-------
EPA/600/9-89/001
January 1989
PROCEEDINGS
OF
STORMWATER AND WATER QUALITY MODEL
USER GROUP MEETING
October 3-4, 1988
Denver, Colorado
Edited by
James C.Y. Guo^-, Ben R. Urbonas?, and Thomas 0. Barnwell, Jr
^-Department of Civil Engineering
University of Colorado at Denver
Denver, Colorado 80204
^Denver Urban Drainage and Flood Control District
Denver, Colorado 80211
^Center for Exposure Assessment Modeling
Environmental Research Laboratory
Athens, Georgia 30613
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATHENS, GEORGIA 30613
-------
DISCLAIMER
The information in this document has been funded in part by
the United States Environmetnal Protection Agency. Papers descri-
bing EPA-sponsored research have been subject to the Agency's peer
and administrative review, and the proceedings have approved for
publication as an EPA document. Mention of trade names or commer-
cial products does not constitute endorsement or recommendation
for use by the U.S. Environmental Protection Agency.
ii
-------
ABSTRACT
This proceedings includes twenty-two technical papers on
topics related to the development and application of computer-
based mathematical models for water quality and quantity
management. These papers were presented in the Stormwater and
Water Quality Model Users Group Confernece held in Denver,
Colorado on October 3 and 4, 1988.
The contents of this proceedings may be divided into the
following subjects:
Revisions and Modifications on the EPA Models
Administrative Concerns
Applications and Experiences
Latest Developments of Computer Applications
Field Observations and Related Studies
A number of papers presented critical reviews on the modeling
concept, numerical approach, and comparisons with the field
observations. Revisions and modifications on the EPA SWMM model
presented in the proceedings are helpful in the enhancements of
the model capability and user-friendliness.
Although the application of computer model was the main
theme in the meeting, many other subjects such as spreadsheet
use, statistical sensitivity of measured data, AUTOCAD
enhancement in data management, and mapping data base
application, were also presented and discussed.
As a result of this meeting, the consensus is that there are
needs in the stormwater quality and quantity computer models to
improve the modeling techniques and prediction reliability.
These are important topics for the future.
-------
CONTENTS
PAGE
FOREWORD iii
ABSTRACT iv
ACKNOWLEDGMENT vii
User-Defined Conduits in the EXTRAN Block of SWM*. 1
Moira L. Yasenchak and Terence J. McGhee,
Tulane University, New Orleans, Louisiana 70118.
Revised Runoff Block of SWMM. 10
James C.Y. Guo, University of Colorado at Denver,
Denver, Colorado 80204 and
Ben R. Urbonas, Urban Drainage and Flood Control
District, Denver, Colorado 80211.
SWMM-4. 21
Wayne C. Huber and Robert E. Dickinson,
Department of Environmental Engineering Sciences,
University of Florida, Gainesville, Florida 32611
Improvements to Surcharge Calculations in EXTRAN. 33
Laura K. Belvin,
Brown and Caldwell, Seattle, Washington 98119.
Urban Runoff Modelling for Administrative Purposes. 43
William P. Ruzzo, Wright Water Engineers, Inc.,
Denver, Colorado 80211.
Modelling Studies for the City of Austin Storrawater 52
Monitoring Programs.
George C. Chang, John H. Parrish, and Channy Soeur,
Environmental Protection Department,
City of Austin, Texas 78701.
Application of swtM in the New Orleans Area. 62
Terence J. McGhee and Moira L. Yasenchak,
Tulane University, New Orleans, Louisiana 70118.
Use of SWMM/EXTRAN and TR-20 to Develop Regional Stormwater
Detention Plans in the Washington D.C. Region. 73
Brian W. Mack, Thomas S. George and John P. Hartigan,
Camp Dresser & McKee, Annandale, Virginia 22003.
IV
-------
The Application of QUAL-II to Explore Wasteload Allocation 81
Alternatives.
Angelo S. Liberti, Rhode Island Department of
Enviromental Management, Provence, RI 02908,
Raymond M. Wright, Department of Civil and Environmental
Engineering, U. of Rhode Island, Kinston, RI, and
Kevin Scott, Metcalf and Eddy, Inc., Wakefield, MA 01880.
Frequency Analysis of Trace-Level Water Quality Data with a 92
Time-Varying Censoring Level.
S. Rocky Durrans, Merrick and Company, Denver, CO 80222.
Application of the HSPF Model to Water Management in Africa. 102
Robert C. Johanson, School of Engineering,
University of the Pacific, Stockton, CA 95211.
Multi-Model Micoro-Computer Based Wet Detention Basin Design 119
Methodology.
Sidney L. Harrell, Engineer, Department of Natural
Resources and Community Development, Releigh, NC 27611.
Modeling and Field Evaluations of Urban Wet Detention Ponds. 129
Jy S. Wu, Department of Civil Engineering, University
of North Carolina at Chalotte, North Carolina 28223.
Hydrologic Data Automation Using AUTOCAD. 142
James Chang, Kiowa Engineering Corporation, Denver,
Colorado, and James C.Y. Guo, Department of Civil
Engineering, University of Colorado at Denver.
