
V
I
1.7
PP
V
S
2.3
Pp
V
SW
2.1
2.0
1.8
Pp
SN
PP
ST
Pp
S
PP
S
SW
ST
PP
R
SW
ST
V
1.9 1.5 1.5 1.8
Pp
S
SW
21
i.e
1 .8
19.
9.3
13.8
10.7
11.2
10.6
11.2
9.1
9.4
9.2
10.2
9.6
8.7
7.9
8.1
7.6
8.1
25.
10.8
17.5
14.2
14.2
1.3.6
13.9
10.3
10.8
12.0
12.3
17.6
11.4
10.
9.9
9.5
9.9
1.8 18
1.7 1.6
develo nt tvne
N
pp
V
S
ST
SW
lout—density
cluster
conventional
17
I
PP
V
Pp
V
I
6
PP
V
S
8
7
PP
V
SW
7
PP
Sr
PP
SW
7
10
PP
S
Pp
S
SW
ST
S
PP
R
SW
PP
S
SW
ST
V
7
7
L
2
L
L
L
50 51 36
L 2 32 26
72
45
59
64
61
60
76
425646
5171 62
53
54
51
52
6
S
6
6
58
-------
FIGURE 6-6. EFFECTS OF CONTROL MEASURES AND DEVELOPMENT TYPES ON RUNOFF
Peak Flow: Percent change from baseline (no controls, conventional development)
baseline — co on value + no control effect
Total Flow: Percent change from baseline
Developoent types
(0 =
pp porous pav ent
v — vegetative Cover
on—lot storage
swales
R = roof—drain disconnection
I impervious soils
CL = cluster developoent
Lo—D low—density developoent
+10
Controls (0 = none)
sw I st P
SW
st St
sw SW
V
V
SW
CL
+ 10
0
—10
—20
—30
—40
—50
—60
—70
—80
—90
—10
—20
—30
—40
-50
—60
—70
-80
—90
+ 10
0
—10
—20
—30
—40
—50
—60
—70
-80
-90
V St Sw P I
+10
0
Pp
pp pp
pp R V
pp pp pp pp pp R v
P R R St SW St
SW SW SW
pp
V
I
St
pp
SW
V
V
CL
—20
—30
—40
—50
—60
—70
-80
—90
Lo-D
59
-------
FIGURE 6-7. EFFECTS OF CONTROL MEASURES AND DEVELOPMENT TYPES
ON SOLIDS RUNOFF
Peak Solids: Percent change fros baseline
SW
V
Develo xnent types
(0 = conventional)
Total Solids: Percent change fros base1 ne
+10
—10
St S —20
____________ pp pp -30
pp pp pp V SW V -40
pp pp pp v R Sw I -50
pppp S S S V V -60
St S SW Sw -70
SW st —80
St —90
v
F
pp — porous pavement
V = vegetative cover
St — on—lot storage
Sw iwales
S roof—drain disconnection
I — impervious soils
CL cluster development
Lo-d = low—density development
+10
Controls (0= none)
SW St
pp pp pp pp pp sw
SW SW
S St I
V
+10
—10
-20
-30
—40
—50
—60
—70
—80
—90
60
-------
TABLE 6-5. TWO-WAY ANALYSES OF PEAK FLOW EFFECTS
control measure
.
pp
PP
5
pp
PP
pp
pp
SW
PP
V
V
V
pp
pp
pp
5
SW
ST
development type
PP
V
N
ST
SW
I
V
I
N
SW
5T
N
ST
V
type
effect
low density 0 .001 0 .003 .001 0 0 .002 0 .003 .001 .005 .005 .005 .009 .005 .004 —.074
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
cluster
.016 —.00 .013 .012 .013 .013 .010 —.008—.015 .001 .008 .017 .008 .004 .00: .004 0 .019
conventional
control measure .047 .00 .048 .045 .043 .035 .041 0 0 .00 .004—.02 027.025 .018 .025 —.02 .088
effect: (cOllimon
value)
residuals peak flow = common fa1ue + control measure effect
x .001 + Development type effect
0 + control x development effect
16 0 + residual
14
12
8
4
+0
-0
a
0
a
00
0000
000
000
000
00000000000000000000000
50
—4
-8
—12
00
0
—16
median 0.0
large remaining residuals for some cells in the “conventional” development
type that remain unaccounted for. Roughly, the residuals are too positive for
small positive effects and too negative for negative effects. Because of
this, it is important to note that small differences in effects from one con-
trol to another are not important because the “error” is larger than such
differences.
Briefly, the two—way analysis demonstrates that the low-density develop-
ment reduces peak and total flow more than any other single factor. (The
reduction is about 87 percent.) A cluster—townhouse development lowers flow
and solids by about 15 to 20 percent. Though no on—site control measure is
as effective as low—density development, porous pavement is the clearly domi-
nant single control. As Figure 6—6 illustrates, porous pavement alone
reduces peak flow by 33 percent and total flow by 40 percent. Combined with
the additional measures of swales, storage, or roof drain disconnection,
additional reductions of about 10-20 percent are attainable, but -- and this
61
-------
is significant — - there is a diminishing rate of return. With three controls,
or even four or five, an additional reduction of only a few percent is
achieved.
Because the two—way n de1 is additive, we can use the effects ratio to
construct “production function” curves using marginal percent effectiveness
to incrementally select those control measures that have the largest impact
in flow and solids reduction (see Figure 6-6). (The ranking of controls is
generally consistent across all of the suxmnary measures of storm events
used: peak and total flow and peak and total solids, as well as across all
developmen t types.)
Incremental Reduction
Incremental Control Measure in Peak Flow
porous pavement 33
+ storage 22
+ swales —2
+ roof—drain disconnection 0
+ vegetative cover 0
Incremental Reduction
in Total Flow
porous pavement 40
+ storage 23
+ swales 0
+ roof—drain disconnection 0
+ vegetative cover 4
porous pavement
+ roof—drain disconnection
+ swales
+ storage
+ vegetative cover
porous pavement
+ storage
+ swales
+ roof-drain disconnection
+ vegetative cover
Incremental
in Peak
Reduction
Solids
39
9
—1
3
—3
Incremental
in Total
Reduction
Solids
25
18
-10
0
10
It is important to notice that the effects of control measures are riot
generally additive: use of porous pavement reduces peak flow by 33 percent
(effect .002), and swales reduce peak flow by ten percent (effect .035), but
the effect of porous pavement and swales is a reduction of 53 percent —— not
an additive change. Also, as den nstrated, control measure and development
type effects are multiplicative. Under different storm volume conditions
around one inch (from .7 to 1.3 inches), the results are about the same.
62
-------
For extremely low storm volumes of about .1 inch, so little runoff is produced
that control effects are hard to distinguish, but they are not necessary in
any case. For extremely high volumes of very short duration (1.80 inches in
ten minutes!), all control measures seem to fail equally (this is also to be
expected). These low and high volume conditions seem to bound the middle
storms that represent conditions for which on—site control measures would be
designed. The marginal ranking of effectiveness therefore seems to hold for
a large range of conditions that are of interest.
CONCLUS IONS
On-Site Controls
• The simulated—reduction of flow/solids via on—site control measures
shows a maximum reduction effect of 50 percent.
• Much of this reduction can be achieved through the use of just one
control, porous pavement. This is an important finding, especially
coupled with the case—experience with porous pavement discussed in
Appendix D.
• Additional controls such as swales, on—lot storage, or roof-drain dis-
connection, have a much smaller effect alone than porous pavement.
Used with porous pavement, there is a diminishing rate of effectiveness.
Development Types
• The density of development is the largest single factor in reducing
runoff.
• The interaction between control measures and development types is
multiplicative . Therefore, one gains multiplied benefits from porous
pavement and a low-density development. This has important consequences
for planning runoff control in new developments.
• Cluster development modestly controls runoff.
Use of Simulation Techniques
• Two—way analyses of development and control types can be made via a
simulation xxdel to yield “production function” rankings of control
measures. Of course, the ultimate validity of the rankings can only
be verified by field measurement. (For example, data on porous pave-
ment runoff is lacking.) However, this section does demonstrate the
feasibility of the simulation method in the study of control measure
comparisons.
63
-------
SECTION 7
STOCHASTIC MODELS
INT )DUCTION
The underlying theme in the preceding discussion of accumulation of
pollutants, washoff, and the effects of control measures is the stochastic
nature of these processes. This is nowhere more apparent than in the
storm events that drive the washoff itself. Any single observed storm is
but one from a potentially infinite variety of storms. How then can any
single, idiosyncratic storm event successfully represent the full range
of conditions that will actually be encountered? Engineers have countered
with the use of artificially constructed “design storms,” but McPherson (1977)
has cogently argued that the idea of such a synthetic storm is inherently
faulty, in that its artificial creation gives it properties that would
never be observed in nature.
What a decision maker must face is not the occurrence of any particular
“design storm,” but rather a set of chances of pollution events of varying
sizes. The gaii is a gamble against these outcomes through bets made in the
form of control strategies. It is in this larger sense that the use of a
design storm is perhaps weak: it ultimately conceals from the decision
maker information on the distribution of states of nature.
Another way to incorporate these stochastic effects is to place them
directly into the models that are proposed. For example, this was done
in the simulation of control options in Section 6 by randomly sampling
storms from a given distribution as if they had actually been observed. In
this section we present three models that also confront the issue of stochasti—
city head on by using probability distributions of storm volumes and
pollutant surface loadings to find the probabilities of overflow and washoff
(Flatt and Howard, 1978).
The first model demonstrates that the use of “design storms” does in
fact represent an implicit cost-minimization decision problem in sewer
design based on storm frequencies.. In contrast, the design storm approach
does not address the cost in tradeoffs in storm frequency ‘ ersus sewer design
explicity . This may be considered good or bad.
A second, related model proceeds to attack the inverse problem; given
a designed system, what is the chance of overflow events? Here the question
is not so much one of cost—minimization of sewer construction (which is
already in place), but rather a consideration of prediction of adverse effects.
64
-------
Finally, the third model is one to be used in the planning of a new
residential development where neither of the other models, which assume
conventional sewers, apply. Using the roughest assumptions about the
distribution of runoff volumes and initial load, the distribution of washoff
solids from an area is estimated.
All three of the models suffer front mathematical difficulties inherent
in the nature of probabilistic calculations: unless the functional forms
are quite simple (normal, exponential, log—norma:1), the resulting computations
are usually intractable. Nonetheless, even with these simplistic assumptions
we can demonstrate the use of probabilistic models in decision-making; they
can provide general guidelines for planning before detailed computer models
are required (Hydroscience, 1976).
t’ 3 [ L #1: STORM DRAINA( DESI(
Storm drainage systems conduct unwanted runoff to the nearest acceptable
discharge point. While it might be desirable to provide channels adequate
for the greatest storm for which there are local records, such a design would
usually require prohibitively large expenditures. Consequently, systems of
modest capacity have been installed in urban areas of the nation. How
have the capacities of these systems been determined? In discussing storm
sewers, Engineering Manual #37, Design and Construction of Sanitary and Storm
Sewers (1970) , states that
.An economic balance is necessary between the cost of structures
and the direct and indirect costs of possible damage to Property
and inconvenience to the public over a long period of years
The average frequency of rainfall occurrence used for design determines
the degree of protection afforded by a given storm sewer system...
However, without a definition of damage and a procedure for estimating it,
these statements are non-operational. In acknowledging the difficulty
the Manual continues
But in practice, cost—benefit studies usually are not conducted
for the ordinary urban storm drainage project. Judgement supported
by records of performance in other similar areas is usually the
basis of selecting the design frequency.
In engineering offices the range of design frequencies used is usually two
to fifteen years for residential areas (with five years most commonly reported)
and ten to fifty years for commercial and high-value districts.
Such guidelines are not new. Advice similar to that above can be found
in standard engineering texthooks. For example, writing more than 50 years
ago, Metcalf and Eddy (1914) stated that
Few problems have afforded the sewer deigner more misgivings than
the determination of the quantity of stormwater for which storm
drains or combined sewers should provide. The chief reason for this
65
-------
lies in the fact that the problem is indeterminate, and that the
information which may be available and the formulas which may be
used only serve to aid his judgement, upon the soundness of which
the correctness of final solution very largely depends. In fact,
it is a difficult task to say when the solution of such a problem
is correct within the usual meaning of the term...
On the topic of benefit-cost analysis, they concluded
Practically, howeverL such computations are of little significance.
Local circumstances and conditions, physical and financial, have
usually a controlling effect upon the extent to which such drains
can be designed to care for extreme maximum rainfalls.
Unlike the Manual, tcalf and Eddy explictly included financial capability
as an important controlling factor in selecting the degree of protection.
They also pointed out that legal responsibility of the community is a
consideration since any damage from flooding must be borne by members of the
community. The courts have held that
• . .rainfalls are differentiated for judicial purposes into ordinary,
extraordinary and unprecedented classes. Ordinary rain storms
are those which may be reasonably anticipated once in a while, and
unprecedented storms are those exceeding any of which a reliable
record is extant. The usual rule in determining the responsibility
of a city was stated many years ago by the New York Court of
Appeals, 32 N.Y. 489 as follows: ‘If the city provides drains and
gutters of sufficient size to carry off in safety the ordinary
rainfall, or the ordinary flow of surface water, occasioned by the
storms which are liable to occur in this climate and country, it
is all the law should require’. (Eng. Rec. June 18, 1912).
What constitutes an “ordinary” storm? Is a storm which is expected to occur
on an average once in ten years ordinary? In some localities the testimony
of an engineer is sufficient evidence of what is an extraordinary rainfall.
However, no two engineers acting independently would always reach the
same conclusions as to storm events. Therefore the question of ordinary
and extraordinary still remains.
The second half of the 19th century witnessed intensive efforts in
rainfall data collection and analysis in this country. The relationship
between the intensity of a rain and its duration was studied for the needs
of storm drain design. Some of the original work was done by Talbot (1899),
who, making use of U.S. Weather Bureau records at 499 stations plotted storm
intensities against durations on a cross—section paper. Two envelope curves
were drawn, one giving what might be called the very rare rainfalls, and
the other the ordinary maximums. The curves were in both cases rectangular
hyperbolas. Their equations were determined to be
i 360 (7—1)
t+30
66
-------
for the curves of rare occurrence, and
i = 105 (7—2)
t+1 5
for those of frequent occurrence, where i is the rainfall intensity in
inches per hour, and t is the duration of the storm in minutes.
Since Talbot constructed these curves, many cities, public agencies, and
engineering firms have developed similar equations as a basis of storm drain
design. For example, Boston has used i = l50/(t+30), a relationship very
much like equation (7—2). Such expressions serve as a first step in developing
designs of storm drainage system but the degree of protection is still based
on judgement, financial capability and legal responsibility.
A Decision Model
To put questions of protection into perspective, we have formulated
a model of the design problem. For a given urban area with known specified
topography, geology, and land use patterns, let the storm sewer cost be
denoted as K(i), damage caused by inundation of property as D(i), where i
represents the rainfall intensity. Let the density function of rainfall
intensity be f(i). One decision problem is to select a sewer system with
capacity adequate to accommodate runoff caused by rainfall intensity less
than or equal to i and minimize the total cost (T is dummy variable of in-
tegration)
Z = K(i) f(i) D(T) dT (7—3)
3-
Rouscuip (1939) has shown that there are substantial economics of scale in
storm sewer construction. We therefore approximate the cost function by
K(i) = Property damage depends on the depth of inundation, which is
influenced by the rainfall intensity and storm duration. For demonstration
purposes only, we may assume that the damage is a constant value for all i.
With these simplifications equation (7—3) becomes
Z = ai + D{ 1—F(i) } (7—4)
Optimum condition, dZ/di = 0, implies that
Df(i) — c i = 0 (75)
If we assume f(i) = l.2e 2i and if we specify, based primarily on Rousculp
(1939), that G2 = 2770, = 0.27, D = 7400, we can solve the equation to obtain
i = 2.66 inches/hour
K(i*) = 2770 x 2.660.27 = $3607/acre
l_F(i*) = e ” 2 x 2.66 = .041
67
-------
T* = l/(l_F(i*)) = 24.3 years
Where T* is the design storm frequency.
As discussed above, the financial capability of a community plays an
important role in sizing the sewer system. To include this consideration,
appropriate constraints can be incorporated in the model.
Data Problems
In equation (7-5) (optimum condition), the computation of i* requires
the estimation of parameters D - , and . Any errors associated with
these estimates can cause changes in the optimum solution, and in turn alter
the design and construction costs. To explain the argument, we differentiate
equation (7-5) to obtain
= { (c + c )i - f(i) D} / { Df’ (i) - -l)i 2 } (7-6)
This equation shows thatchanges (ti) in i are a function of parameters’
values, their changes, and functional forms used to describe rainfall
intensity and sewer costs. The changes in i cascade down to design via the
rational formula Q = CiA. Changes in i also affect the selection of C.
Hence t Q = (AC ÷ Ai C )t i. (7-7)
To test the effects of errors in parameters and the misspecLfication of
functional (or model) form on the solution, sensitivity analysis is
frequently performed.
Rainfall data are often only a short string of observations out of a
large population; the observed events are unlikely to repeat themselves in the
basin. However, population parameter estimates based on observed data may be
non—representative of events arriving during the life of the project.
Kirby (1974) has shown that moment estimates are constrained by the number
of observations in the data. Slack, et al.,(l975) , have demonstrated, by
Monte—Carlo experiment, that there is a substantial difference between the
estimated skewness of the observed data and that of the population.
They have also shown that transformations to facilitate analysis, parti-
cularly the log transform, are dangerous because small errors of estimation
in log space are vastly amplified when transforming back to the raw data
space. It appears from all these studies that robust distributions
to model hydrologic phenomena should be used rather than complicated
distributions which require the estimation of higher moments.
There is another difficulty in using rainfall data for storm sewer
design: records axe available only for a limited number of locations. It
is unlikely that a specific project area will possess a rain gage. Trans-
ferring information to the project site from other areas introduces errors
In an actual application D is a function of depth of inundation.
68
-------
in estimation. Moreover, rainfall data are often measured in hourly
intervals, and only in limited areas are more refined data collected. To
be useful for storm sewer design, hourly data must be decomposed on the
basis of shorter time intervals. This task can also cause error in design
event estimation.
MODEL #2: HYDRAULIC CAPACITY OF A TREATMENT SYSThM
The dry—weather wastewater flow of a city, since it is composed of
domestic wastewater, commercial/industrial waste, and infiltration,
varies in quantity depending on the season, day of the week, and the hour
of a day. In large cities peak dry—weather flow varies from two to four
times the daily average,while in small communities the ratio may be
higher. Interceptor sewers and treatment systems are customarily designed
with a capacity of peak dry—weather flow, whereas in a combined sewer
system the facilities are used to accommodate some storm runoff. To
some extent the ability of a given system to convey and treat stormwater
depends on the quantity of dry flow being discharged at the time the
storm occurs. During peak dry flow any surface runoff reaching these
facilities will cause overflows to occur. On the other hand, these
facilities will have capacity for more stormwater if rainfall occurs when
the flow of dry flow is small. Inasmuch as rainfall shows little tendency
to favor certain hours of the day, the probability of surface runoff
coinciding with peak dry-weather flow is balanced by the probability of
rainfall during periods of low dry flow. Hence the average capacity of
the system for storm runoff can be estimated on the basis of the mean dry—
weather flow and the system’s design capacity. If the interceptor and
treatment system are sized to take an average of twice the dry—weather
flow, then in the long run a quantity of storm runoff equal to the dry-
weather flow will be treated during the periods of rainfall. If the system’s
capacity is three (or n) times average dry flow, the storm runoff treated
will increase to two (or n—i) times average dry flow. To control all or
almost all stormwater requires a large system —— too expensive to be borne
by most communities. In reality overflows are permitted and raw sewage is
discharged along with the storm water.
On the basis of the long run average posited above, McKee (1947)
analyzed the Boston sewer system to determine the percentage of time that over-
flows occur, the proportion of sanitary sewage escaping, and the number
of separate and distinct storms that cause overflows. With a capacity
of twice the dry—weather flow, on the average 2.68 percent of total
sewage would be lost through overflows occurring five to six times a month
between June and November. From his analysis a capacity larger than three
times dry flow would not be economically justified. By following the same
procedure Stanley (1947) studied the sewer systems of Chicago and
Minneapolis and reached a similar conclusion.
In essence the method employed by both McKee and Stanley is the
estimation of the probability distribution of the sum of dry—weather flow
plus storm runoff -- Pr (D+Q), where D stands for dry flow, and Q for
69
-------
storm runoff. Overflow occurs when the sum of flows exceeds the capacity, aD
The probability of the event is Pr{tD+Q)>aD}. By field measurements on an
existing system, the probability of (D+Q) can be estimated. However, for the
long—run case, described above, the problem can be simplified to one of esti-
mating Pr{Q > (a-l)D}. Using the rational formula, the overflow probability
becomes Pr{i > (a-l)D/AC}, in which A is the sewered area that has a runoff
coefficient C.
In a storm initial precipitation wets the ground surface and fills de-
pressions. Runoff occurs only after the total precipitation exceeds the ini-
tial requirement. Hence rainfall records that are to be used in estimating
runoff must be corrected accordingly. 2 For each storm excessive rainfalls
(observed rainfall minus depression storage), intensity, duration, volume,
and antecedent dry period are estimated. These attributes for all storm
events in a given period (e.g., five years or more) are summarized statisti-
cally and fitted by an appropriate function. For example, a gamma distribu-
tion of the form
f(i) = Vr(r) (X/i)rleA (7—8)
is often used to describe intensity frequency. Here the parameters r/X and
r/A 2 represent the mean and variance of intensity and can be estimated by the
method of moments. Using this formulation, the probability of overflows
then becomes
Pr{i > (a-l)D/AC} = ;(a_l)D/ACf(.)d. (7—9)
To illustrate the evaluation of overflow probability, let r/X = O.025”/
hour, n A 2 = 0.0068”, a = 3, D = 4500 gallons/day (= O.007”/hr./acre), C = 0.7,
unit area A = 1 acre, then
(a—l)D/AC = 2 x 4500 x 12/(7.5 x 43,560 x 24 x .7) = O.020”/hour
and
0.02 3.67 —0.0908 —3.673i
Pr(i > 0.02) = 1 — - 0 0 2 (3.673i) e d i 0.35
(7—10)
This calculation implies that on the average the overflow takes place for
about one—third of the storm events. For a = 4, the system can reduce the
overflow probability to about 21 percent in this example.
Since the rate of overflow is CiA - (a-l)D, the expected value of this
quantity can be estimated as ff(i)(CiA - (a-l)D}di. For the case of a = 3,
the expression becomes 0 1 2 f(i) (O.7i — O.014)di. If complete mixing between
sewage and storm water is assumed, the sewage that overflows (as a fraction
of the total overflow) is {ciA - (a-1)D} ÷ (CiA+D). For the case of a = 3,
1 The coefficient a is the multiplier on dry weather flow and signifies
design capacity.
2 For the Boston area, an allowance of 0.03” is appropriate for the depth
of composite storage during summer and fall months.
70
-------
expected percent of sewage overflow during storm events is
0.71 — 0.014
0.02 :i 0.7i ÷ 0.007 di (7—11)
To estimate analytically the amount of a pollutant in the overflow
mixture, two quantities must be known: (i) the deposit or accumulation
rate of sewage in the sewer system; and (ii) the accumulation rate of
dirt and other contaminants on the land.
MODEL #3: STOCHASTIC WASHOFF
We would like to formulate a model that gives an estimate, in the
simplest way possible, of the average kind of washoff event we would
expect to see, given a local storm distribution. Such a model could pro-
vide the planner with the probability of washoff event as well as the
typical variation to be expected. The following is a proposed demonstration
of the method to be used to derive such a model. It cannot be used for
“real” planning because, as shown in Section 5, the functional forms
used probably do not adequately describe the washoff process. However,
it can be viewed as a demonstration of the usefulness of the stochastic
point of view.
Washoff is assumed to follow the equation
L = Lo(1-exp(-krt)) (7-12)
Where L = washoff, lbs/unit area
r = runoff, inches/hour
t = time, hours
Lo = initial loading on surface, lbs/unit area
k = constant, about 1—1.2 after unit conversion,
(from Sartor and Boyd, 1972)
We want to find the probability distribution of load under first flush
conditions (t = 0 to about 0.5 hours). First, we transform to a log scale:
£nL = inLo ÷ £n(1 ’-exp(-krt)) (7-13)
71
-------
Now, given that we want to find the distribution of L in the early stages
of the storm, what are the values of l-exp (-krt)? Here k is about 1.2,
r is less than 0.5 (in/hr) and t is less than 0.5. The product krt ranges
from 0 to 0.3. In this interval 1-exp(-X) is approximately equal to X
(For X = .05, 1 — exp(—X) = .049 for X = .15, 1 — exp (—X) = .139.)
For this first flush interval we therefore replace the function 1 - exp(-krt)
by krt. We are left with
£n(L) = £n(Lo) + £n(krt) (7—14)
= £n(Lo) ÷ £n(kV) (7—15)
where V = runoff volume.
The probability distribution of the load during first flush is then the
convolution of two probability distributions Ln(Lo) and Ln(kV). Now Lo,
the initial surface loading, is log-normally distributed (see Section 5).
Hence n(Lo) is normal (Figure 5-3). How is £ n(1cV) (the log of runoff
vo1ux in the initial half-hour of a storm) distributed? This is uncertain,
but if we plot the runoff volumes from Coiston (1974) in a stem-and-leaf
display, there is a rough log—normal shape (Figure 7-1) . Given all these
assumptions, the probability of the washoff is then simply the sum of
two norma ]. distributions that have been log—transformed. The mean washoff
is then (in log scale) the sum of the means of in(Lo) and in(KV):
average 2 n(L) = average n(Lo) and average 2 n(kv) (7-16)
P N EXAMPLE
Using estimates for k, mean Lo, and V, we can now find expected
washoff. One difficulty is obtaining values for runoff volume, V.
The sources used here is Coiston (1974), samples from a 1069—acre basin
in North Carolina (433 ha), of which 833 acreas (377 ha) are residential, in-
dustrial, or commercial, and 30 percent is impervious. For this basin two
distinct log-normally distributed sets of first-flush (less than 15 minutes)
volumes are apparent. The first, smaller volume group has a mean first-
flush flow of 4.65 cfs (.l6m 3 /sec) (see Figure 7—1). For these 14 storms
the mean suspended solids concentration is 1024 mg/L (see Figure 7-2).
Assuming the Colston study’s value of 18 lbs/acre suspended solids loadings
for developed areas,
4.65cfs
Zn(L) = 2 n(l8 lbs/acre x 833 acres) + [ 1.2 x
320.7 acres impervious
9 )1(L) = 5.56
L 260.89 lbs
3 -x x 2
This is derived from the Taylor series expansion of 1—e x — -.,, + —
kOther common distributions (exponential, for example) could also be used
here. The choice of a log-normal is suggested by the mathematical convenience
and a brief look at the Coiston data.
72
-------
Which is converted to mg/i: mg/i = 267 lbs/mm
cfs
260.89
for 15 minutes = 15 4 65 X 267 = 999 mg/i
Taking the second, larger—flow group of storms, mean flow is 85.25 cfs
(see Figure 7-1). Here, to get a good estimate of the washoff, it is
necessary to use a larger value of k (=2.0), than in the preceding
set of storms. This gives a predicted mean suspended solids value of
1879 mg/i for initial periods of the storm, compared with an observed
average concentration of 2268 mg/i. It can be seen that the predicted
average concentration is sensitive to the value of “k” selected, but
there is no real experimental data on which to base any justification
for the value used. We have only the data of Sartor and Boyd (1972)
for street surface runoff (values in converted units for the calculation
above range from 1.2 to 2.0). Data for other particle sizes and land
surfaces are nonexistent.
Considering the unreliability of the loading, volume, and “k” values,
it must be considered fortuitous that the model is accurate in the two
cases tested; however, it does in any case give correct order—of—magnitude
estimates which could be used at a general planning level. What are
the implications for stormwater planning? First, the estimated average
first-flush loading of suspended solids in log scale is made up of two
components -- initial load and runoff volume. Perhaps more importantly,
the variance —— how uncertain we are of what the true expected load is —-
is also the sum of the variances for initial load (Lo) and runoff volumes
(V) . To reduce the uncertainty, one can reduce either the variance of La,
or that of V. We have seen in Section 5 of this report that even after
accounting for solids loading by land use or climate there is often a
large remaining variation in load due to wind, traffic, or other micro—
climatic effects. But these may be beyond control for, as the San Jose
study shows, even with street cleaning in a single location, loading is
subject to large variations. In contrast, the variance of runoff volume
is possible under direct control of planners through the management
techniques discussed in Section 6.
73
-------
FIGURE 7-1. FIRST-FLUSH VOLUMES F )M
COLSTON, 1974 (FT 3 /SEC)
11 2 11 2
10 0 10
9 9
8 8
7 7
6 3 n14 6 3 n=8
4 4
3 292 3 29
2 30886 2 849
10 18
09 0
low volumes high volumes
unit = 1 cfs unit = 10 cfs
mean 4.65 mean = 88.5
FIGURE 7-2. SUSPENDED SOLIDS FROM
COLSTON, 1974 (MG/fl
3400
2710
11 0
10
92
8
7 7
6 6 n14 6 7
53 5
4 4 4
340 3 5
243 2 6
117 1 1
0 57 0 184
low volumes high volumes
unit = 100 mg/ unit = 100 mg/2,
mean = 1024 mg/2 mean = 2268
74
-------
SECTION 8
ESTIMATING COSTS OF ON-SITE CONTI DL MEASURES
As pointed out earlier in this report, the lack of appropriate cost
relationships makes it difficult to prepare meaningful cost-effectiveness
analysis of stormwater control options. In this section we explore method-
ologins for obtaining generalized cost functions. Ideally, we would like
to develop functional relationships between costs and certain factors
which are available at the preliminary planning level. These factors
typically include general site characteristics, such as slope, size, and
intensity of development and meteorological information such as the design
storm. At this level of analysis engineering design data (such as the
size and length of drainage pipes and the maximum discharge) would not be
available. Thus cost estimates at this level will of necessity be crude
relative to estimates made later in the design process. If properly devel-
oped, however, they would be useful in evaluating area—wide stormwater
management strategies and comparing alternative combinations of measures.
Our analysis has focused initially upon the evaluation of traditional
drainage system costs. This choice is based upon two considerations:
1) In most areas conventional storm sewer systems represent the
current practice against which options such as on-site storage
will inevitably be compared. Thus it makes sense to treat the
conventional system as a baseline for cost evaluation purposes.
2) Since conventional storm sewer systems are widely applied, there
is nxre knowledge and data on costs for this system than for other
measures. Cost estimating approaches can be tested against a con-
ventional system before extending them to measures for which cost
data are limited or unavailable.
PREVIOUS STUDIES
In order to establish cost functions of the type desired, it is neces-
sary to establish an appropriate data base against which to evaluate func-
tional forms. Two paths might be followed. For developing a data base it is
possible to use actual or bid costs of a suitable sample of projects, or to
develop a series of cost estimates of hypothetical projects designed from
scratch. Each approach has its own strengths and weaknesses. Utilizing
actual project data has the advantage of encompassing in a suitable sample
the range of conditions and factors found in practice. On the other hand,
great care must be taken to insure that reported costs actually incorporate
the full range of project expenses of interest and no others. In addition,
75
-------
there may be a number of project—specific factors which influence costs and
add to unexplainable variance in the data set, thereby reducing the preci-
sion of estimated cost functions. A data set developed by synthetic costing
techniques eliminates this type of random variation in the cost data, but in
the process the assumptions and design philosophy of a single individual or
group become imbedded in the data set. This may bias the data in ways which
are not predictable a priori . In addition, synthetic cost studies may often
miss cost areas which are important in real applications. For example, a
recent synthetic cost study of sewage treatment plants produced estimating
procedures which predicted costs substantially lower than those found in
practice (EPA, 1976).
Given any particular data set, we find that cost functions may be
developed by empirically fitting the data to arbitrary functional forms so
as to produce a best fit by some criteria. Usually the techniques of linear
or non—linear regression analysis are employed for this purpose. Alterna-
tively, a particular functional form might be specified based upon the
physical relationships of the variables in the system.
It is possible, and often desirable, to combine the approaches dis-
cussed above, but the existing cost studies have taken separate paths. Grigg
and O’Hearn (1976) developed a cost function based upon a simplified n del of
the hydraulics of runoff and estimated the parameters of the n del from
synthetic costs developed for a single drainage area. Rawls and Knapp (1972)
developed a data base of actual projects and fit a variety of linear and
non—linear xrvdels to develop estimating equations. Functional forms were
apparently chosen rather arbitrarily, in order to obtain the greatest explana-
tory power.
