c/EPA
United States
Environmental Protection
Agency
Office of Water
Nonpoint Source Branch
Washington DC 20460
EPA440/5-87-001
September 1986
Water
Methodology for Analysis
of Detention Basins
for Control of
Urban Runoff Quality
-------
-------
TABLE OF CONTENTS
Section
ACKNOWLEDGEMENTS
FOREWORD
1.0 INTRODUCTION 1
1.1 General 1
1.2 Organization of Report 2
2.0 METHOD OF ANALYSIS 3
2.1 General 3
2.2 Rainfall 4
2.3 Flow - Capture 4
2.4 Flow - Treatment 5
2.5 Volume - Capture 7
3.0 RECHARGE DEVICES 14
3.1 General 14
3.2 Analysis Method 15
3.3 Example Computations 17
3.4 Validation 22
3.5 Discussion 22
4.0 SEDIMENTATION DEVICES 25
4.1 General 25
4.2 Analysis Method 26
4.3 Validation 32
4.4 Example Computation 39
4.5 Discussion 43
5.0 GENERAL PERFORMANCE PROJECTIONS 44
6.0 REFERENCES 51
APPENDIX - DATA ON INPUT PARAMETERS A-l
1.0 General A-l
2.0 Rainfall Statistics A-l
3.0 Runoff Coefficient A-5
4.0 Settling Velocities . A-8
-------
LIST OF TABLES
Table Page
1 SIZE RELATIONSHIPS FOR NURP DETENTION BASINS 35
2 OBSERVED PERFORMANCE OF WET DETENTION BASINS 36
ii
-------
LIST OF FIGURES
Figure Page
1 AVERAGE LONG-TERM PERFORMANCE: FLOW-CAPTURE DEVICE 6
2 LONG-TERM PERFORMANCE OF A DEVICE WHERE REMOVAL
MECHANISM IS SENSITIVE TO FLOW RATE 8
3 AVERAGE LONG-TERM PERFORMANCE: VOLUME DEVICE 11
4 EFFECT OF PREVIOUS STORMS ON LONG-TERM EFFECTIVE
STORAGE CAPACITY 12
5 SCHEMATIC ILLUSTRATION OF RECHARGE DEVICE 16
6 DETENTION BASIN PERFORMANCE - LONG-TERM AVERAGE
REMOVALS BY PERCOLATION - COMPARISON OF STATISTICAL
AND SIMULATION EVENTS 23
7 EFFECT OF SETTLING VELOCITY AND OVERFLOW RATE ON
REMOVAL EFFICIENCY 28
8 FLOW-REMOVAL RELATIONSHIPS FOR EXPONENTIAL
APPROXIMATION 28
9 ILLUSTRATION OF QUIESCENT VS DYNAMIC RESIDENCE TIME
IN A STORM DETENTION BASIN 31
10 COMPARISON OF OBSERVED VS COMPUTED REMOVAL
EFFICIENCIES .38
11 REGIONAL DIFFERENCES IN DETENTION BASIN PERFORMANCE 46
12 EFFECT OF DEPTH (VOLUME) ON PERFORMANCE 47
13 EFFECT OF RUNOFF COEFFICIENT ON PERFORMANCE 48
14 DETENTION BASIN PERFORMANCE 49
iii
-------
ACKNOWLEDGEMENTS
The EPA Nationwide Urban Runoff Program (NURP) supported the preparation of this
manual as well as the local studies that monitored the performance of the type of urban runoff
control measures that are addressed herein. The contribution of the EPA Project Officer, Dennis N.
Athayde, in supporting and encouraging the development of the analysis techniques described in
this report was critical to the effort.
The basic probabilistic methodology that was adapted to the two specific control techniques
for urban runoff addressed here, was conceived and formulated by Dr. Dominic M. DiToro
(Manhattan College and HydroQual Inc.) and further developed by Dr. DiToro and Dr. Mitchell
Small (currently Carnegie Mellon University). This basic groundwork was partly independent, and
partly supported by EPA's NURP program and an earlier contract for which Mr. Athayde was also
the Project Officer.
The adaptation of these basic probabilistic analyses to the specific urban runoff control
measures addressed here, the analysis of the NURP program data, and the preparation of this
report are the work of Eugene D. Driscoll (Woodward-Clyde Consultants). Dr. DiToro provided
technical consultation, and David Gaboury (Woodward-Clyde Consultants) assisted in the analysis
of settling velocities. Dr. Philip E. Shelley (EG & G) assisted in the analysis of the basic NURP
program data.
The contribution of the following individuals and agencies, who provided feedback on the
use of these techniques for local planning purposes, is also acknowledged.
EPA Headquarters
WASHCOG, Washington, D.C.
SEWRPC, Milwaukee, WI
State WRA, Annapolis, MD
State DNR, Raleigh, NC
- Carl Meyers, James Meek,
Patricia Bubar, Stuart Tuller,
Norman Whalen
- Thomas Schuler
- David Kendjiorski
- Bruce Harrington
- Robert Holman
iv
-------
FOREWORD
The principal focus of EPA's Nationwide Urban Runoff Program (NURP) was to develop
and transfer information that would be of practical utility to planning agencies in determining the
need for, and approaches to the control of, pollutant discharges from urban stormwater runoff.
One of the specific objectives was to assess the performance characteristics of control techniques,
and for those indicated to be feasible candidates, to provide data and analysis procedures to guide
and support planning decisions.
This report describes an analysis methodology and presents graphs and example
computations to guide planning level evaluations and design decisions on two techniques for
urban runoff quality control. The control techniques addressed, recharge or infiltration devices,
and wet pond detention devices (basins that maintain a permanent pool of water), were shown by
the NURP studies to be the most consistently effective at pollutant reduction of any of the Best
Management Practice (BMP) approaches considered.
The underlying theory and mathematical computations are relatively sophisticated, but the
application procedures have been reduced to a number of simple, easy to use steps that do not
require expertise in mathematics or statistics. The time required to perform an analysis is quite
short, so that the relatively large number of alternatives that should be examined for a planning
level analysis can be readily made with a very nominal investment in time and resources.
A condensed summary of the technical details of the analysis methodology is presented in
Section 2. Those interested in the theoretical development are referred to the sources cited for this
aspect. The fundamental equations have been solved for the range of values the controlling
parameters can assume, and are summarized in a series of easy-to-use graphs. These graphs are
used in the manual computations of performance. Computer programs in BASIC programming
language, which execute efficiently on personal (micro-) computers, have been developed.
Interested parties should contact the EPA Project Officer.
The actual performance data developed by the NURP program have been summarized in
the NURP Program Final Report (December 1984), along with an analysis of cost effectiveness
and an illustration of these procedures for a general planning analysis for a region. Such material
is repeated here only to the extent that it supports the objective of mis report to describe, illustrate,
and validate the analysis procedure.
-------
-------
1.0
INTRODUCTION
1.1 GENERAL
Best Management Practices (BMPs) receive consideration for control of nonpoint source
pollutant discharges (in this case, urban runoff) because of the favorable influence they are
expected to exert on receiving water quality by reducing the mass loading of pollutants that would
otherwise be carried into such waters by storm runoff. Studies conducted under the NURP
program indicated detention and retention basins to be the most effective and reliable of the
techniques examined for control of urban runoff pollutant loads. The principal mechanisms that
influenced pollutant removals were either subsurface infiltration, or sedimentation.
A detention device installed at a specific location is necessarily of a fixed size and capacity.
Storm runoff, on the other hand, is highly variable. Any installation, therefore, will exhibit
variable performance characteristics, depending on the size of the storm being processed, and in
general, will perform more poorly for the larger storms than for the smaller ones. When
performance is influenced significantly by the storage volume available, results obtained will be
modified by residual stormwater from prior events that still occupies the basin when the next event
occurs. Since storm intervals are variable, this factor frequently has a significant influence on
performance. For detention devices such as wet ponds, which maintain a permanent pool of water,
there is a further complication to the ability to describe performance. For many storms in all
basins, and for virtually all storms in large basins, the effluent displaced during a particular event
represents, in fact, a volume contributed to by the runoff of some antecedent event
The performance of any control device that treats urban runoff should therefore be
characterized in such a way that the variability and intermittent nature of storm runoff is recognized
and accounted for. It is also desirable that the analysis procedures used provide a basis for making
reasonable projections of performance under conditions other than those tested An obvious
alternative set of conditions relates to the effect on pollutant removal of basins of different sizes;
however, the important factors include performance over all storms for an area in contrast to those
monitored in a test program, and performance in areas where storm patterns are different.
The methodology presented in this report is based on a probabilistic technique that
accounts for the inherent variability of the situation it addresses. The analysis has a planning
orientation rather than a research one, consistent with the principal focus of the NURP program.
The basic objective of the analysis that has been structured is to provide a basis for establishing
"first order" design specifications (size, detention time), in terms of a long-term average removal of
urban runoff pollutants. A secondary objective for a useful planning tool is that it be sufficiently
simple, fast, and economical to apply, so mat a large number of alternative scenarios are practical to
-------
examine. The methodology presented meets both these requirements, and by comparison with
actual performance data and/or projections from more elaborate simulation models, is indicated to
provide sufficiently accurate performance projections for the intended purposes.
There are other analysis methods available that can accomplish the same objective. EPA's
Storm Water Management Model (SWMM), and the Storage, Treatment, Overflow Runoff Model
(STORM) are both well documented simulation techniques that have seen extensive use. They
have, hi fact, been used in some of the validation tests of the probabilistic method, where adequate
performance data were not available for comparison. Since these simulators can avoid several of
the simplifying assumptions of the probabilistic approach, the estimates they provide are likely to
be somewhat more accurate projections. The only real restriction to their use is a practical one. The
user must have convenient access to a computer on which the program is installed, and preferably
experience in the use of the programs.
Although other approaches are available to a user, the methodology presented in this report
is believed to have several advantages. It permits an analysis to be performed without the need for
access to a computer. Analyses are simple enough to perform that there is no practical constraint to
examining a large number of alternative conditions of interest These factors and the organization
of the computations (input requirements and output format) emphasize the utility for planning
purposes.
1.2 ORGANIZATION OF REPORT
Section 2 describes the probabilistic methodology and discusses the rationale and use of
the performance graphs, and the equations on which they are based.
Section 3 addresses recharge devices and presents a description of the methodology, an
example problem, validation tests, and a discussion of the application of the methodology and some
limitations and practical considerations.
Section 4 addresses wet pond detention basins using the same format
Section 5 presents results of a series of analyses using the methodology, illustrating
differences in size/performance relationships as influenced by regional differences in rainfall
characteristics. These generalized results may be used as an initial screening indication, to be
further refined by use of specific local parameters in the analysis.
An Appendix provides information to assist the user in estimating values for parameters
used in the methodology.
-------
2.0
METHOD OF ANALYSIS
2.1 GENERAL
Performance estimates for the stormwater control devices addressed in this report are
computed using probabilistic analysis procedures conceived and formulated by DiToro, and
developed by DiToro and Small (2,3,4). These procedures provide a direct solution for the long
term average removal of stormwater and pollutants for several different modes of operation of a
control technique. The variable nature of storm runoff is treated by specifying the rainfall and the
runoff it produces in probabilistic terms, established by an appropriate analysis of a long-term
precipitation record for an area.
Long-term average reduction in mass loading is considered an appropriate measure of
performance for several reasons. It recognizes the highly variable nature of storm runoff, which
for a basin of fixed size, will result in higher removal efficiencies during some storm events and
lower efficiencies in others. In addition, characterizing basin performance in this manner provides
a direct tie-in with the methods adopted by NURP for characterizing the intermittent and variable
impacts of storm runoff on water quality and for evaluating significance in terms of protectiveness
or impairment of beneficial uses.
For assessing performance, the specification of the size or design capacity of a control
device is often ambiguous, because the rate and volume of individual storm runoff events vary so
greatly. This is influenced by regional differences in rainfall patterns, by the size of the drainage
area the device serves, and by the land use distribution of this area, which determines the degree of
impervious cover and the amount of runoff that any particular storm generates. For the procedures
used in this report, variable rainfall/runoff rates, volumes, durations, and intensities are specified as
a MEAN and COEFFICIENT of VARIATION (CV = STANDARD DEVIATION / MEAN). A
meaningful measure of device size or capacity is then the ratio of its volume or flow capacity to the
volume or flow rate for the MEAN storm runoff event. This permits a convenient generalization of
the analyses performed and allows results to be readily applied to various combinations of local
conditions.