Distributed Rainfall-Runoff Modelling Based on Digital
Map Database.
Lynn E. Johnson and Charles Huffman,
Department of Civil Engineering, University of
Colorado at Denver, Denver, Colorado 80204.
PC-Synop, A rainfall Analysis Tool. 161
Eric W. Strecker, Eugene D. Driscoll, and Gary Palhegyi,
Woodward-Clyde Consultants, Oakland, CA 94607.
Computer Aided Planning of Drainageway Improvements Made Easy
With Lotus 1-2-3. J
Michael B. Cooke and R. Perm Gildersleeve, Jr.
Greenhorne & O'Mara, Inc., Aurora, Colorado 80014.
Hyetograph Compositing Effects on Urban Runoff Modeling. 183
Michael P. Jansekok and Ben R. Urbonas,
Urban Drainage and Flood Control District,
Denver, CO 80211.
Flood Hydrograph for Ungaged Watersheds. 196
Wolney Carstens Cunha, Stewart Environmental
Consultants, Inc., Ft. Collins, CO 80522.
-------
Unit Hydrograph Procedures for Arid Lands. 208
George V. Sabol, Consulting Engineer, Brighton, Colorado,
Joe M. Rurnann, Davar Khalili, and Teresa A. Dominguez,
Flood Control District of Maricopa County,
Phoenix, Arizona.
Determination of Designated Floodway Boundaries Around 217
Islands in Stream Channels.
J.F. Harp, Civil Engineering Department,
University of Oklahoma, Norman, Oklahoma 73019.
Gulf Coast Flood Routing. 224
Ronald L. Rossmiller and Kenneth R. Wright,
Wright Water Engineers, Inc., Denver, Colorado 80211.
LIST OF ATTENDEES 232
vi
-------
ACKNOWLEDGMENT
The Stormwater and Water Quality Model Users Group appreciates
the help of interested members in making arrangements for sharing
and discussing their professional modeling experiences.
This particular meeting was organized by Dr. James C.Y. Guo
and Dr. William C. Hughes of the University of Colorado at Denver
and Mr. Ben R. Urbonas of the Denver Urban drainage and Flood
Control District. Mr. Thomas 0. Barnwell, Jr., of EPA1s Center
for Exposure Assessment Modeling has rendered great help in con-
ducting and reporting the meetings.
vii
-------
-------
USJER-DEF1N.ED CONDUITS IN THE.. EXTRAN BLOCK_.OF SWMM
by: Moira L. Yasenchak and Terence J. McGhee
Tulane University
Mew Orleans, LA 70118
ABSTRACT
The drainage system in the city of New Orleans is extremely
complex, containing conduits ranging from small circular pipes to
very large open canals with complicated cross-sectional
geometries. Invert slopes are low (sometimes zero) and permit
branching of flow in the downstream direction as well as flow
reversal during runoff events. Both large and small conduits are
often surcharged and the entire flow must be pumped since the
city is surrounded by levees.
This network has been simulated using EPA's Storm Water
Management Model. SWMM provides a variety of alternative routing
techniques, ranging from a quasi-steady state storage routing
procedure in RUNOFF to a finite difference solution of the Saint
Venant equations in EXTRAN. The solution technique employed in
EXTRAN is most suitable for New Orleans, but this block permits
use of only six standard conduit shapes which do not alwa3"s
correspond to those which exist in the system.
Since the system is not satisfactorily simulated by
TRANSPORT, EXTRAN has been modified to permit the use of any
shape whatsoever - whether it be mathematically definable or not.
Computational changes were limited to the subroutines INDATA,
DEPTHX, and HYDRAD. The revised version of EXTRAN will accept and
run data sets prepared for the standard program with no changes
whatsoever and has been used to assess the effects of
approximating unusual sewer shapes by the standard sections of
EXTRAN. This paper presents the modifications made in the model,
the results of application of both the standard and the modified
version to unusual shapes, and a discussion of the significance
of the modification for the city of New Orleans.
-------
THE NEW ORLEANS DRAINAGE SYSTEM
In the city of New Orleans the drainage system is
complicated and diversified. Some of the problems that must be
handled in the system include non-standard conduits, flow
reversal, surcharging, branching of flow, and multiple pump
stations. The system contains conduits ranging from small
circular pipes to large open canals that have complex cross-
sectional geometries. Flow reversal and surcharging are common
during runoff events. Additionally, the entire flow must be
pumped since the city is surrounded by levees.
CAPABILITIES OF SWMM
The complex New Orleans' system has been simulated using
EPA's Storm Water Management Model (1) with reasonable success,
although certain features are not well modeled (2). Among these,
the complicated shapes of many of the major canals are not
adequately represented by the computational blocks of SWMM.
Within RUNOFF, only circular conduits and gutters are
supported. The hydraulic calculations are based on Manning's
equation and the continuity equation - a procedure which neglects
the possibilities of flow reversal, branching, and downstream
control.