The Rawis and Knapp data base consisted of 70 projects from 23 areas
located across the United States. The project data obtained included the
design storm frequency (T) in years, average slope (S) in percent, runoff
coefficient (C), number of inlets and martholes (M), smallest pipe diameter
(DB) in inches, largest diameter (DE) in inches, outlet capacity (Q) in ft 3 /
sec, total length of drains (LT) in feet, total drainage area (A 1 .) in acres,
developed area (Ad) in acres, and total cost in 1963 dollars (CT). The
individual variables n st highly correlated with the total costs were the
total and developed drainage area, maximum pipe size, outlet capacity,
length of drains, and number of inlets and manholes. For all of these
variables r > 0.5. Noticeably absent from the data set were data on the
magnitude of the design storm or soil characteristics, although these might
be reflected to some extent in the other variables.
Rawls and Knapp estimated t%1O primary types of cost n de1s, and both
were fit by nonlinear techniques. The first was a nodel designed for
estimating costs from preliminary plans:
CT = 58,273.0 + 8.73 (T 0 ° 4 S 089 C° 64 D 5 023 Q 073 Ad 071 ), R 2 0.886 (8—1)
76
-------
The second model was for estimating costs from detailed designs:
CT = 413 + 0.72 LT(DB° 92 + DE° 92 ) — 39,640 + 31 eO d, R 2 = 0.910 (8—2)
It should be noted that both models incorporate engineering design variables
such as pipe size, maximum discharge, and number of manholes and inlets.
Thus while these fits, as measured by R 2 , are good, they are not appropriate
for the planning level that we have in mind.
Rawls and Knapp also estimated simpler linear cost functions separately
for observations from three separate states, and they found that they could
explain a substantial proportion of the variation in costs in California and
Texas. This result indicates that regional effects were important and might
be related to the omission of rainfall and soil factors from the data base.
Grigg and O’Hearn developed a functional form for cost estimation from
an idealized model of a single storm sewer draining a small basin. The
following relations were employed:
Rational Formula: Q = CIA (8-3)
Manning’s Formula: = 1.49 ( ) 2 ” 3 S 1 ” 2 ( 4 4) ) (8-4)
rainfall intensity: I = (8—5)
and pipe cost/ft: C = .0741 D 1663 (from pipe cost data) (8-6)
where T = design storm frequency, in years
n = Manning’s roughness coefficient,
I = rainfall intensity, in inches/hour,
D = pipe diameter, in inches,
t = rainfall duration, in minutes,
C = pipe cost, in dollars/foot,
K,x,b = constants dependent upon region,
and other terms are as defined previously.
Combining these relations yields the cost expression:
C = 0.741 aL 663 TO 624 X (8—7)
p
.. .375
/ nKCA
where a = 15.96 b 1 2 -
\ t s...
77
-------
The final expression for total costs was determined by adding a factor
for adjusting pipe costs to account for design, incidental costs, and other
construction costs as a fraction of pipe costs (E) and a utilization factor,
U = U 0 (l_eYT), expressing the fraction of maximum possible pipe length
actually used in a given system. Thus:
CT (.0741) (l+E) (a 663 )TO 624 X(l_e T)U 0 (8—9)
The authors did not estimate the function for general topographic and
rainfall conditions, but considered only a single site with different degrees
of percent imperviousness and different design periods. They synthesized cost
for drainage systems and fit these costs to functions of the design periods
for each level of impervious area. Their final relationship was a simplified
expression of the form:
C = a.(1—e 0 ) (8-10)
where a. is a constant evaluated separately for each level of impervious area.
While the approach yielded good fits for the synthesized costs of the
particular area under consideration, the adequacy of the general model
(equations 8—3 to 8—6) for estimating costs for different areas, slopes, and
design storms was not really evaluated.
REANALYSIS OF RAWLS AND KNAPP DATA
A model similar to that developed by Grigg and O’Hearn has been used to
reevaluate the Rawls arid Knapp data. This reanalysis has focused on develop-
ing a cost relationship suitable for preliminary analysis when no engineering
data on pipe sizes, appurtenances, maximum discharge or pipe length are
available. The model is outlined below:
Rational Formula: Q = CIA (8-11)
General Flow Formula: = D 2 a (8-12)
b
Cost per unit length of pipe: C b 1 D 2 (8-13)
System length: LT = c 1 A 2 (8-14)
Combining these expressions yields:
/ -a b 2 /a 2
C = b ( cIAs ‘ (8-15)
p l\ a 1 j
78
-------
and,
( a \b 2 /a 2
CT b 1 c 1 a 1 ) A 2 (8-16)
This expression can be multipled by a factor (1+E) to account for addi-
tional costs. Finally, a factor can be introduced to account for the degree
of development in the drainage basin served. We will use the term (A/Ad)c3
which is the ratio of area served (A) to area developed (Ad). The parameter
c 2 is assumed to be less than zero, since the larger the ratio the less exten-
sive would be the pipe network required to serve a given area. The expression
for total costs becomes;
CT = (l+E)blC 1 (CI?) ( ) A (8-17)
which can be simplified to:
dddd d
CT = d 1 C 2 3 A 4 s 5(A) 6 (8—18)
where the di’s represent the appropriate combinations of parameters and are
to be estimated.
Examination of the derivation of the total cost equation leads to the
following predictions about the signs of the parameters:
d 2 ,d 3 ,d 4 > 0 (8—19)
and
d 5 ,d 6 < 0 (8—20)
That is, costs should increase with increasing runoff coefficient, rain-
fall intensity, and drainage area and should decrease with increasing slope
and ratio of total to developed drainage area.
A summary of the Rawls and Knapp data is presented in Table 8-1. One
necessary variable, the design storm intensity, was not supplied directly by
these data. Our initial approach was to replace this variable with an
estimate: the 15—minute storm intensity, 115, associated with the design
frequency, T. The 15-minute interval was selected as being representative of
the drainage basins in the data set. The 15—minute storm intensities are
readily obtainable from historical records such as the graphs developed by
Yarnell (1935). Planners should thus have relatively easy access to this
information. A second, more accurate, measure of the design storm intensity,
I, was inferred from the original data by using the figures on peak discharge,
79
-------
TABLE 8—i. DATA SET SUMMARY, RAWLS PiND KNAPP
(Number of Observations 67*)
Standard
Mean Median Deviation Hinges
Slope (%) 1.36 1.1 1.14 .4; 1.95
Runoff Coefficient .49 .50 .086 .45; .5
Drainage area (acres) 119 59 211 36; 110
Total Cost (1963 $) 82,030 53,811 112,642 32,484; 86,795
Ratio of Total area
to developed area 1.50 1 1.25 1; 1.46
15—minute storm
Intensity (in/hr) 3.99 4.24 1.05 2.84; 5
Design Storm
Intensity (in/br) 3.63 2.95 3.3 1.75; 4.13
System Length (ft) 2,953 2,600 1,774 1,701; 3,574
*
Three systems from the original data set are not included because of
incomplete data.
runoff coefficient, and drainage area together with the rational formula,
i.e., I = Q/CA. The implications of these alternative definitions of storm
intensity area discussed subsequently.
The parameters d 1 to d 5 of equation (8—18) were estimated by taking the
natural logarithms of both sides of the expression and using regression
analysis on the resulting linear model. Results for this basic model are
presented below:
in C = 7.86 — .134(ln S) + .764(ln C) + .530(ln I )
T (.067) (.411) (.357) 15
(8—21)
+ .696(ln A) - . 356(ln
R 2 = .571
S = .636
e
Number of observations = 67
where the numbers in parentheses are the standard errors of the coefficients.
Taking the exponentials of both sides, the equation becomes
CT = 2,594 s_134c764: 15 530A696( j) 356 (8—22)
-------
This equation is in 1963 dollars. Using the ENR Construction Cost Index to
update these costs to March 1977 requires multiplying by the ratio
( 23 ) = 2.79.
All of the coefficients have signs predicted by the theoretical analysis,
although only slope, area, and the ratio of total to developed area are signi-
ficant at the five percent level. The standard error of estimate indicates
that roughly 68 percent of model predictions should fall between 53 and 190
percent of the true value. Thus there is considerable unexplained variation
remaining, which is not surprising given that the data are drawn from all
parts of the country, and detailed design information has been omitted from
the model.
The R 2 given is not directly comparable to the Rawls and Knapp values
because it is computed on the log of the dependent variable. A comparable
measure can be obtained by using actual and predicted cost values after a
retransforination to the original form of the data. This value is R 2 = .65,
which is still much lower than the Rawls and Knapp results. But with the
exclusion of design information from the analysis, the results appear to be
quite encouraging.
Another way of looking at the unexplained variation in the model is
through the stem-and-leaf plot of residuals in Figure 8-1. The distribution
is generally symmetric at about 0; there is one very extreme negative
residual (model overpredicts) corresponding to a San Diego System.
We will now examine the implications of using the 15-minute storm instead
of the actual design storm intensity. Recall that 115 was selected as an
independent variable because it is fairly representative for the drainage
areas of interest and can be obtained most directly by planners, sLice they
do not need to know any detail about a particular drainage area.
In order to examine the potential effect of replacing I by 115, we can
look at a generalized intensity duration relationship (Butler, 1964) :1
I = (8—23)
where T = design storm frequency (years),
t = storm duration (minutes) (generally equal to time of travel if
rational method used)
K,x,b = parameters (usually location specific).
1 This relationship has been presented previously; it is repeated here for
the reader’s convenience.
81
-------
FIGUBE 8-1. RESIDUALS FROM EQUATION 8-21
1.4
1.3
1.2 2
1.1 2
1.0 02
.9 9
.8 523
.7
.6 77509
.5 10
.4
.3 926
.2 525923373
.1 5509
.0 693 Median = .03
-.0 39 Hinges = -.30, .34
—.1 686813
—.2 2087470
—.3 030
—.4 172
-.5 219
—.6 9
—.7 45
—.8 913
—.9
—1.0
—1.1 7
—1.2
—1.3 3
—1.4
—1.5
—1.6
—1.7
—1.8
—1.9
—2.0 0
using this expression, the relationship between I and 1j5 for a constant
design storm frequency is:
15 = 153 or I = I 153 (8-24)
82
-------
Thus
ln( 1 15 ) = ln(I) ÷ a 3 ln(15) — a 3 ln t (8—25)
Two points about equation (8-25) are worth noting. First, t depends
upon the slope and area, thus the use of 115 in place of I would mean that
some of the slope and area effect in equation (8—18) would be incorporated
into the intensity effect by the estimation procedure. Second, a 3 varies
with location, so for the model estimated from national data it can be con-
sidered a random term. According to equation (8-25), this makes the inde-
pendent variable 115 subject to random error, and as a consequence the esti-
mated effect of the variable will be lessened relative to that of the true
storm intensity (see Theil, 1971). The extent of this effect depends upon
the variar.c introduced by a 3 .
Improving the Estimate of I
In order to further examine the implications of this analysis, a second
regression model was specified, this time using our estimate of the actual
design storm intensity, I, in place of 115. This procedure assumes that the
outlet is designed only to handle the runoff from the design storm as pre-
dicted by the rational formula. The results of this regression are:
ln C = 7.88 - .l58(ln S) + l.044(ln C) + .550(ln I)
T (.050) (.359) (.113)
(8—26)
+ .774(ln A) — .41l(ln - )
(.073) (.153) A
= .680
S = .549
e
The inclusion of the more accurate estimate of I had the effect that our
preceding analysis had suggested: each of the coefficients became larger in
absolute value. In equation (8—26) all of the independent variables are
statistically significant at the five percent level. Overall, there is a
substantial improvement in the R 2 and a corresponding decrease in the standard
error of the estimate compared with equation (8—21).
Since it appears that a more accurate estimate of the storm intensity
contributes substantially to the predictive ability of the model, it is use-
ful to consider how this additional piece of information could be made avail-
able at the planning level. One approach would be to compute travel time for
representative classes of development types, slopes, and areas and present
this data in graphical or tabular form. The planner could then use this
information, together with design storm curves, to estimate a value for I in
the cost equation.
83
-------
Regressions on Cost/Foot
It is well known that the total cost of collection or distribution net-
works is heavily influenced by the length of the network. This variable has
been ignored in the earlier regressions; in essence it was treated as a more
detailed design parameter and the area variable used as a surrogate for length
in the equations. However, in many types of developments utilities closely
follow lot frontages, and fairly accurate rules of thumb are known for
frontages associated with these developments. Therefore an alternative
approach to equation (8—18) is to specify the dependent variable as cost per
foot, Cp, in 1963 dollars. The basic model in equation (8—18) otherwise
remains the same. The results of two regressions utilizing C as the depend-
ent variable are reported below:
in C = 1.559 — 0.0944(ln S) + l.142(ln C) + .451(ln I)
p (.044) (.314) (.099) (8—27)
+ .437(ln A) — .0524(ln
(.063) (.134) A
R 2 = .52
S = .480
e
in C = 1.528 — .0759(ln S) + .916(in C) + .45(ln 115)
p (.058) (.353) (.307)
(8—28)
+ .372(ln A) — .0046(in - )
(.075) (.164) A
R 2 = .37
S = .547
e
“flie two regressions differ only -ira the rainfall intensity jerm. I is used in
equation (8-27); 115 in equation (8-28). While the explained variation (R 2 )
appears to be smaller than in the earlier regressions, this occurs only
because the dependent variable has been defined differently. If, for example,
values of in C are computed for each observation using equation (8-27) and
the log of len th is added to each value, the resulting term is a predicted
log of total costs. 2 The variation of these values can be divided by the
total variation of these values can be divided by the total variation in
log CT to yield an R 2 comparable to the earlier regressions. The resulting
value is .90, much higher than the earlier regressions and comparable to the
detailed models of Rawls and Knapp. Thus considering cost on a per foot basis
2 ln(C) ln(length) = ln(C x length) = ln(CT)
84
-------
can result in substantially more accurate regressions, providing rules of
thumb are accurate enough so that the system length will be known virtually
without error.
Evaluation of Predictions on Coefficients
According to the simple model upon which our analysis is based, the cost
per foot should reduce to the following formula:
C = e 1 (CIA) 2 S 3 ( 4 ) 4 (8—29)
or
in C = in e + e(ln CIA) + e 3 (ln S) + e 4 (ln ). (8—30)
Thus the relationship predicts that the same coefficient should apply to the
variables C, I, and A. This prediction was tested by running a regression
for equation (8—30):
in C = 1.0823 + .43l(in CIA) — .094(ln S) — .0594(in - ) (8—31)
p (.0576) (.0492) (.134) A
R 2 = .478
S = .493
e
Despite the constraints imposed in equation (8-31), the R 2 is not much
less than the comparable R 2 in eauation (8-27), where C. I. and A were
allowed to take on separate coefficients. The reduction in R 2 is not signi-
ficant at a five percent level, nor are the coefficients for S andA
changed much. These results further support the validity of the underlying
conceptual model.
Regional Effects
A number of site or region specific factors may influence the collection
system cost. For example, soil characteristics, in so far as they influence
the costs of installing pipe, could be an important factor aside from their
influence on the value of C. Then too, there may be systematic differences
in engineering practice or the nature of design storms used in different
parts of the country. The published data did not permit a detailed evaluation
of these influences. However, we have attempted to introduce some control for
these influences by introducing dummy variables designating region of the
country. Using census definitions, the cities in the data set were assigned to
one South, West or North (Northeast and North Central). One example of the
analysis is given by equation (8—32), which represents an extension of
equation (8-26) incorporating regional effects as dunmiy independent variables:
85
-------
in CT = 7.560 + .782(ln C) + .504(ln I) + .766(ln A) — .141(ln S)
(.40) (.12) (.073) (.057)
A (8—32)
— .452(ln —) + .286(D ) + .293(D
(.17) Ad ( 187 )S ( 238 )W
R 2 = .693
S = .547
e
where D = .dunimy variable for South, 1 if city is in South, 0 otherwise.
D = dummy variable for West, 1 if city is in West, 0 otherwise.
These results can be interpreted to say that total costs were e 286 =
1.33 higher in the South than the North and e 293 = 1.34 higher in the West
than the North, all other factors being equal. Nevertheless, neither of the
two dummy variable coefficients are significant at a five percent level,
nor does equation (8—32) result in statistically significant improvement in
over equation (8—36) at a five percent level. Thus the analysis does not
support the existence of a regional effect. This does not mean that the
factors discussed above are unimportant, although that is one possible
interpretation. Since the regionalization was very crude, it may easily have
blurred important distinctions between areas, and therefore the lack of
statistical significance may be due to the lack of precision in the variables.
Use of Regression Equations
The analysis up to this point has focused on explaining the variation in
costs for a readily available sample of drainage systems. The results suggest
that relatively simple models can be developed for predicting conventional
collection costs, but the use of the specific equations estimated here must
be approached with some caution. If the sample used is not representative of
the types of systems for which we wish to predict costs, equations based on
the sample are likely to give imprecise or possibly biased estimates of costs.
In order to see whether such limitations are important for this data set, we
have examined predictions from our estimated equations against independently
reached estimates for a typical low density subdivision in the Boston area.
In addition, we have compared descriptive statistics on the sample to char-
acteristics of typical new subdivisions.
Costs in the Boston area--
A typical low-density subdivision in the Boston area might have the
following characteristics:
area = 50 acres
lot size = 1 acre
length of pipe = 4000 feet
slope = 5 percent
runoff coefficient = .25
design storm = 5.5 inches/hour
86
-------
According to design data from our consultant, BSC Engineering, the costs
for the storm sewer system would be approximately $60,000. We can compare
this figure with the following predictions where costs have been adjusted to
March 1977:
equation (8—21) CT = 75,927
equation (8—26) CT = 70,967
equation (8—27) CT = 111,570
equation (8-28) CT = 118,007
equation (8—31) CT = 175,332
equation (8—32) CT = 68,243
The models generally over—predict total costs, although the range is
considerable. In this specific case equations (8—21) . (8—26) and (8—32)
which are based on total cost. are better predictors than the equations esti-
mated on a cost per foot basis. despite the fact that the latter fit the
sample data much more successfully.
Table 8—2 compares descriptive statistics on selected variables in the
sample, with values which would be typical of new, small residential sub-
divisions in the Boston area. This table shows that there are substantial
differences between the two sets of numbers. Slopes in the sample tend to be
flatter than typical Boston area values and the runoff coefficients are high
relative to values for single family home developments, although more typical
of higher density uses such as townhouses. It is particularly interesting to
look at the drainage density and cost per foot figures. Systems in the sample
utilize less pipe per acre of land but with higher unit costs than the new
subdivisions. This result suggests that many of the systems in the sample
represent parts of interconnected collection systems, whereas many new sub-
divisions at the urban fringe have independent collection systems for
stormwater. That is, each system handles runoff for the specific subdivision,
not for upstream users, and discharges either to a watercourse or large trunk
lines. There are, of course, other explanations for the differences as well.
If many of the sample systems were existing built—up areas, then the cost of
line installation would be much greater than when land is first developed.
Whatever the reason, it is apparent that our caution in using the estimated
regressions for predicting new subdivision costs is well founded. Given the
differences between the sample and the subdivisions of interest, it is evident
that predictions based on these regressions involve extrapolations which tend
to magnify the estimation errors of the models even if the specifications
th xrtselves are perfectly appropriate. The results of the cost comparison also
suggest that the simple equations (8—21) and (8-26) may be more resistant to
the problem than the models which rely on more detailed information, but it
is impossible to make any generalization based upon this single cost
comparison.
ADAPTING COST FUNCTIONS FOR NON-CONVENTIONAL MANAGEMENT SYSTEMS
The cost functions developed for conventional collection systems can be
extended to handle certain kinds of additional management options. These op-
tions include systems which either add an additional cost component to the sys-
tem and/or directly impact one of the underlying variables in the equation.
87
-------
TABLE 8-2. COMPARISON OF DRAINAGE AREA CHARACTERISTICS
Typical values:
Sample Values Boston area
Median Hinges Developments*
Slope (%) 1.1 .4; 1.95 1—10
Runoff Coef. .5 .45; .50 .25 SF (large lot)
.5 Townhouse
.75 Apt.
Drainage Area
(acres) 64 36; 110 30—200
Drainage System
density
(ft/acre) 54 36—83 80—200
cost/ft
(March 1977
dollars) 60 50; 78 15—20
*
BSC Engineering, Inc.
The use of a storage basin to serve the entire subdivision is an example
of the first case. In this case it can be assumed that the role of the col—
lection system is unaffected and that the final pipe in the subdivision dis-
charges to the basin. Thus the cost function would be written:
CS = CT (C,I,A,Ad,S) + CB (8-33)
where Cr, is the total collection cost function discussed previously and CB is
the cost of providing a storage basin.
This cost can vary widely depending upon the nature of the site, the value of
land, and the design of the storage system. Benjes (1976) and Sullivan, et
al., (1977) have presented cost functions for stoi mwater storage facilities
which relate capital cost to the storage volume of the facility. However.
the types of facilities covered in these studies are for metropolitan scale
management, not for small subdivisions. Engineering Science, Inc. (1973) has
developed cost estimates for sedimentation basins that are probably more
representative of the simplest types of storage facilities that might be
constructed. These costs are related to the size of an earth dam and outlet
structure of specific dimensions, not directly to volume of the facility. By
making specific assumptions about the slope and other characteristics of the
site, however, it is possible to convert these costs to functions of storage
volume.
Increased on—site infiltration is an example of a management system which
affects the underlying variables in the collection cost function. A simple
88
-------
technique of this type is the discharging of the roof gutters to the lawn
rather than to the driveway or collection system. This would appear as a
decrease in the value of C for the cost equation and would reflect a decrease
in the maximum size of pipes in the collection system. For example, according
to equation (8—26), a reduction of ten percent in the runoff coefficient
would result in approximately the same percentage reduction in total costs.
Stormwater management measures that fundamentally change the type of
transport system, for example, by using swales in place of sewers, could not
be treated as simply as the previous two cases. While the basic conceptual
analysis that underlies equations (8-11) to (8—18) would still apply, specific
equations estimated from data on conventional systems would be inapplicable.
A new set of cost data would be needed to estimate functions similar in form
to those for the conventional system.
89
-------
SECTION 9
INSTITUTIONAL AND POLITICAL ISSUES
INTRODUCTION
While the technical problems in stormwater management may be formidable,
the ultimate challenge may yet prove to be that of implementing proposed
solutions at a local level. Most of the measures being considered as
options for on-site control of stormwater are, by their nature, different
from methods generally practiced. Implementing these measures will there-
fore require that local governments and regulatory agencies alter standard
procedures and require new types of information and cooperation from devel-
opers. Since any change involves a certain risk, there will be some
understandable reluctance on the part of these governments to make changes.
This reluctance may be further enhanced by pressures from groups that feel
threatened by changes in the existing procedures. In this section we
examine the possible difficulties that can arise in introducing innovative
stormwater management measures.
REGULP IDRY SYSTEMS AFFECTING INNOVATION IN RES IDENTIAL DEVELOPMENTS
There is not much experience in introducing innovative stormwater
control measures at a local government level. As a consequence, it is
not possible to draw generalized conclusions about difficulties that may
be encountered and ways of overcoming such difficulties from the limited
number of examples that currently exist. Therefore, the first approach taken
in this study was to examine the pertinent literature on similar innovations
in residential development for which there is more extensive experience.
Two types of innovations were selected: the introduction of Planned Unit
Development (PUD) regulations in zoning and subdivision ordinances; and the
modification of building codes to allow innovative construction techniques.
Both of these examples resemble the stormwater case in key respects: they
are tied to the process of developing new residential subdivisions; their
responsibility rests almost exclusively with local boards and agencies; and
they represent examples of innovations which, proponents argue, promise
substantial benefits for the “general public.” There is also a more direct
connection: PtJD ordinances allow the possibility of using residential
development layouts to manage stormwater (see Section 6).
By examining the studies of innovation in zoning subdivision and
building codes, it is possible to develop a general appreciation for the
local factors that will influence the acceptance of stormwater control
measures and how these factors may vary according to geographical location,
90
-------
socio—economic differences, and other considerations. This discussion can
then be used as a general background against which the two case studies that
follow can be assessed. These studies involve subdivisions featuring
innovative stormwater management plans designed by BSC Engineering Inc.
In both cases towns in southeastern Massachusetts accepted the plans only
after considerable difficulty.
PLANNED UNIT DEVELOPMENT REGULATIONS
Planned unit developments (PUD) are large scale residential developments
which cluster housing on part of the site, leaving part for open space. These
developments are a fairly recent innovation in site design. They allow the
developer to minimize utility costs and in some cases, to mix the types of
housing provided, Zoning and subdivision regulations are the main tools used
by government to guide development including PUDs. Regulations specifically
aimed at PUD have developed from these two traditions. We will start with an
overview of zoning and subdivision regulations and then discuss PUD5 in more
detail.
According to the National Commission on Urban Problems (NCUP), every
state has enabling legislation which allows the use of zoning and subdivision
regulations, and more than 10,000 local governments have adopted them (1968).
A survey conducted in 1968 showed that 90 percent of all cities and towns with
populations over 5,000 had zoning ordinances, and 83 percent had subdiviison
regulations (Manvel, 1968). These regulations were first widely used in the
1920s and still reflect patterns established at that time, but they have been
tailored to meet the objectives of each local government.
A zoning ordinance typically assigns a use category such as residential
or industrial to a geographic area, establishes maximum allowable density,
regulates building bulk, and includes a zoning map which establishes
districts of uniform uses. It is administered by a self-executing, non-
discretionary permit process which also provides for zoning appeals for
variances and for amendments for rezoning. Significant differences exist
among various regions of the country in the distribution of zoning powers
among levels of government (NCUP, 1969). In the West, Midwest and South both
counties and municipalities have zoning powers. This is also true in some
northeastern states, but in the six New England states the counties have no
zoning powers.
Density limitations, usually expressed as minimum lot sizes, are
particularly inflexible aspects of zoning ordinances. The 1968 survey showed
that 94 percent of governments with zoning regulations had minimum lot size
provisions. Over one—fifth prohibited lots of less than one—quarter acre.
New England townships are stricter; over fifty percent disallowed less than
quarter acre lots (Manvel, 1968).
A typical subdivision regulation covers site design and relationships,
insures that utilities tie into those of adjoining property, allocates costs
of public facilities such as sewers, and sometimes provides for dedication of
land for schools or parks. It is administered through the provision of
general design standards which are applied by the local planning commission
91
-------
or governing body to preliminary and final plans submitted by the developer.
Subdivision approval is a bargaining process between the municipality and
the developer. Standards frequently established include width and alignment
of streets, dimensional requirements for lots, street grading and paving
requirements, standards for curbs, gutters, drainage and sidewalks, and
utility systems. Subdivision control is generally more flexible than zoning
and involves negotiations on each separate tract. Since subdivision control
relates to vacant land, counties play a larger administrative role than with
zoning, particularly for unincorporated territory. However, municipalities
also exercise this authority, frequently with extra—territorial control.
Both zoning and subdivision regulations are characterized by local
responsibility. While state enabling legislation is required, they reflect
local policies administered by local officials. They presuppose land
development on a small scale, lot-by—lot, and are structured for administra-
tive convenience, zoning to prevent change in established neighborhoods
and subdivision control to protect the public interest in servicing land
to be developed in the future. In discussing the failure of the local
approach to planning to promote coordinated land use, Lamb (1975) observed
that local government structures lead to ad hoc decision making based upon
immediate political, economic and social pressures. In addition, local
officials may be major landowners themselves or be highly dependent on
campaign contributions from large property holders, and thus have an interest
in preserving the status quo.
Recent changes which have occurred in these regulations have included a
widespread reduction in permitted residential density, an increase in the
number of subjects regulated (for example, landscaping), and innovative
administrative changes that allow the locality to adopt a wait and see
approach (as opposed to self-executing standards) leaving the initiative to
the developer. The planned unit development or PUD is an example of this
change in administration which has only been widely adopted since 1963.
The PUD technique applies requirements to an entire project rather than
on a lot—by—lot basis. It also requires discretionary public review of site
plans proposed by the developer, sometimes combining zoning and subdivision
control into a single process. Approval of a planned unit development
depends on the fulfillment of certain conditions. These generally include
minimum size, permitted uses, maximum density, provision of open space and
public facilities and maintenance of control of the area during and after
development. Implementation of PUDS usually occurs through zoning amendments
which require legislative action, through approval of special permits by
the board of adjustment, or through planning commission authority to approve
special exceptions or conditional uses; local provisions vary widely.
In 1968 the National Commission on Urban Problems reported that 45
percent of localities with zoning ordinances provided for specialized treat-
ment of PODs (Manvel, 1968). As of 1973, six states (New Jersey, Pennsyl-
vanja, Connecticut, Kansas, Colorado and Nevada) had enacted PUD enabling
laws based on the Model Act published in 1965 by the Urban Land Institute
and the National Association of Homebuilders (Bangs, 1973). Six other states
92
-------
had included PUD-like techniques in their planning and zoning enabling
legislation.
A 1971 survey conducted by Burchell indicated that PUD projects tend to
be located in states with high urban growth and those with specialized climate
or topographic conditions conducive to recreation housing (Burchell, 1972)
Burchell attempted to relate the existence of municipal PUD ordinances and
developer inquiries regarding PUD with various socio—economic characteristics.
He was unable to determine any significant relationships other than the
availability of sufficient vacant land to accommodate large—scale projects
and the use of a municipal planning consultant. Burchell concluded that a
significant portion of the variation in incidence of PUD legislation
enactment was unexplained by the study and could probably be attributed to
the variation in sophistication of local officials and in developer
promotional ability.
Sophisticated approaches such as PUDs require sophisticated administra-
tors. Many urban fringe jurisdictions which are experiencing development
pressures do not haveadequate staff to handle large scale development. The
typical reaction of local governments such as these is either to defer to
the developer, to impose traditional unsophisticated controls or to try to
prevent development completely.
Jan Krasnowiecki, a leading theoretician of land use law, indicates that
there is a growing tendency in PUD ordinances to limit permissible densities
to the same level as permissible under applicable standard zoning because
municipalities are afraid that PUD represents growth (Krasnowiecki, 1973).
He also feels that an increasing amount of detail is beginning to find its
way back into PUD ordinances, so that a developer may be turned down for a
“legitimate” reason or so that he must build very expensive housing. This
conclusion is supported in Table 9—1 which shows the results of an American
Society of Planning Officials study of PUD ordinances. Local officials are
not comfortable with the resonsibility of a negotiated project, and many do
not have the necessary professional staff, so they prefer to rely on
ordinance provisions.
In discussing the problems encountered by large—scale developments, the
Urban Land Institute ranks high the conflicts and uncertainties caused by
public agency regulations (Urban Land Institute, 1977). In many localities,
developers must proceed through a maze of regulations with different officials
reviewing the plan for street layout, sanitary facilities, building location,
etc. Often the process of obtaining permits turns into prolonged negotiations
involving considerable uncertainty as to the final standards to be imposed
on the project. Delays and uncertainty have major cost implications including
extended loan payments and higher interest rates due to increased risk. As
stated by the National Commission on Urban Problems, “Institutional delays
are not uncommon in communities which prefer no development at all and the
disapproval of a single official, perhaps on personal whim, can destroy the
proposal entirely or set it back for months. Even where PUD is theoretically
available to developers, some prefer to build conventional subdivisions solely
to avoid the added dangers and burdens which administrative processing can
93
-------
TABLE 9 -1. ORDINANCE DESIGN STANDARDS FOR PUDs
Percent of Ordinances
Design Elements With Specific Standards
Uses permitted 79.0
Density 77.8
Minimum parcel size 92.6
Usable public open space 46.9
Private open space 33.3
Maximum site coverage 51.9
Building spacing 44.4
Building bulk and height 46.9
Building architecture 4.9
Location of window walls 7.4
Quantity of parking spaces 74.1
Location of parking spaces 24.7
Perimeter requirements 34.6
School and recreation site 25.9
Streets and utilities 48.1
Landscaping 33.3
Signs and street lighting 35.8
Screening and fencing 38.3
View protection 16.0
Source: So, Mosena, and Bangs, 1973.
dedication
94
-------
impose.” (NCUP, 1968, p. 227). This statement is supported by a survey
conducted by the American Society of Planning Officials of 300 members of
the National Association of Home Builders concerning the PUD review process.
In comparing the processing time for PUD developments with conventional
subdivisions, 69 percent of those responding felt that PUD developments
were significantly slower (ASPO, 1973).
As one developer was quoted as saying, “Some agencies make procedural
requirements so complex that only the ircst determined developer would attempt
to go the PUD route.” (ASPO, 1973, p. 14)
The following case examples will illustrate this point more fully.