Analysis procedures for computing size-performance relationships for three operational
modes are presented in this section. A particular stormwater control device may incorporate one or
more of these modes. Estimating performance for specific devices (for which examples are
presented in later sections of the report) requires selecting and combining the procedures for the
modes that are appropriate, or adapting the procedures to the specific circumstances dictated by the
nature of the device.
-------
2.2 RAINFALL
A long-term record of hourly precipitation data, available from the U.S. Weather Service
for many locations, may be separated into a sequence of discrete storm "events" for each of which
volume, duration, average intensity, and interval since the preceding event can be readily
determined. The full set of values for each of these parameters may then be statistically analyzed to
determine the mean and standard deviation, as well as the probability distribution of the set of all
values for a parameter. A NURP publication (1) documents a computer program (SYNOP) that
computes these statistics (and other information) from a USWS hourly precipitation record.
Appendix Section 2 provides a tabulated summary of storm statistics for gages in various
parts of the country, developed from analysis of rain gage data by the SYNOP program. Appendix
Section 3 presents information for estimating runoff coefficient. This information is provided to
assist the user in estimating appropriate values for local analyses.
Analysis of a number of rainfall records indicates that the storm parameters that are used in
the analyses described in this report are well represented by a gamma distribution. This distribution
has accordingly been incorporated in the probabilistic analysis procedures described in this report.
2.3 FLOW - CAPTURE
This procedure addresses the condition where a device captures 100% of all applied flows,
up to its capacity QT, and bypasses all flows in excess of this. No consideration is given to what
happens to the "captured" fraction, other than that it no longer discharges with the uncontrolled
fraction. Some examples include the following: in a Combined Sewer Overflow situation, the
amount of the total wet weather flow that is carried away from the overflow point by an interceptor
sewer and conveyed to a downstream sewage treatment plant can be considered to have been
"captured," or removed from the overflows that would otherwise occur. A recharge device that
diverts a portion of the runoff by causing it to percolate into the ground has captured some fraction
of the surface runoff that would otherwise completely flow into a surface water body.
Whether or not further consideration must be given to the storm runoff so captured is not
addressed here. The technique simply determines the long-term average reduction (or capture) in
stonnwater volumes processed by the device, and the pollutant loads associated with them.
For storm flows that are gamma distributed, and a device that captures all inflows up to a
rate, QT, the long-term fraction not captured is given (3) by:
-------
where:
= fraction not removed by Row-Capture device
= l/CV^ (reciprocal of square of CV of runoff flows)
= Gamma function for T}
E = q/QR - QT/QR
q = runoff flow rate for an event
QR = mean storm runoff flow rate
QT = flow rate capacity of device
Transformed for numerical integration by Laguerre quadrature, this performance equation
becomes:
r/i"2 e-ri< QT/QR) "
fpc G? Z
,-fi (QT/QR)
W. f"Yy 1 /*)\
j I V.A ij 12.)
3=1
\vhcrc*
f(Xj) = Xj (Xj/ri + QT/QR) ri'1
x;, w; = abcissas and weights for Laguerre quadrature
This equation has been solved for a range of values for normalized treatment capacity
(QT/QR), and variability of storm runoff flows (CVq). Results are presented in Figure 1 which
illustrates the effect of the above variables on long-term control efficiency of a device with this
mode of operation.
2.4 FLOW - TREATMENT
This procedure addresses the performance of a device under variable input flows, when the
treatment or removal efficiency for a pollutant varies with the rate of applied flow. It differs from
-------
100
o
z
D
CC
z
<
m
cc
o
u.
O
UJ
CC
I-
z
HI
o
cc
UJ
0_
-Coefficient of Variation
0.5 0.75 1.00 1.25 1.50 1.75 2.00
0.2 0.3 0.4 0.6 0.8 1.0 2.0 3.0 4.0 6.0 8.0 10.0
FLOWRATE CAPACITY
RATIO:
MEAN RUNOFF FLOW
(OT/O-R)
Figure 1. Average long term performance:
flow-capture device
-------
the previous case in that the entire runoff flow is-processed. An example would be a sedimentation
basin which is less efficient at higher flow-through rates than it is at lower ones.
For variable runoff flows entering a treatment device that are gamma distributed and
characterized by a mean flow and coefficient of variation (CVq), the long-term average fraction of
total mass removed is:
r+1
r-
]
(3)
where:
= long term average fraction removed
= fraction removed at mean runoff rate
= 1/CV^ (reciprocal of square of CVq)
= coefficient of variation of runoff flow rates
M
Z = maximum fraction removed at very low rates
A graphic solution to this equation is presented by Figure 2 and illustrates the effect on
long-term performance caused by variability of stormwater flows. The analysis assumes that
removal efficiency of the device is an exponential function of flow, thus:
FRACTION REMOVED = 1 - exp ( Q/k) (4)
While not exact, this relationship appears to approximate many removal relationships
adequately, and is appropriate for a planning level analysis.
2.5 VOLUME - CAPTURE
This procedure addresses devices whose effectiveness is a function of the storage volume
provided. This mode of operation is illustrated by a basin that captures runoff flows until it is filled
and thereafter passes (untreated) all additional stormwater. The captured stormwater runoff is then
removed from the basin in some manner once runoff ceases, in preparation for the next event
The analysis does not consider what happens to the captured volume; it simply assumes it
to be removed from the total discharge processed by the device. Off-line detention basins for
-------
100
Nl
CO
H-
2
LU
LU
LU
CC
LU
O
cc
LU
LL
> c
cc g
i-U i-
J "S
< s
> *
i I
LU
CC
Coeffici«nt of variation
of Runoff Volume CVq /
REMOVAL AT MEAN RUNOFF FLOW
MAXIMUM REMOVAL AT VERY LOW FLOW.
(expressed as percent)
(RM/Z)
Figure 2. Long term performance of a device where removal
mechanism is sensitive to flow rate
-------
CSOs, which pump captured overflows back to the sewer system for processing at the treatment
facility, provide one example of this mode of operation. Another example is a recharge basin,
which (in addition to operating as a Flow-Capture device, Section 2.3) removes captured runoff
volumes through percolation.
is:
For storm volumes that are gamma distributed, the fraction not captured, over all storms,
r
r v
q=o
00
I A [A + ] exp [-r2Aj dA
A =o
dq
(5)
where:
rl
cv
q
A
v
and r2 = l/CVd2
coefficient of variation of runoff flow rates
l/CVq2
coefficient of variation of runoff durations
storm runoff flow rate
average interval between storm midpoints
basin effective volume, divided by mean
storm runoff volume (VE/VR)
fraction of all volumes NOT captured by basin
The double integral cannot be evaluated analytically. A numerical technique using a
Laguerre quadrature to approximate the integral with a weighted polynomial is applied. The basic
equation transformed for solution using quadratures is:
f =
r
r/2
G (rf) G(r,) ^
1 2 k =
(6)
-------
where:
fl = l-]r' t--1-expt-r, r2 V/Xk]
n = number of orders used in integration
X;, Xjp Wj, Wjj. = abcissas and weights for Laguerre Integration
(from any handbook of mathematical functions)
This integral has been solved for a range of values of V (=VE/VR) and values for
coefficient of variation in a range typically observed for rainfall/runoff. Results are plotted in
Figure 3, which may be used instead of the equation.
From this figure, the average long-term performance of a volume device may be estimated
based on the basin volume relative to the mean storm volume and the variability of individual event
volumes being processed. However, the relationship is based on "effective" basin volume (VE)
which may be quite different than the physical storage volume of the basin (VB). In the original.
CSO application, DiToro and Small (3,4) present a procedure for approximating the effective
volume, based on an emptying rate ratio (E):
E = (7)
where:
*
A = average interval between storms (hours)
Q, - rate at which basin empties (cu ft/hour)
AH = volume removed between storms, on average (cuft)
VR = runoff volume from mean storm (cuft)
The effect of the emptying rate ratio on the fraction of physical basin volume which is
effective is described by Figure 4. As indicated, in cares where the volume which can be removed
in the average interval between storms is small relative to the storm volume which enters on
average, much of the available volume may be occupied with carryover from prior storms each time
it rains. In such cases, effective volume may be considerably smaller than the physical storage
volume provided.
10
-------
LL
LL
O
cc
CD
(T
O
LL
O
O
LU
cc
UJ
O
cc
UJ
Q.
TOO
90
80
70
60
50
40
30
20
10
0.1
Coefficient of Variation
of Runoff Volume CVVR
1.25
2.00
J I I I I i l l
0.2 0.3 0.4 0.5 0.6 0.8 1.0
2.0 3.0 4.0 5.0 6.0 8.0 10.0
RATIO- EFFECTIVE BASIN VOLUME
RAT'°- MEAN RUNOFF VOLUME (VE/VR>
Figure 3. Average long term performance:
volume device
11
-------
0.1
1.0
VR
2.0 3.0
I" STORAGE VOLUME (EMPTY)"]
L MEAN RUNOFF VOLUME J
4.0
5.0
Figure 4. Effect of Previous Storms on Long-Term Effective
Storage Capacity
12
-------
The expression A& may be thought of as the volume emptied from the basin during the
average interval between storm events. The smaller this quantity is relative to VR, the average
volume entering the basin during storms, the more likely it is that the basin will still contain leftover
runoff when a storm begins, and the smaller will be the effective volume. When this ratio, E, is
less than about 2, the effective volume becomes quite small compared with the physical volume
provided, especially for the larger basins.
13
-------
3.0
RECHARGE DEVICES
3.1 GENERAL
Recharge devices may take a variety of forms, including porous pavement, infiltration
trenches, percolating catch basins, or larger basins which occupy land set aside for the purpose.
There are no fundamental differences in the devices, either in the way they control storm runoff, or
in the procedure for analyzing performance. The differences are in details such as the size of the
basin, the configuration, and the size of the catchment area routed through a particular unit.
Given a specific surface area provided for percolation, and a unit infiltration rate defined by
soil characteristics, an overall "treatment rate" can be defined for a specific device. When storm
runoff is applied to the device at rates equal to or less than this rate, 100% is intercepted. At higher
applied rates, the fraction of the runoff flow in excess of the treatment rate overflows to a surface
water.
If the device also provides storage volume, the volume stored can be retained for
subsequent percolation. Overflow to surface waters (runoff that "escapes" the device) occurs only
when the available storage is exceeded. Long-term average removal is the net reduction in
overflows over the long-term sequence of storms of different size, with different intervals between
successive storms.
Performance will obviously vary with the basin size in relation to the area served, with the
soil percolation rate, and with tne characteristics of local storm patterns.
The analysis procedure described in this section permits one to either (a) evaluate the
potential for a specific recharge installation to reduce pollutant loads from a particular drainage
area, or (b) develop a general relationship on size or areal density for different levels of pollutant
control. Examples of a site-specific approach are presented below; generalized analysis results are
presented and discussed later in Section 5.
Level of control is expressed as a long-term average removal of storm runoff flows. The
tacit assumption is that the urban runoff which is caused to percolate into the ground is "removed"
as a discharge to surface water bodies, as are the pollutants which are present in the runoff. Any
percolated waters which eventually reach surface waters through groundwater flow are assumed to
14
-------
percolated waters which eventually reach surface waters through groundwater flow are assumed to
have had pollutants of interest removed by relevant soil processes (filtration, biological action), and
hence are ignored by the analysis. The validity of this assumption will be influenced by the type of
pollutant of interest and local conditions.
As with any model or computation, judgment is required in interpreting the results of this
analysis, and in evaluating the overall suitability of recharge devices in a local area. Apart from the
factors used in the analysis, considerations such as soil type, slope and stability, depth to water
table, etc., will be important determinants of suitability at any site.
It should be noted that the analysis does not address eventual blockage of the soil. The
rates assigned should be typical values which can be maintained naturally or by maintenance
programs. Neither does the analysis speak to the issue of contamination of the ground water
aquifer. Such considerations must be addressed in any actions or decisions related to
implementation of this control approach.
The input data requirements for use of the analysis procedure consist of the following:
Rainfall - mean and coefficient of variation of rainfall intensity. These statistics are
developed by the SYNOP program. (See the Appendix for further discussion on
this procedure and for a summary of data for a number of cities in different
regions of the country.)
Urban Catchment - area and runoff coefficient (ratio of runoff to rainfall).
Device Size - surface area provided for percolation, and storage volume.
Percolation Rate - rate of infiltration provided by local soil - usually reported in
inches per hour or gallons per day per square foot. A "Treatment Rate" is
defined as the product of the unit percolation rate and the surface area over which
percolation occurs.
3.2 ANALYSIS METHOD
Figure 5 illustrates the operating principles involved and siimmarizes the terminplogy. The
illustation is for the general case; for specific recharge device designs, only the configuration is
different. For example, porous pavement would be represented as having a negligible storage
volume; an infiltration trench would have the storage area filled with coarse aggregate, and available
storage volume reduced to the void volume contained within the gravel or crushed stone.
It is assumed that the device is at the "downstream" end of the urban drainage area it
serves, i.e., all runoff from the defined catchment area is routed through the basin.
15
-------
RAINFALL |
Mean and Coefficient of Variation
for Intensity and Volume
RUNOFF
Rate = OR
QR = RVIA
DRAINAGE AREA
Area = A
Runoff Coefficient = R
Percolation Area r?:V-
Ap
Percolation Rate
of Soil
P -
Runoff "Captured"
by Percolation
OVERFLOW
(not captured)
Treatment Rate
QT
Percolation Area (Ap)
QT= x
Percolation Rate (P)
Figure 5. Schematic illustration of recharge device
16
-------
Long-term performance characteristics are defined as a function of the ratio between the
"treatment capacity" (QT) of the device and the runoff rate (QR) from the average storm. It is
strongly influenced by the inherent-variability in the rate of runoff for different storms ~ which is
characterized by the coefficient of variation of runoff flow rate (CVq).
If there were no variability, i.e., if all runoff entered the device at the mean runoff rate,
then performance during any event and long term average performance would be the same and
would be equal to the treatment capacity provided relative to the applied rate. If treatment capacity
were made equal to runoff rate (QT/QR = 1), 100% removal would be achieved. However, where
treatment rate is fixed by design and runoff rate is variable, performance is reduced. The greater
the variability, the poorer the performance, on average, because of the increasing number and
magnitude of events which produce rates greater than the mean runoff rate.
3.3 EXAMPLE COMPUTATIONS
The performance of recharge devices can be projected using the performance curves
presented in Section 2. The examples presented in this section illustrate the use of these curves.
3.3.1 Porous Pavement
A. Given
A shopping center has an area of 1 acre. It is all paved surface and runoff coefficient is
estimated to be 0,9. Configuration and slopes are such that porous pavement can be
installed as part of the catchment paved area and intercept all runoff produced.
The controlling rate of percolation (either porous pavement or the soil below it) is 1
inch/hour.
Storage volume in pores of pavement is assumed negligible.
The site is near Baltimore, Maryland, and rainfall statistics for the area are estimated (from
tables in the Appendix) to be:
Mean Coef. of'Variation
Volume (V) inch , 0.40 1.48
Intensity (I) in./hr . 0.069 1.21
Duration (D) hour 6.0 1.01
Interval (A) hour 82.0 1.03
17
-------
B. Required
Estimate the long-term average percentage of storm runoff that would be captured if porous
.pavement, equal to 10% of the total area of the catchment, were installed.
C. Procedure
Step 1 - Select appropriate performance curve to use for estimate.
Porous Pavement provides no significant amount of storage volume.
Therefore, the device does not capture any volume, and Figures 3 and 4 do
not apply.
Percolation rate, and hence treatment rate (QT) is independent of applied
flow rate. Thus, the treatment rate does not depend on flow and Figure 2
does not apply.
Mode of operation corresponds to that described for FLOW - CAPTURE
devices described in Section 2.3. Therefore Figure 1 describes
performance.
Performance estimates are based on QR, QT and CVq.
Step 2 - Compute mean runoff rate (QR) in cubic feet per hour.
QR = a) * (Ry) * (AREA) * (DIMENSION CONVERSION)
= 0.069 * 0.9 * 1 * 43560/12
= 225 CFH
Step 3 - Compute treatment rate (QT) in cubic feet per hour.
Percolation rate (P) is 1 in./hr = 0.083 ft/hr
Treatment rate QT = Rate (P) * Area (Ap)
If 10% of the 1-acre catchment area is installed as porous pavement:
Ap = 43,560 * 0.10 = 4,356 sq ft
QT = P * Ap = 0.083 * 4,356 = 362 CFH
18
-------
Step 4 - Compute Design Ratio (QT/QR).
QT (from step 3) = 362 CFH
QR (from step 2) = 225 CFH
QT/QR = 362/225 = 1.6
Step 5 - Estimate Long-term Removal.
In Figure 1, enter horizontal axis at QT/QR =1.6
Extend a line vertically until it intersects the curve for the coefficient of
variation (from rainfall statistics for intensity, CVg = 1.25 approximately)
Extend a line horizontally from this point, and read removal efficiency as
approximately 72%
3.3.2 Recharge Basin
A. Given
For a 10-acre residential development, the runoff coefficient is estimated at 0.25. All
stomrwater runoff from the area is to be routed to a recharge basin.
Minimum basin depth must be at least 2 ft to penetrate a relatively impervious surface, soil
and reach a layer with good drainage properties. The subsoil has a percolation rate of 2.5
in./hr.
Rainfall statistics for the area are :
Mean Coef. of Variation
Volume (V) inch 0.53 1.44
Intensity (I) inVhr 0.086 1.31
Duration (D) hour 7.2 1.09
Interval (A) hour 85.0 1.00
Space constraints limit the basin to a bottom dimension of 25 by 50 ft, or a maximum
percolation area of 1250 sq ft.
B. Required
Estimate the long-term average reduction in storm runoff that can be obtained from a
recharge basin with the minimum (2 ft) depth.
19
-------
C. Procedure
Step 1 - Select appropriate performance curve(s).
Figure 1 applies in this case because treatment rate is based on percolation
rate, and is independent of applied flow
Figure 2 does not apply for the above reason
Figures 3 and 4 also apply in this case because storage capacity is provided
by the device
Step! - Compute runoff parameters for mean storm flow rate (QR) and volume (VR).
QR = (I) * (Rv) * (Area) * (43,560/12)
= 0.086 * 0.25 * 10 * 3630 = 780 CFH
VR = (V) * (Ry) * (Area) * (43,560/12)
= 0.53 * 0.25 * 10 * 3630 = 4807 CF
CVq=l,31 and CVV =1,44
Step 3 - Compute treatment rate (QT) and the design ratio for treatment (QT/QR).
Percolation rate (P) =2.5in./hr = 0.208 tVhr
Percolation area (Ap) = 1,250 sq ft
QT = P*Ap = 0.208 * 1,250 = 260 CFH
QT/QR = 260/780 = 0.33
Step4 - Compute basin effective volume and the design ratio for storage (VE/VR).
For the minimum (2 ft depth) basin, physical basin volume (VB) is:
VB = 1,250ft2 * 2ft = 2,500cuft
VB/VR = 2,500/4,807 = 0.52
Emptying Rate ratio (E)
E=A*Q/VR
20
-------
A is the average interval between storms = 85 hr
Q is the emptying rate of flow = QT = 260 CFH
E = 85 * 260/4,807 =4.6
From Figure 4, enter horizontal axis at VB/VR = 0.52; extend a line vertically to intersect
curve for E = 4.6; then horizontally to read VE/VR on vertical axis. Estimate that effective
volume VE is essentially the same as physical volume for this case.
VE/VR = VB/VR = 0.52
Step 5 - Estimate performance of recharge basin.
Removal accomplished by infiltration is estimated from Figure 1 for the
conditions
QT/QR = 0.33 and CVq =1.31
% Removed (FLOW) = 24%
Removal accomplished by storage is estimated from. Figure 3 for the
conditions
VE/VR = 0.52 and CVV=1.44
% Removed (VOLUME) = 35%
(This efficiency applies not to the overall runoff from the drainage area, but
to the fraction that escapes the percolation process.)
Overall removal accomplished by the combined infiltration/storage process
may be computed directly from the fractions NOT removed by each
process.
Fraction not removed by infiltration
fq = !-(% Removed AGO) = 0.76
Fraction not removed by storage
fv = 1 - (% Removed 7100) = 0.65
% Removed (overall) = (1 - [ fq * fv ]) * 100%
= (1-[0.76* 0.65]) * 100%
= 51%
21
-------
3.4 VALIDATION
Although several of the NURP sites included recharge devices, the data obtained were not
sufficient in either scope or extent to provide a suitable basis for use as a validation test for the
probabilistic procedure described above.
An examination of the reliability of the performance estimates provided by the procedures
presented in this report was conducted by comparing projections for a range of conditions with
those produced by an established deterministic simulation model. The model "STORM" was used
to generate runoff for a hypothetical urban drainage area, using a long-term (approx. 20 years)
hourly rainfall record. This runoff record was then processed by the Storage-Treatment block of
the SWMM model, and from the long-term output produced by the simulation, the average percent
reduction was computed.
This computation was performed for a variety of basin sizes and soil percolation rates.
Figure 6 compares these results with those produced by the probabilistic analysis
procedures.
3.5 DISCUSSION
The procedures described for estimating performance of recharge devices on the basis of
size, local soil conditions, and rainfall patterns provide estimates that compare quite favorably with
those produced by accepted simulation techniques. They are simple to use and permit examination
of the wide variety of alternatives usually desirable in planning activities.
The procedures described provide a basis for quantifying the performance capabilities of a
variety of recharge devices, using information that will normally be readily available. However,
the suitability of recharge/infiltration systems will vary with location and must be determined on the
basis of local conditions.
The possibility of contributing to undesirable impacts on ground water aquifers by
enhanced recharge to protect surface waters must be considered on a local basis. Situations have
been identified where it has been concluded that the contaminants (and their concentrations)
normally present in urban runoff, and which reach the aquifer following percolation, do not
constitute a problem or a significant cause for concern. In these situations the practice is
encouraged. There are, however, othtr situations where there are legitimate concerns with the
appropriateness of this approach.
22
-------
100
NJ
CO
Detention Basin
Long Term Average Removal
By Percolation
Birmingham, Ala. - Rainfall
Rv = 0.25
Basin Volume = 10 x Percolation Area
Statistical Method
Storm" Model
Simulation
0.1 inch/hr (Soil Percolation Rate)
I » I I I
0.05 0.10 0.5 1.0
PERCOLATING AREA AS % OF CONTRIBUTING CATCHMENT AREA
Figure 6. Detention basin performance - long term average removals by
percolation - comparison of statistical and simulation methods
-------
The approach may be unsuitable for areas with steep slopes and unstable soils, or areas
with water supply wells in sufficiently close proximity to recharge areas.
A tacit assumption in the analysis is that the water table is far enough below the percolation
surface that a significant interaction with the temporary mound of ground water, which may form
during an event, does not take place.
A further consideration is that percolation rates assigned in the analysis are represenative of
long-term conditions, and that significant soil blockage with use either does not occur or is
accounted for. Historical experience with recharge basins and with land application of waste
waters indicates that progressive blockage is not generally a problem when the soil can be "rested"
between applications. The intermittent nature of storms, and the fact that in most areas of the
country storm periods occur less than 10% of the time automatically provides such rest periods that
help maintain soil permeability.
24
-------
4.0
SEDIMENTATION DEVICES
4.1 GENERAL
Detention basins that receive storm runoff, but that have negligible losses through
infiltration, must rely principally on sedimentation processes for pollutant removal. Under some
conditions, and to some extent, reductions attributable to other processes may influence removal of
specific pollutants (e.g., natural die-off of coliform bacteria, and algal uptake of soluble nitrogen
and phosphorus).
Of the variety of configurations and operational modes that have been used, stormwater
detention basins that maintain a permanent pool of water, often referred to as "wet ponds," are
generally considered to be the most effective for pollutant reduction.
Nine such devices in various parts of the county were actively monitored during the NURP
program, as the local agencies' choice of a preferred control approach.