In 1968, a developer proposed a PUD for an area of a New Jersey
township zoned for one—half acre single family residences ( House and
Home , 1971). The proposed development included mutli—family dwellings
and commercial and office buildings. The town planning board
rejected the plan because the higher density did not appeal to
them, nor did the concept of preservation of open land, which the
board viewed as a tax loss. So the developer decided to apply
for a conventional single family development. However, many
objections were raised since the town, in fact, wanted no new
development at all. The developer went to court and finally got
approval for the single family subdivision. At that point the
town decided that it wanted a PUD after all. So a PUD ordinance
had to be passed which took six months and involved numerous
public meetings to explain the concept. After the ordinance
was approved, it took another year involving negotiations concerning
tax revenues, school loads, road layout, and provision of utilities,
for the final plan to be approved. At least four years were spent
on the whole process.
In 1965, a cluster development was proposed in a small town
fifty miles from Manhattan (Raymond and May Associates, 1968).
The proposed plan had lot sizes half the size of the zoning
ordinance requirement but designated half the site for open
space. It had a central sewage system and less road mileage
than a conventional development. Town law permits the town
board to authorize the planning board to modify the zoning
ordinance to permit this type of development. Townspeople were
hostile to any form of development, especially homeowners adjacent
to the site who were afraid that the smaller lot sizes would
attract undesirables. The town board rejected the proposal on the
basis that the reduction in lot sizes did not meet the general
character of the area. The developer countered by submitting a
conventional grid system plan meeting the zoning regulations.
This was rejected by the planning board partially on the basis
of the lack of a central sewage system, even though there were
no regulations requiring such a system. Just prior to the
decision a new planning board member had been appointed who was
95
-------
an adjacent homeowner and an opponent of development on the
site. The developer took the decision to court but lost his
case.
BUILDING CODE REGULATIONS
A building code is a set of specifications designed to establish minimum
safeguards in the erection and construction of buildings. It is a checklist
that a unit or structure must satisfy before it is approved for construction.
Codes have been used since colonial times; today almost every community has
one in force. Manvel found that 80 percent of all cities and towns of over
5,000 population have building codes (1968). Like zoning ordinances,
building codes are formulated and enforced through the police powers of
state governments which have traditionally been delegated to and exercised
by local governments.
The typical building code is composed of three parts: definition of
terms; licensing requirements; and standards. Each code defines key work
areas such as what constitutes plumbing work, and authorizes, through
licensing requirements, who is to do what work. These provisions serve to
guarantee certain work for certain groups such as building trade unions.
Building codes are based on standards which are technical specifications
that determine whether a product, subsystem or complete housing system meets
minimum requirements for health and safety. Most standards specify how
an element of a housing unit must be built rather than how it should perform.
If a new technology does not conform to the specifications of the standard,
it cannot be approved, even if its performance is better. Some codes
contain “equivalency” clauses which allow the local code enforcement
official to determine if a material or process is of equivalent quality to
that specified in the standard. However, this determination is generally
difficult for a local official because of lack of technical knowledge and
experience.
Product standards are developed by the various building trade
associations representing specialized product groups such as the Cast Iron
Pipe Institute or the National Forest Products Association. One cause of the
lack of performance standards is the resistance of product manufacturers to
establish standards that open the market to newcomers.
Standards are incorporated into local codes through direct lobbying
by industry representatives at the local level or through one of the model
code associations. There are four major model code groups for the
building construction industry, each generally drawing membership from
one region of the country. The Building Officials’ Conference of America
(BOCA) (Basic Building Code) dominates in the Northeast, the International
Conference of Building Officals (ICBO) (uniform Building Code) in the West,
and the Southern Building Codes Conference (SBCC) (Southern Standard
Building Code) in the South. Both BOCA and ICBO compete for city members
in the North Central region. The National Building Code, published by the
96
-------
American Insurance Association, has been adopted by 1600 communities in all
regions of the country except the West. The first three model code groups are
associations of building officials. In the West and South over the past
fifteen years, regional model codes have been generally adopted by all
municipalities (Falk, 1973). In other areas of the country, there is more
variation among localities.
The building official insures that the building code standards are met.
He is responsible for approving building plans, inspecting construction of
buildings and issuing permits of occupancy. As mentioned above, he generally
has special approval power over new technologies. Local building code
officials are not well paid and frequently lack technical training, job
security and on—going educational opportunities. Most officials come out
of the local construction industry. These factors tend to make them con-
servative toward new technologies arid building code reform (Field and Rivkin,
1975).
There are also other reasons why local control of building codes has
had adverse consequences for the acceptance of new technologies. The lack
of a uniform set of substantive requirements means either that separate
designs must be prepared for each locality or that the product must meet the
most stringent local requirements, both of which cause cost increases. Some-
times local enforcement causes an inspection problem. - For example, sophis-
ticated production techniques used closed wall construction which cannot be
visually inspected. In reaction, local ordinances may ban manufactured
housing entirely or require the producer to remove the walls off the site,
nullifying any cost saving.
Furthermore, as mentioned earlier, most building codes are written in
specification terms rather than performance terms so that a new technology
which does not conform to the specifications cannot be approved. Any
radical innovation, such as housing manufacturing or industrialization, which
affects major portions of the housing unit is resisted if it means loss of
jobs or sales to subcontracting firms or individuals. In addition, when a
material producer proposes changes that affect other firms in the industry,
for example, the plastic industry t s proposed use of plastic pipe in place of
cast iron pipe, resistance appears. Building codes are a part of this
resistance because, as discussed above, their standards are developed by
building trade associations and their provisions are influenced by lobbying
by industry representatives. Through the efforts of building supply
manufacturers, dealers and building trade unions, certain specific technol-
ogies (e.g., plastic pipe) have been specifically prohibited by some locally
enacted codes.
The 1968 National Commission on Urban Problems survey investigated the
effect of building codes on fourteen building characteristics or components
which involved building practices that resulted from recent technological
developments. Most of the practices considered were acceptable under the
four regional model codes (Manvel, 1968). The survey showed that one practice
was outlawed by over 60 percent of all governments with building codes, four
were outlawed by over one-third, three by one-quarter, and two by one-fifth,
97
-------
while the other four showed a smaller percentage of prohibition. Very little
difference in prohibition was observed between those governments whose codes
were reported to be based on regional models and those with locally developed
codes.
Thus the establishment of model codes for local adoption has not resolved
the problem of uniformity across municipalities even though model codes are
revised from time to time to accept new products and methods. Most localities
do not adopt a model code -without substantial amendments, and furthermore,
many localities do not adopt the periodic revisions without several years
delay. The National Commission on Urban Problems study results showed that
only 15 percent of all municipalities and New England-type townships with a
population greater than 5,000 had in effect a national model building code
which was reasonably up to date (NCUP, 1968). The rest used their own code
or one based on a state code, failed to keep their model code up to date or
had no code. In addition, local inspectors often interpret the code in a
manner which differs from the language of the model code and from interpreta-
tions of the identical language by inspectors in neighboring localities.
Field and Rivkin, in their book The Building Code Burden , (1975)
analyze in detail the effects of building codes on market aggregation and on
innovation, and relate them to certain parameters such as the vitality of the
construction industry and the form of lcoal government.
They report the results of a second study made of the same 14 innovative
building practices studied by the National Commission on Urban Problems
(see p. 9-15). Each city was scored in terms of the number of code prohibi-
tions among the 14 items. Locally-based codes were more prohibitive than
those based on model or state codes, although many items were still specifi-
cally prohibited even in the latter cities. Code revision indicates
progressiveness; the more recent the code revision, the less prohibitive the
code. Again communities with locally based codes were the least likely to
be current. A high proportion of the largest cities used locally drafted
rather than model codes. Field and Rivkin suggest that this pattern occurs
for two reasons: 1) model codes do not typically cover high rise construc-
tion; and 2) local interest groups such as unions and subcontractors use
stronger influence where a higher volume of construction is at stake. On
a national basis, the older, more industrial sections of the country were more
likely to use locally drafted codes since steady or declining construction
activity probably encourages efforts to preserve local jobs and sales. Thus
the younger, generally faster growing western and southern regions tended
to adopt model codes, whereas the slower growing northeast and north central
regions had a greater incidence of locally drafted codes.
Since market aggregation is important for the growth of innovative
building technologies such as manufactured housing, Field and Rivkin
aggregated the data from their study of the 14 innovative building technol-
ogies to create state—level code prohibition scores. The highest prohibition
scores occurred in areas where locally—based codes were more common, the
mid—atlantic states, the Midwest, and the northern Midwest. Variation in
code provisions from community to community reflects localism, the degree to
98
-------
which decision making takes place within the community and subject to its
influences. Field and Rivkin rated states according to the degree of within-
state dissimilarity in acceptance of the 14 innovative items. The regions of
least dissimilarity were the west and central plains states. Generally, the
more industrial the states, the stronger the local construction industry and
the greater its influence on local code design. Also, the ICBO code associa-
tion appeared to be a strong influence in the West. City government form
may influence the degree of localism. The city manager form of government
is prevalent in the West and the professional orientation of city managers
may make them more willing to accept the ICBO code association recommenda-
tions. The mayoral form of government is more common in the South and North,
perhaps reflecting local attitudes toward professionalism in government.
Further analysis of this data combined with information from a 1970 study of
housing manufacturers indicated that the restrictiveness of codes was
related to the vitality of the construction industry and to the extent to
which political influence could be exerted on local building officials. Slow
growing cities with mayoral forms of government had the most restrictive
codes; city manager cities had the least restrictive codes.
The discussion so far has examined the widespread problem of restrictive
local building codes. It is instructive to review an example of federal
involvement in an attempt to modify this regulatory system. In 1969 the
Department of Housing and Urban Development initiated Operation Breakthrough.
One of its objectives was to develop ways to overcome the perceived con-
straints of local building codes on the growth of the manufactured housing
industry which uses innovative technology and requires national markets.
The strategy selected was to encourage state legislation authorizing state-
wide regulation of manufactured housing which would pre—empt locally enacted
building codes under the states’ police power. When the program was
announced, no states had mandatory state—wide building code regulatory
systems applicable to all forms of manufactured housing. In 1973 there were
28 states with such laws, primarily in the West, Northeast and Southeast
(Falk, 1973). The greatest degree of uniformity among state standards exists
in the West, and the West also has the beginning of a centralized approval
system for new technologies. Those which are accepted by the ICBO Research
Committee are also acceptable by regulation in some states and carry weight
in other states. Thus is the West, the prerequisites for an aggregate
regional market have developed. The situation is not as favorable in other
regions of the country. Attempts to achieve reciprocity or uniformity of
substantive requirements between states have not been successful. Thus
Operation Breakthrough has created statewide markets for manufactured housing,
but not a national market.
SUMI4ARY OF LITERATURE REVIEW
The regulatory systems which we have reviewed hinder the application of
innovative technologies in the residential construction industry. Although
regulations are, of course, necessary to preserve the health, welfare and
environment of a community, their substance and administration vary widely.
Certain factors appear to influence the degree to which these regulations
inhibit innovation. These are summarized here. These same factors are
99
-------
likely to influence the regulation of techniques to control stormwater runoff.
Growth Rate
A fast growing area provides a fertile market for the building industry.
If there is plenty of work, then there is less need to restrict newcomers to
the market, so innovations are less likely to be resisted.
Community Attitude Toward Growth
A “no growth” or “slow growth” community will attempt to use the
regulatory system to keep new development out or to raise costs in order to
increase the tax base.
Influence of Project Manufacturers and Unions
Because of the way the building industry is structured, with many inter-
related firms and individuals, and because these groups have access to the
regulation development process, they tend to have a strong conservative
influence on regulations which can be used to maintain their sales and jobs.
Innovations which do not disrupt their positions will be easier to implement.
This factor is related to the first, since an expanding economy will ease
the pressure to maintain a market share, while a contracting one will increase
the pressure.
Degree of Professionalism in Government
The degree of professionalism in government is likely to affect both the
content and administration of regulations. The more professional government
will turn to national models for the design of its regulations. Such a
government will also be less likely to allow personal interest or political
favors to interfere with regulation administration.
City Size
This factor is important, as it relates to the first factor, since a
large city may have more work available and therefore more permissive regu-
lations. A larger city is also more likely to have an adequate staff with
the technical expertise to administer sophisticated regulations. However,
a large city, especially one located in an older industrial area of the
country, may be slow growing and therefore exacerbate the problems caused
by competition among firms.
Local 1 egulatory Officials
The experience, training, interests, and attitudes of the local regula-
tory officials determine to a large extent the acceptability of innovative
techniques, since these individuals play such an important role in the
existing regulatory process.
100
-------
Geography
In the East and North, regulations seem to be more restrictive, while
in the Midwest, South and West, less so. Rural areas by definition are the
least developed and therefore likely not to have constraining regulations,
although when faced with development may apply unsophisticated controls.
Suburban areas around major metropolitan centers are likely to be under the
most pressure and therefore be the most restrictive.
It is important to note that the factors identified above are not
strictly separable. For example, western municipalities in the surveys cited
tend to be faster growing, have a more positive attitude toward growth, and
are often regarded as having greater professionalism in government than
northeastern cities. Similarly, southern cities are faster growing and are
less influenced by strong unions than northeastern cities. Thus it is not
appropriate to identify a specific effect with a single factor 7 the casual
linkages are far too cou lex. It is clear however, that there is consider-
able variation among municipalities in their acceptance of innovation in the
residential development and construction industry, and that this variation is
at at least partially explained by a number of community characteristics.
Without exception, it is anticipated that the factors discussed will
bear on municipalities’ acceptance of innovations in stormwater management.
Like building codes, new management techniques will require that municipal-
ities take a more flexible approach in the implementation of technical stan-
dards and may entail the replacement of existing products, thereby displacing
some local suppliers and labor. Like PUD ordinances, some stormwater control
measures would require consideration of the layout of the tract as a whole
rather than a lot-by-lot approach and would require flexibility in subdivi-
sion standards for road pavements, curbing and other elements. Thus if past
experience in these related areas is any guide, there will be considerable
difficulty in obtaining acceptance of new stormwater management concepts.
Our two case studies describe the types of difficulties that did arise in two
towns. These are relatively rural communities in southeastern Massachusetts.
Both tend to be “slow growth” in outlook and have small government organiza-
tions with little in the way of specialized technical expertise. Thus we
would predict apriori that they would be highly resistant to new innovations.
CASE STUDIES 1
The two case studies describe the process of gaining acceptance of
innovative drainage and roadway designs for two residential developments in
Southeastern Massachusetts. Both developments are located in affluent
bedroom suburbs of Boston, have similar populations, lie within the coastal
plain of Massachusetts Bay and share the same vehicular corridor to Boston.
Despite these similarities, they provide an interesting comparison since
their sites have completely different geomorphological characteristics and
This subsection is based on a document prepared by BSC Engineering
Inc. under subcontract to Meta Systems Inc.
101
-------
and the communities in which they are located have contrasting land—use
regulatory systems.
Trout Farm—-
Trout Farm is a unique residential development of clustered homes in
Duxbury, Massachusetts. It was the first project to receive full approval
under Duxbury’s innovative Impact Zoning Ordinance which was adopted in 1973.
This zoning by-law permits alitxst complete flexibility in design of the
development. Density is negotiated under a ceiling of two dwelling units
per acre, depending ontheproposal’s sensitivity to the site constraints
and off—site socio—economic impacts.
The ordinance was the by-product of a comprehensive community planning
exercise which included the development of a socio—economic nodel of the
community. This i*del is used to weigh additional housing demands versus
available community infrastructure capacity. Site development features,
housing types, bedroom mixes and housing densities are compared to their
projected impacts. Tax revenue/services costs (benefit/cost) analysis is
made for the project and becomes a part of the approval process for a
proposed development.
Under the Duxbury Protective Zoning By-Law, the approval process
requires submission of plans at three stages and concurrent review at each
stage by all municipal agencies. The multi-agency review/negotiation
period specified in the by-law includes a 60—day approval period for a
special permit and is designed to be completed in less than a year.
Geonorphology--The 42 acre Trout Farm site is a white pine forest with
little forest understory. The site is roughly divided by a small ground-
water fed stream of very high quality which discharges to a small, shallow
cranberry bog reservoir no longer in agricultural use. The valley of the
stream is covered by wild fields with a mixed succession of soft and hard-
wood trees and shrubs reclaiming the once cleared fields.
The soils are exceptionally well drained glacial outwash deposits.
Depth to the groundwater table ranges from zero at the brook to greater
than 20 feet near some residences. The soils generally are classified
as Merrimac sandy loam and Carver coarse sand under the USDA Soil Conserva-
tion Service classification system. They are highly permeable soils with
pernEabilities in excess of 6 inches per hour. Actual percolation tests
in carefully constructed test holes with adjacent observation trenches showed
percolation rates of 30-45 seconds per inch with a roughly 1 horizontal
to 5 vertical gradient through the unsaturated soils and a saturated
gradient (slope of the phreatic surface) of .0l-.03 ft./ft.
Because of the sensitive nature of the brook —— a stream of two— to
three—foot width and six-inch depth, primarily groundwater fed -- and the
characteristics of the soils, it was decided to design a completely infiltra-
tive drainage system. The objective was to increase the infiltration charac-
teristics of the site by developing it. The approved plan allows for 105
102
-------
dwelling units of various types, bedroom counts and ownership arrangements.
Generally, the housing is set in the higher wooded areas of the site and the
lower elevation fields and stream are preserved as common open space. All
the roadways have a stabilized stone surface and all the roofs are piped
through downspouts to leaching pits. Other than the roof leaders there
are no drainage structures, pipes or outfalls anywhere in the project. The
history of the site plan approval was as follows:
• Late in 1973 the first stage was completed. Preliminary
Qualification and Site Analysis submissions were presented and
preliminary approval was received to proceed with the Tentative
Plan submission.
• Planning for the site was carried out and the complex graphic and
text submission required for the second approval stage was prepared.
The Tentative Plan for a special permit exception was submitted
on June 4, 1974.
• Based upon initial comments during review meetings, supplemental
text and graphics were prepared and submitted on July 29, 1974.
This text nearly equaled the initial submission.
• A public hearing was held at the High School Auditorium on
August 29, 1974 and the special permit exception issued on
November 26, 1974, nearly six months after submission -— not
two months as specified in the by-law.
• The permit provided general conditions on which the project
could proceed. These conditions set forth further planning
requirements to be approved prior to preparation of the final
SITE PLAN . Studies and meetings continued, and on May 27, 1975
an amended special permit was issued.
• During this time the housing market was deteriorating. Market
acceptance of condominiums was particularly poor, and it was
decided to modify the design by changing most of the attached
housing to a “zero—lot line” clustered subdivision. The physical
layout of units remained essentially the same except lot lines
were drawn through the party-walls of attached units. This
change greatly enhanced the market acceptance of the units, even
though the lots were small and restricted to their natural pine
needle cover.
• On March 5, 1976 the final site plans were submitted and reviewed
by agencies and their consultants.
• Final permits were issued in May of 1976, nearly three years after
the beginning of planning.
• Sewage disposal permits and curb opening approvals were not issued
until October of 1976.
The process, which was intended to take less than a year, actually lasted
almost three years. why did the delays occur in the streamlined multi-agency
review and one—stop permit process? Clearly, since this was the first
103
-------
proposal to go through the process, some slowness due to inexperience was
to be expected.
The multi-agency nature of the review process itself, however, caused
certain difficulties. At each stage, the submitted plans were distributed
to all the local boards and, in turn, to their consultants for review. The
planning board included both a planner and an engineer; the board of health
had an engineer and also forwarded plans to a state agency for comments;
the water department had a consultant; the conservation commission included
engineers and hydrologists; and other boards did their own review. As one
might expect due to the number of people involved, many of the review
comments conflicted with each other and negotiations were required to
resolve the various opinions. These negotiations did not concern housing
density and infrastructure demand as envisioned in the town ordinance. For
example, a lively debate took place between the town board of health and
the state environmental agency. Frequently, the developer had to take a
mediating role in order to keep the permit approval process moving. In
addition the uniqueness of the roadway was a source of considerable concern,
particularly from engineering consultants with a more conservative or tradi-
tional point of view. Questions asked repeatedly were, “Where have you
done this before?” or “Can you prove it will work over time?” What
finally won the approval of the infiltrative roadways were not technical
argwnents but the design of a politically satisfactory package in which
granting approval would not make existing residents liable for future
maintenance or repair costs for the roadway. This will be discussed further
below.
Bowker Woods--
The Bowker Woods project in Norwell is a conventional subdivision in
which the homes are clustered on oversized lots. The zoning in Norwell
permits only conventional subdivisions of single-family detached residences
on one-acre lots. Lots must have 150 feet minimum width at the building
sethack line and a shape which permits the enclosure of a 150-foot diameter
circle in which the house and septic system must be placed. No wetland area
in any lot can be counted toward the minimum one—acre lot area.
All new roads in Norwell must have a right-of-way depth of 50 feet and
meet strict horizontal and vertical alignment criteria. The pavement width
must be 26 feet and must have curbs and 5—foot sidewalks on each side.
Minimum permissible water main size is ten inches. Pavement section require-
ments are as follows: 1 1/2 inch bituminous concrete top course; 3 1/2 inch
bituminous concrete binder course; .05 gal/sy MC 700 tack coat; 12-inch
gravel base course —— 95 percent compaction; 23-inch non—frost susceptible
material -- 97% compaction.
Geomorphology--The 36-acre Bowker Woods site was a former wood lot for
a box-mill and had been nearly completely cut-over during the early part of
this century. A strong second growth of mixed deciduous trees occupies the
steeply sloping upland areas. The site is rather gently sloping along an
old wood road, but rises sharply to the north and west with slopes over
25 percent. The southeast corner and a significant part of the parcel’s
frontage are upland red maple swamp.
104
-------
The steeper portions of the site are very dense stony till deposits
or dense silty till deposits with high groundwater levels. In fact, ground-
water seeps from the surface of the ground during the spring. The wetlands
areas are unsuitable for development. Only the narrow band of land along
the old woods road contains some ice-contact deposits of permeable soils.
The SCS soil classifications of the till deposits are Essex —— very stony
coarse sandy loam, Norwell —— extremely stony sandy loam, and Gloucester ——
very stony find sandy loam. The glacial retreat deposits are Brockton and
Merrimac sandy barns with permeabilities in the six inches per hour range.
The parcel fronts on an existing public way, Bowker Street -- a desig-
nated historic way. It has very poor vertical and horizontal alignment,
narrow (l6—ft) pavement width, small culverts with little or no cover, and
no drainage. Ice and frost conditions in winter require the highway depart-
ment to make the road one—way. However, residents have no desire to change
its character because it controls traffic speeds and maintains the town’s
original country flavor.
Prior to the involvement of BSC Engineering Inc. in early 1974, a 27-lot
conventional subdivision planned by others had been proposed for the Bowker
Woods area. This proposal was denied because of adverse impact on the site.
Planning was begun by BSC Engineering for a 19—lot development with a unique
roadway design which required relief from some of the town’s road cOnstruction
standards. This resubmission in October 1974 also was not viewed favorably
primarily due to sewage disposal considerations. A detailed soil investiga-
tion was then conducted to determine the soil characteristics and their
suitability for on-site sewage disposal systems. This study indicated that
the site was severely constrained from a geological standpoint.
The area of suitable soils was mapped and the regulated sethack from the
wetlands was plotted to determine the net buildable area in terms of sewage
disposal systems. It appeared that the developer could reasonably expect
to receive ten sewage disposal permits, since approximately ten lots could
be placed on suitable soils.
Ten was also the number of lots which could be located on the site
if it were simply divided while meeting the town zoning requirements and
fronting on Bowker Street. The planning board could not legally deny such
a proposal. Because of these considerations, it was possible to use the
ten—lot number to negotiate for significant waivers in lot configuration and
roadway standards. The planning board agreed to allow gerrymandering of the
lot boundaries and a narrow, ecologically sensitive roadway design generally
following an old cart path, as long as the developer agreed to go through a
detailed subdivision review process which otherwise would not have been
required. The ten—lot development plan was approved on October 27, 1975.
Municipal Financial Risk Considerations
In both Duxbury and Norwell town officials showed concern regarding the
future performance of the unique roadway designs included in the two
development proposals. BSC Engineering had not designed or built such a road
105
-------
previously, nor had a similar installation been developed by others in the
area. Thus adequate performance could not be proved from past experience.
The approach used to gain approval for the new roadway designs was to be
open about their experimental nature and to request that town officials
become part of this advance in technology. However, care was taken to insure
that neither elected officials nor other residents of the coimnunity would
risk any future public expenditures.
In both Trout Farm and Bowker Woods the roadway right-of-way geometry
meets towr . standards. However, the ways are private and their maintenance
cost is borne by the homeowners in the development. Therefore, if in the
future the roads completely failed, requiring the towns to take action, no
land—taking would be required. Furtheri ore, if reconstruction were required
at some point, betterments would be char d to the abutting homeowners,
thereby passing on to them road construction charges that would have been
included in the cost of their lots had a nxre traditional roadway design
been used initially.
In both cases during the negotiation period, BSC Engineering illustrated
how the roadways could be converted to nore conventional street designs
completely within the dedicated private ways. In Trout Farm catch basins
and leaching pits could provide drainago for an impervious surface. In
Bowker Woods, where the soils do not permit total leaching disposal,
conventional basin and pipe design was demonstrated which discharged to a
retention pond, and outlet structures were designed assuming an impervious
roadway of conventional width. The public was granted rights of access
to the retention pond.
These agreements concerning future financial risk and the discussions
regarding alternative roadway designs were all part of the public record.
Their validity is upheld by the covenants placed upon the land and in the
language in the homeowners’ association agreement.
REASONS FOR PROJECT SUCCESS
As noted in the first part of this section, many innovative developments
are discouraged or abandoned when faced with complex and lengthy approval
processes. The two projects that have been described in the case studies
managed to endure through such processes as well as through a severe economic
decline. Residential developments were produced which were different from
any others in their coninunities. What was unique about these projects, their
developers and the municipalities which permitted success? Following are some
of the major reasons for their uniqueness:
• The developers in both instances were known in their communities.
The Trout Farm owners are a group of local businessmen who have other
sources of income and who had a strong desire to do something creative.
The Bowker Woods developer is an old Boston real estate firm with an
excellent reputation that had the ability and willingness to continue
through all of the negotiation and approval steps required.
106
-------
• The developers owned the properties outright and were not under time
pressure to obtain the approvals and close on the property.
• Local officials and municipal agency personnel wanted to see something
creative done and were willing to risk the political consequences if
the projects turned out poorly and became a source of complaints and
ridicule.
• The high—income nature of the developments meant that the developers
could afford to do proper planning.
• The developers had a reputable planning team that was willing to do
some unusual design work.
107
-------
APPENDIX A
ANALYSIS OF LOADING AND WASHOFF DATA
INTRODUCTION
Two basic strategies adopted in gathering data for use in quantifying
urban stormwater pollutant loadings are discussed below. One strategy has
concentrated on measuring the acci nulation and composition of dust and dirt
on street surfaces. Street accumulation measurements have been taken to pro-
vide bases for evaluating street sweeper effectiveness and for calibrating
the accumulation/washoff functions typically included in various urban runoff
models. These studies are plagued by the difficulties involved in estimating
deposition rates from street solids measurements in the presence of a variety
of et— and dry—weather deposition and removal mechanisms. Uniform sampling
and data reduction procedures for measuring street solids have not been
defined. Further, both pollutant loadings scoured from off-street surfaces
during rainstorms and possible non—conservative behavior of some pollutants
in stormwater collection and transportation systems are ignored when street
solids measurements are used to estimate loadings to urban waterways.
The second strategy is essentially end-of-pipe sampling. Measurements
of flow and concentration provide the basis for estimation of loadings.
Difficulties are associated with the temporal variability of flow and concen-
tration typical of urban storm events. Such variability can cause sampling,
measurement, and data interpretation problems. Possible seasonal or longer-
term effects suggest that such studies should be carried out for long periods
to provide a basis for estimating characteristic loadings.
BRIEF DISCUSSION OF PREVIOUS STUDIES
Currently, the n st widely quoted and applied study on accumulation rates
of solids and associated pollutants on street surfaces is that performed in
1974 by TJRS Research Company (tJ RS, 1974). Their results have been published
in at least three forms (tTRS, 1974; Singh, 1977; and Bradford, 1977) and have
been adopted by authors of various manuals dealing with methodologies for
urban runoff assessment (McElroy, ét al. , 1976). The URS study consisted of
compilation and analysis of existing data (as of 1974) on street solids
accumulation rates and composition.. The data were drawn primarily from four
studies —— APWA, 1969; AVCO, 1970; and tJRS, 1972 —— and are listed in the
1974 UPS reports.
A total of seven “independent” variables were considered: climate
(region); land use; average daily traffic, type of landscaping beyond
108
-------
sidewalks; type of street surface; days since last rain; arid days since last
sweeping. “Dependent” variables included solids loading rate (lbs/curb—itiile/
day) and solids composition (micrograms/gram), comprising a total of 16 con-
stituents and 153 observations.
The authors anticipated and observed a great deal of variance in solids
loading rates and composition within each independent variable category. This
variance was attributed to “natural” variability and to differences in sample
collection and analysis methodologies. The latter factor was undoubtedly
important, especially because of the methods used to aggregate samples, and
because, although most of the observations were based upon street accumula-
tion measurements, some were based upon direct runoff measurements. The
analysis was also hindered by missing values and by collinearity in the inde-
pendert variables.
The data were grouped according to independent variable categories, and
averages and standard errors were computed for each group. For each depen-
dent variable t—tests were employed to test for significant differences
between each sub—group and the overall data set. In a formal statistical
sense these tests were not done properly, since in each case the data from
the tested group were contained in both populations being compared. The tests
should probably have been done between each group and the remaining data.
Also, the comparisons are incomplete: t—tests to consider whether two popula-
tion means differ assume that the means are drawn from a distribution with a
common variance; there were no checks made to see if this was in fact the
case. Failing this, the t-tests can tell us little about the differences in
population means. In fact, we might expect the sub—population variances from
different land use categories to be different, and there is a wide range in
the sample variances that tIPS presented.
EXPLORATORY DATA ? NALYSIS
The exploratory process partially relieves us of the burden of having
a specific explanatory model of the data at hand. Thus we start with a dis-
play of the data. As described in Section 5, the graphical technique used
is called a stem—and-leaf display (Tu.key, 1977). The basic purpose of such
displays is to arrange raw data into roughly numerical order -- like a histo-
gram —- so that one can easily answer questions such as: what is the largest
value? the allest? what does the distribution of values look like? Unlike
a histogram, however, the display retains s of the identity of the original
data by using the actual digits of the data values to construct the display.
This makes it easier to see which data value is located where in the histogram.
How then do we construct the display? We first scan the list of the
orj ina1 data (see Figure A—i) to determine a suitable scale for the values.
In this case units of tens seem appropriate. Taking the first data value,
400 lbs/curb—mile/day (112.7 kilograms/curb—kilometer/day), we cut it to units
of tens (the ones place is truncated). Four hundred becon s 40 (tens). Next,
40 is separated into two parts: the right-most digit, 0 and the rest, 4. The
0 part is called the leaf (because it projects from the left—hand side), and
the 4 is called the stem . We show the separation of the stem from the leaf
by drawing a vertical line between them -- 4** 0; the asterisks indicate that
109
-------
there are really two digits in the leaf (400 really could become 4100), but we
show only one of them -- the tens digit. The display is built by placing the
stems on the left and the leaves on the right; Figure A-2 shows the step-by-
step construction of the display. At each step the number placed into the
display is marked with an arrow. The next data value is 600; it becomes 6**l0,
and its placement on the display is shown by Step 2 of Figure A-2. Value 390
becomes 3** 19 (Step 3), 210 becomes 2**Il, 170 becomes l** 17, 19 becomes 0**Il,
and so forth. Data value 32 appears on the same line as data value 19 and
becomes 0** 113.
FIGURE A-i. RAW DATA: RESIDENTIAL LOADING RATES
(LBS/CURB-MILE/DAY)
(URS, 1974)
400 600 390 210 170 019 032
121 148 081 062 121 135 148
019 020 096 153 060 022
032 035 024 033 041 028
070 092 2700 690 260 860
220 372 659 418 70 85 24
77 238 18 34 103 93 40
770 950 205 950 100 67 93
33 11 8 3 295 31 165
13 69 17 27 18 6 8
39 45 22 12
Number of observations = 71
FIGURE A-2. STEPS IN CONSTRUCTING STEM-AND-LEAF DISPLAY
Step 1 Step 2 Step 3 Step 10
6** 6** 0 6** 0 6**
5 5 5 5
4** 0 . - 4** 0 4** 0 4** 0
3 3 3 9+ 3 9
2** 2** 2** 2** 1
1 1 1 1 247
0 0 138 +
(value (value (value (value
placed 400) placed = 600) placed = 390) placed = 081)
110
-------
When there are too many values to fit on one line, stems can be separated.
As an example, we could have one stem for leaves 0, 1, 2, 3, and 4 and the
other for 5, 6, 7, 8, and 9. If we were to do this for the data points repre-
sented by O** 1138, we would have: . 8 . Working through all the data
O** 13
points leads to a complete stem-and-leaf presentation (Figure A-3).
As discussed in Section 5, log-transformations lead to improved
representations. Thus, Figures A-4 through A—ll show the stem-and-leaf dis-
plays of the distribution of loadings for the constituents Pb, cadmium,
organic nitrogen, orthophosphate (O-PO 1 ), and nitrate. Exceptions to the log-
normal pattern are usually found where the sample size is small (less than
20); this is to be expected from statistical considerations. In most cases
log-transformations do a good job of making the data more symmetrical. We
can conclude from this that the reported mean values and standard deviations
are not good values to use for summarizing loading rates. Use of log-trans-
forms is suggested, along with medians.
ANALYSIS BY TWO-WAY TABLES
The model described above assumes that the processes that account for
the observed loading on streets are numerous and otherwise unanalyzable; that
is, a black box. To continue the exploratory analysis, we specify a set of
factors that we presume contribute to the final loading level. Other
researchers have attempted to specify such categories as land use or perhaps
traffic density as intuitively plausible factors; we would expect industrial
and commercial land to have patterns of accumulation that differ from low—
density residential. Nevertheless, it may well be that other factors, includ-
ing a generally high “residual” variability, may mask these effects. This was
found to be the case in the URS study: high variability within categories --
as measured via t-tests —- hampered the attempt to distinguish differences in
accumulation among land use categories. As we have seen, this is due in part
to the use of raw data; log-transformed values do not behave quite so badly.
Another approach is to use medians instead of a least squares analysis because
medians are insensitive to extreme fluctuations in the data. To perform the
analysis, we disaggregate the data into a model of the form:
LOADING = common value + land use effect
(over all categories) + traffic effect
+ climate effect
+ residual
where the residual is a normally distributed source of error. We are in
effect making the categories into “dummy variables.”
The first step in applying the technique is to find the effect of being
in a particular land use -— for example, across all climates (Table A-l). For
residential land uses only, what is the typical loading? For this row the
typical value as represented by the median is (1.70 + 1.59)/2 = 1.65. Simi-
larly, the commercial, light industry, and industry row medians are 1.61, 1.98,
and 1.76, respectively. Proceeding, we subtract the row medians from each of
the data values to obtain a new table (Table A—2). Just as with land uses,
we now find the climate fit, only this time we naturally look down each column.
lii
-------
FIGURE A-3. RESIDENTIAL LOADING RATES
(LBS/CURB-NILE/DAY)
(URS, 1974)
loading rate x 100
27** 0 Milwaukee
10* *
55
9
6
8
7
7
95
6 0
5
4 01
97
3
69
2 1230
756
1 2423400
8696797879696
131223323422134210031121003421
Number of observations = 71
Results : upper hinge: 205 (57.80 kg/curb-km-day)
median: 70 (19.74 kg/curb-km-day)
lower hinge: 27 ( 7.61)
spread: 178 (46.52)
Note : median: the data value half—way in from either end; half the
data lies above, or below this data point.
hinge: upper (lower) -- half-way from the high (low) data
extreme to the median, i.e., one—quarter of the way in
from either end.
spread: difference between upper and lower hinges, i.e., contains
one-half of the data.
112
-------
FIGURE A-4. LN (RESIDENTIAL LEAD WADING, MICROGRAMS/GRAM) FIGURE A-S. LM (RESIDENTIAL N0 WADING, MICROGRAMS/GRAM)
CURS, 1974) ( UPS, 1974 )
loading x 0.01
loading x 0.01
6
8 12120
• 9767666988
7** 335035034304
• 77686898
6 44310541 10
• 868
5** 9 3 +(Lawrence, Maaa.)
• 79
4 1 8**
9
3** 7
8 • 68678
2 6** 4042205221
798
1 5
upper hinge- 7.60
n ’53 median— 7.00 upper hinge— 6.70
lower hinge— 6.35 n20 median— 6.30
lower hinge- 6.00
FIGURE A-6. RESIDENTIAL CADMIUM WADING (MICROGRAM/GRAM)
(uPS, 1974) FIGURE A-7. LW (COMMERCIAL ORTHOPHOSPNATE WADING MICROGRAMS/GRAM)
loadingx o.1 loading x 0.01
8 088
7 2 5
4** 20
6 001 3 9274
2** 9733
5 2455 1
+0** 583
4 0235 —o 9
—1 3
3 01345556 n15
upper hinge 3.70
2 03456778 median— 2.70
lower hinge- 0.50
1 0111334667
0 000034678
n— 50 upper hinge — 4.3
median — 2.7
lower hinge — 1.1
-------
FIGURE A-8. Iii (CC*ViERCIAL LEAD WADING. MICROGRAMS/01A14) FIGURE A-9. LOG (C0I8 ERCIAL NITRATE WADING, MICROGRAI4S/GRN4S)
CURS, 1974) CURS, 1974 )
loading x 0.01
loading x 0.01
2
6
l’ 4222403
565677
0
6
n— 14
4 30 upper hinge— 1.35
• 5559658 Radifl— 1.10
3 430001 lower hinge— 0.60
• 75
2
upp.r hing.— 3.70
n— 17 medjan 3.50
Lower hing.- 3.00
I-a
FIGURE A-10. LN (INDUSTRY AND LIGHT INDUSTRY COO LOADING, ICR0GRAMS/GRAZ4) FIGURE A-li. LN (INDUSTRY AND LIGHT INDUSTRY LEAD WADING, MICROGRAMS/GRAM)
CURS. 1974) CURS. 1974 )
loading x 0.01 loading x 0.01
8 1
5 34415 . 968965
4 20 7 0020324
3 148 . 686
2 8 6 14
1 . 857
5
nll 4 1
8
3
upper hinge— 5.35 2**
median— 4.20
lower hing. 3.60 n—24
upper hinge— 7.60
median— 7.00
lower hinge— 6.10
-------
For a particular climate, what is the typical value? (This time we are not
using the original data values, but those with land use effects subtracted
out.) In Table A-2 notice that we also find the median of all land use
effects themselves by looking down the column labelled “Row Median.” This is
the “median of the row medians” -- the overall loading common to all land uses
and climates. Again, we subtract the column medians from the respective table
values above them and subtract the overall effect from each of the row medians
(Table A—3).
TABLE A-i. LOG-TRANSFORMED LOADING DATA
Land Use Northeast
Climatic Region
Southeast Northwest
Southwest
Row
Median
residential 2.13
1.70
1.59
1.43
1.65
commercial 2.01
1.61
1.20
1.61
1.61
light
industry 2.13
2.02
1.77
1.93
1.98
industry 2.71
1.64
1.59
1.87
1.76
TABLE A-2.
TRANSFORMED LOADING
DATA MINUS
ROW MEDIAN
Land Use Northeast.
Climatic Region
Southeast
Northwest
Southwest
Row
Median
residential 0.48
0.05
—0.06
—0.22
1.65
commercial 0.40
0
—0.41
0
1.61
light
industry 0.15
0.04
—0.21
—0.05
1.98
industry 0.95
—0.12
—0.17
0.11
1.76
column
medians 0.440
0.020
—0.190
—0.025
1.705
Overall
Value
If we had used means as typical values, we would now be done; however,
since we are using medians the process is not quite complete. If we now cal-
culate row medians, we see they are not quite zero; this indicates that we
should go through the whole process once more, but this time “polishing” the
fits by subtracting only the new, small row medians. We must then do the
same for the column medians. We repeat, moving from polishing rows to columns
until the medians for both are relatively (within round-off error) close to
zero. We are left, in this case after four steps, with row effects along the
right—hand side, column effects along the bottom, and the common, overall
value in the lower right-hand corner (Table A-4). What numbers are left in
the cells? Those not accounted for by climate, land use, or overall value —-
the residuals.
115
-------
TABLE A-3. TRANSFORMED LOADING DATA WITH ROW AND COLUMN MEDIANS SUBTRACTED
Climatic Region Row Row
Land Use Northeast Southeast Northwest Southwest Fit Median
residential 0.040 0.030 0.130 —0.195 —0.055 0.035
commercial —0.040 —0.020 —0.220 0.025 —0.095 —0.030
light
industry —0.290 0.020 —0.020 —0.025 0.275 0
industry 0.51 —0.140 0.020 0.135 0.055 0.078
Column Fit 0.440 0.020 —0.190 —0.025 1.705 —0.003
Common
Value
TABLE A-4. LOG (MEDIAN SUSPENDED SOLIDS LOADING, LBS/CURB—MILE/DAY)
LAND USE VERSUS CLIMATIC REGION
(UBs, 1974)
Original Data Region
Land Use Northeast Southeast Northwest Southwest
Row
Median
residential 2.13 1.70 1.59 1.43 1.65
commercial 2.01 1.61 1.20 1.61 1.61
light
industry 2.13 2.02 1.77 1.93 1.98
industry 2.71 1.64 1.59 1.87 1.76
Common Value+
The Two-Way Fit: Loading = Land Use (row) Effect + Region (column)
Effect + Residual Row
Effect
residential 0.01 —0.01 0.13 —0.25 —0.08
commercial —0.01 0.01 —0.15 0.04 —0.18
light
industry —0.29 0.02 0.02 —0.04 0.22
industry 0.43 —0.22 —0.02 0.04 0.07
Column Effect 0.44 0.03 —0.23 —0.01 1.76
Common
Value
Tables A-S through A-b give the results of the two-way table analysis as
applied to a series of comparisons between landscaping, climatic region, land
use, and traffic density for various constituents along with a stem—and-leaf
display of the residuals for each analysis. The results are discussed in Sec-
tion 5 of the main text. Following the tables are the diagnostic plots for
all of the two—way tables, Figures A-12 to A-20. These attempt to uncover
116
-------
possible non—linear interactions between the factors analyzed in the corre-
sponding table; the results are discussed in Section 5 of the main text.
TI BLE A-5. ‘l O-WAY FIT, LOG (MEDIAN SUSPENDED SOLIDS, LBS/CURB-MILE/DAY)
LANDSCAPING VERSUS CLIMATIC REGION
CURB, 1974)
Original Data
3
7
0010101
3
9
1
upper hinge= 0.04
n13 median= 0.00
lower hinge=—0.07
Landscapinq
Climatic Region
Northeast
Southeast
Southwest
Northwest
grass
trees
landscaped
buildings
none
row
madian
2.714
1.833
1.431
1.484
1.672
1.519
absent
absent
2.170
1.544
absent
1.455
2.182
1.778
1.371
1.204
2.176
1.661
1.401
1.455
The Fit
Landscaping
grass
trees
landscaped
none
row
Climatic Region
Northeast
Southeast
Southwest
Northwest
bui ldincis
effect
0.450
0.016
0.013
0.013
—0.131
0.131
absent
absent
0.000
—0.211
absent
0.078
—0.005
0.006
0.004
—0.190
0.604
0.189
co1 nn effect 0.097
—0. 020
1.563
(common
value)
—0.364 0.003
Residuals x 1000
4** 5
3
2**
1
0**
—l
_2**
—3
117
-------
TABLE A-6. TWO-WAY FIT, LOG (MEDIAN SUSPENDED SOLIDS, LBS/CURB-MILE/DAY)
LAND USE VERSUS LANDSCAPING
(URS, 1974)
Original Data
The Fit
Landsca
Land Use grass trees
Residential —0.139 —0.288
Co erciaj. 0.000 0.289
All Industrial 0.267 absent
ing
landscaped
buildings
none
row
effect
0.128
—0.128
absent
0.138
0.000
—0.267
0.0
—0.476
0.245
3
2**
1
O**
—l
—3
86
23
00
32
86
—0.045 1.936
(common
value)
Land Use
Residential
Coi eercia1
All Industrial
Landscapina
1.842
1.505
2.493
grass
trees
landscaped
buildings
none
row
median
1.512
1.613
absent
2.130
1.398
absent
2.029
1.415
1.869
1.936
1.460
2.181
column effect 0.045
—0.136 0.066
Residuals x 1000
upper hinge 0.130
n=10 median= 0.0
lower hinge=-0.130
118
-------
TABLE A-7. TWO-WAY FIT, LN (NEDIAM COD, MICROGRAMS/GRAM X 0.001)
CLIMATIC REGION VERSUS LAMD USE
(URS, 1974)
Original Data
2
00000
6
4
upper hinge 0.20
nlO median 0.0
lower hinge-0.0
Climatic Region
Northeast
Southeast
Southwest
Northwest
Land Use
Residential
Commercial
All Industrial
Row
median
3.56
4.87
4.14
3.64
6.13
3.00
4.82
absent
3.22
2.89
5.36
absent
3.56
3.00
4.82
3.64
The Fit
Land_Use
Climatic Region Residential Commercial All Industrial Row
Northeast
effect
0.0
2.57
0
—0.04
Southeast
1.87
0
—0.45
-0.60
Southwest
—0.68
0
0.20
1.22
Northwest
0
absent
absent
0.04
column effect
0
—0. 34
0
Residuals x 100
3**
5
2
8
3.60
(coimnon
value)
+0
119
-------
TABLE A-S.
LAND
Original Data
TWO-WAY FIT, U ( DIAN COD, MICROGRANS/GRAM x 0.001)
USE VERSUS AVERAGE DAILY TRAFFIC VOLU (ADT)
(URS, 1974)
Land Use
Residential
Cossnercial
All Industrial
—
absent
absent
S403
less than
500
500-5000
5000-15000
greater
than 15000
row
median -
3.820
5.509
3.951
6.020
4.1S1
3.469
3.211
absent
3.820
5.509
4.777
The Fit
Traffic
Land Use less than 500—5000
Volume
5000—15000
greater
row
Residential
500
than 15000
effect
absent
0.00
0.000
0.974
-0.294
comnercial
absent
0.00
0.380
—0.974
1.395
All Industrial
0.00
absent
—0.094
absent
0
column effect 1.223
—0.066 0.065 1.390
4.180
(common
value)
Residuals x 100
10
9 7
8
7
6
5
4
3 8
2
1
0* 0000
—l 9
2
3
—4
5
6
7
8
9 7
—10
upper hinge 0.19
n—8 median 0.00
lower hings -0.05
120
-------
TABLE A-9. TWO-WAY FIT, LN (MEDIAN LEAD, MICROGRAMS/GRAM X 0.1)
LAND USE VERSUS CLIMATIC REGION
(URS, 1974)
Original Data
The Fit
Climatic Region
Land Use Northeast Southeast Southwest - row effect
Residential 0.383 —0.589 0.00 .489
commercial
0.0
0.118
—0.384
0.755
Industrial
—0.025
0.00
0.00
0.00
3* *
2
+0
—0
—2
_3**
—4
_5**
upper hinge= 0.059
n 9 median 0.0
lower hinge—0.205
Land Use
Residential
Coninercial
All Industrial
r.l , . . ..-4,.
4.394
4.277
3.497
Northeast
Southeast
Southwest
row median
4.443
5.416
4.543
5.416
5.298
4.927
4.44 3
5.298
4.543
column effect —1.021 0.00 0.384
Residuals x 1000
4.543
(common
value)
8
1
0000
2
8
8
121
-------
TABLE A-1O. TWO-WAY FIT, LN (MEDIAN LEAD, MICROGRAMS/GRAM X 0.1)
LAND USE VERSUS AVERAGE DAILY TRAFFIC VOLUME
(URS, 1974)
Original Data
The Fit
Avera
Land Use less than 500
e Daily Traffic
Volume (ADT)
500—5000
5000-15000
greater
row
Residential
than 15000
median
0.147
0.000
—0.485
absent
—0.069
Cmeiaercial
absent
0.000
0.000
—0.229
1.043
All Industrial
—0.146
-0.213
0.324
0.230
0.000
4**
3
2**
1
0**
—l
—3
4**
2
3
4
000
4
12
8
Averaae
Land Use
Residential
cozxs ercial
All Industrial
Daily Traffic Volume (ADT)
4.284
absent
4.060
less than 500 500—5000
5000—15000
greater
than 15000
row -
median
4.575
5.687
4.431
4. 369
5.966
5.247
absent
5.371
4.787
4.369
5.687
4.609
LtP N EFFECT —0.403 0.035 0.314
Residuals x 1000
4.609
(common
value)
—0.052
upper hinge 0.140
n10 median 0.0
lower hinge=—0. 210
122
-------
FIGURE A-12. DIAGNOSTIC PLOT, RESIDUALS FROM TWO-WAY FIT,
LOG (MEDIAN SUSPENDED SOLIDS, LBS/CURB-MILE/DAY)
LP.ND USE VERSUS CLIMATIC REGION
x
40
30
20
010 X
0
— XX
K x
0_ XX
d X
-I0
K
-20
X
-30 X
l I I I I I i 1 1 I
-4 -2 0 2 4 6 8
C0MF RIS0N VALUE x 1,000
123
-------
FIGURE A—13. DIAGNOSTIC PLOT RESIDUALS FROM TWO-WAY FIT,
LOG (MEDIAN SUSPENDED SOLIDS, LBS/CURB-MILE/DAY)
LAND USE VERSUS AVERAGE DAILY TRAFFIC VOLUME
10 X
8 x X
6- x
4-
2
O 0 x
o -2
0 X
‘--4
-6 -
x
-12
-14 x
-16 -
(‘268) (off scale)
I I I I l l I I I I I
-14-12-10-8-6-4-2 0 2 4 6 8 10 12 14 16
COMPARISON VALUE xl000
124
-------
FIGURE A-14. DIAGNOSTIC PLOT RESIDUALS FROM TWO-WAY FIT,
LOG (MEDIAN SUSPENDED SOLIDS, LBS/CURB-MILE/DAY)
LANDSCAPING VERSUS AVERAGE DAILY TRAFFIC VOLUME
4 x
3
2
x
01
.:: 0 xxx xx
d X x
0
U i x
x
-4
-6
I I I I I
-30 -20 -10 0 10 20 30
COMPARISON VALUE x 1000
125
-------
FIGURE A-15. DIAGNOSTIC PLOT RESIDUALS FROM JO-WAY FIT,
LOG (MEDIAN SUSPENDED SOLIDS, LBS/CURB-MILE/DAY)
LANDSCAPING VERSUS CLIMATIC REGION
x
10—
o
XX
—I
0
Cl)
U i
-20
I I I I I I I I I I I I I
-10 -5 0 5
COMPARISON VALUE x 100
126
-------
FIGURE A-16. DIAGNOSTIC PLOT RESIDUALS F DM TWO-WAY FIT,
LOG (MEDIAN SUSPENDED SOLIDS, LBS/CURB-MILE/DAY)
LAND USE VERSUS LANDSCAPING
x
2—
1—
0
x
COMPARISON VALUE x 100
127
-------
FIGURE A-17. DIAGNOSTIC PLOT RESIDUALS FROM TWO-WAY FIT,
LN (MEDIAN COD, MICROGRAMS/GRAM X .001)
CLIMATIC REGION VERSUS LAND USE
3-
x
2
“ Cl
-J
x
x
LU x
I I I I I
-2 -1 0 1 2 3
COMPARISON VALUE x 10
128
-------
FIGURE A-18. DIAGNOSTIC PLOT RESIDUALS FROM TWO-WAY FIT,
LN (MEDIAN COD, MICROGRAMS/GRAM X .001)
LAND USE VERSUS AVERAGE DAILY TRAFFIC VOLUME
10- x
5-
x
__j A-
4 ‘ X x
U)
w
—io (46,9.7)
I I I
-10 -5 0 5
COMPARISON VALUE x 1000
129
-------
FIGURE A-19. DIAGNOSTIC PLOT RESIDUALS FROM ‘:NO—WAY FIT,
LN (MEDIAN LEAD, MICROGRAMS/GRAM X .1)
LAND USE VERSUS CLIMATIC REGION
6
5
4
x
3
2
1
0
I — x
*
—i-i
0
w
x
-5 -
-6 X
I I I I I I I I I I
-16 -14 -12 -10-8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18
COMPARISON VALUE x 100
130
-------
FIGURE A-20. DIAGNOSTIC PLOT RESIDUALS FROM O-WAY FIT,
LN (MEDIAN LEAD, MICROGRAMS/GRAM X .1)
LAND USE VERSUS AVERAGE DAILY TRAFFIC VOLUME
r
5
4
3 X
x
2
1
0
xx x
x
x
U)
I ii
-5 x
I I I I .1 I I I I I
-5-4-3-2-1 0 1 2345678
COMPARISON VALUE x 100
131
-------
ANALYSIS OF WASHOFF: THE ENVIREX-DOT DATA
Developing the Data-Base
In the analysis of individual storms the relationship of suspended solids
to flow was studied by comparing cumulative suspended solids to cumulative
flow. A conservative estimate of the quantity accumulating between observa-
tions was made by simple linear interpolation in which the observations used
to define the intervals were those of the pollutant parameter. However by this
method, the flow always had more observations than were used in a regression;
some information was not used. For instance, a time interval in a zinc versus
flow relationship may be defined by the one-half hour between zinc observa-
tions, while a plot of suspended solids versus flow for the same storm permits
15—minute intervals to be used. The zinc versus flow plot will show a linear
relationship over the one—half hour interval, but the suspended solids versus
flow plot may show a significant non-linear relationship. Though we may sus-
pect that zinc behaves similarly to suspended solids, we are unwilling to
assume this a priori ; consequently, the conservative method of analysis chosen
does not reflect this information and may be in error. Where this type of
error seemed likely the results were eliminated from this report.
As discussed in the main body of Section 5, several additional explana-
tory plots were made of washoff constituents versus cumulative flow. The next
three graphs display this relationship for iron, TOC, chlorides, total solids,
arid zinc.
132
-------
FIGURE A-21. CUMULATIVE Fe AND TOC VERSUS CUMULATIVE FLOW
MILWAUKEE STORM #4
x FE
0 TOC
w
0
0
F-
-J
0
300
100
poo opcx qooo ooo $00,000 $20,000 4Q000
CUMULATIVE FLOW (ft’)
133
-------
FIGURE A-22. CUMULATIVE C TS VERSUS CWIULATIVE FLOW
MILWAUKEE STORM #4
o TS
TS
U)
.0
U)
I—
-J
0
Li
>
-J
0
x CL
2O OOO 40,000 64000 80,000 C0,000
120,000 144000
CUMULATWE FLOW (ft 3 )
134
-------
FIGURE A-23. CUMULATIVE ZINC VERSUS CUMULATIVE FLOW
HARRISBURG STORM #7
U,
o
z
U i
>
j 04_
4 Ob*srvsd
o Linsor modsi
o P4os-Iinsor siodsi
0
/-;--
I I I
2000 4000 6000 8000 10,000 2,000
CUMULATIVE FLOW (ft 3 )
135
-------
APPENDIX B
DEVELOPING PRODUCTION FUNCTIONS FOR RUNOFF Q)NT OL 1IEAS’JRES
DETAILS OF THE HYDROLOGIC MODEL
As described in the main text, any simulation ucdel must divide the
land area considered into hypothetical “blocks” (see Figure B—i). Each block
(or cell) is selected so as to be “relatively” independent of its neighbors;
this simplifies bookkeeping, but is not strictly necessary. For our interests
a block consists of a pervious and an impervious section, each of which
behaves differently in its response to rainfall. In the simulation ni del this
means that it takes different anvunts of time for water to flow from the
impervious part of a block to our observation point A than from the pervious
part. Our observer at point A notes a difference in the lag time —- hence
the flow contributions to the hydrograph -- from the pervious and impervious
parts of a block. If we keep account of this lag time for each block and
for each rainfall interval, we can construct an output hydrograph at point A
as the sum of the contributions from pervious and impervious areas.
We further divide the pervious and impervious sub—sections as follows:
1) pervious areas that receive runoff from impervious areas (lawns that are
drained into by runoff from rooftops or driveways); 2) all other pervious
areas; 3) impervious areas that drain to pervious (roofs, driveways); 4) im-
pervious areas that do not drain to pervious areas; and 5) impervious main
road segments (the latter category is designed to deal separately with the
section that might be porous pavement). For each sub-section we calculate
the runoff from a cell using the runoff curve method of the Soil Conservation
Service (U.S. Soil Conservation Service, 1972).
2
runoff = ( —l)
where P = potential maximum runoff (taking into account storage and excess
precipitation);
S = retention factor derived from a “curve number” formula;
that is,
= 1000 — 10 (B—2)
136
-------
where CN = curve number based on soil type and antecedent moisture conditions.
The curve number ranges from 0 (completely pervious) to 100 (com-
pletely impervious).
FIGURE B-i. HYPOTHETICAL DEVELOPMENT BLOCK
BLOCK I
IMPERVIOUS PERVIOUS
I OBSERVATION
I POINT
J A1 _____
TREATMENT
The additional information required is the lagtime from the cell to the entry
point of the transport/storage system:
0.8 0.7
TRAVEL (i,j) = 1 (S+1 ) (B—3)
1900 Y (SCS,1972)
where TRAVEL(i,j) = lag, in hours, from cell i, sub-section j to transport
system.
1 = length from farthest point in section to transport system
y = average slope, (percent)
S = as defined above (retention factor).
Flow in cell (i,j) at time K appears in the transport system at time
K + TRAVEL(i,J). To obtain the final hydrograph, we simply sum the flows
over all times and cells during the storm, lagging for travel time:
N
FL (time K) = (Pervious area flow (i,j,K-TRAVEL(i,j)) (B-4)
i=l
÷ Impervious area flow (i,j ,K-TRAVEL(i,j)))
The final step entails routing this flow through the transport system using
a version of Manning’s equation. This simulation model is quite close in
form to others using a simple lag-calculation; for example, the ILLUDAS
i del (1974).
137
-------
DESIGN OF THE MODEL PI GRAM: DISSOLVED SOLIDS WASHOFF
Solids are first accumulated on the cell sections differentially (accord-
ing to type). Solids flux is computed by either 1) exponential washoff from
impervious areas based on runoff intensity or 2) erosion from pervious areas
based on the USLE erosion method (discussed below). Solids from impervious
surfaces can wash to pervious ones in the same cell. Solids flow is then
lagged using the methods described above.
INODRPOBATING CONTROL MEASURES IN THE MODEL
Since all natural features of the site have been incorporated in para-
meters of the model, runoff controls must be represented through changes in
these parameters. The parameters for each cell i and sub—section j are:
area
curve number (CN)
storage of water within the cell
loading rate of solids accumulation
decay rate of solids accumulation
USLE factor—erosion number (See discussion later in this appendix)
slope of cell
hydraulic length of cell
transport hydraulics:
depth of each section of transport system
slope of each section of transport system
length of each section of transport system
bottom width of each section of transport system
Manning’s coefficient of each section of transport system
antecedent days since last rain or sweeping
maximum storage and treatment capacity
The control measures and their corresponding parameter changes are given in
Table B-i.
In addition, the model has the capability to test a variety of other
situations; one that was specifically examined was a change in soil type from
relatively pervious to relatively impervious. This was done by changing curve
numbers for pervious sub-sections from low to high values.
Three residential development layouts were chosen for simulation:
1. “conventional” development, with quarter—acre lots;
2. “low-density” development, one-acre lots;
3. “cluster—townhouse” development, with four units per townhouse,
and clusters of three to five townhouses per cell.
These different development patterns were selected so as to incorporate a
broad range of alternatives, from those that develop almost all the area of
site (alternative 1, conventional) to those that leave most area untouched
138
-------
TABLE B-i. !4 DDEL PARAMETERS CHANGED BY CONT )L MEASURES
Control Measure Parameter Changes
base case (no controls) no changes
porous pavement 1. area that is road (sub—section 5)
becomes part of pervious area
(sub—section 1)
2. curve number = curve number of
surrounding soil
3. storage-unchanged (possibly could
add a slight amount)
4. loading/decay rates: unchanged
5. slope: unchanged
6. hydraulic length: unchanged
7. transport hydraulics: unchanged
swales all parameters unchanged except
roughness and infiltration in trans-
port system. Roughness is increased;
infiltration is increased, by a factor
of 1.5
roof—gutter disconnection area = impervious roof area now
connected to pervious area
all other parameters unchanged
on—lot storage increase storage in pervious sub-
sections of cell to 0.07 inches
vegetative cover lower curve numbers to about 30—40
for pervious subsections of cell
lower USLE erosion factor in these
subsections, by one—half
street sweeping (not tested) reduce antecedent days of accumulation
(alternative 2). In addition, alternative 3 (townhouse) lets us test whether
one can put the same number of total units onto an area as in conventional
developments, but with less impact on the hydrology. All information on
standard sethacks, housing unit areas, and general placement of units was
taken from the Real Estate Research Corporation (1974).
To summarize, we have the following sets of alternatives to be simulated:
1) three development types: conventional, low-density, cluster
2) two soil regimes: highly pervious, highly impervious
139
-------
3) six control measures:
a. none
b. porous pavement
c. swales
d. roof—gutter disconnection
e. on—lot storage
f. vegetative cover
There are nine combinations of control measures:
b,e
b,c
d, f
b,c,d
b,d,f
b,e,f
b,c,d,e
b,c,d, f
b,c,de, f
total: 15 control alternatives
4) a randomized set of storms with a given distribution over time.
The Development Layouts
For each of the three site layouts seven or eight simulation model cells
were formed. Cells were drawn roughly the same size in each layout. The
total area covered is about 26.4 acres (10.75 ha). Schematic drawings of the
sites are given in Figures B-2 to B-5.
Development Number One: Conventional Design——
Some assumptions made in the conventional design were:
1) House sethack 30 feet (7.92 m);
2) Driveway 12 feet wide (3.17 in);
3) House unit is 45 x 35 feet (11.89 x 9.25 in);
4) Main road is 20 feet wide (5.28 in)
Parameter values for each of the cells and their sub—sections are given in
Table B-2. Recall that sub-sections are defined as:
1) pervious, draining impervious area (receives runoff from sub-
section 4); lawns, open field, etc.;
2) pervious, not draining impervious area;
3) impervious draining to transport system (roof drains, if connected);
4) impervious draining to pervious (driveway, rooftop if not connected
to transport system);
5) impervious main road.
Curve numbers are taken from the SCS bulletin 58, pages 2-5. Hydraulic length
and slope are from topographic maps. All solids buildup/erosion factors are
default values (as used in SWMM); the formulation of the USLE factor is de-
scribed below.
140
-------
FIGURE B-2. BOWKER WOODS DEVELOPMENT SITE
FIGURE B—3. BOWKER WOODS QUARTER-ACRE DEVELOPMENT (“CONVENTIONAL”)
N
500’ I
500’
141
-------
FIGURE B-4.
BOWKER WOODS LOW-DENSITY DEVELOPMENT
FIGURE B-5.
BOWKER WOODS CLUSTER- T(MNHOUSE DEVELOPMENT
500’— 1
500’ I
142
-------
TABLE B-2. MODEL PARAMETERS FOR CONVENTIONAL DEVELOPMENT
Sub- Area Curve
Cell I Sectioni acres Nt ber
Storage Average hydraulic USLE ero-
inches Slope ,% length,f .et sian factor
solids deposit*
and prosion factors
1
soil type—A
* Units 10
1
2
3
4
5
0.2926
3.1191
0.3961
0.0345
0.2066
49
49
95
95
95
0
0
0
0
0
3.5
3.5
3.0
3.0
2.5
250
400
70
60
75
.75
.75
0
0
0
0
0
5
5
5
.01
.01
.01
.01
.01
0
0
0
0
0
2
soil type—D
$ Units — 9
1
2
3
4
5
0.3696
1.0245
0.3203
0.0310
0.1653
84
84
95
95
95
0
0
0
0
0
4.0
4.0
3.0
3.0
3.0
250
400
75
75
75
.75
.75
0
0
0
0
0
S
5
5
.01
.01
.01
.01
.01
0
0
0
0
0
3
soil type—A
• Units — 9
1
2
3
4
5
0.7231
1.2237
0.3564
0.0310
0.2066
49
49
95
95
95
0
0
0
0
0
5.0
5.0
4.0
3.5
3.0
300
450
70
60
75
.75
.75
0
0
0
0
0
5
5
5
.01
.01
.01
.01
.01
0
0
0
0
0
4
soil type. .A
* Units 9
1
2
3
4
5
0.4660
2.6498
0.3564
0.0310
0.2158
49
49
95
95
95
0
0
0
0
0
4.0
4.0
4.0
4.0
3.0
150
200
70
60
75
.75
.75
0
0
0
0
0
5
5
5
.01
.01
.01
.01
.01
0
0
0
0
0
5
soil type—A
• Units — 12
1
2
3
4
5
0.8597
1.2341
0.4752
0.0413
0.2296
49
49
95
95
95
0
0
0
0
0
4.0
4.0
4.0
4.0
4.0
150
200
70
60
75
.75
.75
0
o
0
0
0
5
5
5
.01
.01
.01
.01
.01
0
0
0
0
0
6
soil type—C
• Units — 19
1
2
3
4
5
0.6428
4.1246
0.4994
0.0655
0.3673
79
79
95
95
95
0
0
0
0
0
4.0
4.0
4.0
4.0
3.0
200
350
70
60
75
75
.75
0
0
0
0
0
5
5
5
.01
.01
.01
.01
.01
0
0
0
0
0
7
soil type
* Units — 0
1
2
3
4
5
0
5.6918
0
0
0
79
79
95
95
95
0
0
0
0
0
—
4.0
—
800
.75
.75
0
0
0
0
0
5
5
5
.01
.01
.01
.01
.01
0
0
0
0
0
*solids deposit and erosion factors are dimensionless
143
-------
Development Number Two: Low-Density Design--
Table B-3 gives the parameter information for the low-density (l+acre)
layout. The number of units placed on the large lots is only 16, roughly
corresponding to the density proposed in the actual plan by BSC Engineering
(ten units). This means that the percent area that is converted to imper-
vious is much smaller for this type of development than for the conventional
development (road and house surface area is smaller).