This section presents a procedure for projecting performance of such devices, and a
comparison of results with observed performance of the NURP detention basins. A wide variety
of concepts and configurations is represented by the wet ponds that were studied, ranging from
oversized storm drains to natural ponds and small lakes. The size of the devices relative to the
contributing drainage area varied over a wide range; the common elements for all were, the
maintenance of a permanent pool of water and sedimentation as the principal pollutant-removal
mechanism.
The input data requirements for analysis of sedimentation devices are essentially the same
as for recharge devices described in the previous section, but with the following exception. In this
case the "treatment rate" is determined not by soil percolation rates, but by the settling velocity of
the particulates present in the urban runoff. Represenative values for settling velocity can be
assigned to urban runoff on the basis of a significant number of settling column tests conducted
during the NURP program.
25
-------
4.2 ANALYSIS METHOD
The probabilistic computations and performance curves presented in Section 2 can be
applied to \yet ponds (with appropriate adaptation and interpretation) to reflect the nature of the
treatment process that occurs in detention basins of this type.
A basic aspect of such a system is that part of the time (while runoff inflows occur),
stormwater is moving through the basin, and sedimentation takes place under dynamic conditions.
During the considerably longer dry periods between storm events, sedimentation takes place under
quiescent conditions.
4.2.1 Removal Under Dynamic Conditions
Characterization of the performance of sedimentation devices has been extensively
analyzed over the years because of the important role such devices play in both water treatment and
wastewater treatment systems. A method of analysis which is particularly suitable is presented by
Fair and Geyer (5). Removal due to sedimentation in a dynamic (flow through) system is
expressed by the following equation:
where:
R = fraction of initial solids removed (R * 100 = % Removal)
vs = settling velocity of particles
Q/A = rate of applied flow divided by surface area of basin (an "overflow
velocity," often designated the overflow rate)
n = a parameter which provides a measure of the degree of turbulence or
short-circuiting, which tends to reduce removal efficiency
One value of this model is that it provides a quantitative means of factoring into the
analysis an expression for impaired performance due to short-circuiting (since many stormwater
retention basins will not have ideal geometry for sedimentation). Fair and Geyer suggest an
empirical relationship between performance and the value of "n," which is: n = 1 (very poor); n =
3 (good); n > 5 (very good). In addition, when a value of n = °° is assigned (ideal performance),
the equation reduces to the familiar form wherein removal efficiency is keyed to detention time.
26
-------
R = * - exP - TTT or ( 9 )
R = 1 - expf- ktl (10)
where:
k = vs/h (sedimentation rate coefficient)
h = average depth of basin
t = V/Q residence time
V = volume of basin
The two expressions are equivalent. To use them, one must be able to identify an
appropriate value for either settling velocity, or for the rate coefficient (k), which will ultimately
depend on the settling velocity of the particulates present.
Solving equation (8) for a range of overflow rates and particle settling velocities and
plotting the results as shown by Figure 7, indicates the wide range in removal that can be expected
either (a) at a constant overflow rate for particles of different size, or (b) at different rates of flow
for a specific size fraction. Both of these variable factors are present in urban runoff applications.
The effect of a range of particle settling velocities is addressed by performing separate computations
for a number of settling velocities and then using weighted mass fraction to compute net removal.
Storm sequences result in variable overflow rates, each event producing a different average
rate, and hence, removal efficiency. The probabilistic analysis procedure described in Section 2.4
(Flow-Treatment), and summarized by the design performance curves in Figure 2, is the relevant
analysis to apply. This analysis makes the following assumptions:
The short-term variability of flows (within storm events) is small compared with
the variability of average flows between storms. To the extent that this is not the
case, Figure 2 will overestimate long-term performance.
Storm flows and pollutant concentrations are independent. If flow rate and
concentration are negatively correlated (high flows produce lower
concentrations), performance will be better than indicated. For positive
correlations, performance will be poorer than indicated.
27
-------
100
1.0 10
OVERFLOW RATE Q/A (ft/hr)
100
Figure 7. Effect of settling velocity and overflow rate on removal efficiency
100
EXPLANATION
Exact solution
eqn. (8) with
n = 3
Approximate
solution
eqn. (9)
0.5 1.0 1.5 2.0 2.5 3.0
OVERFLOW RATE Q/A (ft/hr)
Figure 8. Flow-removal relationships for exponential approximation
28
-------
« Removal efficiency is an exponential function of flow.
Available data on stormwater retention basins" are not suitable to provide empirical
estimates of flow rate/removal relationships. The relationship represented by equation (8) has been
used instead. Removal fractions for a range of settling velocities representative of urban runoff, as
computed by equation (8), are presented in Figure 8 as a semi-log plot on which the exponential
approximation, equation (9), would plot as a straight line. For a site-specific analysis (for each
settling velocity separately), the straight line approximation would match the exact solution at the
point corresponding to the mean overflow rate (QR/A), and the slope would be adjusted to give the
best match over the range of rates expected to span the bulk of the important storms. The intercept
of this fitted line (Q/A = 0) provides the estimate for the factor Z in equation (3). For example, in
the sample illustration shown in Figure 9, the overflow rate for the mean storm is 1.5 ft/hr. For the
size fraction represented by a settling velocity of 0.3 ft/hr, removal at the mean flow rate (RM) is
0.18 and Z is estimated to be 0.8. Over the range of overflow rates of interest, the exponential
approximation is within about 10%.
Long-term average removal of a pollutant under dynamic conditions can, therefore, be
estimated from the statistics (mean and coefficient of variation) of runoff flows, basin surface area,
and representative particle settling velocities for urban runoff.
4.2.2 Removal Under Quiescent Conditions
For much of the country, the average storm duration is about 6 hours, and the average
interval between storms is on the order of 3 to 4 days. Thus, significant portions of storm runoff
volumes may be detained for extended periods under quiescent conditions, until displaced by
subsequent storm events. The volume of a basin relative to the volumes of runoff events routed
through it is the principal factor influencing removal effectiveness under quiescent conditions.
The probabilistic computation described previously in Section 2,5 (Volume-Capture), and
summarized by design performance curves in Figures 3 and 4, is used to estimate removals under
quiescent conditions. This analysis assumes that physical volumes are removed from the basin
during the dry periods between storms, as in the recharge basin analysis presented in the preceding
section, where captured volume percolates. However, for sedimentation devices that maintain a
permanent pool of water, some modification is required because there is no loss of stored volume
between runoff events. Instead, it is the particulates in the detained volume that settle out under
quiescent conditions. The modification required is to express this condition in terms of the
parameters of the design performance curves.
The term Q. may be thought of as a "processing rate." For a recharge device, it is the rate
at which volume is removed from the basin by percolation through the bottom and sides. For a
sedimentation device, it may be thought of as a particle removal rate. Using this interpretation, the
term Q. A in equation (7) can be considered to represent that portion of the basin volume from
which solids with a selected settling velocity have been completely removed. Instead of the TSS
concentration of the entire volume diminishing with time under quiescent settling, the concentration
is assumed to remain constant, while the remaining volume with which this concentration is
associated diminishes with time. The solids removal rate is then:
29
-------
a = vs* A (ii)
where:
vs = particle settling velocity (ft/hr)
A = basin surface area (square feet)
4.2.3 Combining Dynamic and Quiescent Effects
The procedures described above can be used to compute separate long-term removal
efficiencies under dynamic and quiescent conditions. Since each type of condition prevails in a
detention basin at different times, the overall efficiency of a basin is the result of the combined
effect of the two processes at work. The simple model used to integrate these effects is illustrated
by Figure 9.
. Five identical storms with an interval between event midpoints (A) of 3.5 days are routed
through a basin, assuming plug flow. Each storm has a duration of 12 hours (0.5 day), and a
volume which is 25% of the basin volume (VB/VR = 4). The plotted lines track the residence/
displacement pattern in the basin for the leading edge, midpoint, and trailing edge of Storm #1. The
shading highlights the fraction of the total residence time when dynamic conditions prevail. For
this simplified case, and for actual conditions where both storm volumes (VR) and intervals (A)
fluctuate, the fraction of time under dynamic conditions is estimated by:
Fraction of residence time
under dynamic conditions =D/A ' (12a)
Fraction under quiescent conditions = 1 - (D / A) (12b)
where:
D = mean storm duration
A = mean interval between storm midpoints
\
This simple schematic illustrates several relevant features of the operation of this type of
device. When the basin is as large as that indicated (which is not uncommon for current practice),
the outflow volume during an event represents a different parcel'of water than that for the storm that
causes it to be displaced. Assessing performance by comparing paired influent and effluent loads
for individual storms is less appropriate than the comparison of overall influent and effluent loads
for a long-term sequence of storm events.
All runoff volumes which enter the basin undergo the dynamic removal process one or
more times before discharge. For the large basin illustrated, this is broken up into four different
periods of displacement. For a basin with a volume small enough that the runoff passes all the way
30
-------
00
<
CO
O
D
O
QC
I
UJ
CJ
o
0.25
100
3.75
5 @ A= 3.5 Days-
7.25
10.75
-*"|o.FJ-<- *Jo.5H*- -»M0.5|-*- *-p.5J-*-
14.25
TIME AFTER START OF STORM NO
For Storm Midpoint Volume
Total Residence Time = 14.0 Days
Dynamic Time: (0.25) + (3 x 0.5) + 0.25 = 2.0 Days 2/14 = 0.14
Quiescent Time: 14.0-2.0 = 12 Days 12/14 = 0.86
D/A- 0.5/3.5 = 0.14
0.86
Figure 9. Illustration of quiescent vs. dynamic residence time
in a storm detention basin
31
-------
through, there would be only one such period of dynamic removal. Performance efficiency is
affected simply on the basis of the "overflow rate" that the basin size provides.
The quiescent removal process then operates on (a) those portions of the total runoff
volume that remain in the basin during the dry interval that follows an event, and (b) on that
fraction of the influent pollutants that remain in the water column after operation of the dynamic
process. In the situation illustrated, the average runoff volume is exposed to four different periods
of quiescent settling, amounting to an extended period under this condition. In a very small basin,
the relative effect of the quiescent removal process may be insignificant, simply because such a
small fraction of the total runoff remains in the basin at the end of each storm.
The removal efficiency for the basin under the combined effect of both dynamic and
quiescent processes can be computed by applying the removal efficiency of either the dynamic or
quiescent process to the pollutant fraction remaining after the operation of the other. If the
fractions not removed by the dynamic and quiescent processes operating independently are frj> and
fq, respectively:
COMBINED % REMOVAL = 100 [ 1 - (fD * fq) ] (13 )
It should be noted that in the larger basins, either process operating alone will be capable of
high degrees of removal. One might consider the quiescent process to be the dominant one in large
basins, because high paniculate reductions can be produced even if there were no removal during
dynamic periods, and because the quiescent periods provide the conditions in which the removal
processes other than sedimentatipn can come into play. In small basins, the dynamic process will
be the dominant one because only small fractions of the runoff will remain in the basin subject to
the quiescent process.
4.3 VALIDATION
Performance data from nine wet pond detention basins monitored during the NURP
program have been analyzed and used to test the reliability of the probabilistic methodology. These
devices coyer a wide range of physical types, and also provide a wide range of basin sizes relative
to the contributing urban drainage area.
For the calibration effort, monitored data on storm runoff rates and volumes entering a
detention basin are analyzed to define their statistical characteristics. For long-term performance
projections,, long-term rainfall records for the area in question are used, and the statistical properties
of runoff are estimated from the rainfall record. The settling velocity of particulates in urban runoff
is estimated from data obtained from settling column tests performed by a number of the NURP
projects.
In addition to producing a fairly extensive data base on pollutan:s entering and leaving
detention devices, another critically important contribution of the NURP effort was data to support
32
-------
estimates of the settling velocity of particles in urban runoff. Any analysis methodology for
sedimentation, including that adopted for this analysis, requires information of this nature for use
either directly (equation 8) or in surrogate form, as with a reaction rate (equation 10).
4.3.1 Settling Velocity of Particles in Urban Runoff
Settling tests were conducted by a number of NURP projects on samples of urban runoff.
Results from these tests, and from a similar set of tests reported by Whipple and Hunter (7), have
been analyzed to derive information on particle settling velocities in urban stormwater runoff. The
analysis procedure used for reducing settling test data and a detailed discussion of the overall
analysis results, which are summarized briefly below, are presented in the Appendix.