Additional changes from the conventional layout are:
1) House sethack is 60 feet (15.85 m);
2) A housing unit is 50 x 40 feet (13.20 x 10.56 m).
Development Number Three: Cluster Development——
Finally, Table B—4 gives information for the “cluster” development (see
also sketch Figure B-5). Each cluster is made up of groups of townhouses --
four units per townhouse. There are 17 clusters for a total of 68 units, so
we are constructing the same total number of units as for the conventional
layout. Further modifications include the cul—de-sac roads that lead into
each cluster; this adds to the surface area taken up by roads. Other mod.i—
fications are:
1) House sethack 30 feet (7.92 m);
2) Each townhouse unit occupies 600 square feet (41.9 m 2 , 4 units!
townhouse);
3) Cul—de—sac road is 20 feet wide (5.28 m).
TRANSPORT SYSTEM HYDRIiIJLIC PARAMETERS
As described above, it is assumed that the network of cells is connected
by a “transport system.” This can be either a conventional sewer system
(pipes) or open channels. The network drains all cells in a simple upstream-
to—downstream order: a cell cannot drain to an adj acent cell, but only to the
transport network. A hydrograph is computed by routing and accumulating the
flows as they appear in sequence from upstream to downstream along each
section of the network that drains an additional particular cell. For
example, in the conventional development the upstream—to—downstream order of
drainage is Cell Number 4 to transport, then Cells 3, 5, 6, 2, 1 and finally
Cell 7. The output is the final output hydrograph after routing through the
transport network that drains Cell 7 and all the flow in the network before
Cell 7. Notice that in this case the transport system has seven sub—parts.
The parameters required for this routing (which uses Manning’s equation)
are, for each sub-part:
1) Whether circular pipe or 5) length of pipe or channel sub-
trapezoidal channel flow; section, feet;
2) depth of channel or pipe 6) diameter of pipe or bottom
(feet); width of channel, feet;
3) side—slope of channel 7) which cell number is drained
(not needed if pipe); by this sub-section.
4) Manning’s coefficient;
144
-------
ThBLE B-3. MODEL PAR iETERS FOR LOW—DENSITY DEVELOPMENT
rsfl I
Sub- Area Curve Storage
aersc
Average hydraulic USLE ero-
slnns I lannth.fsst sinn fa. tnr
solids deposit
and erosion factors
1
2
3
4
5
1
2
3
4
5
1
2
3
4
3
1
2
3
4
•units —0 S
1 0.5670 49
2 2.2014 49
3 0.1625 95
4 0.0248 95
5 0.0746 95
1 0.3788 84
2 1.0146 84
3 0.1084 95
4 0.0166 95
5 0.0599 95
1 0.5165 7
2 0.6538 79
3 0.1645 95
4 0.0248 95
5 0.0976 95
1 0.5165 79
2 0.7245 79
3 0.1645 95
4 0.0248 95
5 0.0804 95
0.4821 49
0.9603 49
0.1645 95
0.0248 95
0.0757 95
.5510 49
.7815 49
.1166 95
.0248 95
.0757 95
0 49
6.9903 49
0 95
0 95
0 95
0 0
9.6 79
0 - 0
0 0
0 0
0 3.5 100
0 3.5 200
0 3.0 100
0 3.0 100
0 2.5 75
0 4.0 150
0 4.0 250
0 3.0 100
0 3.0 100
0 3.0 75
0 5.0 200
0 5.0 300
0 4.0 100
0 3.5 100
0 3.0 75
0 4.0 250
0 4.0 300
0 4.0 100
0 4.0 100
0 3.0 75
4.0 200
4.0 300
4.0 100
4.0 100
4.0 75
4.0 200
4.0 300
4.0 100
4.0 100
3.0 75
4.0 700
.75 0 .01 0
.75 0 .01 0
0 5 .01 0
0 5 .01 0
0 5 .01 0
.75 0 .01 0
.75 0 .01 0
0 5 .01 0
0 5 .01 0
0 5 .01 0
.75 0 .01 0
.75 0 .01 0
0 5 .01 0
0 5 .01 0
0 5 .01 0
.75 0 .01 0
.75 0 .01 0
0 5 .01 0
0 5 .01 0
0 5 .01 0
.75 0 .01 0
.75 0 .01 0
0 5 .01 0
0 5 .01 0
0 5 .01 0
.75 0 .01 0
.75 0 .01 0
0 5 .01 0
0 5 .01 0
0 5 .01 0
.75 0 .01 0
.75 0 .01 0
0 5 .01 0
0 5 .01 0
0 5 .01 0
.75 0 .01 0
.75 0 .01 0
0 5 .01 0
0 5 .01 0
0 5 .01 0
1
soIl type—A
* units — 3
2
soil type-D
• units • 2
3
soil type-C
• units — 3
4
soil type—C
I units 3
S
soil type-A
a units — 3
6
soil type-A
I units — 2
7
soil type-A
• units — 0
a
soil type-C
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4.0 1000
145
-------
TABLE B-4. MODEL PARAMETERS FOR JNBOUSE DEVELOPMENT
call I
Sub— Area Curve
t.tlan Sent tJ nh.i
Storage Averag, hydraulic USIZ .ro-
inchea slone • % length • f..t sion
solids deposit
and .rosion factors
1
1
0.0899
49
0
3.5
0
.01
0
2
0.5957
49
0
3.5
350
5
.01
0
Soil type—A
3
4
0.4089
0.1338
95
95
0
0
3.0
3.0
150
100
0
0
5
5
.01
.01
0
0
Cluat ,re-4
5
0.1116
95
0
2.5
100
2
soil typeD
1
2
3
4
0.6313
0.8616
0.2134
0.0206
84
84
95
95
0
0
0
0
4.0
4.0
3.0
3.0
250
350
150
100
.75
.75
0
0
0
0
5
5
5
.01
.01
.01
.01
.01
0
0
0
0
0
Clust.r e -3
5
0.1837
95
0
3.0
75
3
soil type—p
1
2
3
4
0.6211
1.4757
0.2204
0.1141
9
49
95
0
0
0
0
5.0
5.0
4.0
3.5
150
250
100
100
.75
.75
0
0
0
0
5
5
5
.01
.01
.01
.01
.01
0
0
0
0
0
C lustre-4
5
0.1096
0
3.0
75
0
4
1
2
0
3.719
0
0
4.0
450
.75
.75
0
0
5
.01
.01
.01
0
0
0
soil type.D
3
4
0
0
95
0
0
0
5
5
.01
.01
0
C
IC lusters0
5
0
5
0
—
—
5
1
2
0
2.5069
49
49
0
0
—
4.5
—
700
.75
.75
0
0
.01
.01
.01
0
0
0
soil type-P
3
4
0
0
95
95
0
0
—
—
—
—
0
0
5
5
.01
0
0
Cluster.—0
5
0
95
0
—
—
0
5
6
1
2
0
7.4875
79
79
0
0
—
4.5
—
550
.75
.75
0
0
.01
.01
.01
0
0
0
.oil type-C
3
4
0
0
95
95
0
0
—
—
—
—
0
0
5
5
.01
.01
0
0
C lu ster e -0
5
0
95
0
—
—
7
1
2
0.5510
3.1035
79
79
0
0
4.5
4.5
250
350
.75
.75
0
0
.01
.01
.01
0
0
0
soil type-C
3
4
0.3489
0.0183
95
95
0
0
4.0
3.5
150
100
0
0
5
.01
0
0
IC lusts re-6
5
0.1377
95
0
3.0
100
0
5
146
-------
Table B—5 gives the values for these parameters used in the simulations.
Figure B-6 gives sample output from the program. A complete listing of the
program is supplied in Appendix C.
TABLE B-5. TRANSPORT SYSTEM PARAMETERS (CONVENTIONAL DEVELOP ’1ENT)
Depth of Channel
or Pipe
Side
Slope
Manning’s
Coefficient
Length
Diameter
(bottom width)
Channel
Number
.75
—
.015
200
.75
4
.75
—
.015
300
.75
3
.75
—
.015
350
.75
5
.75
—
.015
350
.75
6
.75
—
.015
200
.75
2
.75
—
.015
250
.75
1
1
3
.025
350
2
7
LIMITATIONS OF THE MODEL
Like any computer model, the control measure simulation model must bear
the inherent weaknesses of its parts. These fall into the usual categories:
data uncertainties and the substitution of simple models for physical pro-
cesses instead of the “correct” ones.
Hydrology
• simple classification into pervious/impervious areas.
• inadequate physical modeling in the use of Manning’s equation for
overland or sewer flow instead of kinematic wave equation.
• no provision for complex detention of water besides initial
abstraction in SCS method.
• use of SCS lagtime equation.
• use of SCS methods applied to extremely small time increments, for
which little verification has been made.
Solids
• exponential washoff model “K” factor uncertain.
• no disaggregation into particle size classes.
• use of USLE—type erosion model, not generally used for time increments
within a storm.
Control Measures
• need to more adequately model effects of porous pavement, vegetative
cover, etc; what other parameters need be included to do this?
As a result, the model can be used only to demonstrate comparative control
measure effectiveness. It cannot be used to accurately model a real water-
shed, especially without calibration. On the other hand, the deficiencies
described above could quite easily be remedied; the existing model is already
comparable to a model like ILLUDAS and gives quite reasonable results.
147
-------
FIGURE B-6. SAMPLE OUTPUT FROM COMPUTER PRO(RAM
CONVENTIONAL QUARTER-ACRE; CONTROL OPTIONS SELECTED=SWALES
FLOW & SOLIDS--> (-- CUMULATIVE RAIN
3.5 1.7 0.0
0.0 +S P 1
** MONTE CARLO SIMULATION **
Is P1
Is P1
O.12+S P I
Is P I
Is NOTE : P I
Is* P I
FLOW INCREMENTS= 0.005 cfs p I
FLOW MAX= 0.500 cfs I
P I
SOLIDS INCREMENTS= 1.200 lbs
SOLIDS MAX= 120 .ooo lbs p I
RAIN(CUM.)= 0.175 inches I
0.42+S4 (* P I
RAIN (MAX CUM.)= 3.500 inches p I
P I
P I
IS’4 p I
0.574S P I
Is* P I
Is P I
Is P I
P I
0.72+S44 * * P I
I M 44M3 * 4**1 **5 P I
P I
I 4*4** S I
P I
0.87+* S P I
Is P
I** S P I
IS P I
1*5 P I
1.02.S P I
I*s P I
o.o5 0.1 o.i5 0.2 0.25 0.3 0.35 0.4 0.45 0.5
0.0 12.0 24.0 36.0 48.0 60.0 72.0 84.0 96.0 108.0 120.0
FLOW & SOLIDS-—> <—- CUtIULATIVE RAIN
PRECIPITATION 0.0174 0.0174 0.0694 0.1041 0.1388 0.5726 1.1453 1.4403
-------
1GURE B-6. (CONTINUED)
STORM: 2 ANTECEDENT DRY PD: 200 PRECIP DURATION:
23 EVENT DURATION:
36 ANTE MOISTURE COND:
2 CASE: 1
A
I
P
CP
QI
SQl
FSX
SSI
CI
QD
SQD
P50
SSD
CD
FSR
SSR
1
0.01
0.017
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2
0.052
0.069
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
3
0.104
0.174
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
4
0.035
0.208
0.000
0.000
0.002
0.002
0.275
0.0
0.0
0.0
0.0
0.0
0.0
0.0
5
0.069
0.278
0.000
0.001
0.028
0.030
0.259
0.0
0.0
0.0
0.0
0.0
0.0
0.0
6
0.416
0.694
0.003
0.004
0.189
0.219
0.242
0.0
0.0
0.0
0.0
0.0
0.0
0.0
7
0.469
1.163
0.005
0.008
0.144
0.363
0.140
0.0
0.0
0.0
0.0
0.0
0.0
0.0
8
0.312
1.475
0.009
0.017
0.445
0.609
0.227
0.0
0.0
0.0
0.0
0.0
0.0
0.0
9
0.104
1.579
0.058
0.075
2.323
3.131
0.177
0.0
0.0
0.0
0.0
0.0
0.0
0.0
10
0.035
1.614
0.115
0.190
1.959
5.091
0.075
0.0
0.0
0.0
0.0
0.0
0.0
0.0
11
0.035
1.649
0.135
0.325
2.230
7.320
0.073
0.0
0.0
0.0
0.0
0.0
0.0
0.0
12
0.035
1.683
0.113
0.438
1.521
8.841
0.059
0.0
0.0
0.0
0.0
0.0
0.0
0.0
13
0.0
1.683
0.086
0.525
1.617
10.458
0.083
0.0
0.0
0.0
0.0
0.0
0.0
0.0
14
0.0
1.683
0.067
0.592
1.113
11.572
0.073
0.0
0.0
0.0
0.0
0.0
0.0
0.0
15
0.0
1.683
0.053
0.645
1.179
12.751
0.097
0.0
0.0
0.0
0.0
0.0
0.0
0.0
16
0.0
1.683
0.038
0.683
0.788
13.538
0.092
0.0
0.0
0.0
0.0
0.0
0.0
0.0
17
0.0
1.683
0.027
0.710
0.834
14.372
0.137
0.0
0.0
0.0
0.0
0.0
0.0
0.0
18
0.0
1.683
0.019
0.729
0.557
14.929
0.130
0.0
0.0
0.0
0.0
0.0
0.0
0.0
19
0.0
1.683
0.013
0.742
0.590
15.520
0.194
0.0
0.0
0.0
0.0
0.0
0.0
0.0
20
0.0
1.683
0.009
0.751
0.394
15.914
0.183
0.0
0.0
0.0
0.0
0.0
0.0
0.0
21
0.035
1.718
0.007
0.758
0.418
16.331
0.274
0.0
0.0
0.0
0.0
0.0
0.0
0.0
22
0.0
1.718
0.005
0.763
0.279
16.610
0.259
0.0
0.0
0.0
0.0
0.0
0.0
0.0
23
0.017
1.735
0.003
0.766
0.295
16.906
0.368
0.0
0.0
0.0
0.0
0.0
0.0
0.0
24
0.0
1.735
0.080
0.846
15.456
32.361
0.851
0.0
0.0
0.0
0.0
0.0
0.0
0.0
25
0.3
1.735
0.057
0.903
0.209
32.570
0.016
0.0
0.0
0.0
0.0
0.0
0.0
0.0
26
0.0
1.735
0.073
0.976
19.162
51.753
1.158
0.0
0.0
0.0
0.0
0.0
0.0
0.0
27
0.0
1.735
0.052
1.028
0.148
51.901
0.013
0.028
0.028
0.080
0.080
0.013
0.0
0.0
28
0.0
1.735
0.037
1.064
13.571
65.472
1.637
0.037
0.064
13.571
13.651
1.637
0.0
0.0
29
0.0
1.735
0.026
1.090
0.105
65.577
0.018
0.026
0.090
0.105
13.756
0.018
0.0
0.0
30
0.0
1.735
0.018
1.106
9.602
75.178
2.314
0.018
0.108
9.602
23.357
2.314
0.0
0.0
31
0.0
1.735
0.013
1.121
0.074
75.252
0.025
0.013
0.121
0.074
23.431
0.025
0.0
0.0
32
0.0
1.735
0.009
1.13].
6.793
82.046
3.271
0.009
0.131
6.793
30.224
3.271
0.0
0.0
33
0.0
1.735
0.006
1.137
0.052
82.098
0.036
0.006
0.137
0.052
30.277
0.036
0.0
0.0
34
0.0
1.735
0.005
1.142
4.806
86.904
4.623
0.005
0.142
4.806
35.083
4.623
0.0
0.0
35
0.0
1.735
0.003
1.145
0.037
86.941
0.050
0.003
0.145
0.037
35.120
0.050
0.0
0.0
L.
0.0
1.735
0.002
1.147
3.400
90.341
6.535
0.002
0.147
3.400
38.520
6.535
0.0
0.0
VOLUME
AND SOLIDS
EVENT TOTALS INCLUDING STORED
DISCHARGE VOLUME = 1.15
DISCHARGE SOLIDS 48.88
REMOVED SOLIDS 41.46
STORAGE/TREAThENT EFFICIENCY
0.459
Define variables:
P=precipitation
CP=cumulative precipitation
QI=incremental flow
SQI=cumulative flow
FSI=incremental solids flow
SSI=cumulative solids flow
CI=concentration
QD=flow stored
SQD=cumulative flow stored
FSD=solids stored
SSD=cumulative solids stored
CD=concentration stored
FSR=flow not stored
SSR=solids not stored
-------
THE SIMULATION MODEL: A FLOWCHART
The computer model is a straightforward encoding of the simple rain-
washoff—transport components.
The first routine, Main, initializes various control options that affect
the simulation generally: the number of subdivision layouts; the number of
control option bundles to be simulated; the time increment used; and whether
bnte Carlo sampling for storms is to be employed. Main’s other function is
to call a series of subroutines that do the actual work of the simulation.
They are as follows:
1) SETCEL —— reads in parameters that describe each watershed “cell”
(area, S( runoff curve number, storage on-site, pollutant buildup
and washoff rates, slope, and hydraulic “length”); computes time—lag
for runoff from that cell to the transport system.
2) SAMPLE —- is called next if 1. nte Carlo sampling has been selected.
Given a mean storm volume supplied by the user, SAMPLE makes pre-
liminary calculations to distribute that volume over an empirically
supplied storm distribution.
3) SETSTM -- uses one of two methods to develop a cumulative hyetograph
(synthetic storm).
a) reads in actual storm data from cards
b) if I bnte Carlo method is used, allocates the volume randomly
according to the supplied rainfall distribution.
4) CONTRL —— reads what control measures have been selected; determines
effects on the model parameters.
5) STORM - - using the Soil Conservation Service Method, determines
antecedent moisture conditions, updated runoff curve values for
each cell, and updated pollutant loadings given the washoff/buildup
equation.
6) CELL -- calculates the runoff from all cells and solids washoff from
impervious cells, using the SCS Method.
7) TRANSP —- transports the resulting flow from all cells via Manning’s
equation to a treatment/storage point.
8) 1 )tY E -- is called by TRANSP and does the actual work of calculating
time delays in the transport system. Pipe or channel flow can be
selected for a series of channels.
9) TREAT -- routes the flow after transport through a storage/treatment
system and determines whether its capacity is exceeded.
10) RESULT - — prints a table of flow, solids flux, and washoff over time.
11) GRAPH -- plots a cumulative hyetograph, a hydrograph, and a polluto-
graph (solids flux).
12) UNCNTL —— undoes effects of control measures, to prepare for next
simulation.
150
-------
Routines 3—12 are executed first for each control measure bundle selected and
then for a series of random draws from the user—supplied storm distribution.
Finally, if a series of subdivision layouts is chosen, routines 1-12 are run
through in sequence. The result is a simulated “experimental design” of a
number of control measures and layouts observed under randomly varying storm
conditions.
Description of Routines
MAIN
1.0 read number of layout cases to be simulated;
2.0 read unit conversion factors;
3.0 loop over number of cases;
3.1 read number of cells, storms, control groups, whether sampling
to be done, time increment (variables NC, NS!IO, NCO, SAMPL,
TINC);
3.2 write out the values of these variables;
3.3 read n del parameters for antecedent moisture factors, runoff
decay constant, initial abstraction;
3.4 read storage/treatment parameters (maximum storage, treatment
levels);
3.5 write out 3.3 and 3.4 vali s;
3.6 read title for this case;
3.7 call SETOEL;
3.8 if SAMPL is true, call SAMPLE (VOLUME);
3.9 for each control group;
3.9.1 Call CONTRL;
3.9.2 For each STORM;
3.9.2.1 call SETSTM (SEED, VOLUME);
3.9.2.2 call STORM;
3.9.2.3 call cELL;
3.9.2.4 call TRANSP;
3.9.2.5 call TREAT;
3.9.2.6 call RESULT;
3.9.2.7 call GRAPH;
3.9.3 end loop over storms;
4.0 end loop Over control groups;
5.0 end loop over cases;
6.0 end.
151
-------
SAMPLE
1.0 read number of antecedent days, (NTA); number of times periods in
storm (NTD), storm total volume;
2.0 get random number from uniform distribution between 0 and 1;
3.0 adjust total volume based on random number and set volume increment
for storm;
4.0 read distribution for synthetic storms;
5.0 return to MAIN.
SETSTM (SEED, VOLUME)
1.0 write header message;
2.0 get 100 random numbers from uniform distribution using random
start SEED;
3.0 put volume increments into “bins”, distribution (1),
distribution (NTDJ based on the random numbers, giving rainfall
distribution;
4.0 find cumulative precipitation based on this rainfall distribution;
write cumulative precipitation;
5.0 return to MAIN.
ROU’IE (PIPE, DB, SIDE, DEPTH, MANNING, SLOPE, LENGTH, Q, VEL, TI)
1.0 find trapezoidal or pipe flow hydraulic radius through this channel;
1.1 set DEPTH=full depth;
1.2 calculate flow using Manning equation based on this depth;
1.3 if equal to flow, return as radius;
1.4 if not, decrement and iterate to 1.2.
2.0 if no proper depth found, triple depth and try again;
3.0 using final hydraulic radius, compute cross-section using flow
and Manning equation;
4.0 compute velocity as flow/area;
5.0 compute TI=travel time=length/velocity;
6.0 return to TRANSP.
TRANS P
1.0 zero out flow, solids arrays;
2.0 loop over time intervals (1 to NTD); loop over each cell (1 to NC);
loop over each sub—section;
2.1 get time lag to transport for this sub-section;
2.2 add flow lagged by this an unt into flow from this cell;
152
-------
3.0 loop over cells J=l to NC;
3.1 get cell number that is drained by this part of transport
system (starting upstream), from array CONECT, NODE=CONECT (J);
3.2 get channel characteristics for this part of channel:
DB1 = DB(NODE);
Z = SIDE(NODE);
Si = SLOPE(N ODE);
Y = DEPTH(NODE);
AL = LENGTh (NODE);
N MANNING’S (NODE);
PIPE2 = PIPE(NODE);
3.3 adjust N$ if swales;
3.4 find area of channel or pipe;
3.5 find hydraulic radius; con ute maximum discharge;
3.6 loop through flow for this cell and see if 0 or exceeds
maximum; set to maximum if latter;
3.7 find average flow Q1;
3.8 route using this flow:
call ROUTh (PIPE2,DB1,2,Y,N$,Sl,AL,Q1,VEE.,TI);
3.9 shift flow, and solids according to lag TI;
4.0 save this flow and move to next downstream node;
5.0 save final flows, solids;
6.0 return to MAIN.
GRP PH
1.0 set maximum flow, rain, solids;
2.0 find graph increments;
3.0 write title, top heading;
4.0 loop over time intervals (1 to NLST);
4.1 blank out line;
4.2 find graph positions for flow, solids, rain;
4.3 write out time, line;
5.0 write trailer;
6.0 return to MAIN.
X NTRL
1.0 read control option card;
2.0 save old parameter values in arrays SAVE, SAVE2,SAVE3, AND SAVE4;
3.0 look for valid option numbers;
153
-------
3.1 if valid, branch to proper control option and adjust
parameters;
3.1.1 porous pavement;
3.1.2 vegetative cover;
3.1.3 roof—drains disconnection;
3.1.4 on-lot storage;
3.1.5 swal.es;
3.1.6 soil type impervious;
3.1.7 unused;
3.2 all others, write invalid option message, keep looking;
4.0 if 0 option, print ‘end of controls’ message and return to MAIN.
UNCNTL
1.0 restore old values of parameters from arrays SAVE, SAVE2, SAVE3,
and SAVE4;
1.1 loop through all cells Ci. to NC);
1.2 loop through all sub—sections (1 to 5);
2.0 return to MAIN.
RESULT
1.0 write heading;
2.0 for each time period in STORM (1 to NLST);
3.0 write out re suits time, rain, cumulative rain, flow, cumulative
flow, solids flux, cumulative solids, concentration, flow stored,
cumulative flow stored, solids stored, cumulative solids stored,
storage concentrations;
4.0 compute and write out total volume, solids, retroved solids,
storage/treatment ratio;
5.0 return to MAIN program.
SETCEL
1.0 for each cell,
1.1 for each of 5 sub—sections
1.1.1 read cell#, sub—section#, area, curve number, storage,
solid buildup rate, decay rates, USLE factor, slope,
length;
1.1.2 calculate lag-time to transport;
1.1.3 convert to number of time-increments;
1.1.4 calculate maximum load possible in exponential solids
accumulation equation = buildup/decay;
154
-------
1.1.5 write out cell parameters;
1.1.6 find maximum time lag; accumulate total area;
1.2 read number of channels in transport system;
1.2.1 for each channel
1.2.1.1 read whether a pipe, depth, slope, side length
(channel only), Manning’s coefficient, length,
bottom-width or diameter, cell number this
channel connects
2.0 return to MAIN program.
STORM
1.0 if not using Monte Carlo method, read antecedent rain conditions,
number of time increments in storm, cumulative precipitation;
2.0 compute antecedent moisture condition factors: multiplier for
antecedent moisture conditions 1, 2, or 3 (U.S. Soil Conservation),
1 = 2.4 (dry soils); 2 = 1.0 (average antecedent condition),
3 = 0.420 (saturated);
3.0 compute retention factor S = 1000/curve number — 10;
4.0 adjust retention factor based on antecedent moisture: SS*ante_
cedent moisture factor (more saturated = lower S value).
5.0 find change in solids accumulation, using formula for load,
k 2 -Krt)
6.0 return to MAIN program.
CELL
1.0 for each cell;
1.1 for each time increment (1 to NTD) compute excess precipita-
tion: test = cumulative precipitation — storage—runoff
curve retention factor;
test 2
1.2 compute runoff, Q = test+S , where S = retention factor
(see routine STORM); add runoff from impervious areas to
pervious areas;
1.3 find increment in flow this time period by subtracting flow
of last period;
1.4 find solids washoff, from impervious areas w(l_ekt);
1.5 find erosion, FS = erosion = USLE*(Q 2 /TINC) 056 ;
1.6 find this time period precipitation;
1.7 add solids flux from impervious to pervious areas;
155
-------
2.0 end loop around time increment, and cells;
3.0 return to MAIN.
TREAT
1.0 set initial conditions on storage;
2.0 I DU’IE flow into storage;
2.1 accumulate volume;
2.2 check if storage filled (VSMAX);
2.3 accumulate volume in storage, VS;
3.0 I )UTE through treatment;
3.1 check for treatment capacity, QTMAX;
3.2 check for solids removed;
3.3 compute ratios, solids treated/total solids;
4.0 return to MAIN.
Erosion Estimate
The method used to estimate erosion in the simulation model is based
upon a modification of the Universal Soil Loss Equation (USLE) (Wischmeier
and Smith, 1972), introduced by Williams (1972).
In place of the rainfall erosivity factor in the USLE, Wi].liains used
peak runoff rate and total storm runoff volume as measures of the erosivity
of a given storm. Empirical calibration based upon data from 18 Texas water-
sheds gave the following result.
w = 95 (Qq ) 56 KCPL (B-5)
where W = sediment yield per event (tons);
Q = volume of runoff (acre-ft);
= peak runoff rate (cfs);
K = soil erodibility factor in USLE;
C = cover factors in USLE;
P = practice factor in tJSLE;
= length/slope factor in USLE.
Values of the four USLE factors are tabulated in Wischmeier and Smith (1972)
for specific watershed conditions.
This relationship is applied to each time increment in the simulation
to estimate erosion rates from pervious areas. With appropriate unit trans-
formations, equation (B-5) is equivalent to
156
-------
.56
W’ = 4.75 x 10 L i KCPL (B-6)
where W’ = sediment yield (lbs/time increment);
Q. = runoff volume (acre—inch/time increment);
= time increment size (hours).
The following USLE factors are assumed as being typical of a grassed area
with a two percent slope, a slope length of 200 feet, and a soil of inter-
mediate erodibility:
K = .25
C = .01
p=l
L = .3
S
KCPL is the term “USLE” in the model.
S
The above scheme is expected to give only order of magnitude estimates of
erosion rates for stable watershed conditions. Use of Williams’ formulation
on an incremental basis is actually not appropriate, since it has been
derived on a total storm event basis. Simulations indicate, however, that
erosion rates are generally insignificant compared with solids accumulation
and washoff from impervious areas. bre detailed simulations of the erosion
process that consider the generation and transport of sediment particles in
overland flow are not justified, provided that pervious areas are well-
vegetated. The above procedure may not be adequate for predicting erosion
losses during construction periods in which bare soil is exposed. Under
such conditions it may be desirable to resort to a more complex simulation
approach (Donigian and Crawford, 1976) or to use the Williams’ formulation
on a total storm event basis.
157
-------
APPENDIX C
LISTING OF COMPUTER PROGRAM
FORTRAN IV Si RELEASE 2.0 MAIN DATE : 78198 10/01/10
C MAIN PROGRAM - EDISON SIMULATION
0001 COMMON P1,P2,TINC,FAC(3),NC,PLA5T,XKW,NLMAX,ATOT,FIA(5),NSTO,
ININ ,NSIN,NCIN,A(iQ,5),ST OR(1O,5),CN(1O,5),W55(1O,5),TLAG(1O,5),
2USIE(i0,5),SKR(iO,5),NTD,CP(150),NTS,NLST,W(1Q,5),Q(10 0,1O,5),
3F5t 100,10 ,5) ,VSMAX ,VThAX,QSMAX,QTPiAX,KTS,KTT,UNIT( 10) ,QI(150,10),
4FSI(150,10),QD( 15 0),FSD(150),F SR( 150),IAMC,NOUT,S(iO,5),SS,VS,IS,
5F 43.( 10) ,JCASE ,NTA
0002 COMMON /CHAN/ DB(1O),SLOPE(iO),LENGTH(1O),DEPTH(1O),S IDE(1O),
1 AN(10),DI STRB(10 0),c ONECr(1O), PIPE(1O)
0003 COMMON /SWTCH/ OPTIONt1O),SAMPt,SWALES
0004 COMMON 1 1 1Th TITLE(20)
0005 REAL*8 SEED
0006 REAL KTS,KTT
0007 INTEGER*2 OPTION,CONECT
0008 LOGICAL*1 PIPE,SWALES,SAIIPL
0009 NIH 5
0010 NOUT = 6
0011 NSIN : 5
0012 NCIN : 5
0013 READ(NIN,iO10) NCASE
0014 1010 FOPFIAT(4 15,4X,Li,F1O.0)
0015 READ(NIN,i02) (UNIT(I),I:1,6)
0016 102 FORMAT(10F6.0)
0017 00 100 ICASE i,NCASE
0018 READ(NIM ,1O10) JCASE,NC,N STO,NCO,SAMPL,TINC
C
NCO NUMBER OF CONTROL OPTION COMBOS
C
0019 WRITE(NOUT,2O4) JCASE,t4C,NSTO,NCO,TINC
0020 204 FORtIAT( ’l ’, CASE : ‘ , 13, NO. CELLS = ‘, 13 , ’ NO. STORMS ‘, 13,
I • NO. CONTROL GROUPS: , 13/
2 TIME INCREMENT ‘,F6.3,’ KOURS ’/)
0021 READ(NIN,103) P1 ,P2. ( FAC(J ),J:1 ,3) ,PLAST,XKW,( FIAtK) ,K1,5)
0022 READ{NIN ,103) VSMAX,QSMAX,KTS ,VTMAX ,QTtIAX,KTT
0023 WRITE(NOUT,201) P1,P2,(FAC(J),J :1,3),PLAST,XKW,(FIA(K),K1,4)
0024 201 FORMAT(// • MODEL PARAMETERS /(5X ,1OF1O.3))
0025 WRITE(NOUT,202) VSMAX,QSMAX,)KTS,VTMAX,QTMAX,KTT
0026 202 FCRMAT(// ’ STORAGE/TREATMENT PARAtIETERS’/(5X,1OF1O.3))
0027 WRITE(6 ,203) (UNIT(IJ,I :i,6)
0028 203 FORMAT(// UNIT PARAMETERS ’/(5X,1OFIO.3))
0029 QSMAX : SMAX*TINC
0030 QTMAX = QTMAX*TINC
0031 103 FORtIAT(5X ,12F6.0)
0032 READ(5,1090) TITLE
0033 1090 FOPMATt2OA4)
0034 CALL SETCEL
0035 IF(SAMPL) CALL SAMPLE(VOLUME)
RANDOM SEED FOR THIS SET OF STORMS
0036 SEEDO.3711248
0037 DO 210 J1,NCO
0038 CALL CONTRL
0039 DO 200 IS i,NSTO
0040 IF(SAMPL) CALL SETSTtI(SEED,VOLUTIE)
158
-------
FORTRAN IV G1 RELEASE 2.0 MAIN DATE 78198 10/01/10
0041 CALL STORM
0042 CALL CELL
0043 CALL TRANSP
0044 CALL TREAT
0045 CALL RESULT
0046 CALL GRAPH
0047 200 CONTINUE
C* RESTORE ENVIRONMENT
0048 CALL UNCNTL
0049 210 CONTINUE
0050 100 CONTINUE
0051 END
159
-------
FORTRAN IV Gi RELEASE 2.0 SETCEL DATE 78198 10/01/10
0001 SUBROUTINE SETCEL
C SETSUP WATERSHED PARAMETERS
0002 COMMON P1,P2,TINC,FAC(3),NC,PLA5T,XKW,NLMAX,ATOT,FIA(5),NSTO,
1NIN,NSIN,NCIN,At lO,5) ,STOR(10 ,5) ,CNU O,5),W SS(10,5),TLAG(1O,5),
USLE(10,5),SKR 1O,5),NTO,CP(150),NTS,NLST,W(10,S),Q(100,10,5),
3FS(100,10,5),VSrIAX,VTMAX,QSMAX,QTMAX,KTS,KTT,UNIT(10),QI(1S0,10),
4FSI(150,10) ,QD(150),FSD(l50),FSR(150),IAMC,NDUT,S(10,5),SS,VS,IS,
5F41( 10) ,JCASE ,NTA
0003 COMMON /CHAH/ DB(10),SLOPE(10),LEHSTH(10),DEPTH(10) ,SIDE(i0),
1 AN(10),OI5IRB(10O) CONECT(1O),PIPE(1O)
0004 REAL*4 LENGTH
0005 INTEGER*2 CONECT
0006 LOGICAL 1 PIPE
0007 ATOT 0.
0008 TLM 0.
0009 WRITE(NOUT ,201)
0010 201 FORrIAT(// ’ WATERSHED PARAHETEPS)
0011 DO 100 JJ 1 ,NC
0012 00 200 II 1,5
0013 REAO(S,2000) J,K,A( J,K) ,CH(J,K),STOR(J,K) ,SLR,SKN,SKS,USLE(J,K),
1 SLOPE1,ALEN
0014 2000 FOPMAT(2 15,F10.0 ,8F5.O)
0015 IF (A(J,KLLE. 0.) GO TO 105
0016 S2:1000./CN(J,K)—10.
0017 TLA6(J,K):((S2+1)**0.7)* ALEN/t1900*SQRT(SLOPE1)))
0018 WRITE(6,1010) TLAG(J,K)
0019 1010 FORMAT(0TLAG ,F11.4)
0020 ATOT ATOT . A(J,K)
0021 TLAG(J,K) = TLAG(J ,K)/TINC
0022 TLII AtIAX1(TLM,TLAG(J,K))
C
JUMP ON TYPE OF LAND IN CELL
C* 1,2:PERMEABLE. 3 ,4 ,SIMPERMEABLE(BUILDLJ.’ OCCURS)
C
0023 60 TO (120,120,110,110,110) ,K
0024 120 WSS(J,K) = 0.