The analysis of 46 separate settling column tests indicates the following:
There is a wide range of particle sizes, and hence settling velocities in any
individual urban runoff sample.
The distribution of settling velocities can be adequately characterized by a
log-normal distribution.
There is substantial storm-to-storm variability hi median (or other percentiles of)
settling velocity at a specific site. The range indicated is about one order of
magnitude in observed values for any percentile of the distribution in a specific
storm. Uncertainty in the coefficient of variation of the site-averaged settling
velocity distribution (95% confidence interval) is smaller, but still appreciable
(about a factor of 5).
No significant differences between site-to-site mean distributions have been
identified The within-site variability is on the same order as potential site-to-site
differences.
Assuming the data available for analysis are representative, the foregoing
indications, with regard to storm-to-storm and site-to-site differences, support the
pooling of all available data to define "typical" characteristics of particle settling
velocity distributions in urban runoff, and the assumption that such results are
generally transferrable to other urban runoff sites. Appendix Figure A-5 illustrates
best estimates (at present) for the distribution of particle settling velocities in urban
runoff from any site. For the calibration tests and subsequent projections,
computations are performed for five size fractions haying the following average
settling velocities (based on the distribution shown by Figure A-5):
Size % of Particle Mass Average Settling
Fraction in Urban Runoff Velocity (ft/hr)
1 0-20% 0.03
2 20-40% 0.3
3 40-60% 1.5
4 60-80% 7.
5 , 80-100% 65.
33
-------
4.3.2 NURP Performance Results
A total of thirteen detention basins were monitored by various NURP projects. Of these,
nine may be classified as "wet basins," which maintain-a permanent pool of water. Performance
characteristics of these basins have been analyzed and used to compare observed removals to those
predicted using the methodology described earlier.
The detention basins studied under the NURP program encompass a wide variety- of
physical types. They include oversized sections of a storm drain installed below street level (Grace
Street sites), ponds or small lakes on streams which drain urbanized areas (Unqua Pond, Lake
Ellyn), flood control basins (Traver), a converted farm pond (Westleigh), and a golf course pond
through which storm drains from an adjacent urban area were routed (Waverly Hills site). In spite
of this diversity, these different detention devices may be compared by the ratio of the size of the
device relative to the connected urban drainage area, and the magnitude of the storms which are
treated.
Table 1 summarizes such size relationships for the NURP basins, which are arranged in
order of increasing performance expectations. Based on the analysis presented in the previous
section, one should expect that lower overflow rates (QR/A) and higher volume ratios (VB/VR)
would tend to produce better removal efficiencies by sedimentation. Therefore, these ratios are
used in Table 1 as qualitative indicators of performance. The wide range provided by the NURP
data set is apparent Basin #1 has an average overflow rate during the mean storm of about six
times the median .settling velocity (1.5 ft/hr) of particles in urban runoff. Further, less than 5% of
the mean storm volume remains in the basin after the event, to be susceptible to additional removal
by quiescent settling. At the other end of the scale, the mean storm displaces only about 10% of the
volume of Basin #9, and the average overflow rate is a small fraction of the median particle settling
velocity.
Table 2 summarizes the observed overall average performance of the NURP detention
basins over all monitored storms. Removal efficiency is determined from the sum of pollutant
masses entering and leaving the device for all storms. At some sites, there were an appreciable
number of events for which monitoring data were only available for either inflows or outflows. In
such cases, a reduced data set (consisting of only those events for which both inlet and outlet data
were available) was used in the computation. The qualitative indications of relative performance
suggested by the ranking (based on size) are supported by the tabulated results. However, the
variability in actual performance results tends to confuse the picture somewhat, such that the
performance relationships may be better seen in the illustrations presented in the following section.
4.3.3 Calibration Results
The probabilistic methodology was used to compute the expected removal by
sedimentation of a number of pollutants. The surface area and volume of each of the nine detention
devices was determined from the project reports. The statistics (mean and coefficient of variation)
of runoff flow rate and volume were computed from monitoring data for storms entering the basin.
A value of n = 3 was arbitrarily assigned for the shortcircuiting factor for all of the analyses which
follow.
34
-------
90352T1 CON-1
Table 1. SIZE RELATIONSHIPS FOR NURP DETENTION BASINS (BASED ON
MONITORED STORMS)
Detention Basin Size
Code
No.
1
2
3
4
5
6
7
8
9
Project and Site
Lansing
Grace Street N.
Lansing
Grace Street S.
Ann Arbor
Pitt-AA
Ann Arbor
Traver
Ann Arbor
Swift Run
Long Island
Unqua
Washington, D.C.
Westleigh
Lansing
Waverly Hills
Northern Illinois
Lake Ellyn
Approx.
Average
Average
Basin
Depth
(Ft)
2.6
2.6
5.0
4.1
1.5
3.3
2.0
4.6
5.2
Relative to Mean
Monitored
Overflow
Rate -
QR/A
(ft/hr)
8.75
2.37
1.86
0.30
0.20
0.08
0.05
0.09
0.10
Storm
Volume
Ratio
VB/VR
0.045
0.17
0.52
1.16
1.02
3.07
5.31
7.57
10.70
Relative to
Size of Urban
Catchment (Surf
Area/Drain Area
X 100%)
0.0095%
0.035%
0.09%
0.31%
1.15%
1.84%
2.85%
1.71%
1.76%
35
-------
TABLE 2. OBSERVED PERFORMANCE OF WET DETENTION BASINS
REDUCTION IN PERCENT OVERALL MASS LOAD
Site
No.
1
2
3
U
5
6
1
8
9
Project
and
Site
Lausir.g
Grace St. N.
Lansing
Grace St. S.
Ann Arbor
Pitt-AA
Ann Arbor
Traver
Ann Arbor
Swift Run
Long Island
Unqua
Washington, D.C.
Westleigh
Lansing
Waverly Hills
NIPC
Lake Ellyn
No.
of
Storms
18
18
6
5
5
8
32
29
23
Size Ratios
QR/A
8.75
2.37
1.86
0.30
0.20
0.08
0.05
0.04
0.10
VB/VR
0.05
0.17
0.52
1.16
1.02
3.07
5.31
7.57
10.70
Average Mass Removals - All Monitored Storms (Percent)
TSS
(-)
32
32
5
85
60
81
91
84
BOD
14
3
21
(-)
4
(TO
69
COD
(-)
(-)
23
15
2
C=7)
35
69
0
TP
(-)
12
18
34
3
45
54
79
34
Sol.P
(-)
23
(-)
56
29
71
70
e
TKN
(-)
7
14
20
19
(-)
27
60
N02+3
(-)
1
7
27
80
(-)
66
T.CU
(-)
(-)
57
71
T.Pb
9
26
62
82
80
95
78
T.zn
(-)
(-)
13
5
(-)
26
71
71
u>
ON
Notes: (-) Indicates apparent negative removals.
Indicates pollutant: was not monitored.
-------
Because of the wide variability in particle settling velocities, and their important effect on
removal by sedimentation, independent removal efficiency computations were performed for
separate size fractions and results combined for the overall removals indicated. All five size
fractions (Section 4.3.1) were assigned for TSS, total lead, and total P computations. For the other
heavy metals (Cu, Zn), for TKN, and for BOD and COD, it was assumed that there would be no
significant association with the largest size fraction, and computations were performed using four
size fractions.
Most analyses of pollutant concentrations measured the total quantity, and did not
distinguish between soluble and paniculate fractions. Sedimentation computations are based on the
paniculate or settleable fraction. However, overall removal is expressed in terms of total quantities
of pollutant, which is both the most relevant way to express results for control decisions as well as
the basis for reporting observed results to be used for comparison with computations. For the
analysis, therefore, it is necessary to assign the fraction of the total concentration or load which is
settleable. For TSS, total P, and total lead, there is a reliable basis for doing so. Suspended solids
are particulates by definition. Data developed through the NURP program indicate that lead
consistently exhibits very high paniculate fractions. Thus, although no specific measurements of
soluble and paniculate forms were made at detention basin sites, a paniculate fraction of 0.9 can be
assigned to lead with confidence. All but one of the sites (Basin #6) monitored both total and
soluble phosphorus, and the actual paniculate fraction for the site was used in the computation. A
settleable fraction of 0.6 was assigned for Basin #6, guided by results from the entire NURP data
base.
f
For these three pollutants, for which reliable estimates of paniculate fractions are available
and for which a significant fraction of the total is settleable, the comparison between observed
removal efficiency and removals computed by the methodology described earlier is presented in
Figure 10. There are a few obvious outliers; however, in general, predictions are within 10% to
15% of observed performance results. Additional confidence is derived from the fact that both
observed and computed results span the entire range of performance possibilities, from less than
5% to 10%, to 90% or better.
Four significant outliers were identified and investigated. In all cases, actual monitored
percent removal was much less than that projected. ~
Site #4 (see Table 2) shows almost no TSS removal, although a substantial
(~60%) removal is projected. At this newly installed basin, the project report
indicates that significant bank erosion at the outlet structure occurred during the
test program. Lead was not monitored, but observed/predicted Total P removals
compare quite favorably at this site.
Site #5 data show almost no Total P removal, although about 50% reduction is
projected. On the other hand, both TSS and lead projections compare favorably
with observed data. The basin is a shallow, vegetated area, characterized by the
local project as a wetland. The possibility of the basin outlet discharging
phosphorus from internal sources, rather than influent runoff, is suggested.
37
-------
Basin Performs
cr
a?
Q
UJ
>
GC
UJ
w
m
O
% REMOVAL OF POLLUTANTS
DURING MONITORED STORMS
Above
Expectation
Below
Expectation
20
40 60
COMPUTED %R
80
100
Figure 10. Comparison of observed vs. computed removal efficiencies
(site numbers given for outlierssee text)
38
-------
Site #9 shows Total P removal projections that are significantly in excess of
observed removals. However, as with Site #5, projected removals compared
quite favorably with observed performance for both lead and TSS. This rather
large basin, actually a five-acre lake, supports significant algal growth. The
observed significant reductions for soluble phosphorus and nitrogen are
attributed to algal uptake, since they could not have resulted from sedimentation.
Conversion of soluble.nutrients to algal cells would tend to add a source of TSS
and Total P to basin outflows that are not associated directly with the paniculate
forms entering with the stormwater. Such processes tend to reduce the apparent
sedimentation efficiency.
Site #6 is a natural pond (with surrounding park) in a stream system draining an
urban area, and it supports an appreciable population of ducks fed by local
residents. Lead and Total P removals compare favorably to projections.
Removal of TSS is appreciably less than projected. A comprehensive analysis
of removal efficiency for coliform organisms was conducted at this site. This
was not incorporated into the methodology calibration due to the lack of similar
data at other sites. It is instructive to note, however, that despite the duck
population, average removals for the monitored storms were on the order of
90% for total coliforms, fecal coliforms, and fecal strep.
4.4 EXAMPLE COMPUTATION
A. Given
A 10-acre residential development has a runoff coefficient (Rv) estimated at 0.25. All
stormwater runoff from the area is to be routed to a wet pond detention basin.
Space constraints limit the basin dimensions to 25 by 50 ft, or a surface area of 1250 square
feet. The basin will have an average depth of 4 feet Physical storage volume is 5000 cubic
feet(CF).
Rainfall statistics for the area are:
mean -coef. of variation
Volume (V) inch 0.53 1.44
Intensity (I) in./hr 0.086 1.31
Duration (D) hr 7.2 1.09
Interval (A) hr { 85.0 1.00
Particle settling velocities as tabulated in Section 4.3.1 are assumed to apply for this site.
B. Required
Estimate the long-term average reduction in total suspended solids (TSS) in storm runoff that
can be obtained from the specified basin size.
39
-------
C. Procedure
Step 1 - Select appropriate performance curve to use.
Figure 1 does not apply because removal efficiency by sedimentation varies with
flow through rate, as illustrated by Figures 7 and 8
Figure 2 applies for removal under dynamic conditions
Figure 3 and 4 apply in this case because storage capacity is provided by the device,
and removal by sedimentation also occurs during quiescent conditions between
storm events
Step 2 - Compute runoff parameters for mean storm - flow rate (QR) and volume (VR).