0025 SKR(J,K) = 0.
0026 W(J,K) 0.
0027 60 TO 200
0028 110 SKR(J,KI = SKN + SKS
0029 WSS(J,K) = SLR/SKR(J,K)
0030 W(J,K) = WSS(J ,K)
0031 GO TO 200
0032 105 CONTINUE
0033 WSS(J,K):0.
0034 W(J,KI0.
0035 TLAGtJ ,K)0.
0036 200 CONTINUE
0037 F41(J ) 0.
0038 IF AtJ,4).GT.0.) F41(J) :A(J,4)/A(J,1)
0039 100 CONTINUE
0040 NLMAX TLM
0041 READ(5,1020 NCHAN
0042 1020 FORMAT(15)
0043 REAO(5,1021 J (PIPEtJ),DEPTHLJ),SLOPE(J),SIDE(J),APq(J),LENGTHfJ),
1 OB(J),CONECT(J),J1,NCHAN)
0044 1021 FORMAT(4X,L1,6F10.0,15)
0045 1025 FORPIAIC OCHANNEL CHARACTERISTICS’/lO( lOX, CHAP4
1 DIAtI—BWID= ,F9.3,2X,DEPTH ,F9.3,2X, ’SLOPE ,F6.3,
2 2X,SIDE SLOPE ,F6.3,2X,LEN ,F8.2,2X,MANNINGS ,
3 F7.4))
0046 RETURN
0047 END
160
-------
FORTRAN IV Gi RELEASE 2.0 STORM DATE = 78198 10/01/10
0001 SUBROUTINE STORM
C INITIATES STORM AND UPDATES WATERSHED VARIABLES
0002 COMMON P1 1 P2,TINC,FAC(3),NC,PLAST,XKW,NLMAX,ATOT,FIA(5),NSTO,
1NIN ,NSIN ,NCIN,A(1 0,5),STOR(1O,5),CN(1O,5), WSS( lo,5),TLAS( lo,5),
2USLE(1O,5),SKR(1O,5),NTD,CP(150),NTS,NLST,W(10,5),Q(10 0,10,5),
3FS(100 ,10 ,S) ,VSIIAX ,VTtIAX,QSMAX,QTMAX,KTS,KTT,UNIT(10),QI(150,10),
4FSI(150 ,10) ,QD(150),FSD(15 0),F SR(] .5 0),IAtIC,NOUT,5(1 0,5),SS,V5,IS,
SF4lt 10) ,JCASE,NTA
0003 COMMON /SWTCH/ OPTION(10),SAMPL,SWALES
0004 INTEGER 2 OPTION
0005 LOGICAL*1 SWALES,SAMPL -
C INPUT STORM DATA
0006 IF( .NOT.SA?IPL) READ(NSIN,101) NTA,NTD,(CP(I),I=1,NTD)
0007 101 FORMAT(5X,2I5,13F5.O/(15X,13F5.O))
C COMPUTE ANTECEDENT MOISTURE CONDITIONS
0008 TA = NTA*TINC
0009 IF(TA.GT.120.) GO TO 100
C
C* ANTECEDENT MOISTURE CONDITIONS 1,2, OR 3
C
0010 IF(PLAST.LE.P1) GO TO 100
0011 IF(PLAST.LE.P2) GO TO 120
0012 IAIIC 3
0013 GO TO 140
0014 120 IAtIC = 2
0015 GO TO 140
0016 100 IAMC 1
0017 140 PLAST CP(NTD)
C UPDATE S VALUES IN EACH CELL
0018 DO 300 J 1,NC
0019 DO 200 K 1,5
0020 IF(A(J,K).EQ.O) GO TO 200
0021 52 1000./CN(J ,K) — 10.
0022 S(J,K) FAC(IAMC)*S2
0023 200 CONTINUE
C UPDATE SOLIDS ACCUMULATION ON IMPERVIOUS AREAS IN EACH CELL
C LINEAR BUILDUP WITH FIRST-ORDER DECAY
0024 DO 300 K 3,5
0025 IF(A(J ,K).EQ.0) GO TO 300
0026 W(J,K) WSS(J,K) + (W(J,K)-WSS(J,K))*EXF(-SKR(J,K)*TA)
0027 300 CONTINUE
0028 RETURN
0029 END
161
-------
FORTRAN IV 81 RELEASE 2.0 CELL DATE 78198 10/01/10
0001 SUSROUTINE CELL
C COMPUTES RUNOFF, SOLIDS DYNAMICS FOR EACH CELL AND TItlE INCREMENT IN GIVEN STO
0002 COtIIION P1,P2,TINC,FAC(3),NC,PLAST,XKW ,NLMAX,ATOT,FIA(5),NST0 ,
1HIN,NSIN,NCIP4,A(10,5),STOR(10,5),CN(10,5),WSS(10,5),TLMt10 ,5) ,
ZUSLE(10,5),SKR(10,5),NTD,CP(150),NTS,NLST,W(10 ,5) ,Q(100 ,1 0 ,S) ,
3FS(100,10;5),VSt IAX,VTMAX,QSMAX,QTMAX,KTS,KTT,UNIT(1O),QI(150,10) ,
4FSI(150,10),QD(150),FSD(150),FSR(150),IAMC,N0UT,S(10,5),SS,’ ’5 ,IS ,
5F41( 10 ),JCASE,NTA
0003 REAL*4 CQ(5) ,DQ(5)
0004 DO 600 J 1,NC
C INITIALIZE WORKING VECTOR DQ
0005 00 100 K = 1,5
0006 100 DQ(K) = 0.
0007 CPL 0.
C COMPUTE EXCESS PRECIPITATION
0008 00 300 I 1,NTD
0009 00 200 KK 1,5
0010 K = 6-KK
0011 IF(A(i,K).LE.0) GO TO 200
0012 FS(I ,J,K) = 0.
0013 Q(I,J,K) = 0.
0014 TEST = CP(I)_FIA(K)*S(J,K)-STOR(J,K)
0015 IF(K—1) 215,215,216
C PERVIOUS AREA ACCEPTING RUNOFF FROM IMPERVIOUS APES
0016 215 TEST = TEST + CQ(4)*F41(J)
0017 216 IF(TEST) 210,210,220
0018 210 CQ(K) = 0.
0019 GO TO 230
0020 220 CQ(K) = A(J,K)*TEST*TEST/tTEST+S(J,K))
0021 230 Q(I,J,K) = (CQ(K)-DQ(K))
0022 DQ(K) = CQ(K)
0023 200 CONTINUE
C COMPUTE SOLIDS WASHOFF FROM IMPERVIOUS SURFACES
0024 DO 400 K = 3,5
0025 IF(A(J,K).LE.0) GO TO 400
0026 RUN = Q(I,J,K)/A(J ,K)
0027 OW W(J,K)*(1._EXP( XKW*RUN))
0028 W(J ,K) = W(J,K) - OW
0029 FS(I,J ,K) UNIT(1)*DW*A(J,K)
0030 400 CONTINUE
C COMPUTE EROSION FROM PERVIOUS AREA
0031 DO 450 K = 1,2
0032 IF(A(J,K).LE.0) GO TO 450
0033 FSt I,J,K) = UHIT(2)*USLE(J,K) ( (Q(1,J,K)*Q(I,J,K)/TINC) .56)
0034 450 CONTINUE
0035 PREC (CP(I)-CPL)
0036 CPL = CP(I)
C ADD SOLIDS INFLUX FROM IMPERVIOUS AREA
0037 IF(A(J,1hLE.0) GO TO 460
0038 TEMP:Q(I,J,4)+ PPEC*A(J,1)
0039 IF (TEMP.EQ.0) GO TO 460
0040 F5(I,J,1)FS(I,J,1)+FS(I,J,4)*Q(I,J,1)/TEMP
0041 460 CONTINUE
C END LOOP AROUND STORM EVENT
0042 300 CONTINUE
C END LOOP AROUND CELLS
0043 600 CONTINUE
0044 RETURN
0045 END
162
-------
FORTRAN IV Gi RELEASE 2.0 TREAT DATE 78198 10/01/10
0001 SUBROUTINE TREAT
C STORES AND/OR TREATS POLLUTOSRAPH BELOW POINT A
C IGNORES OVERLAP OF STORMS
0002 COMMON P1,P2,TINC,FAC(3),NC,PLAST,XKW,NLMAX,ATOT,FIA(5),NSTO,
1NIN,NSIN,NCIN,A(1O,5),STOR(10,5),CN(1O,5),W 55(1O,5),TLAG(1O,5),
ZLJSLE(1O,5), SKR(1O,5),t41 0,CP(15 0),NT S,NLST,W(1O,5),Q(1 0 0,1O,5),
3FS( 100, 10,5) ,VSMAX ,VTHAX ,Q SIIAX,QTMAX ,KTS ,KTT,UNIT( 10 ) ,QI( 150,10),
4F$I(15 0,1O),QD(15 0),FSD(15 0),F SR(15 0),IA MC,NOtJT,5(1O,5), SS,V5,IS,
5F41( 10) ,JCASE,NTA
0003 REAL KTS,KTT
C SET INITIAL CONDITIONS
0004 VS 0.
0005 SS 0.
0006 DO 200 I = 1,NLST
0007 QD(I) 0.
0008 FSD(I) 0.
0009 FSR(I) = 0.
0010 IF(Q1(I,1).LE.O.) GO TO 200
0011 160 VN VS + QI(I,1)
C CHECK FOP FILLED STORAGE FACILITY
0012 IFtVN—VSHAX) 110,120,120
C NOT FILLED
0013 110 SS SS + FSI(I,1)
0014 VS VS + QI(I,1)
0015 GO TO 200
C FILLED
0016 120 DV VSMAX — VS
0017 VS = VSMAX
0018 SS SS + DV*FSI(I,1)/QI(I,1)
0019 Q SB QI(I ,1)-DV
0020 FSB QSB FSI(I,1)/QI(I ,1)
C CHECK TREATMENT CAPACITY
0021 QD(I) QSB
0022 IF(QSB—QTMAX) 130,130,140
0023 130 FSD(I) FSB*(1.-KTR)
0024 FSR(I) = FSB — FSD(I)
0025 GO 10 200
C TREATMENT CAPACITY EXCEEDED
0026 140 FTP QTIIAX/QSB
0027 FSD(I) FSB U1.—FTR KTP)
0028 FSRtI) = FSB FTR KTR
0029 200 CONTINUE
0030 RETURN
0031 END
163
-------
FORTRAN IV Si RELEASE 2.0 RESULT DATE = 78198 10/01/10
0001 SUBROUTINE RESULT
C PRINTS RESULTS FOR LAST STORM
0002 COMMON Pi,p2,TIWC,FAC(3),UC,PLAST,XKW,NLMAX,ATOT,FIA(5),NSTO,
1NIN,NSIN,NCIN,A(10,5),STOP(10,5),CN(i0,5),WSS(1O,5),TLAGf 1O,5),
2USLE(10,5),SKR(10,5),NTD,CP(150),NTS,NIST,W(i0,5),Q(i00,10,5),
3F5(1 0 0,1O,5),V$HAx,VTtIAX,QS MAX,QTMAX,KTS ,KTT,UP4IT(i0),QI( 150,10),
4F 5 1(150,i0),QD(150),FSD( 150),FSR(150),IAMC,NOUT,S(10,5),SS,VS,IS,
5F41(10 ),JCASE,NTA
0003 REAL KTS,KTT
0004 WQITE(6,101) IS,NTA,NTD,NLST,IAMC ,.JCASE
0005 101 FCRMATr1” $TOptl : ’,15,’ ANTECEDENT DRY PD: ’,IS , ’ PRECIP DURATION’
1,’: ,I5, EVENT DURATION: ‘,IS , ’ ANTE MOISTURE COND ,15,
2’ CASE: , 15)
0006 WPITE(6,103)
0007 103 FORNAT(/ I P CP QI SRI FSI SSI ’,
1’ CI QD SQO FSD $50 CD FSR SSP
2)
0008 DO 200 I = NTD,NLST
0009 200 CP(I) = CP(NTD)
0010 POLO 0.
0011 SQ l = 0.
0012 SSI 0.
0013 SQO = 0.
0014 SSD 0.
0015 SSP 0.
0016 DO 100 I = 1,NLST
0017 CD = 0.
0018 CI 0.
0019 IF(QD(I)) 110,110,120
0020 120 CD UNIT(5)*FSD(I)/QD(I)
0021 110 IF(QI(I,1).LE.O.) GO TO 130
0022 CI UNIT(5)*FSI(I,i)/QI(I,1 )
0023 130 SQl SRI + QI(I,1)
0024 SSI SSI + FSI(I,i)
0025 SQO = SQO + Q0(I)
0026 5 50 SSD + FSD(I)
0027 SSR SSR + FSR(I)
0028 p CPU) - POLD
0029 WRITE(6,102)I ,P,CP(I),QI( I,1),SQI,FSI( I,i),SSI,CI,QD(I),SQD,
1 FSD(I),SSD,CD,FSRU),SSR
0030 102 FORMATLI4,14F8.3)
0031 POLD CP(I)
0032 100 CONTINUE
C STORED VOLUME AND SOLIDS
0033 VS VS + SQO
0034 FR SSR + KTS*SS
0035 SS SS0 + (i.-KTS)*SS
0036 EF = FR/SSI
0037 WRITE(NOUT,104) VS,SS,FR,EF
0038 104 FORMAT(// EVENT TOTALS INCLUDING STORED VOLUME AND SOLIDS’/
1’ DISCHARGE VOLUME = ‘,F10.2/ ’ DISCHARGE SOLIDS ‘,F10.2/
2’ REMOVED SOLIDS = ‘,Fi0.2/’ STORAGE/TREATMENT EFFICIENCY
3F10.3)
0039 END
164
-------
FORTRAN IV 61 RELEASE 2.0 CONTRL DATE 78198 10/01/10
0001 SUSROUTINE CONTRL
C
C* CALCULATES EFFECTS OF CONTROL OPTIONS
C
0002 COMMON P].,P2,TINC,FAC(3),NC,PLAST,XKW,NLMAX,ATOT,FIAI5),NSTO,
1NIN,NSIN,NCIN,A(10,5),STOR(1O,5),CN(10,5),WS5(10,5),TLAG(].0,5),
JJSLE(1O,5),SKR(10,5),NTD,CP(15 0),NTS,NLST,W(10,5),Q(1 00,10,5),
3FS(10 0,10,5),VSMAX,VTFIAX,QSMAX,QTMAX,KTS,KTT,UNIT(1O),QI(15O,10),
4FSI(150,10),QD(150),FSD(150),FSR(150),IAtIC,NOUT,S(10,5),SS,VS,IS,
5F41( 10) ,JCASE,NTA
0003 COMMON /SWTCH/ OPTION(10),SAMPL,SWALES
0004 COMMON /SAV/ SAVE(10,5),SAVE2(10,5),SAVE3(10,5),SAVE4(10,5)
0005 INTEGER 2 OPTION
0006 LOGICAL*1 SWALES,SAMPL
0007 SWALES .FALSE.
0008 READ(5,1000) OPTION
0009 1000 FO MAT(10I5)
0010 DO 50 11,NC
0011 DO 50 J:1,5
0012 SAVEtI,J) CNtI,J)
0013 SAVE2(I,J) A(I,J)
0014 SAVE3(I,J):USLE(I,J)
0015 SAVE4(I,J):STOR(I,J)
0016 50 CONTINUE
0017 DO 220 I 1,10
0018 IF(OPTION(I).LE.0) GO TO 250
0019 K:OPTION(I)
0020 GO TO (110,120,130,140,150,160,170),K
C
UNUSED OPTION
C
0021 WRITE(6,1010) OPTION(I)
0022 1010 FORMAT(0 NO SUCH OPTION NUt’IBER**, 1X,I5)
0023 GO TO 220
C
POROUS PAVEMENT
C
0024 110 DO 112 J 1,NC
0025 112 CN J,5) CNL),1)*0.8
0026 60 TO 220
C
VEGETATTIVE COVER
C
0027 120 DO 122 J 1,NC
0028 CN(J,1) 3O
0029 CN(J,2) 30.
0030 USLE(J,1) USLE(J,1)*0.5
0031 122 USLE(J,2)USLE(J,2)*0.5
0032 GO TO 220
C
C* ROOF DRAIN DISCONNECT
C
0033 130 00 132 J 1,NC
165
-------
FORTRAN IV 61 RELEASE 2.0 COHTRL DATE = 78198 10/01/10
0034 AtJ,4)A(J,4)+AtJ,3)*0.8
0035 132 A(J,3)A(J,3)*0.2
0036 60 TO 220
C
ON-LOT STORAGE
C
0037 140 DO 142 J1,NC
0038 142 STOR(J,1)0.07
C
0039 GO TO 220
C
C* SWALES
C
0040 150 SWALES.TRUE.
0041 GO TO 220
C
C* SOIL TYPES
C
0042 160 00 162 J:1,NC
0043 CN(J,1):85.
0044 162 CN(J,2):85.
C
C
0045 170 CONTP UE
0046 220 CONTINUE
0047 250 CONTINUE
0048 WRITE(6,1030)
0049 1030 FORMAT(0**END OF CONTROL MEASURE CHANGES**’)
0050 RETURN
0051 END
166
-------
FORTRAN IV Gi RELEASE 2.0 UNCNTL DATE 78198 10/01/10
0001 SUBROUTINE UNCNTL
C* RESTORES ORIGINAL ENVIRONMENT
0002 COMMON P1,P2,TINC,FAC(3),NC,PLAST,XKI4,NLMAX,ATOT,FIA(5),NSTO,
1NIN,NSIN,NCIN,A(10,5),STOR(10,5),CN(10,5),WSS(10,5),TLAG(10,5),
ZUSLE(1O,5),SKR(1O,5),HTD,Cpt15 0),HTS,NLsT,W(1O,5’),q(10 0,10,53,
3F 5(100,10,5),VSMAX,VTMAX,QSMAX,QTMAX,KTS,KTT,UHIT(10),QI(1S0,1O),
4FSI(150,10),QD(150),FSD(150),FSR(150),IAPIC,NOUT,S(10,5),SS,VS,IS,
5F41(10),JCASE,NTA
0003 COMMON /SWTCH/ OPTION(10),SAMPL,SWALES
0004 COMMON /SAV/ sAvE(1o,s’i,sAvEZ(1o,5),SAVE3(10,5 ,SAVE4(1O,5)
0005 INTEGER*Z OPTION
0006 LOGICAL*1 SWALES,SAMPL
0007 DO 50 I 1,NC
0008 DO 50 J:1,5
0009 CH(I,J3 5AVE (I,J)
0010 A(I,J) SAVE2(I,J)
0011 USLE(I,J):SAVE3(I,J)
0012 STOR(I,J) SAVE4(I,J)
0013 50 CONTINUE
0014 SWALES .FALSE.
0015 RETURN
0016 END
167
-------
FORTRAN IV Si RELEASE 2.0 GRAPH DATE = 78198 10/01/10
0001 SUBROUTINE GRAPH
C
0002 COtl?ION P1,P2,TINC,FAC( 3) ,NC,PLAST,XKW,NLFIAX,ATOT,FIA(5),NSTO,
1NIN,NSIN,NCIN,A(10,5),STOR(10,5),CN(10,5) ,WSS(10,5),TLAG(10,5),
2USLE(10,5),SKR(10,5),t4TD,CP(150),NTS,NLST,W(10,5),Q(100,10,5),
3F 5( 100,10,5) ,V t1AX,VTl1AX,QS?1AX,QTt1AX,KTS,KTT,UNIT( 10 ) ,QI( 150 ,10
4FSI(150,i0),QD(150),FSD(1S0),FSR(150),IAMC,NOUT,S(10 ,5),SS,VS,IS,
5F41( 10),JCASE,NTA
0003 C0 1NON /SWTCH/ OPTION(10),SAMPL,SWALES
0004 COIIIION /TITL/ TITLE(20)
0005 INTEGER*2 OPTION
0006 LOGICAL*1 SWALES,SAMPL
C
C
0007 LOGICAL 1 QSYM/*/,PSYM/P/,SSYN/S/,BLANK/ /,DSYM/ 2/
0008 LOGICAL*1 QLINE(121),FLAG
0009 INTEGER*2 QPOS,SPOS,PPOS
0010 REAL*4 PINC(3),SINC(11),QINC(11)
C
0011 FLAG:.FALSE.
C* GET SCALES
C
0012 QMAX:O.5
0013 SMAX:120.
0014 PMAX3.5
0015 QINCRQtIAX/100*10.
0016 SINCR:SIIAX/100*10.
0017 PINCRPMAX/20.*10.
0018 TINCR:0.0
0019 QSCALE:QMAX/99
0020 SSCALESMAX/99
0021 PSCALE:PMAX/20
0022 QINC(1)0.
0023 SINC(1)0.
0024 PINC(3):0.
0025 PINC(2):PINCR
0026 PINC(1):PINCR+PINCR
0027 DO 25 1=2,11
0028 QINC( I )QINCR+QINC( I—i)
0029 SINC( I ):SINCR+SINC( I—i)
0030 25 CONTD&IE
C
C* WRITE TOP HEADER
C
0031 WRITE(6 ,1000) TITLE
0032 1000 FORMAT(’1 ’,1OX,20A4)
0033 WRITE(6,2000)
0034 2000 FORT1AT( OCONTROL OPTIONS SELECTED ‘
0035 DO 30 1 :1,10
0036 KOPTION(I)
0037 IF(K.EQ.0) GO TO 30
0038 FLAS:.TRUE.
0039 60 TO 131,32,33,34,35,36,30,30,30,30),K
168
-------
FORTRAN IV 61 RELEASE 2.0 GRAPH DATE 76198 10/01/10
0040 GO 10 30
0041 31 WRITE(6,2001)
0042 2001 FORMAT(26X,’POROUS PAVEMENT)
0043 GO TO 30
0044 32 WRITE(6,2002)
0045 2002 FOThAT(26X,VEGETATIVE COVER)
0046 GO TO 30
0047 33 WRITE(6,2003)
0048 2003 FORIIAT(26X,ROOF-DRAINS DISCONNECTED)
0049 GO TO 30
0050 34 WRITE(6,2004)
0051 2004 FOPMAT( 26X, 0N-LOT STORAGE)
0052 GO TO 30
0053 35 WPITE(6,2005)
0054 2005 FORHAT(26X,SWALES)
0055 GO TO 30
0056 36 WRITE(6,2006)
0057 2006 FORMAT(26X,’IMPERVIOUS SOILS)
0058 30 CONTINUE
0059 IFLNOT.FLAG) WRITEt6,2010)
0060 2010 FOPMAT(26X,WONE)
0061 WRITE(6,1005)
0062 WRITE(6,101S1 PINC
0063 1015 FORMAT(109X,2(F3.1,7X),F3.1)
0064 WRITE(6,1006)
0065 1005 FOt MAT(15X,’FLOW & SOLIDS-—>’,65X,<—— CUMULATIVE RAIFP)
0066 1006 FOPHAT(1OX,+,24(’....+))
C
LOOP OVER TIME INTERVALS
C
0067 DO 100 11,NLST
C
C* BLANK LINE
C
0068 DO SO J1,1Z1
0069 50 QLINE(J):BLANK
C
C* GET POSITIONS
C
0070 IF(QI(I,1).GT.QMAX) QI(I,1) QMAX
0071 IF(FSI(I,1).GT.SMAX) FSI(I,l):SMAX
0072 QPOS INT(QI(I,1)/QSCALE)+1
0073 SPOS INT(FSI(I,1)/SSCALE ).1
0074 PPOS INT( CP I uPSCALE )+1
0075 PP0S 122-PPOS
C
0076 DO 77 J1,QPOS
0077 77 QLINE(J) QSYt1
0078 QLINE(SPOS):SSYM
0079 QLINE( PPOS ):PSYFI
C
C
0080 IFtI.EQ.1) IIRITE(6,1045) TINCR,QLINE
169
-------
FORTRAN IV Gi RELEASE 2.0 GRAPH DATE 78198 10/01/10
0081 IF((I—(I/S)*5LEQ.0) GO TO 75
0082 WRITE(6,1050) QLINE
0083 00 TO 80
0084 75 WRITE(6,1045) TINCR,QLINE
0085 1045 FOPtIAT(4X,F5.2, ‘+ ,121A1,
0086 1050 FORt1AT(9X,I ,121A1, I)
0087 80 CONTINUE
0088 TINCRTINCR+TINC
C
END OF LOOP OVER TItlE DELTAS
C
0089 100 CONTINUE
C
Cs WRITE TRAILER
C
0090 WRITE(6,1006)
0091 WRITE(6,1010) QINC
0092 1010 FORMAT(9X,11(F3.1,7X))
0093 WRITE(6,1020) SINC
0094 1020 FORMAT 7X,11(FS.1,5X))
0095 WRITE(6,1005J
C
0096 WRITE(6,1090) Q$CALE,QtIAX,SSCALE,SPIAX,PSCALE,PtIAX
0097 1090 FORtIAT(//10X,FLOW INCREMENTS ‘,F1O.4/
1 1OX,FLOW MAX ,F10.4/
2 1OX,50110S INCREMENTS: ,F10.4/
3 1OX,’SOLIDS MAX: ,F1O.4/
4 1 OX,RAIN(CU?1. ) : ,F10.4, INCHES /
5 1OX,’RAIN (MAX CLJtI.) ,F10.4,’ INCKES )
C
0098 RETURN
0099 END
170
-------
FORTRAN IV 81 RELEASE 2.0 TRANSP DATE = 78198 10/01/10
0001 SUBROUTINE TPANSP
C CONSTRUCTS HYDROSRAPH AND POLLUTOGRAPH AT POINT A
0002 C0T Th10N P1,P2 ,TINC,FAC(3),NC,PLAST,XKW,NIHAX,ATOT,FIA(5),NSTO,
1NIN,NSIN ,NCIN ,A(10 ,5),STOR(1 0,5),CN(10,5),WSS(1O,5),TL.AG(1O,5),
2USLEt1O,5),SKR(1O,5),NTD,CP(15 0),NT5,HLST,Wj l O,5),Q(10 0,10,5),
3FS(10 0 ,1O ,5) ,VSMAX,VTMAX,QSMAX,QTMAX,KTS,}(TT ,UNIT(10),QI(15 0,1O),
4F5I(150,1O),QD(15O),FSD(15O) FSR(15O),IAMC,NDUT,S(1O,5),SS,VS,IS,
5F41( 10) ,JC&SE,NTA
0003 COMMON /CHAN/ DB(10),SLOPE(] .0),LENGTH(10),DEPTH(10),SIDE(10),
1 AN(1O),DISTRB(1 00),CONECT(10) ,PIPE(10)
0004 COMMON /SWTCF4/ OPTION( 10) ,SANPL,SWALES
0005 REAL*4 QOUT(100,10),Q5(1 00,1O),FSOUT(100,10),F5(100,10),N$,LENGTH
0006 INTEGER*2 DPTION,CONECT
0007 LOGICAL*1 PIPE,PIFEE
0008 LOGICAL 1 S 4ALES,SAMPL
0009 NLST NTD + NLF1AX
0010 DO 10 1:1,100
0011 DO S J :1,NC
0012 FSI(I,J) :Q.
0013 QOUT(I,J):0.
0014 FSOUT(I,J) :O.
0015 5 QI(I,J) :0.
0016 10 CONTINUE
0017 DO 100 I : 1 ,NTD
0018 DO 100 J : 1 NC
0019 DO .5 K:1,5
0020 IF (A(J,K).LE.0) GO TO 95
0021 L = TLAG(J,K)
0022 L:I+L
0O’3 QI(L,J) :QI(L,J)+Q(I,J,K)
0024 FSI( L,J)FSI( L,J)+FS(I,J,K)
0025 95 CONTINUE
0026 100 CONTINUE
C
C* LOOP THROUGH LIST OF CELLS
C* ARRAY CONNECT HOLDS UPSTREAM-DOWNSTREAM CONNECTIONS
ADD HYDROGPAPHS BY STARTING WITH UPSTREAM END, FLOWS.
CALL TRANSPORT AND GET OUTPUT HYDROGRAPH, QOUT
C* ADO THIS TO NEXT Q(TItIE, CELL) TO GET NEXT INPUT HYDROGRAPH,
C QIN. ETC.
C
0027 MAXLST :0
0028 DO 140 J1,NC
0029 NODE :CONECT(J)
C
0030 DB1:DB(NDDE)
0031 Z:SIDE(NODE)
0032 S1 :SLOPE(NDDE)
0033 1:DEPTH( NODE)
0034 ALLENGTH(NODE)
0035 N$AN(NODE)
0036 IF(SWALES) N$N$*1.5
0037 PIPE2:PIPE( NODE)
171
-------
FORTRAN IV 61 RELEASE 2.0 TRAHSP DATE = 78198 10/01/10
0038 IF(J.EQ.1) NODE1NODE
C
0039 IF(.NOT.PIPE2) GO To 110
0040 AREAY*( DB1+Z*Y)
C
P1 IS WETTED PERIMETER
C
0041 P1:DB1+2.*Y*SQRT( 1+Z*Z)
0042 GO TO 120
C
CIRCULAR CR055-SECTION
C
0043 120 QCON 1.486/N$*AREA**1.667 (SQRT(S1))/P1* .667
0044 110 APEA 3.14159*DB1/4.