QR = (I) * (Rv) * (Area) * (43,560/12)
= 0.086 * 0.25 * 10 * 3630 = 780 CFH
VR = (V) * (Rv)* (Area)* (43,560/12)
= 0.53 * 0.25 * 10 * 3630 = 4807 CF
Assume that the variability of runoff parameters is the same as for the corresponding
rainfall parameters.
CVq = 1.31 and CVV = 1.44
Step 3 - Compute the removal under DYNAMIC conditions.
The overflow rate during the mean storm (QR / A) is
QR/A = 780/1250 = 0.62 ft/hr
Each of the selected size fractions will have a different removal efficiency at the mean
flow. Use the appropriate settling velocity in equation (8), or scale from Figure 8 to
estimate RM, the removal at the mean oveflow (QR / A = 0.62).
Fit a straight line approximation for each removal curve in Figure 8 so that it intersects the
exact curve at the mean overflow rate (QR/A = 0.62). Estimate the removal efficiency at
very low rates (Z in equation 3) from the point where the fitted line intersects the vertical
axis.
Then, for each size fraction, use the values obtained above in equation (3), together with
the estimate of coefficient of variation of runoff flows to estimate the long-term average
removal (RL).
Alternatively, if estimates of "Z" are 100% for all size fractions (a reasonable estimate in
this case), the long-term average removals (RL) can be scaled directly from Figure 2.
40
-------
Since the size fractions are mass weighted, the overall TSS removal will be the average of
the five size fractions.
Results using the graphic approach are as follows:
Size Average Settling RM (%) RL (%)
Fraction Velocity (ft/hf) (Fig. 8^ (Fig. 2*)
1 0.03 5 5
2 0.3 40 23
3 1.5 90 77
4 7. 100 100
5 65. 100 100
OVERALL AVERAGE REMOVAL =61
fraction NOT removed fD = (100 - 61 )/100 = 0.39
Step 4 - Compute the removal under QUIESCENT conditions.
Basin Volume ratio (VB/VR)
(VB/VR) = 5000/4807 = 1.04
The long-term average removal efficiency is defined by Figure 3. This is based on the
coefficient of variation of runoff volumes (estimated at 1.44 in Step 2) and the "Effective"
Volume ratio (VE/VR), rather than the volume ratio computed immediately above, which
is based on physical size of the basin.
The desired ratio (VE/VR) is scaled from Figure 4 using'the ratio VB/VR = 1.04
computed above, and the Emptying Rate ratio.
E = A * Q/VR
A is the average interval between storms = 85 hr
VR is the mean storm runoff volume = 4807 CF
Q. is the solids removal rate as defined by equation (11) in Section 4.2.2, and is the
product of basin surface area (1250 sq ft) and the settling velocity (ys).
Q = vsA
41
-------
Each of the five size fractions has a different settling velocity, and therefore different
values for Q,, E, the effective volume ratio VE/VR, and finally the quiescent removal
efficiency. The table below lists the results of the foregoing procedure for estimating
removals under quiescent settling.
SIZE FRACTION Q E VE/VR %REM
NO. Vs(ft/hr) (=VsA) (=AQ/VR) .(Fig. 4) (Fig. 3)
1 0.0 38 0.7 0.50 35
2 0.3 375 6.6 LOO 54
3 1.5 1875 33.2 1.04 56
4 7 8750 154.7 1.04 56
5 65 81250 1436.7 1.04 56
OVERALL AVERAGE REMOVAL = 51
fraction NOT removed fQ = (100 - 53) / 100 = 0 .49
Step 5 - Compute the COMBINED removal under both dynamic and quiescent conditions.
Overall removal accomplished by the combination of dynamic and quiescent processes is
computed directly from the fractions NOT removed by each process.
»
Fraction NOT removed by quiescent settling f^ = 0.49
Fraction NOT removed by dynamic settling fD = 0.39
% Removed (overall) = [1- (fQ * fv) ] * 100%
= [1- (0.49*0.39)] *100%
= 81 %
A careful examination of the results is instructive. As the following summary table
indicates, the quiescent process has a lesser effectiveness for the removal of particles with
the higher settling velocities, compared with dynamic removals. This is not because the
process provides less efficient sedimentation. It is a result of the fact that for a basin
volume about equal to the mean storm runoff volume (VB/VR = 1.04), a significant
percentage of storm event runoff volumes are greater than the basin capacity. The
indicated quiescent removals reflect the fact that some fraction of the total runoff does not
remain in the basin to undergo quiescent settling.
The efficiency and importance of the quiescent process is reflected by its significantly
higher effectiveness in removing the slower settling fractions.
42
-------
SIZE FRACTION % REMOVAL % REMOVAL % REMOVAL
NO. Vs(ft/hr) DYNAMIC QUIESCENT COMBINED
1 0.0 5 35 38
2 0.3 23 54 65
3 1.5 77 56 90
4 7 100 56 100
5 65 100 56 100
ALL 61 51 81
4.5 DISCUSSION
On the basis of the comparisons between observed and predicted performance (presented
in Figure 10) the analysis methodology described earlier appears to provide sufficiently reliable
estimates of performance for use in planning activities. More refined computations, which do not
require some of the approximations and assumptions used in the probabilistic methodology, are
certainly possible. SWMM and some other deterministic models have this capability, and it would
be interesting and useful to compare projections. It should be noted however, as a close scrutiny of
observed performance (Table 2) will indicate, that because of either limited data sets or complex
site-specific factors, or both, actual observed performance does not conform to a consistent pattern.
It is suggested that other, more refined computations are likely to reflect similar levels of
uncertainty when compared with actual performance data.
The discussion of the outliers in the comparison between observed and computed
performance serves two purposes. First, by identifying site factors that can reasonably be expected
to cause anomalous results, it adds credibility to the analysis methodology. Second, it highlights
the fact that competing processes are at work in wet pond detention basins that may enhance or
degrade removal of specific pollutants.
It is tempting to consider an extension of this methodology (or other analysis
methodologies) to incorporate biological or other processes that are also obviously at "work in at
least some stormwater detention basins. The available data were considered inadequate to support a
meaningful extension of the analysis at this time, although the means for doing so are clear.
Biological or other decay mechanisms are typically expressed as rate coefficients with units of the
reciprocal of time (e.g., I/day). Such rates, for which reasonable estimates can be derived from the
literature or specific studies, can be converted to a psuedo-settling velocity (or vice-versa per
equation 10). With additional data, this would be a worthwhile effort due to the significance of
mechanisms other than sedimentation in stormwater basins.
43
-------
5.0
GENERAL PERFORMANCE PROJECTIONS
The analysis methodology described in Section 2 provides a basis for relating the size of a
retention basin to its average performance as a stormwater quality control device, accounting for the
intermittent and highly variable character of urban stormwater runoff. The calibration results
presented indicate that performance projections, while not precise, are quite adequate
approximations for use in planning activities. Because the calibration analysis covered a very wide
range of physical basin types and sizes relative to the hydraulic loads applied, it is reasonable to
consider the model suitable for use in a generalized analysis.
A generalized analysis is desirable because it addresses the following issues:
Transfer-ability: If information derived from a limited set of site specific
monitoring data can be extended to other areas and other situations, its value is
greatly enhanced. Transferrability of data and information was an important
objective of the NURP effort.
Adjustment: Monitoring programs appropriately emphasize conditions of higher
stress which maximize the information content of a set of data. In this context,
the storms monitored were consistently biased toward more severe events.
Thus, for all test sites, the average of monitored storm events was significantly
larger than the long-term average for all storms each particular basin can expect
to treat As a result, long-term performance will be better (perhaps appreciably)
than performance under test conditions.
Utility: NURP's emphasis was on planning tools, as oppo'sed to a design or
research emphasis. Accordingly, the information which can be developed
should be structured in a format which assists planning activities.
In the results presented below, the analysis methodology is applied using rainfall
characteristics as the basic input because long-term records are available for all areas of the country.
Rainfall is converted to runoff parameters by applying a runoff coefficient, estimates of which are
available from both NURP data and prior literature.
There are regional and local differences in rainfall patterns. Depending on the size and
development of an urban area, runoff coefficients will vary. Feasible local options for basin
surface area and depth will vary. Further, soluble fractions of certain pollutants may vary from site
44
-------
to site, as may typical particle sizes and settling velocities in urban runoff. Because of the
foregoing, local analyses using site specific conditions are the most appropriate approach. Some
general perspectives are possible, however, provided that it is recognized that local factors may
modify results.
There are local differences in rainfall patterns within a region; however, based on rainfall
records for 50 or more cities analyzed under the NURP program, fairly typical regional rainfall
characteristics can be assigned (see Appendix Figure A-2). Detention basin performance for these
rainfall patterns, for basins which have an average depth of 3,5-feet, and catchments which have a
runoff coefficient of 0.2 are illustrated by Figure 11. The comparisons are based on TSS removal.
The depth value shown is an average value: in effect, it defines the relationship between surface
area and volume and is typical of the units in the NURP data base which has been analyzed. The
runoff coefficient used is estimated, based on NURP data analyzed, to be fairly typical of the
average for a large urbanized area. This figure, therefore, illustrates the order of differences in
performance characteristics which can result from regional differences in rainfall patterns.
In Figure 11, and the other figures which follow, basin size is expressed as a
(percentage) ratio between the surface area of the basin and the contributory urban drainage area.
For example, an area ratio of 0.10% on the horizontal axis reflects a basin with a surface area of
0.64 acres serving a 1-square-mile (640-acre) urban drainage area. The performance relationships
could alternatively be expressed in terms of basin volumes, although depth would also have to be
shown in such a case because performance depends on both area and volume provided.
Figure 12 illustrates the effect of increasing average basin depth, and hence volume,
using the Rocky mountain area rainfall statistics. Comparisons are based on TSS removal. Note
that, for basins which provide area ratios in the order of 0.10%, doubling the volume (7 versus 3.5
foot depth) may improve removal efficiency as much as 20%. However, for relatively large
basins, increased depth improves performance only marginally.
Since detention basin performance depends on runoff, rather than the rainfall which must
be used for long-term projections, the runoff coefficient assigned (ratio of runoff to rainfall) is quite
important. The value of 0.2 assigned in Figure 12 is estimated to be a representative value of an
average for broad urbanized areas, and hence useful in providing an estimate of overall areawide
requirements. However, the procedure may also be used to identify detention basin requirements
for smaller, specific urban areas. In such cases, the runoff coefficient may either be lower (low
density residential areas) or higher (commercial, very high density residential). The significant
effect of runoff coefficients on performance in shown by Figure 13, using rainfall characteristics
typical of the Northeast, and TSS removal for the comparison.
A set of detention basin performance charts may be developed using the NURP analysis
methodology, and appropriate local factors, to provide a working guide for planning decisions.
The previous performance charts were based only on TSS removal to simplify the comparisons
which were made. For planning activities, however, estimates of removals for other pollutants of
interest would be desireable.
45
-------
100
EXPLANATION
Basin Depth = 3.5 ft
Runoff Coeff. = 0.20
RM = Rocky Mt. (Denver)
NW = Northwest
NE = Northeast
SE = Southeast
Short Circ. Param. - 3
I
.05 .1 .5 1
BASIN SURFACE AREA AS % OF CONTRIBUTING CATCHMENT AREA
Figure 11. Regional differences in detention basin performance
-------
100
LL
Ut O
o z
< 2
OC DC
^ Z
> <
< 03
5 °c
1 =>
LU a.
I- O
O _i
Z <
O >
-J O
20
3.5 Basin Average Depth (ft)
ROCKY MOUNTAIN AREA
Rainfall Stats.
Mean
Cv
Volume (inch)
Intensity (in/hr)
Duration (hr)
Delta (hr)
0.200
0.040
4.000
100.000
1.60'
1.00
1.20
1.00
Runoff Coeff. (Rv) = 0.20
Short Circ. Param. = 3
I
05
.0.1
0.5
1.0
5.0
BASIN SURFACE AREA AS % OF CONTRIBUTING CATCHMENT AREA
Figure 12. Effect of depth (volume) on performance
-------
100
.p-
00
EFFECTOR Ru-NE RAIN .