0045 P13 .14159/DG1
0046 H0
0047 Q1:0.
0048 DO 125 K:1,NLST
C
CHECK FOR ZERO FLOWS AND OVERFLOW
C
0049 IF(QI(K,HODE).EQ.0) GO TO 125
0050 IFtQI(K,NODE).LT. .00001) GO TO 125
0051 N:N+1
0052 IF(QI(K,NODE).LE.QCON) GO TO 124
0053 WRITE(6,1011) NODE,QI(K,NODE)
0054 1011 FORMAT(OOVEPFLOW IN CHANNEL , 15, FLOW ,F10.4)
0055 QI( K ,NODE ):QCON
0056 124 Q1 Q1+QI(K,NODE )+QOUT(K,NODE1)
C ACCLJ?IULATE FLOWS
C
0057 QI(K,HODE ):QI(K,NODE)+QOUT(K,NODE1)
0058 F5I(K,NODE) =F5I(K,NODE )+FSOUT(K NODE1)
0059 125 CONTINUE
C GET AVERAGE VOLUtIE
C
0060 IF(N.GT.0) Q1Q1/N
ROUTE USING THIS AVERAGE V
C
0061 CALL R0UTE(PIPE2,0B1,Z,Y,N$,S1,AL,Q1,VEL,TI)
0062 C:VEL/(VEL+l.7)
0063 TI:TI*C
0064 CS:1.
oo& 5 IF(TI.GT.0.) C 51. —(1.—C)**((TINC+0.5*TI)/(1.5*TI))
0066 QQ:O
0067 FF:0
0068 00 127 K2,NLST
0069 KK:K 1
0070 QOUT(K,NODE)( 1-CS )*QQ+C5*QI(KK,NODE)
0071 FSOUT(K,NODE ):(1-CS)*FF4CS *FSI(KK,NODE)
0072 FF:FSOUT(KK,NODE)
0073 127 QQ:QOUT(K,NODE)
0074 QOUT(1,NODE) 0.
172
-------
FORTRAN IV Gi RELEASE 2.0 TRANSP DATE 78198 10/01/10
0075 FSOUT(1,HODE) O.
0076 DO 130 K 1,NLST
0077 F5(K,NOOE) FSOUT(K,NODE)
0078 130 Q5(K,NODE):QOUT(K,NODE)
0079 KC TI/TINC+O.5
0080 DO 135 K=1,NLST
0081 KI K+KC
0082 FSOUT( KI ,NODE ) F5( K ,NODE)
0083 135 QOUTtKI,NODE) Q5 K,NODE)
0084 IF(KC.LE.0) GO TO 138
0085 DO 137 K 1,KC
0086 FSOUT(t(,NODE):0.
0087 137 QOUT(K,NODE) Q.
0088 138 CONTINUE
C
END OF LOOP OVER CELLS
C
0089 IF(MAXLST.LT.KC) MAXLST KC
0090 NODE1 NODE
0091 140 CONTINUE
0092 NLST NLST+MAXLST
0093 WRITE(6,8050) NLST,MAXLST
0094 8050 FORNAT(’OEND TI 1E ,I5, MAX TRAVEL TIME ,I5)
0095 DO 200 I 1,HLST
0096 QI(I,1) QOUT(I,NODE)
0097 FSI(I,1 FSOUT(I,NODE
0098 200 CONTINUE
0099 RETURN
0100 END
173
-------
FOQTRAN IV 61 RELEASE 2.0 ROUTE DATE 78198 10/01/10
0001 SUBROUTINE ROUTE(PIPE,DB,SIQE,DEPTH,MANING,SLOPE,LENGfl4,Q,VEL,TIJ
C
0 02 REAL*4 tIANING,LEt4GTH
0003 LOGICALS1 PIPE
C
0004 TI 0..
0005 VEL O.
0006 DYDEP T H/40.
0007 P 4 :0
Cs
Cs N IS INDEX FOR OVERFLOW CONDITION
0008 IF(.NOT.PIPE) GO TO 20
C
CS TRAPEZOIDAL CHANNEL
0009 5 AREA:OEPTHS(DB+SIOE*DEPTH)
Cs P IS WETTED PERIMETER
0010 P:DB+2 . *OEPTH*SQPT( 1.SIDE*SIDE)
C
C
0011 Q1:(1.486/MANING)*APEA S*1.667*(SLOPE S*O.5)/(Ps*0.667)
0012 IF(Q.GE. Qi) GO TO 10
0013 Q2 :q l
C
Cs ITERATE ON DEPTH
C
0014 0EPTH :OEPTH DY
0015 IF(DEPTH.LE.O) RETURN
0016 GO TO S
0017 10 IF(N.NE.1) GO TO 15
0018 DEPIH:OEPTH S3.
0019 WRITE(6,1000) DEPTH
0020 1000 FORMAT(0**FLOW EXCEEDS CHAHHEL,DEPTH: ,F8.4)
0021 GO TO 5
C
Cs GOT FINAL HYDRAULIC RADIUS
C
0022 15 DEPTh DEPTH+DY- Q2-Q )/( Q2-Q1 ) DY
0023 AREA:DEPTHS( DB+SIDE*DEPTH)
0024 P :DB+2.*Y*SQRT( 1.SIDE*SIDE)
0025 Q1 (1.486/MANING)*AREA* 51.&67*( SLOPE**0.5)/( P 5*0.667)
0026 GO TO 40
C
C* CIRCULAR PIPE SECTION
C
0027 20 TH :2.*3.14159
0028 DTH:TH/40.
0029 25 AREA :(TH—SIN(TH))*(DB*DB)/8.
0030 N :N+1
0031 P:TH*DB*0.5
0032 Q1:U.486/IIANING)*AREA**1.667*(SLOPE**0.5)/(P**0.667)
0033 IF(Q.GE.Q1) GO TO 28
0034 Q2 :Q1
0035 TH:TH-DTH
174
-------
FORTRAN IV 81 RELEASE 2.0 ROUTE DATE 78198 10/01/10
0036 IF(TH.LE.0) RETURN
0037 GO TO 25
0038 28 IF(N.HE.1) GO TO 30
0039 DB 3*DB
0040 WRITE(6,1015) 08
0041 1015 FOPtIAT(’O** PIPE DIAtI TRIPLED, CHANNEL OVERFLOW, DIAtI ‘,F8.4)
0042 GO TO 25
C
C GOT FINAL HYDRAULIC RADIUS
C
0043 30 TH TH+DTH-(Q2—Q)/tQ2-Q1)*0TH
0044 AREA:(TH-SIN(TH) )*DB*DB/8.
0045 P TH DB*0.5
0046 Q1 ( 1.486/MAHING) AREA**1.667*tSLOPE**0.5)/P**0.667
0047 DEPTH DB (1.-COS(TH/2.))*0.5
VELOCITY
0048 40 VEL Q1/AREA
TI TItlE SHIFT IN HOURS
0049 TI:LENGTH/V EL/3600.
0050 IF(N.EQ.1) WRITE(6 ,1055)
0051 1055 FORhAT(’O**CHANHEL OVERFLOW AT SOIIE POINT IN TIPIE**’)
0052 RETURN
0053 END
175
-------
FORTRAN IV Si RELEASE 2.0 SAMPLE DATE 78198 10/01/10
0001 SUBROUTINE SAMPLE( VOLUME)
0002 COMMON P1,P2,TINC,FAC(3),NC,PUST,XKW,NLMAX,ATOT,FIA(5),NSTO,
1NIH,NSIH,NCIN,A(1O,5),STOR(1O,S),CN(10,5),WSS(10,5),TLA (10,5)
2USLEt1O,5),SKR(1O,5),NTO,CP(15 0),NTS,NLST,W(10,5),Q(100,10,5),
3FS(1O0,10,5),V$t1AX,VTMAX,QSMAX,QTM .X,KTS,KTT,UNIT(iO),QI(150,10),
4FSI(150,1O ),QD(15 0),FSD(i50 ),FSR(i5O),IAMC,NOUT,S(10,5),55,V5tI51
5F41( 10 ),JCASE,NTA
0003 COMMON /CHAN/ DB(1O),SLOPE(10),LENGTH(1O),DEPTH(1O),SIDE(10),
1 AH(10),DISTRB(100),CONECT(1OhPIPE(1O)
0004 INTE tR 2 CONECT
0005 LOGICAL*1 PIPE
C
RETURNS RANDOM VOLUME INTERVAL AND SITRIBUTION
0006 REAL*8 SEED
0007 REAL*4 RESULT(1)
0008 READ(5,1005) NTD,NTA,VOLUtIE
0009 1005 FORtiAT(215,F1O.O)
0010 SEED:0.817585
0011 CALL GGUB(SEED,1,RESULT)
C
0012 REStJLT( 1 ):RESULT( 1 )/1O.
0013 VOLUtIE:(VOLUME+(PESULT(i)—0.05) )/iO0
0014 WPITE(6,1000) VOLUtIE
0015 1000 FORMAT(ORAHDOM VOLUME , F1O.4)
0016 READ(5,1010) (DISTRB(I), 11,NTD)
0017 1010 FOR 1AT(16F5.0)
0018 RETURN
0019 END
176
-------
FORTRAN IV Gi RELEASE 2.0 SETSTM DATE 78195 10/01/10
0001 SUBROUTINE SETSTh( SEED ,VOLUME)
0002 COMMON P1,P2,TINC,FAC(3),NC,PLAST,XKW,NLMAX,ATOT,FIA.(5),NSTO,
1NIN,NSIN,NCIN,A(1O,5),ST O P(1o,5),CN(1o,5),wss(1o,5),TLAG(1o,5),
2U LE(10,5 ),SKR(10,5),HTD,Cp(150),HTS,NLST,W(1O,5),Q(10O,10,5),
3F 5(100,10,5),VSMAX,VTMAX,qstjAx,qmAx,KT S,}crT,uNIT(1 0),QI( 15 0,1o,,
4FSI(150,1O),QD(150),F 5D(15 0),FS P(150),IAMC,NOUT,S(10,5),SS,V5,IS,
5F41( 10) ,JCASE,NTA
0003 COMMON /CNAN/ DB(10),SLOPE(1O),LENGTH(10),OEpTH(1O), SIDE(1O),
1 AN(10),DI STRB(1 0 0),CONECT(1O),pIpE(1 o)
0004 LOGICAL *1 PIPE
0005 INTEGER*2 CONECT
0006 REAL*4 LENGTH
0007 REAL*4 PESULT( 100)
0008 REAL*8 SEED
C
C
C
C* SEE IF WERE USING CARDS OR MONTE CARLO
C
0009 WRITE(6,1000)
0010 1000 FORMAT( ,0** MONTE CARLO SIMULATION **)
C
GET RANDOM NUMBERS FROM UNIFORM DIT.
C
0011 CALL GGLJB(SEED,100,RESULT)
0012 00 10 I 1,150
0013 10 CF(I) 0.
0014 DO 100 I 1,1O0
0015 DO 50 J 2,NTD
0016 IF(RESULT(I).LT.OISTRB(J).AND.RESULT(I).GE.DISTRB(J-1))CP(J) CP(J)
1 +VOLUME
0017 50 CONTINUE
0018 IF(PESULT(I).LE.DISTRB(1)) CP(1)CP(1)+VOLUME
0019 100 CONTINUE
C
0020 DO 200 I 2,NTD
0021 200 CP( I ) CP( I )4CP( I—i)
C
0022 WRITE(6,1010) (CP(I),I 1,NTD3
0023 1010 FORMAT(0PRECIPITATION ‘, 8(2X,F6.4,5X))
0024 RETURN
0025 END
177
-------
APPENDIX D 1
WRFORMANCE OF INNOVATIVE DESIGNS FOR SWRMWATER MANAGEMENT
As discussed in Section 9 of this report, innovative site and roadway
designs were recently implemented in two residential developments in South-
eastern Massachusetts, Trout Farm and Bowker Woods. The roadways and housing
unit layouts were designed to manage stormwater in non—traditional ways. This
appendix describes the actual roadway designs used for these two sites and
assesses their performance after one year of construction and use. Also
assessed are the infiltration characteristics of the sites. For a more
detailed description of the sites, refer to the case studies in Section 9 of
the main report.
I DAD DESIGNS
Given the highly permeable soil conditions at the Trout Farm site in
Duxbury, Massachusetts, the roadway was designed with a 4—inch porous macadam
surface placed upon a 6—inch gravel base. The subgrade preparation conformed
to the Commonwealth of Massachusetts, Department of Public Works Standard
Specifications, Section 170.16, and the gravel base to MDPW Section 405. The
porous macadam consisted of MDPW 1 1/2 inch crushed stone gradation, 2 gallons
per square yard of AC-b liquid asphalt, and 3/8 inch keystone.
The roadway has minimum shoulders for off-grading with clearing limited
to close to the paved surface. The road follows closely the natural contours
of the land; only a few significant cuts and fills were made. Paved surfaces
for two—way roads are 18 feet wide. One-way traffic sections, such as loop
cul—de—sacs, are 12 feet wide. Turn—outs for emergency vehicles were designed
at hydrants and within cul-de-sacs.
At the Bowker Woods site in Norwell, Massachusetts, the generally wet
soils are quite unsuitable for construction purposes. Where the soils did
not conform to the MDPW standards, a gravel base course was placed over the
prepared, excavated subgrade. The macadam surface is 8 inches of 1 1/2 inch
crushed stone, penetrated and choked with keystone as descrthed above. The
gravel base is pitched towards a 12-inch perforated aluminum underdrain which
sits in a 3/4—inch crushed stone filter directly contacting the macadam
surface. The underdrain carries the water infiltrating through the macadam
surface to a small surface retention pond which dries seasonally as the water
table drops.
1 This appendix is based on a document prepared by BSC Engineering, Inc.
under subcontract to Meta Systems.
178
-------
By placing the road in a slight cut just below the original ground level
it is possible to intercept runoff and spring seepage from upland areas,
filter it through the road and the underdrain envelope, and then discharge it
to a small pond. After filtering, settling, and buffering, the runoff flows
to an existing wetland.
As in Trout Farm, minimal clearing was performed. The width of the paved
surface is 16 feet including shoulder turn—outs. The road follows existing
contours and is not centered in the dedicated right—of—way. However, it is
entirely contained within the 50 foot right-of-way layout.
Roadway Performance Observations During Construction and Use
There has now been some actual experience with these roads. They have
been observed both during their construction and during the period of heavy
use when the homes in the developments were being constructed. They have
both gone through one winter. As would be expected, the keystone has become
visually worn in the tire tracks. This is particularly evident on the one
sharp curve in Trout Farm. Snow plowing has not been a major problem.
At only one location in Trout Farm has there been any lack of infiltra-
tive capacity. This is at a sharp turn at the entrance that receives heavy
construction loads under turning stresses. Here there have been puddles
after heavy rains. However, everywhere else the road has remained dry under
all precipitation and snow melt conditions. Although Trout Farm is now
approximately 40 percent built and occupied, there is still no apparent over-
land discharge to the brook which divides the site other than that which
falls directly on the stream or its banks, as it would have been had the site
been left in its natural state.
At Bowker Woods this spring, during the snow melt, small seepage streams
were observed running overland to the road and then disappearing through the
pavement. Even under these conditions very little of the road became wetted;
all of the runoff appears to be intercepted within the area about six inches
in from the edge of the pavement. After flowing through the road and the
settling pond to the wetland, this runoff discharge continues to be very small
and of apparently high quality.
It must be noted that the reduced asphalt content of the roadway pave-
ment does reduce its strength. While this is not a problem for even loads,
such as redi-mix concrete trucks, pavement strength is a direct function of
the gravel base grading and compaction. Where these reduced asphalt pave-
ments have been constructed in a slight embankment above the original ground
level and have little or no end restraint, it has been observed that wheel
loads can depress the edge of the pavement. There seems to be no change in
drainage performance of these edge depressions, however.
Site Infiltration Performance Observations
Since the years 1974 to 1975 winter frost observations have been made at
two sites in Southeastern Massachusetts: an undeveloped coastal plain, pitch
pine and scrub oak forest site in Plymouth; and Trout Farm.
179
-------
At the Plymouth site during the period of February 19 through 21, 1975,
a rainfall simulation was conducted with a cranberry bog irrigation system on
the frozen snow—covered woods. Rainfalls simulated ranged from 13 inches in
4 hours to 17 inches in 24 hours. Frost depths in the wooded sections were
1 1/4 to 1 1/2 inches and were 15 inches in the grassed shoulder of an old
cart path.
These depths are similar to the frost depths which have been measured
during three winters at Trout Farm. There, in the undisturbed portion of the
site, frost depths ranged from 1 1/4 to 1 1/2 inches in 1974—1975 to 1 3/4 to
2 inches in 1977—1978. This observation period has included two of the most
severe winters in recent history. Essentially, in these well-drained outwash
soils, only the fresh humus layers actually freeze.
It is not possible to excavate, grade, or compact soils such as these
without dramatically changing their structure and frost characteristics.
However, this alteration is limited to almost the exact area of disturbance,
even when clearing alters wind, sunlight exposures and microclimates. At
Trout Farm frost in the backfilled areas of the houses was 22 to 24 inches
deep by the end of January. The road base was also frozen to an estimated
depth of 20 inches. However, 3 feet away from the pavement in a graded
shoulder the frost was 7 1/4 inches, and the frost was 2 5/8 inches 6 feet
away. This compared to 1 3/4 to 2 inches in undisturbed areas.
From these observations it can be concluded that narrow roads, minimal
site work, and limited clearing for housing construction greatly enhance
site surficial geologic characteristics. Continuous irregular bands of
undisturbed land left throughout the construction site can provide natural
runoff interceptors. During winter and summer their high infiltration
capacities are sufficient to eliminate overland flow to a watercourse.
Furthermore, as at Bowker Woods, the roads can also act as such interceptors
if they are properly located on the landscape.
180
-------
APPENDIX E
COST ALLOCATION IN MULTIPURPOSE PROJECTS
Unlike traditional stormwater drainage systems, innovative stormwater
management projects are generally multipurpose projects, meeting such objectives
as improved water quality, improved flood control, more pleasing aesthetic
surroundings, lower operation and maintenance costs, etc. As such, it may
be in the interest of more than one organization or government agency to
promote their implementation. If more than one level of government or more
than one government agency becomes involved in promoting and financing
innovative stormwater management techniques, the question of appropriate cost
allocation may become important. This appendix addresses the question of
cost allocation by explaining general principles which could be followed,
outlining some rules which are now used to guide such decisions, and measuring
benefits accruing to multipurpose projects.
COST ALLOCATION IN MULTIPURPOSE PROJECTS
Whenever the financial responsibility for a project is divided, its
total cost must be distributed among the responsible groups. Cost allocation
can be defined as the assignment of costs of a multiple purpose project
to individual purposes. st cost allocation solutions attempt to meet some
combination of the following two objectives (Regan, 1964):
• Efficient use of resources;
• Promotion of incidence and distribution policies
Various arrangements attempt to minimize conflicts among these two objectives.
S. V. Ciriacy—Wantrup (1964), after reviewing numerous discussions of the
subject written by engineers, lawyers, accountants, and economists, reports
that most conclude that cost allocation is more or less arbitrary. Although
this conclusion may be generally valid, there are certain principles which
can form a basis upon which to develop a cost allocation procedure. We will
first examine those principles from economics, law and public finance which
apply to the problem of cost allocation. This review will provide background
for an examination of alternative rules which are commonly used to allocate
costs. We will conclude with a discussion of the difficulties inherent in
applying one of the most desirable cost allocation rules.
181
-------
Principles
In discussing common property resources and users of such resources who
engage in activities which have beneficial or harmful effects on other users
of the common resources (externalities) , economists rely on the principle of
efficiency to determine what the optimum allocation of resources should be.
The most efficient use of resources is that which maximizes the net value of
production. I nald Coase, in “The Problem of Social Cost” (1960), develops
what has become known as the Coase Theorem regarding activities involving
externalities. He demonstrates that the optimal allocation of resources will
occur if the actors are able to negotiate to their mutual advantage at no
cost. Under this condition the same result will be obtained no matter whichof
the joint cost causers is initially charged with the costs. That is, an
arrangement will be negotiated so that either the damages will compensate the
sufferer in order to be able to continue his activity or the potential sufferer
will pay the damage to desist or alter his activity.
However, in the real world negotiations, even market transactions, are
not completely costless (there are selling costs, for example). Especially
where common property such as a water resource is involved, there are such
difficulties as arranging for bargaining among large numbers of people or the
problem of excluding freeloaders, etc. Here, the recommendation of Coase
and others (see Calabresi, 1968, and Turvey, 1963) is that the least costly
alternative measures (market transaction, taxation liability rules), including
no charge at all, be chosen. To put it in efficiency terms, when the costs
of alternative arrangements are taken into account, a rearrangement should
only be undertaken when the resulting increase in the value of production is
greater than the costs involved in bringing it about. Thus the cost of
limiting environmental damage should be made to fall wherever it can be borne
most cheaply.
Many economists, such as those whose ideas are discussed above, avoid
judgements regarding the distributional consequences of assigning liability in
the cheapest possible way. A choice regarding income distribution is gen-
erally felt to be outside the province of economics. Frank I. Michelman,in
his essay, “Property, Utility, and Fairness: Comments on the Ethical Founda-
tions of ‘Just Compensation’ Law” (1967), tackles this weighty problem. He
discusses the situation in which society decides to reallocate resources so as
to increase total welfare (meet the efficiency objective), and consequently
some members of society become less well off than they were before. The
loss incurredby thesepeople is the opportunity cost of the allocation decision.
The question Michelman attempts to answer is whether such costs should remain
where they fall initially or whether compensation should be paid and the
costs explicitly distributed according to the tax structure or some other
principle. He appeals th utilitarian theories of property as a collection of
rules which are accepted as governing the exploitation and enjoyment of re—
sources. These rules form the basis of people’s expectations. Therefore,
he concludes that compensation should be paid in cases where if it were not
paid, it would be critically demoralizing to members of society.
182
-------
In any resource reallocation decision there are efficiency gains to be
had, demoralization costs, and what Michelman terins “settlement costs, ‘ (the
costs incurred in compensating those who are made less well off). Michelman s
test for a reallocation decision is that the measure should be rejected if
efficiency gains are less than demoralization costs or settlement costs and
that compensation should be paid when demoralization costs are greater than
settlement costs (or should be paid up to the point where the cost of compen-
sating the remaining losers is greater than the remaining demoralization
costs) . In applying this rule Michelman uses the Rawisiar concept of fairness:
an arrangement is fair if it is consistent with principles which can be
agreed upon by every member of the group, even the one who is least well
off under any rule. According to this view, compensation practice should
reconcile efficiency with the protection of expectations of fairness.
Musgrave and Musgrave (1976) also grapple with the definition of fairness
in their discussion of the objective of equity as the basic criterion for
designing an appropriate tax structure. An equitable tax structure is defined
as one in which each tax payer contributes his “fair share” to the cost of
government. The Musgraves discuss the approaches to defining “fair share”;
one approach involves the benefit principle, inwhich each tax payer contributes
according to the benfits he receives from public services. Examples of benefit
taxes are fees, user charges, and tolls. The application of this approach
is limited by the difficulties involved in identifying the beneficiaries of
many public services.
The other approach applies the ability-to-pay principle, in which
taxation is independent of the government’s expenditure policy. According to
this definition, a given revenue is needed and each taxpayer contributes in
line with his ability to pay. This requires that equal taxes will be paid
by persons with equal abilities to pay (horizontal equity) and unequal taxes
to be paid by those in unequal positions (vertical equity). Since John Stuart
Mill, equitable treatment has been viewed as involving an equal sacrifice
or loss of welfare. If measured in terms of income, the ability-to—pay
of persons with equal income will be equal. For those with unequal incomes,
the answer is less obvious. The Musgraves describe three rules which could
be chosen to define equal sacrifices: equal absolute sacrifice, in which the
loss of income from each person entails an equal loss of utility dependent
on each one’s shape of the marginal utility of income curve; equal proportional
sacrifice, in which each pays an equal proportion of income; equal marginal
sacrifice, in which the marginal utility of the income given up by each is
the same and which also results in the least total sacrifice. Thus there are
numerous choices to be made even in following what appears to be a straight-
forward allocation principle.
Common Cost Allocation Rules
As mentioned at the outset, it is necessary to divide the cost of a
multipurpose water resources project among the respective project purposes.
For example, costs for flood control must be distinguished from those for
water quality. Before discussing some commonly used rules for cost allocation,
183
-------
we will provide some necessary definitions. Direct costs are the costs of
each distinct physical portion of the project which serve only one purpose.
The separable costs of a purpose are the incremental costs of including that
purpose in the multiple purpose project. This is the difference between the
cost of the multipurpose project and the cost of the project with that parti-
cular purpose omitted. For example, the separable cost of flood control in a
dual-purpose flood control and water quality project is the cost of the dual-
purpose project less the cost of a single—purpose water quality project. Due
to economies of scale and complementary among purposes, total cost will
normally exceed the sum of the separable costs. This difference is the non—
separable cost . Nonseparable costs include joint and coninon costs. Joint
costs occur when a portion of the project contributes to the production of
more than one output. For example, water released from a reservoir for
low—flow augmentation may enhance both water quality and navigation. Connxon
costs are indirect or other fixed costs which must be paid but cannot be
associated with any production operation. The salary of a supervisor in
charge of operating a multipurpose water project is an example of such a cost.
James and Lee (1971) list eight guidelines for selecting a cost allocation
method. They stress that there is no correct method, since the choice of
method is essentially a successful resolution of conflicting interests.
1) The allocation to any purpose should never be less than the
additional cost of including that purpose in the plan nor more
than the total benefits provided to the purpose. The cost
allocated to a purpose such as flood control, which provides a
benefit of 3 and adds 2 to project cost, should be between 2 and 3.
2) The sum of the allocation to all the purposes should equal the
total. project cost.
3) The allocation method should avoid costly data and complex
computations that have no other use. Complex allocation
methods are no more correct than simple ones.
4) The allocation process should be straightforward and easily
understood. Conflicting interests are more likely to accept
compromises they can understand.
5) The amount allocated to each purpose determines the price
charged for project services. The cost allocated to an irri-
gation district will determine how much the district must
charge individual irrigators. If the allocation to the irri-
gation district results in a charge to the individual irrigator
approximately equaling the marginal cost of serving him, the
irrigator is encouraged to use the economically optimum amount
of water.
6) The charges resulting from the allocation should be fairly
constant with time in order to provide stability to the market
for project goods and services.
184
-------
7) As cost allocation helps determine user charges, it affects
income distribution. Where those served by one purpose
are relatively more well to do than those served by another,
the equity of the resulting income distribution is an important
component in evaluating the allocation.
8) Joint facilities should be operated in accordance with the
cost allocation. It is not equitable to allocate most of the
cost to a purpose having a low service priority in facility
operation.
As can be seen, these guidelines are influenced by the principles of effi-
ciency, fairness, and equity discussed earlier.
Once separable costs have been determined, following the guidance of
rule 1 above, nonseparable costs must be allocated to each purpose according
to an appropriate measure. The following vehicles have been used for allocat-
ing such costs (James and Lee, 1971):
1) Equally among the purposes;
2) Entirely to the high-priority purpose within the limit of benefit
the purpose provides;
3) Proportionally to the quantity of use of the facilities for the pur-
pose;
4) Proportionally to the benefit in excess of assigned separable cost to
the given purpose;
5) Proportionally to the excess cost required to provide the service by
some alternate means;
6) Proportionally to the smaller of the excess benefit or the excess
cost of the alternative project.
Different definitions of nonseparable costs have also been used. The most
common two are total financial cost less direct costs for the particular pur-
pose and total financial cost less assigned separable costs. The main advantage
of using direct costs is that the complex and often controversial computational
process required to estimate separable costs can be avoided. However, this
method does favor purposes which have large separable and small direct Costs.
The appropriateness of each of the above six allocation vehicles depends
on the specific situation in which they are used. The principles which were
discussed earlier can be used to evaluate the use of a particular rule for a
specific case.
Efficiency requires that costs be allocated as closely as possible to
actual costs of production, qualified by the cost required to carry the alloca-
tion procedure. The assignment of direct or separable costs to each purpose
follows this principle. Rules 1 and 2 would be very simple to follow, but
would only be appropriate in cases in which each purpose is approximately equal
in scope (rule 1) or in which there is one overriding purpose and other very
minor ones (rule 2). Rule 3 follows one kind of equity principle (see the
185
-------
discussion of benefit principle above), using a measure of facility use upon
which to base charges. This method, called the Use of Facilities Method, is
frequently used by the Soil Conservation Service (U.S. Department of Commerce,
1972). Although such an allocation may not limit the cost of a given purpose
to the benefit provided, it appeals to fairness by making payment proportional
to a visible, physical quantity (low settlement costs). Methods 4, 5, and 6
also incorporate fairness and equity by allocating cost proportionately to
alternative measures of benefits less the costs (direct or separable) pre—
viously assigned to each purpose. The use of net benefits in rule 4 appeals
to fairness and equity; however, there may be difficulties involved in calcu—
lating benefits in a clear and concise manner. More will be said about this
later. Th advantage of rule 5 is that calculation of benefits is avoided by
substituting, as a measure of benefits provided, the cost of the cheapest
alternative project which can provide the same service. Rule 6 combines rules
4 and 5 by allocating cost according to the smaller of the two benefit measures.
This method is the most widely used and has been adopted by the U.S. Inter-
Agency Committee on Water Resources (Kaiper, 1971). It is called the Separable
Costs-Remaining Benefits Method. The steps recommended by the Department of
the Army, Corps of Engineers to carry out this method are as follows (U.S.
Department of Commerce, 1972):
1) The benefits of each purpose are estimated.
2) The alternative costs of single-purpose projects to obtain the
same benfits are estimated.
3) The separable cost of each purpose is estimated.
4) The separable cost of each purpose in the multi—purpose project
is deducted from the lesser of each purpose’s benefits or alter-
native cost. The lesser figure is used since alternative cost
is used in this method only if it represents a justifiable ex-
penditure; that is, if it does not exceed the benefits.
5) From total cost of project all separable costs are deducted to
determine residual costs.
6) Residual costs, designated as joint costs in this method, are
distributed in direct proportion to the remainders found in step 4.
7) To determine the cost allocated to each purpose, the separable
and distributed costs for each purpose are added together, an4 in
the case of power the amount of taxes foregone that was used in com-
puting power costs under steps 2 and 3 is subtracted from that sum.
Benefits
As mentioned above, the most widely used cost allocation rules rely
on the measurement of benefits from a project which are attributable to a cer-
tain purpose. Despite the appeal of this method as reasonable and equitable,
it has serious practical limitations. Information about demand or benefit
schedules is not easily obtainable. We will review here some of the important
problems as they relate to benefits from improved water quality.
186
-------
Most economists generally agree that willingness—to—pay is the appropriate
measure of benfits. The choice facing society is not between clear water and
polluted water, for example, but between various levels of pollution. It is
the incremental or marginal values that are important in making decisions. The
“demand” for water quality (the analog to market demand) is the aggregate
of how much individuals will give up (will pay) to enjoy additional increments
of improved water quality.
The economic theory for valuing benefits is well developed. A complete
theory on the provision and use of public goods, those which are enjoyed
in common, such as the water quality of a stream, has been developed. But
as is well known, these general principles for management of public good
are not so easily applied. The problems of the misallocation of resources
and externalities are not theoretical but empirical ones. For instance, there
is the problem of the lack of a market. As we said, public goods are enjoyed
in common. They are shared and so they are not contained in market trans-
actions and have no market price to use to define demand. The question of
intangible benefits is also complex. A hypothetical demand curve can be de-
rived from aggregating individuals’ willingness—to—pay (for increased increments
of public good, as mentioned above) . One approach to estimating willingness—
to—pay is to calculate the damages that would occur if a project were not
undertaken. However, this method still underestimates psychic benefits
(called “intangibles”). In addition, the benefits to a public service such as
a multipurpose water resources project cannot be limited to the direct recip-
ients of the service. Indirect or secondary benefits, such as increased em-
ployment, accrue to the general public and can be of significant magnitude.
In most cases a complete, thorough analysis is impossible because it is
too difficult to estimate the multitude if impacts of, say, a change in water
quality. It is particularly difficult to isolate specific pollutants and
relate them to a value measurement. The existence of interactions, substitu-
tions, and indirect benefits in most water quality control problems contri-
butes to the difficulty of conducting an adequate analysis as defined by eco-
nomic theory. Furthermore, data needs are immense and the expense and person-
nel necessary for data collection are great. These are the greatest impedjments
to good empirical benefit estimation work. Examples of the types of data used
for the various methods of estimating water quality benefits are survey data,
property sales prices, detailed studies of physical damages, and origin and
destination data from travelers. Many methods use data that must be collected
anew for each study. Table E-l shows alternative methodologies which are
appropriate for measuring different water quality benefits (impacts). Depend-
ing on the use of the water and the surrounding land uses, certain impacts are
of more or less interest to groups of people concerned with water quality.