Rainfall Stats. Mean Cv
Volume (inch)
Intensity (in/hr)
Duration (hr)
Delta (hr)
0.400 1.50
0.080 1.10
6.000 1.00
80.000 1.00
Runoff Coeff. (Rv) = as shown
Basin Depth (ft) = 3.5
Short Circ.Param. = 3
I
0.1
0.5
1.0
5.0
BASIN SURFACE AREA AS % OF CONTRIBUTING CATCHMENT AREA
Figure 13. Effect of runoff coefficient on performance
-------
SOUTHEAST
Rainfall Stats. Mean
Volume (inch)
Intensity (in/hr)
Duration (hr)
Delta (hr)
0.450
0.120
5.000
72.000 1.00
Runoff Coeff. (Rv) = 0.20
Basin Depth (ft) = 3.5
Short Circ. Param. = 3
TKN.BOD.
COD.Cu.Zn
.05 0.1 0.5 1.0
BASIN SURFACE AREA AS % OF CONTRIBUTING CATCHMENT AREA
Figure 14. Detention basin performance
-------
An illustration of such a chart is presented by Figure 14, using Southeast rainfall
patterns, a basin average depth of 3.5 feet, a runoff coefficient of 0.20, and the paniculate fraction
of specific pollutants developed in the calibration analysis. The paniculate fractions for lead (0.9)
and total P (0.67) employed for this projection are typical values for urban runoff, based on the
NURP data base. For TKN, Cu, Zn, BOD and COD, the estimates of paniculate fraction (0.5) are
based on more limited NURP data and are less certain.
In the absence of appropriate local data, the NURP estimates derived from a very large
data base would provide the best estimate. However, where a local monitoring program is
planned, such estimates and performance projections can be refined if the relevant analytical
determinations are incorporated into the monitoring program.
50
-------
6.0
REFERENCES
1. USEPA. 1980. Data Management Procedures Manual. NURP document,
Washington, D.C., March.
2. Hydroscience, Inc. 1979. A Statistical Method for the.Assessment of Urban
Stormwater. for USEPA NonPoint Sources Branch, EPA 440/3-79-023, May.
3. DiToro, D.M. and M.J. Small. 1979. Stormwater Interception and Storage. Journal
of the Environmental Engineering Division, ASCE, Vol. 105, No. EE1, Proc. Paper
14368, February.
4. Small, MJ. and D.M. DiToro. 1979. Stormwater Treatment Systems. Journal of the
Environmental Engineering Division, ASCE, Vol. 105, No. EE3, Proc. Paper 14617,
June.
5. Fair, G.M. and J.C. Geyer. 1954. Water Supply and Waste Water Disposal. John
Wiley and Sons.
6. USEPA, 1982. Detention Basins for Control of Urban Stormwater Quality.
Washington, D.C., Water Planning Division, First Draft, September.
7. Whipple, W., Jr. and J.V. Hunter. 1981. Settleability of Urban Runoff Pollution.
Water Pollution Control Federation Journal, 51(12), pp. 1726-1731, December.
51
-------
-------
APPENDIX
DATA ON INPUT PARAMETERS
1.0 GENERAL
This Appendix presents information on representative values for parameters used in the
computations. It is intended to serve as a reference that will permit the user to make preliminary
estimates for use in a screening analysis, and for comparing local values against those developed
from a broader data base.
2.0 RAINFALL STATISTICS
Long-term rainfall patterns for an area are recorded in the hourly precipitation records of
rain gages maintained by the U.S. Weather Service (USWS). The analysis procedures used in this
manual are based on the statistical characteristics of storm "events." As illustrated by Figure A-l,
the hourly record may be converted to an "event" record by the specification of a minimum number
of dry hours that defines the separation of storm events. Routine statistical procedures are then used
to compute the statistical parameters (mean, standard deviation, coefficient of variation) of all events
in the record for the rainfall properties of interest
A computer program, S YNOP, documented in a publication of EPA's Nationwide Urban
Runoff Program (NURP), computes the desired statistics from rainfall data tapes obtainable from
USWS. It generates outputs based on the entire record, and also on a stratification of the record by
month, which is convenient for evaluating seasonal differences.
Table A-l summarizes the statistics for storm event parameters for rain gages in selected
cities distributed throughout the country. These data may be used to guide local estimates, pending
analysis of specific data based on a site-specific rain gage. The tabulations provide values for mean
and coefficient of variation for storm event volumes, average intensities, durations, and intervals
between storm midpoints. The cities for which results have been tabulated are grouped by region of
the country. Results are presented for both the long-term average of all storms, and for the June
through September period that is often the critical period for receiving water impacts.
Figure A-2 provides initial estimates of stonn event characteristics for broad regions of the
country, based on data in the foregoing table.
A-l
-------
(a) HOURLY RAINFALL VARIATION
Time
(b) STORM EVENT VARIATION
Time
Volume
Duration
Average intensity
Interval between
event midpoints
PARAMETER
For each
V
d
i
a
storm event
(inches)
(hours)
(inch/hour)
(hours)
For all
Mean
V
D
1
A
storm events
Co«f Var
VM
"d
v\ .
n
Figure A-1. Characterization of a rainfall record
A-2
-------
3949C-I
Table A-l. RAINFALL EVENT CHARACTERISTICS FOR SELECTED CITIES
!>
,*>
Annual
Mean
Location
Great Lakes
Champa ign-Urbana, IL
Chicago, IL (3)
Chicago, IL (5)
Davenport, IA
Detroit, Ml
Louisville, KY
Minneapolis, MN
Steubenville, OH
Toledo, OH
Zanesville, OH
'Lansing, Ml (5) (30 yr)
Lansing, Ml (5) (21 yr)
Ann Arbor, Ml (5)
Lower Mississippi Valley
Memphis, TN
New Orleans, LA (8)
Shreveport, LA (9X17 yr)
Lake Charles, LA (10)
Average
Texas
Abilene, TX
Austin, TX
Brownsville, TX
Dallas. TX
Waco, TX
Average
V
0.35
0.27
0.27
0.38
0.21
0.38
0.24
0.31
0.22
0.30
0.21
0.26
0.52
0.61
0.54
0.66
0.58
0.32
0.53
0.27
0.39
0.36
0.33
'
.063
.053
.053
.077
.050 '
.064
.043
.057
.048
.061
.041
.047
.066
.113
.080
.108
.097
.083
.078
.072
.079
.086
0.080
D
6.1
4.4
5.7
6.6
4.4
6.7
6.0
7.0
5.0
6.1
5.6
6.2
6.9
6.9
7.8
7.7
7.3
4.2
4.0
3.5
4.2
4.2
4.0
A
80
62
72
98
57
76
87
79
62
77
62
87
89
89
110
109
99
128
96
1.09
100
1.06
ioa
Coefficient
"u
.47
.44
.59
.37
.59
.45
.48
.28
.52
.24
.56
.42
1.36
1.46
1.39
1.64
1.46
1.52
1.88
2.02
1.64
1.66
1.74
vi
.37
.58
.54
.24
.16
.42
.22
.03
.16
.01
.55
.42
*
1.31
1.40
1.27
1.40
1.35
.24
.53
.43
.23
.40
1.37
of Variation
vd
.02
.06
.08
.40
.02
.08
.08
.39
0.99
0.93
1.10
0.95
1.07
1.24
1.09
1.26
1.17
.01
.06
.20
.00
.08
1.07
VA
1.02
1.12
1.00
1.01
1.07
1.00
0.98
1.00
1.03
1.03
1.02
1.00
1.01
1.02
0.99
0.99
1.00
.45
.44
.50
.32
.36
1.41
V
0.45
0.33
0.37
0.49
0.27
0.36
0.34
0.39
0.29
0.36
0.29
0.34
0.44
0.53
0.49
0.63
9.52
0.42
0.38
0.33
0.38
0.40
0.38
June to
Mean
1
.102
.091
.090
.112
.095
.094
.075
.094
.083
.100
.073
.078
.112
.142
.105
.130
.122
.121
.106
.104
.100
.117
.110
0
4.6
6.2
4.5
5.3
3.1
4.5
4.5
5.9
3.7
4.3
4.2
5.1
4.7
5.0
5.3
5.9
5.2
3.5
3.3
2.8
3.2
3.3
3.2
A
87
67
76
91
64
78
74
88
69
80
71
89
88
65
109
86
87
114
108
101
III
124
112
September
Coefficient
«u vi
.44
49
.42
.32
.43
.40
.34
.28
.43
.23
.39
.25
1.35
1.40
1.50
1.90
1.54
.56
.82
.94
.65
.60
1.71
.22
.37
.37
.14
.32
.31
.26
.27
.37
.11
.25
.13
1.28
1.42
1.27
1.41
1.35
1.32
1.71
1.33
1.24
1.34
1.39
of Variation
vd VA
.01
.00
.04
.22
0.82
.01
.00
.76
.93
0.95
0.98
0.90
1.12
1.34
1.28
1.43
1.29
0.98
1.02
1.30
1.01
1.07
1.08
1.05
1.13
1.02
0.94
1.14
1.04
0.92
0.95
1.06
1.06
1.00
0.98
1.06
1.08
1.09
0.99
1.06
.46
.49
.67
.44
.39
1.49
-------
3949C-2
Table A-l. RAINFALL EVENT CHARACTERISTICS FOR SELECTED CITIES (continued)
Location
Northeast
Caribou, ME
Boston, HA
Lake George, NY
Kingston, NY
Poughkoepsie, NY
New York City, NY
Mineoia LI, NY
Upton LI, NY
Wantagh LI, NY (2 YR)
Long Island, NY
Washington, O.C.
Baltimore, MO (3)
Southeast
Greensboro, NC
Columbia, SC
Atlanta, GA
Birmingham, ALA
Gainesville, FLA
Tampa, FLA
Average
V
0.21
0.33
0.23
0.37
0.35
0.37
0.43
0.43
0.40
0.41
0.36
0.40
,
0.32
0.38
0.50
0.53
0.64
0.40
0.49
Mean
1 D
.034 5.8
.044 6. 1
.067 5.4
.052 7.0
.052 6.9
.053 ,6.7
.088 5.8
.076 6.3
.075 5.6
.126 4.2
.067 5.9
.069 6.0
.067 5.0
.102 4.5
.074 8.0
.086 7.2
.139 7.6
.110 3.6
.102 6.2
Annua 1
June to September
Coefficient of Variation Mean Coefficient of Variation
A
v, wd VA V IDA Vy Vj vd UA
55 .58 0.97 1.03 1.03 0.24 .054 4.4 55 .64 .15 1.00 1.01
68 .67
76 .26
80 .35
81 .31 (
77 .37
89 .34
81 .42
83 .54
93 .35
80 .45
82 .48
67 1 .40
68 1.55
94 1.37
85 1.44
106 1.35
93 1 .63
89 1.47
.02 1.03 1.06 0.30 .063 4.2 73 .80 .20 1.12 1.12
.98 0.91 1.48 0.27 .076 4.5 72 .25 .61 0.86 1.44
.01 0.91 0.98 0.35 .073 5.0 79 .46 .27 I.O& 1.08
).95 0.87 0.95 0.36 .081 4.9 82 .48 .16 0.96 1.00
.04 0.93 0.89 0.30 .076 4.8 75 .51 .28 1.03 0.95
.14 .30 0.99 0.41 .114 4.5 88 .42 .17 1.48 1.03
.06 .09 0.99 0.42 .101 4.6 88 .56 .10 1.23 1.02
.24 .03 1.03 0.34 .091 4.0 74 .59 .08 1.28 0.99
.30 .12 1.72 0.41 .127 3.4 99 .52 .15 1.21 1.57
.18 .03 1.00 0.41 .107 4.1 78 .67 .38 1.10 1.06
.21 .01 1.03. 0.43 .107 4.2 79 .66 .49 1.08 1.08
.44 .11 1.18 0.34 .093 3.6 62 .67 .43 .20 1.19
.59 .'13 1.18 0.41 .153 3.4 58 .59 .68 .25 1.13
.16 .11 0.93 0.45 .100 6.2 87 .43 .27 .31 0.97
.31 .09 1.00 0.45 .III 5.0 76 .47 .35 .18 1.01
.14 .66 1.06 0.65 .161 6.6 70 .41 .13 .65 0.92
.21 .11 1.10 0.44 .138 3.1 49 .70 .28 .28 1.01
.28 1.22 1.05 0.48 .133 4.9 68 1.52 1.34 1.33 1.01
-------
3949C-3
Table A-l. RAINFALL EVENT CHARACTERISTICS FOR SEI ECTED CITIES (concluded)
e
Annual
Mean
Location
Rocky Mountains
Denver, CO (3) 8 YRS
Denver, CO (3) 25 YRS
Denver, CO (13) 24 YRS
Rapid City, SD (3)
Rapid City, SD (12)
Salt Lake City, UT (3)
Salt Lake City, UT (3)
(2 GAGES)
Average (2)
California
Oakland, CA
San Francisco, CA (75)
Southwest
El Paso. TX
Phoenix. AZ
Average
Northwest
Portland. OR (3) 25 YRS
Portland, OR (10) 10 YRS
Eugene, OR (6)
Eugene, OR (15)
Eugene, OR (20)
Seattle, WA (15)
V
0.15
0.15
0.22
0.15
0.20
0.14
0.18
0.15
0.19
0.78
0.15
0.17
0.17
0.17
0.36
0.39
0.63
0.72
0.46
1
.033
.033
.032
.039
.033
.031
.025
.036
.033
.017
.047
.055
.045
.017
.023
.030
.026
.025
,023
D
4.3
4.8
9.1
4.0
8.0
4.5
7.8
4.4
4.3
59
3.3
3.2
3.6
5.4
15.5
10.9
23.1
29.2
21.5
A
97
101
144
86
127
94
133
94
320
515
226
286
277
60
83
73
118
136
101
Coefficient
vv
2.00
.73
.49
.81
.46
.42
.32
1.77
1.62
1.45
1.54
1.38
1.51
.60
.51
.85
.88
.85
.45
,
.58
.07
.13
.63
.09
0.91
1.06
1.35
0.74
0.89
1.12
1.26
1.04
0.85
0.79
0.87
0.88
0.91
0.86
of Variation
*.
.24
.20
.15
.21
.24
0.92
Q.85
1.20
1.03
1.37
1.07
0.97
1.02
.00
.09
.25
.35
.34
.26
VA
1.25
1.15
0.92
1.33
0.95
1.39
0.97
1.24
1.60
0.72
1.43
1.42
1.48
.47
.32
.74
.30
.19
.02
V
0.18
0.15
0.22
0.20
0.25
0.14
0.16
0.18
0.11
0.14
0.19
0.21
0.17
0.15
0.22
0.21
0.28
0.31
0.29
June to
Mean
1
.053
.055
.053
.063
.059
.041
.031
.059
.020
.017
.069
.090
.060
.019
.027
.033
.029
.027
.024
D
3.2
3.2
4.4
3.0
6.1
2.8
6.8
3.1
2.9
11.2
2.6
2.4
2.6
4.5
9.4
6.3
12.0
15.0
12.7
A
82
80
101
75
101
125
164
78
756
830
142
379
425
109
179
167
226
250
159
September
Coefficient
s
.90
.85
.78
.63
.50
.51
.43
1.74
1.65
1.46
1.68
1.51
1.61
.45
.32
.32
.28
.24
.45
,
.44
.51
.53
.36
.46
.13
.06
1.44
0.56
0.70
1.28
1.64
1.16
0.99
1.33
1.01
1.07
1.15
0.92
of Variation
Yd
'
.20
.20
.35
.08
.39
0.80
1.01
1.14
1.00
1.67
1.20
0.84
1.01
0^95
.13
.05
.92
.19
.24
-4
1.26
1.05
0.23
1.20
0.94
1.41
0.98
1.13
1.09
0.75
1.44
1.25
1.26
.64
.20
.49
.20
.11
.04
Average
0.48
.024 20.0 101
1.61 0.84 1.23 1.21 0.26 .027 11.4 188 1.35 I.II 1.20 1.15
-------
ZONE
t
2
3
4
5
6
7
1
9
«RWO
ANNUM
SUMMER
ANNUAL
SUMMER
ANNUAL
SUMMER
ANNUAl
SUMMER
ANNUAl
SUMMER
ANNUAl
SUMMER
ANNUAl
SUMMER
ANNUAl
SUMMER
ANNUAl
SUMMER
RAINFAU STATISTICS
VOLUME UNI
MEAN
OX
042
OJ8
0.40
0.4}
0.41
0-M
0,52
0.33
0.31
0.17
0.17
o.48
0.26
0.14
0.14
0.15
O.U
C.V.
1.46
1.38
1.4S
\SJ
1.47
\S1
1.46
1.S4
1.74
1.71
LSI
1J1
1.61
US
1.42
LSI
1.77
1.74
INTENSITY HN/Hffl
MEAN
0051
0.082
QJK6
0.101
.102
.133
.097
.122
.080
.110
J345
JXD
0.024
0.027
JJ31
Ml
.038
J1SS
C.V.
U1
U9
U2
1J7
1J«
1J4
1.3S
L3S
1.37
1J9
1JM
1.18
O.S4
1.11
0.11
1.13
US
1.44
DURATION IHRI
MEAN
M
4.4
S.9
4.2
12
4.9
7.3
5.2
40
3.2
3.E
2.8
20.0
11.4
4.5
2J
4.4
3.1
£.₯.
IDS
1.14
LOS
1.09
1.22
1.33
1.17
1.29
107
1.08
1J2
1.01
1.23
1JO
0.92
0.80
1.20
1.14
INTERVAl IHRI
MEAN
73
78
77
77
89
68
99
87
108
112
277
42S
101
188
94
12S
94
78
C.V.
1J37
1.07
LOS
1.08
LOS
1.01
1.0Q
1.06
141
1.49
L4«
1.26
1.21
1.1S
1.39
1.41
1.24
1.13
Figure A-2. Representative regional values for preliminary estimates
A-6
-------
From the statistics of the storm event parameters, other values of interest may be
determined.
The ratio of mean storm duration (D), to the mean interval between storms (A), reflects the
percent of the time that storm events are in progress:
% time that it is raining = -&
A
The average number of storms during any period of time is defined by the ratio between the
total number of hours in the selected period and the average interval between storms (A). For
example, on an annual basis:
Avg. number of storms per year = 355 * 24
A
The storm event parameters of interest have been shown to be well represented by a gamma
distribution, and the results listed in Table A-l indicate that the coefficient of variation of the event
parameters generally falls between 1.0 and 1.5. Figure A-3 plots the probability distribution of
gamma distributed variables with coefficients of variation of 1.0, 1.25, and 1.5, in terms of
probability of occurrence as a function of the magnitude, expressed as a multiple of the mean. This
plot can be used to approximate the magnitude of an event with a specified frequency of occurrence.
For example, consider a site where storm events have volume statistics for mean and
coefficient of variation of 0.4 inch, and 1.5 respectively. Figure A-3 can be used to estimate that 1
percent of all storm events have volumes that exceed about 7.5 times the mean (or 7.5 * 0.4 = 3
inches). If the same location has an average interval between storms (A) of 87.5 hours, there will be
an average of:
(365 * 24) / 87.5 = 100 events/year
and the 1 percentile event (3 inches) reflects a storm volume exceeded on average, once per year.
3.0 RUNOFF COEFFICIENT (Rv)
Runoff coefficient is defined as the fraction of rainfall that appears as surface runoff. The
substantial data base developed under EPA's NURP program indicated that runoff coefficient varied
from event to event at any site. Variations were not significantly correlated with storm size or
intensity and can be treated as random. The median value for a site was best estimated by the
percent of impervious surface in the drainage area.
Figure A-4 illustrates the relationship between the median runoff coefficient observed at an
urban site and the percent of impervious area in the catchment
A-7
-------
PERCENT EQUAL OR GREATER
90 80 70 60 50 40 30 20 10 5
2 1 0.5 0,2 0.1
10,
LI CV = 1.25-i,
-:-+.--"-t- 1
10 20 30 .40 50 60 70 80 90 95
PERCENT EQUAL OR LESS
98 99.0 99.9
Figure A-3. Probability distribution for a variable with a
gamma distribution
A-8
-------
oe
LU
o
1.0
.9
.7
.6
.5
.4
.3
7
.£
1
n
o
o
!o°0
0
o r
I
0 Ch
C
D
O
°0 °
b o
o
o
0
<
o
o
o
0
o
0
D
<
C
0
c
1
10 20 30 40 50 60 70 80 90 100
% IMPERVIOUS
Figure A-4. Relationship between percent impervious area and
median runoff coefficient
A-9
-------
This information may be used to guide estimates of the surface runoff routed to a detention
basin during storm events.
4.0 SETTLING VELOCITIES
The settling velocity of particulates in urban runoff is a key determinant of the efficiency of
pollutant removals by sedimentation. Settling velocity measurements were conducted oh
approximately 50 different runoff samples from seven urban sites. These data may be used to guide
estimates in the absence of local settling column study results.
There is a wide range of particle sizes, and hence settling velocities, in any sample- of
stormwater runoff. This range can be described by a probability distribution of pollutant settling
velocities and determined by an appropriate analysis of the data obtained from standard settling
column tests, as described further below. When the settling velocity distributions obtained from the
NURP studies were analyzed, it was found that there were differences between separate storms at a
site, and differences between individual storms at different sites. Site-to-site differences were of the
same order as storm-to-storm variations at a particular site, justifying the combination of all data.
The result of such an analysis, illustrated by Figure A-5, indicated that it is reasonable to make
estimates of "typical" urban runoff settling characteristics and expect that, in an appropriate analysis,
short-term variations will average out This assumption and the relationship shown, proved to work
out quite well in the analysis of the performance of nine different detention basins in different parts
of the country and differing radically in size.
For analysis purposes, the indicated range of settling velocities can be broken down into
five equal fractions that have the characteristics listed in Section 4 of this document
While the "typical" values provided here are considered to be satisfactory for initial
estimates, and for screening analyses, additional settling column studies are encouraged to expand
the data base and improve site-specific estimates. The test procedure is quite simple, and utilizes
equipment arid procedures that have been in general use for many years and frequently applied in
water and waste treatment applications. The only difference is the technique suggested for analyzing
the data to increase its utility for stormwater runoff applications.
The equipment and procedure are shown schematically by Figure A-6. The settling column,
typically lucite and about 6 inches in diameter by 6 feet high, is fitted with a series of sample ports.
It is filled with the runoff sample, then small samples are withdrawn from the ports at scheduled
intervals of time. Concentrations of pollutants of interest are compared with the initial concentration
and the pattern of percent removal versus port depth (H) and time (T) is determined. Since each port
depth and sample time corresponds to a settling velocity, each measurement (expressed as percent
removal) can be interpreted as the percent of the total that have settling velocities equal to or greater
than that characterized by port location and sampling time.
A-10
-------
10*-
100
10
-1
Mean and 95%
Confidence Interval
o
o
UJ
O
CO
10
2
50
10
o
o
LLI
>
2
IU
CO
0.5
I
I
10
20 30 40 50 60 70 80
90
0.1
PERCENT WITH SETTLING VELOCITIES
EQUAL TO OR LESS THAN INDICATED VALUE
Figure A-5. Probability distribution of settling velocities in
urban runoff-typical based on pooled data
A-ll
-------
«»<
M'
a ' H2
, r
i
k i
'3
h
r
i
L
4
o o o o
- o o o o
- o o o o
- o o o o
111 1
SETTLING COLUMN
ELAPSED TIME TO SAMPLE WITHDRAWAL
0 = Data Point - Record % removed based on observed
vs. initial concentration
Settling velocity (Vs) for that removal fraction is determined
from the corresponding sample depth (h) and time (t)
VS=H/T
Observed % removed reflects the fraction with velocities
equal or greater than computed Vs
A probability plot of results from all samples
describes the distribution of particle settling
velocity in the sample
PROBABILITY
Figure A-6. Estimating settling velocity distributions from settling
column tests
A-12
-------
Test results are often somewhat erratic because of the sensitivity of analytical tests
(especially TSS at low concentrations) and thermal currents and other disturbances in the column.
The use of multiple ports and settling times provides data on a range of settling velocities, and
provides duplicate measurements for many settling velocities and therefore an opportunity to average
out variations inherent in the test procedure.
A-13
-------
------- |