Therefore, it is necessary, using an appropriate methodology, to determine
which groups are likely to derive the most benefit from which aspects of im-
proved water quality. 1
1 For a review of recent empirical work pertinent to the estimation of water
quality benefits, see Meta Systems mc, Water Quality Impacts and Socio-’Eco-
nomic Aspects of Reducing Nonpoint Source Pollution From Agriculture (Appen-
dix E), prepared for U.S. Environmental Protection Agency, September, 1978.
187
-------
alternative
cost
TABLE E-1: COMPARISON OF METHODOLOGIES TO MEASURE WATER QUALITY BENEFITS
I-J
0,
co
Methodology
Types
— Benefit_Categories
,
. —
>1
U) a, ‘- .4 ‘ . -4
o 0
t O
4 - $1) “ . 4 “.4
ii .4W 4- 0 - - I
a • 4t) 4’
U 5 (U a ,U ) o 0 O o
o .. ‘W o , . .
° 0 . .L p 4 . 4 a, ‘ i U ‘ 0 “-4
a
a,
.l o Er
)
‘4- 4
time
budget
x
x
bidding
games
x
x
x
travel
costs
x
marginal
costs
medical
costs
& lost
earnings
treatment
roduction
costs
treatment
)roduction
costs
x
net factor
income
yield
change
.
xprice
yield
change
•
xprice
market
study
X
non—dollar
measurement
ranking
ranking
change
in
habitat
input/out-
put model
X
cost to
reproduce
-------
There are trade—offs involved in choosing a u thodo1ogy appropriate for
use in estimating benefits to water quality groups. The major one is the use
of readily available secondary data versus the need for a theoretically valid
model which relates specific pollutants to a value measurement. There are
more data available for certain benefit categories such as household water
supply than for others such as aesthetics. Surveys are expensive and time con-
suming,but there does not appear to be any feasible alternative, especially for
measuring recreation or aesthetic benefits which are two of the major cate-
gories of benefits from many multipurpose water resources projects.
Applicability to Stormwater Management
Since many innovative stormwater management projects will provide multi-
ple benefits, the principles discussed above are generally applicable toa con-
sideration of cost allocation among local government agencies and private inter-
ests and among agencies at various levels of government. However, most of the
on—site control measures applied at a subdivision level are too small indivi-
dually to justify the elaborate cost-benefit calculations that are undertaken
for large multi—purpose water resources projects. It will therefore be nec-
essary to develop cost—sharing rules based upon general notions of costs and
benefits for broad classes of projects.
As long as individual localities bear the sole responsibility for storm—
water management, it is likely that most will place the full cost on the sub-
division developers (and hence the new residents), as is currently the case
for conventional drainage. This procedure is justifiable under the benefits
received principle and promotes efficient land use decisions, since the resi-
dents and developers pay the full cost for developing the site in an environ—
mentally acceptable manner. However, if other levels of government become in—
volved in financing stormwater management facilities, then more complex allo-
cation formulas must be worked out along the lines suggested above.
189
-------
REFERENCES
Section 4
Coiston, Newton V., Jr., characterization and Treatment of Urban Land Runoff .
Cincinnati: U.S. Environmental Protection Agency, EPA-670/2-74-096,
December, 1974.
Fairfax County, Virginia, Erosion Sediment Control Handbook . Fairfax County,
Virginia, December, 1974.
Hittinan Associates, Inc., ? thods to Control Fine-Grained Sediments Resulting
from Construction Activity . Washington, D.C.: U.S. Environmental Protec-
tion Agency, EPA 440/9-76-026, December, 1976.
Poerther, Herbert G., “Practices in Detention of Urban Stormwater Runoff,”
Chicago: American Public Works Association Special Report, No. 43, 1974.
Real Estate Research Corporation, The Costs of Sprawl . Washington, D.C.:
U.S. Government Printing Office. Prepared for the U.S. Council on
Environmental Quality, Department of Housing and Urban Development, and
Environmental Protection Agency, April, 1974.
The Urban Land Institute, the American Society of Civil Engineers, and the
National Association of Homebuilders, Residential Storm Water Management .
Washington, D.C.: The Urban Land Institute and the American Society of
Civil Engineers, 1975.
UPS Research Company, Water Quality Management for Urban Runoff, A Manual .
Prepared for U.S. Environmental Protection Agency under Contract No.
68—01—1846, August, 1974.
U.S. Soil Conservation Service, “Urban Hydrology for Small Watersheds,”
Washington, D.C.: U.S. Department of Agriculture, Technical Release No. 55,
January, 1975.
Section 5 and Appendix A
American Public Works Association, Water Pollution Aspects of Urban Runoff .
Federal Water Pollution Control Administration, January, 1969.
190
-------
AVCO Systems Corporation, Storm Water Pollution from Urban Land Activity .
Wilmington, Massachusetts: AVCO Systems Corporation, Tulsa Study 2,
11O34FKL, July, 1970.
Bradford, W., “Urban Stormwater Pollutant Loadings -- A Statistical Summary
Through 1972,” Journal of the Water Pollution Control Federation . April,
1977.
Coiston, N., Characterization and Treatment of Urban Land Runoff . Washington,
D.C.: U.S. Environmental Protection Agency, EPA-670/2-74-096, December, 1974.
Envirex, Inc., Unpublished Data on Stormwater Quality and Quantity, For the
Department of Transportation, 1978.
Espey, Huston, and Associates, “Maximum Utilization of Water Resources in a
Planned Community,” Vol. 1, Draft Report. Edison, New Jersey: U.S. Environ-
mental Protection Agency, September, 1976.
Hammer, T.R., Planning Methodologies for Analysis of Land Use/Water Quality
Relationships . U.S. Environmental Protection Agency, Contract No. 68—01-
3551, October, 1976.
McElroy, A.D., et al., Loading Functions for Assessment of Water Pollution
from Non—Point Sources . Midwest Research Institute, EPA—600/2-76—151, May,
1976.
Moore, E.W., Thomas, M.A., Jr., Snow, W.B., “Simplified Method for Analysis of
BOD Data,” Sewage and Industrial Waste . Vol. 27, No. 1343, 1950.
Pisano, William C., “Interim Progress Report on Characterization of Solids
Behavior in, and Variability Testing of Selected Control Techniques for
Combined Sewer Systems,” Urban Stormwater Management Workshop Proceedings ,
Edison, New Jersey, December, 1, 1977.
Sartor, J.D. and Boyd, G.B., Water Pollution Aspects of Street Surface Contami-
nants . U.S. Environmental Protection Agency, EPA-R2—72-081, November, 1972.
Shaheen, D.G .., Contributions of Urban Roadway Usage to Water Pollution . U.S.
Environmental Protection Agency, EPA-600/2-7500 4 , 1975.
Singh, R., Statistical Coefficient of Variation of Pollutant Loading Rates .
Paper presented at ASCE Meeting, Dallas, Texas, April, 1977.
Sutherland, R. and McCuen, R., “A Mathematical Model for Estimating Pollution
Loadings in Runoff from Urban Streets,” in Brebbia, C.A., Mathematical
Models for Urban problems . London: Pentech Press, 1976.
Tukey, J.W., Exploratory Data Analysis . New York: Addison-Wesley, 1977.
191
-------
URS Research Co., “Water Pollution Aspects of Street Surface Contaminants,”
Draft of Final Report. Washington, D.C.: U.S. Environmental Protection
Agency, Contract No. 14-12-921, January, 1972.
URS Research Co., Water Quality Management Planning for Urban Runoff, A Manual .
Washington, D.C.: U.S. Environmental Protection Agency, Office of Water
Planning and Standards, Contract No. 68-01-1846, August, 1974.
Whipple, W., Hunter, J., and Yu, S., Characterization of Urban Runoff --
New Jersey . Washington, D.C.: Office of Water Research and Technology,
Project C—534l, June, 1976.
Woodward-Clyde Consultants, Demonstration of Non-Point Pollution Abatement
Through Improved Street Cleaning Practices . First Technical Report.
Edison, New Jersey: U.S. Environmental Protection Agency, April, 1977.
Section 6 and Appendix B
American Public Works Association, Nationwide Characterization Impacts, and
Critical Evaluation of Combined Sewer Overflows, Stormwater, and Non—Sewered
Urban Runoff . Washington, D.C.: U.S. Environmental Protection Agency,
EPA No. 68—03—0283, 1977.
Donigian, A.S., and Crawford, N.H., Modeling Nonpoint Pollution From the Land
Surface . Athens, Georgia: U.S. Environmental Protection Agency, Office
of Research and Development, Environmental Research Laboratory, EPA-600/
3—76—083, July, 1976.
Espey, Huston, and Associates, Maximum Utilization of Water Resources in a
Planned Coninunity , Vol. 1, Draft Report, Edison, New Jersey: U.S. Environ-
mental Protection Agency, September, 1976.
Jackson, T., and Rogan, R., “Hydrology of Porous Pavement Parking Lots,”
Journal of the Hydraulics Division of the American Society of Civil
Engineers , HY12, December, 1974, pp. 1739—1752.
Northeastern University, “Characterization of Solids Behaviors in, and
Variability Testing of Selected Control Techniques for Combined Sewer
Systems,” Washington, D.C.: U.S. Environmental Protection Agency,
EPA Research Grant No. R-804578, Interim Progress Report, 1977.
Poertner, Herbert G., Practice in Detention of Urban Stormwater Runoff ,
American Public Works Association Special Report No. 43, June, 1974.
Real Estate Research Corporation, The Costs of Sprawl: Environmental and
Economic Costs of Alternative Residential Development Patterns at the
Urban Fringe , Washington, D.C.: U.S. Government Printing Office, April,1974.
192
-------
Terstriep, M. and Stall, J.B., The Illinois State Drainage Area Simulator
( ILLUDAS) , Urbana: Illinois State Water Survey. Bulletin 58, 1974.
Thelen, E., Grover, W., Hoiberg, A., and Haigh, T., Investigation of Porous
Pavements for Urban Runoff Control , Washington, D.C.: U.S. Government
Printing Office, U.S. EPA No. 11034 DUY 03/72, March, 1972.
U.S. Soil Conservation Service, U.S. Department of Agriculture, A Method
for Estimating Volume and Rate of Runoff in Small Watersheds , Technical
Report No. 149, Washington, D.C.: U.S. Department of Agriculture, April,
1973.
U.S. Conservation Service, U.S. Department of Agriculture, Urban Hydrology
for Small Watersheds , Technical Release No. 55, Washington, D.C.: U.S.
Department of Agriculture, January, 1975.
U.S. Soil Conservation Service, U.S. Department of Agriculture, “ National
Engineering Handbook, ” Section 4, Hydrology, Washington, D.C.: U.S. De-
partment of Agriculture, August, 1972.
Williams, J.R.., “Sediment Yield Prediction with Universal Equation Using
Runoff Energy Factor,” in Present and Prospective Technology for Predicting
Sediment Yields and Sources , Oxford, Mississippi: Proceedings of the
Sediment-Yield Workshop, USDA Sedimentation Lab, November, 1972.
Wischeimer, W.H. and Smith, D.D.,, “Predicting Rainfall-Erosion Losses from
Cropland East of the Rocky Mountains,” Washington, D.C.: U.S. Department of
Agriculture, Agricultural Research Service, Handbook No. 282, 1972.
Section 7
American Society of Civil Engineers, Design and Construction of Sanitary and
Storm Sewers . New York: American Society of Civil Engineers, Manual of
Engineering Practice No. 37, 1970
Colston, N., Characterization and Treatment of Urban Land Runoff . Washington,
D.C.: U.S. Environmental Protection Agency, EPA-670/2-74-096, December, 1974.
Flatt, P. and Howard, C.D.D., “Preliminary Screening Procedure for Economic
Storage—Treatment Tradeoffs in Stormwater Control,” International Symposium
on Urban Stormwater Management , July, 1978.
Hydroscience, Inc., Areawide Assessment Procedures Manual , Vols. 1 and 2,
Cincinnati, Ohio: U.S. Environmental Protection Agency, Municipal Environ-
mental Research Laboratory, EPA-600/9-76-0l4, July, 1976.
Kirby, W., “Algebraic Boundedness of Sample Statistics,” Water Resources
Research , Vol. 10, No. 2, April, 1974.
193
-------
McKee, J.E., “Loss of Sanitary Sewage Through Storm Water Overflows,”
Journal of Boston Society of Civil Engineers . 1947.
Metcalf, L., and Eddy, H.P., Design of Sewers . Vol. 1., 1st ed., New York:
McGraw—Hill Book Company, Inc., 1914.
Ogden, C.E.., Sewer Design . 1st ed., New York: John Wiley and Sons, 1899.
Rouscuip, J.A., “Relation Qf Rainfall and Runoff to Cost of Sewers,”
Trans. American Society of Civil Engineers , Vol. 104, No. 1473, 1939.
Sartor, J.D., and Boyd, C.B., Water Pollution Aspects of Street Surface
Contaminants . Washington, D.C.: U.S. Environmental Protection Agency,
EPA—R2—72—081, November, 1972.
Slack, J., et al., “On the Value of Information to Flood Frequency Analysis,”
Water Resources Research , Vol. 11, No. 5, October, 1975.
Stanley, W.E., “A Discussion of McKee’s Paper,” Journal of Boston Society of
Civil Engineers , 1974.
Section 8
Benjes, H.H., “Cost Estimating Manual -- Combined Sewer Overflow Storage and
Treatment,” EPA—600/2—76-286, December, 1976.
Butler, Stanley S., Engineering Hydrology . Englewood Cliffs, New Jersey:
Prentice—Hall, Inc., 1964.
Engineering—Science, Inc., Comparative Costs of Erosion and Sediment Control
Construction Activities . EPA—430/9—73—0l6, July, 1973.
Grigg, Neil S. and O’Hearn, John P., “Development of Storm Drainage Cost
Functions,” Journal of the Hydraulics Division . AScE, Vol. 102, No. HY4,
April, 1976, pp. 515—526.
Rawls, Walter J. and Knapp, John W., “Methods for Predicting Urban Drainage
Costs,” Journal of the Hydraulics Division . ASCE, Vol. 98, No. HY3,
September, 1972, Pp. 1575—1585.
Sullivan, Richard H.; Manning, Martin J.; Heaney, J.P.; Huber, W.C.; Medina,
M., Jr.,; Murphy, M.P.; Nix, S .J.; and Hasan, S .M., Nationwide Evaluation
of Combined Sewer Overflows and Urban Wastewater Discharges . Vol. 1,
EPA—600/2—77—064a, September, 1977.
Thiel, Henri, Principles of Econometrics . New York: John Wiley and Sons, 1971.
194
-------
U.S. Environmental Protection Agency, Office of Water Program Operations,
Municipal Construction Division, An Analysis of Construction Cost Experience
for Wastewater Treatment Plants . EPA- .430/9—76—002, February, 1976.
Yarnell, David Q., Rainfall Intensity-Frequency Data . Washington, D.C.: U.S.
Department of Agriculture, August, 1935.
Section 9 and Appendix E
Bangs, Frank S. Jr., “PUD in Practice: The State and Local Legislative
Response,” in Frontiers of Planned Unit Development: A Synthesis of Expert
Opinion , Robert W. Burchell. New Brunswick, New Jersey: Center for Urban
Policy Research, 1973.
Burchell, Robert W., Planned Unit Development, New Communities American
Style . New Brunswick, New Jersey: Center for Urban Policy Research, 1972.
Calabresi,Guido, “Transaction-Costs, Resource Allocation, and LiabilityRules,”
Journal of Law and Economics , April, 1968. Reprinted in Economics of the
Environment , R. Dorfman and N. Dorfman. New York: W.W. Norton Co., 1972,
pp. 194—204.
Ciriacy—Wantrup, S.V., “Cost Allocation in Relation to Western Water Policies,”
in Economics and Public Policy in Water Resource Development , Stephen C.
Smith and Emery N. Castle. Ames, Iowa: Iowa State University Press, 1964.
Coase, Ronald, “The Problem of Social Cost,” Journal of Law and Economics ,
October, 1960. Reprinted in Economics of the Environment , R. Dorfman and
N. Dorfman, New York: W.W. Norton Co., 1972, pp. 100-129.
Falk, David, Building Codes and Manufactured Housing . Washington, D.C.:
U.S. Department of Housing and Urban Development, June 28, 1973.
Field, Charles G. and Rivkin, Stephen R., The Building Code Burden . Lexington,
Massachusetts: Lexington Books, 1975.
“Pezoning for the PUD,” House and Home , February, 1971, pp. 58-63.
James, Douglas L. and Lee, Robert R., Economics of Water Resources Planning .
New York: McGraw—Hill, 1971.
Kaiper, Edward, Water Resources Project Economics . Hartford, Connecticut:
Daniel Davey and Co., 1971.
Krasnowiecki, Jan Z., “Legal Aspects of Planned Unit Development in Theory
and Practice,” in Frontiers of Planned Unit Development: A Synthesis of
Expert Opinion , Robert W. Burchell. New Brunswick, New Jersey: Center for
Urban Policy Research, 1973.
195
-------
Lamb, Charles N., Land Use Politics arid the Law in the 1970’s , Program of
Policy Studies in Science and Technology Monograph 28, Washington, D.C.:
George Washington University, January 1975.
Manvel, Allen D., Local Land and Building Regulation , National Commission on
Urban Problems Research Report No. 6, Washington, D.C.: 1968.
Meta Systems mc, Water Quality Impact and Socio-Economic Aspects of Reducing
Nonpoint Source Pollution from Agriculture . Athens, Georgia: Draft Report,
prepared for the U.S. Environmental Protection Agency, February 1978.
Michelman, Frank I., “Property, Utility, and Fairness: Comments on the Ethical
Foundations of ‘Just Compensation’ Law”, Cambridge, Massachusetts:
Harvard Law Review , Vol. 80, No. 6, April 1967.
Musgrave, Richard A. and Musgrave, Peggy B., Public Finance in Theory and
Practice , 2nd ed. New York: McGraw-Hill, 1976.
National Commission on Urban Problems, Building’ the American City . Washington,
D.C.: U.S. Government Printing Office, 1968.
National Commission on Urban Problems, Fragmentation in Land—Use Planning
and Control , Research Report No. 18, 1969.
Rayn nd and May Associates, Zoning Controversies in the Suburbs . National
Commission on Urban Problems Research Report No. 11, 1968.
Regan, Mark M., “Sharing Financial Responsibility of River Basin Development,”
in Economics and Public Policy in Water Resource Development , Stephen C.
Smith and Emery N. Castle. Ames, Iowa: Iowa State University Press, 1964.
So, Frank S., Mosena, David R., and Bangs, Frank S. Jr., “Planned Unit Develop-
ment Ordinances,” in American Society of Planning Officials , Planning
Advisory Service Report No. 291, May 1973, p. 26.
Turvey, Ralph, “On Divergences Between Social Costs and Private Cost,”
Economics , August 1963. Reprinted in Economics of the Environment , R.Dorfman
and N. Dorfman. New York: W.W. Norton Co., 1972, pp. 130-134.
Urban Land Institute, Large Scale Development: Benefits, Constraints, and
State and Local Policy Incentives . Washington, D.C.: 1977.
U.S. Department of Commerce, National Bureau of Standards, Federal Cost-Sharing
Polcies for Water Resources . Washington, D.C.: prepared for the National
Water Commission, April 1972.
196
-------
GLOSSARY
analysis of covariance: An extension of analysis of variance (see
below) which controls for the influence of variables measured
on a continuous scale.
analysis of variance: Method by which the total variation of a set of
observations (measured by sums of squares of deviations from
the mean) is separated into categories associated with the be-
havior of the variation.
arithmetic mean: The sum of observations in a data set divided by the
total number of observations.
average daily traffic (A.DT): The average number of vehicles passing a
specified point on a roadway during a 24-hour period.
BOD: Biological Oxygen Demand: The quantity of dissolved oxygen used
by microorganisms in the biochemical oxidation of organic
matter and oxidizable inorganic matter by aerobic biological
action. Generally this refers to the standard 5-day BOD test
(BOD5).
box and whisker plot: A graphical display of a data set which has a
“box” to show the range from upper hinge to lower hinge and
“whiskers” or “tails” to show the extreme values.
Central Limit Theorem: In simple form, the theorem states that if n
independent samples have finite variances then their sum will
tend to be normally distributed as n tends to infinity.
COD: A measure of the oxygen-consuming capacity of inorganic and organic
matter present in water or wastewater. It is expressed as the
amount of oxygen consumed from a chemical oxidant in a specific
test. It does not differentiate between stable and unstable
organic matter and thus does not necessarily correlate with
biochemical oxygen demand.
column effect: See two-way table.
collinearity: See multicollinearity.
confidence interval: The interval between two values such that the
probability of an observation falling within an interval is
equal to a specified value.
197
-------
covariance: A measure of the degree of association between two random
variables.
dependent variable: See regression analysis.
deposition: The act or process of settling solid material from a fluid
suspension.
design storm: A hypothetioal or real storm event which specifies the
storm conditions under which storm related structures will
be designed to operate successfully.
detention: Temporarily holding storm water on a surface area, in a
storage basin or within the sewer system.
deterministic: Containing no random elements, hence completely determined
at some fixed point in time.
diagnostic plot: A plot of comparison value versus residuals for
each observation for a statistical fit. It is used to define
systematic patterns in the residuals.
dummy variable: A variable that takes on one of two values. For example:
o and 1.
economies of scale: A situation of decreasing costs per unit of produc-
tion with increasing rates of production.
efficiency: The ratio of the lowest variance feasible to the actual
variance of a statistic.
first flush: The condition, often occurring in suspended solids dis-
charges and contained sewer overflows, in which a dispropor-
tionately high pollution load is carried in the first portion
of the discharge or overflow.
geometric mean: Where n is the number of observations and x 1 , x 2 x
are the observations, the geometric mean G can be
shown as:
,.xx ,x
“ 12 n
grand mean: See two—way table.
heavy metals: Metals that can be precipitated by hydrogen sulfide in
acid solution. For example: lead, silver, gold, mercury and
copper.
hinge: upper (lower): The values greater than 75 (25) percent of all
observations for a variable.
198
-------
histogram: A graphical representation of a frequency distribution by
means of rectangles whose widths are the class intervals and
whose heights are the corresponding frequencies.
hydraulic radius: The cross-sectional area of a stream of water divided
by the length of that part at its periphery in contact with its
containing conduit; the ratio of area to wetted perimeter.
hydrograph: A plot of discharge (or stage) versus time.
hyetograph: A graph of average rainfall, rainfall excess rates, or
volume rates over specified areas during successive units of
time during a storm.
independent variable: See regression analysis.
infiltration: The water entering a sewer systeia and service connections
from the ground, through such means as, but not limited to,
defective pipes, pipe joints, connections or manhole walls.
The movement of water into the ground from the surface.
interaction: Non—additive effects of two or more independent variables
on the value of a dependent variable.
interpolation: A process used to estimate an intermediate value of one
(dependent) variable when values of the dependent variable
corresponding to several discrete values of the independent
variable are known.
Kjeldahl Nitrogen: A measure of organic nitrogen content in water and
wastewater.
least squares: A technique of fitting a curve close to some given
points which minimizes the sum of the squares of the deviations
of the given points from the curve.
loading: 1) AccumulatiOn of potential pollutants in a washoff area
such as a street; 2) Increase of pollutant concentration
in a stream of water.
Manning’s Equation: A hydraulic equation which explains flow velocity
(v) as a function of hydraulic radius (R), slope (s) and an
empirically derived surface roughness coefficient (“Manning’s
coefficient”), n. The equation commonly appears in the form:
1.486 R 213 12
mean: The arithmetic mean of a set of numbers.
199
-------
mean square deviation: A measure of the extent to which a collection
v 1 , v 2 ,...., v 2 of numbers is equal; it is given by the expression:
(1/n) [ (vi 2 + ••• + (v ...) 2]
where V is the mean of the numbers.
median: Half the data lies above or below this value.
Michaelis-Menton Equation: A growth equation derived from the differen-
tial equations used to characterize enzyme-catalyzed reactions.
Monte Carlo Method: A technique which obtains a probabilistic approxi-
mation to the solution of a problem by using statistical
sampling techniques.
xnuiticollinearity: A problem arising in regression analysis when the
same quality is measured by two or more independent variables.
It leads to arbitrariness about the allocation of coefficients
given to the different variables.
non—conservative: A substance which undergoes chemical or biological
transformation in the environment.
non—linear regression: Regression study of jointly distributed random
variables where the function measuring their statistical
dependence is non—linear.
nonpoint source: ny confined and nondiscrete conveyance from which
pollutants are or may be discharged.
pervious: Possessing a texture that permits water to move through
perceptibly under the head differences ordinarily found in
subsurface water.
polish: To use an iteration technique to find the best estimate of row
and column effects in two—way analyses. It is only relevant
where the calculation uses an average value other than the
mean (e.g., median).
pollutograph: A plot of pollutant concentration (or load) versus time.
porous pavement: A pavement designed to have a high infiltration rate.
probabilistic: Pertaining to random variables.
200
-------
PTJD: Planned Unit Development: A large residential development which
locates housing on part of a site, leaving the rest for open
space.
R 2 : Coefficient of Determination: A standard measure of correlation or
explained variation in a regression of two or more variables.
Possible values lie between 0 and 1, with 0 indicating no
correlation and 1 indicating perfect correlation (no unexplained
variation)
random variable: A measurable function on a probability space; usually
real valued, but possibly with values in a general measurable
space.
regression analysis: Given two or more stochastically dependent random
variables, regression functions measure the mean expectation
of one (the dependent variable) relative to the other or
others (the independent variables).
residual: The difference between an observed value and a value predicted
by statistical analysis. Also called “noise” or “unexplained
variation.”
robust: A characteristic of a statistic that is efficient under a
variety of situations.
roof-gutter disconnect: A method of reducing storm water flows by
disconnecting roof gutters, thereby allowing otherwise
contained storm water to soak into pervious lawn areas.
row effect: See two—way table.
runoff: That part of the precipitation which runs off the surface of a
drainage area and reaches a stream or other body of water or
a drain or sewer.
scour: The action of a flowing liquid as it lifts and carries away the
material on the sides or bottom of a waterway, conduit or
pipeline.
sensitivity analysis: Analysis of the sensitivity of model output to
changes in the values of input parameters.
skewness: A measure of the departure of a frequency curve from symmetry.
simulation: The representation of a system by a model that imitates
the behavior of the system.
sinusoidal: Exhibiting periodic behavior analagous to a sine wave.
201
-------
Soil Conservation Service method: A procedure for estimating peak
runoff for a watershed given certain information on soil types
antecedent conditions (i.e., moisture retention factor) and
slope.
solids: Total solids the sum of dissolved and undissolved constituents
in water or wastewater, usually stated in milligrams per liter.
spread: The difference between upper and lower hinges, i.e., contains
one-half of the data.
standard d viation: A common measure of dispersion equal to the square
root of the variance.
stem and leaf diagram: A method of graphically displaying the frequency
distribution of a variable.
stochastic: Pertaining to random variables.
STORM: Storage Treatment and Overflow Model: A computer simulation of
stormwater runoff developed by U.S. Corps of Engineers Hydraulic
Engineering Center.
swale: A wide, shallow ditch, usually grassed or paved.
SWMM: Storm Water Management Model: A model developed by the EPA to
generate detailed simulation of the quality and quantity of
stormwater during a precipitation event.
transformation: A function between vector spaces.
two—way table: A method of evaluating the effects of two (independent)
variables on a third (dependent) variable. The dependent
observations are placed in cells of the table. Bows of the
table represent the values of one independent variable, the
columns represent the categories of the second. The analysis
separates the influence of the independent variable (row effect
and column effect) from the common value or grand mean.
univariate: A single variable.
variance: A measure of dispersion of a set of numbers from a central
tendency equal to:
1
n-i
where x. are the numbers in the set, x is the mean and n is the
total n imber of observations.
202
-------
CON VS RSI ON FACTORS
U.S. CUSTOMARY TO METRIC
U.S. Customary Abbr. Multiplier Abbr. Metric Unit
acre
acre— foot
cubic foot
cubic feet per second
cubic inch
cubic yard
degree Fahrenheit
feet per minute
feet per second
foot (feet)
gallon(s)
(U.S. liquid)
gallons per minute
inch(es)
inches per hour
million gallons per
day
acre
acre—ft
ft 3
c fs
he ctare
cubic meter
liter
cubic meters per
second
cubic centimeter
cubic meter
degree Celsius
meters per second
meters per second
meter(s)
liter(s)
liters per second
centimeter
centimeters
per hour
cubic meters
per second
kilometer
kilometer per
hour
kilogram
0.4047
1,233.5
28.32
0.02832
16.39
0. 7646
(°F—32)/1.8
0. 005080
0.3048
0. 3048
3.785
0. 06309
2.540
2.540
ha
In 3
1
xn 3 /sec
cm 3
in 3
rn/sec
rn/Sec
In
1
1/sec
cm
cm/hr
in
yd 3
f pm
fps
ft
gal.
gpm
in.
in. /hr
rngd
mi
mph
lb
0.04381 m 3 /sec
mile
miles per hour
pound(s)
1.609
1.609
km
km/hr
0.4536 kg
203
-------
U.S. Customary
Abbr.
Multiplier
Abbr.
Metric Unit
square foot
ft 2
0.09290
in 2
square
meter
square inch
in 2
6.452
cm 2
square
centimeter
square mile
m1 2
2.590
km 2
square
kilometer
square yard
yd 2
0.8361
m 2
square
meter
ton (short)
ton
0.9072
metric
ton
metric
ton
yard
yd
0.9144
m
meter
CONVERSION OF CONCENTRATION x DISCHARC TO MASS FLOW RATE
Define:
C pollutant concentration
Q — wastewater discharge
M = pollutant mass flow rate
Then:
M 1b/hr} = 0.224 741 x C{mg/1} x Q{cfs}
M{kg/hr} = 0.101 941 x C{mg/l} x Q{cfs}
M{lb/hr} = 7.936 641 x C{mg/1} x Q{m 3 /sec}
M{kg/hr} = 3.600 000 x c{mg/1} x Q{m 3 /sec}
204
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1, REPORT NO. 2.
EPA—600/2—80—0l3
3. RECIPIENT S ACCESSION NO.
4. TITLE AND SUBTITLE
SELECT TOPICS IN STORMWATER MANAGEMENT PLANNING
FOR NEW RESIDENTIAL DEVELOPMENTS
5. REPORT DATE
March 1980 (Issuing Date)
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Robert Berwick, Michael Shapiro, Jochen Kuhner,
Daniel Luecke, Janet 3. Wineinan
S. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Meta Systems, Inc.
Ten Hoiworthy Street
Cambridge, Massachusetts 02138
10. PROGRAM ELEMENT NO.
1BC822, SOS #2, Task #8
11. CONTRACT/GRANT NO.
R—805238
12. SPONSORING AGENCY NAME AND ADDRESS OH
Municipal Environmental Research Laboratory--Cm.
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, Ohio 45268
13. TYPE OF REPORT AND PERIOD COVERED
Final July 1977-January 1979
14. SPONSORING AGENCY CODE
EPA/600/ 14
15. SUPPLEMENTARY NOTES .
Richard Field, Chief, Storm and Combined Sewer Section
Edison, NJ 08817 Tel. (201) 321—6674
FTS 340—6674
16. ABSTRACT Several aspects of stormwater management planning for new residential
developments are investigated. Areas of research include the evaluation of pollutant
accumulation and washoff data using exploratory statistical techniques; simulations to
compare the relative effectiveness of various control measures and layout patterns
from a small subdivision; formulation of simple stochastic models for stormwater man-
agement planning; estimation of cost models for conventional storm sewer systems; and
evaluation of institutional and political problems in implementing non-conventional
control measures. Analysis of existing data on street surface accumulation and washoff
suggests the modification of functional forms and parameter values in current storm—
water simulation models such as STORM or SW 1 that are used to estimate street loadings
and washoff. Simulation studies, used to evaluate the effect of on-site control mea-
sures and development layout on runoff, indicate that porous pavement and interactions
with subdivision layout are important in controlling runoff. Three simple stochastic
models were developed to illustrate their use as preliminary planning tools. As for
costs, a planning model that predicts quite accurately conventional drainage costs was
developed from an existing data set. Finally, investigation into the institutional
aspects of innovative stormwater controls, including two detailed Massachusetts case
studies, identified several factors as important in the acceptance of innovation:
strength of the housing market; professionalism and technical expertise in government;
aric l oitiv izr nd r ” ’
‘-
17. KEY WORDS AND DOCUMENT ANALYSIS
I. DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
C. COSATI Field/Group
Water pollution, Control simulation, Cost
effectiveness, Surface water runoff,
Mathematical models, Economics, Land use
Residential development,
Street surface pollutant
accumulation
l3B
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
221
20. SECURITY CLASS (This page)
UNCLASSIFIED
22. PRICE
EPA Form 2220—1 (Rev. 4—77)
205
US GOVERNMENT PPtNTIMV VFF(CE oso —657—146(5619
------- |