TRANSPORT performs its hydraulic calculations using a four-
point implicit difference scheme and a dynamic wave approximation
to the Saint Venant equations. This block permits the use of
thirteen standard sewer shapes, up to two user-defined shapes,
and simple pump stations. On the other hand, the effects of
downstream elements or conditions on upstream flows cannot be
modeled due to the fact the calculations proceed in a down-slope
direction.
EXTRAN employs a finite difference solution of the Saint
Venant equations. This technique permits the simulation of
various complex phenomena, including branching of flow in the
downstream direction, pressurized flow (surcharging), reversals
in flow, and a variety of downstream controls. A limitation of
EXTRAN is that it provides only six standard sewer shapes and
allows for no user-defined shapes.
In many applications, the six standard shapes available in
EXTRAN may be sufficient. In other cases, the actual conduits
may be reasonably approximated by the "most similar" standard
shape. However, sometimes the "most similar" standard shape is
not obvious, nor is it known how much error might be introduced
by such an approximation. This is the case within the drainage
system of the city of New Orleans. The system is not
satisfactorily represented hydraulically by TRANSPORT, so EXTRAN
was modified to permit the use of any shape, whether it was
-------
mathematically definable or not.
MODIFYING EXTRAN
The two main guidelines used while modifying EXTRAN were to
change the model as little as possible and to take maximum
advantage of the existing calculation procedures. After
examination of the FORTRAN listing, it was concluded that there
were three obvious methods in which additional conduits could be
handled. First, one could simply increase the number of standard
shapes b\ providing normalized block data for other sections.
Second, one could provide equations to calculate the cross-
sectional characteristics. This procedure would calculate the
width, area, and the hydraulic radius as a function of depth,
based on user-supplied dimensions. Lastly, one could produce
additional normalized block data within the program based upon
user-supplled data.
EXTRAN MODIFICATION DETAILS
This final technique, producing additional normalized block
data, is completely general, while the first two techniques
require that the shapes be defined in advanced. Hence, this
method was selected.
Computational changes were limited to the subroutines
INDATA, DEPTHX, and HYDRAD. The subroutines TRANSX, INFLOW,
HEAD, BOUND, TIDCF, and OUTPUT were only altered in their
internal definition of variables.
Subroutine INDATA
In order to accept user-defined data, INDATA was modified to
check for the number of user-defined conduits. If such conduits
are present, user-supplied dimensions of up to 20 sections are
read. Then, block data for each section is developed so that
subsequent calculations can proceed in the same mariner as for
standard conduits. Changes made in this subroutine arc shown in
Figure 1.
Subr_ou_t_ine DEPTHX
When user-defined conduits are present subroutine DEPTHX
will search the user-defined area, width, and hydraulic radius
block data calculated within INDATA to establish the critical
depth and normal depth. Changes in DEPTHX are shown in Figure 2.
Subroutine HYDR_AD
When user-defined conduits are present, subroutine HYDRAD
will determine the width, area, and hydraulic radius by
interpolation in the user-defined block data in the same manner
-------
i u [5] n
figure 1. Changes in INDATA
-------
as for the standard sections. Changes in HYDKAD are shown in
Figure 3.
Figure 2. Changes in DEPTHX
-------
CALCULATE CRITICAL
DEPTH FOR TRAPEZOID
CALCUUTE NORMAL
DEPTH FOR TRAPEZOID
CALCULATE HRAD FOR
iECTANGULAR SECTION
SEARCH BLOCK DATA
AREA ie WIDTH TABLES
SEARCH USER-DEFINED
AREA tc WIDTH TABLES
I
FIND CRITICAL DEPTH
[SEARCH USER-DEFINED!
[AREA & HRAD TABLES!
SEARCH BLOCK DATA
AREA Jf HRAD TABLES
FIND NORMAL DEPTH
FIND NORMAL DEPTH
Figure 3. Changes in HYDRAD
Testing the Modified Block
To determine whether the modifications to EXTRAN were
-------
correct, a network of rectangular and circular conduits, shown in
Figure 4, was analyzed using the modified version without user-
defined conduits. Next, it was re-analyzed by sequentially
replacing the rectangular and circular conduits one by one with
user-defined shapes, which were actually rectangles and circles.
Figure 4. Test System
Some differences were anticipated, due to the fact the manner of
calculations for user-defined shapes is not identical to that of
standard shapes. However, when the rectangular shapes were
redefined, one-by-one, as user-defined rectangles, the
differences in calculated flows and depths became progressively
larger. The error was assumed to be in the model change, but
after examination no error was found. Attention was then
-------
directed towards the original version, where an error was found.
In EXTRAN the hydraulic radius for rectangular conduits was set
equal to the wetted perimeter, which is an error that has existed
at least since Version III became available in 1981.
Potential Effect of the_. Err_or
The hydraulic radius is the cross-sectional area of flow
divided by the wetted perimeter. This value can vary by more
than an order of magnitude from the wetted perimeter. The
minimum ratio of the perimeter to the hydraulic radius for a
rectangular section is eight. Overvaluing the hydraulic radius
by a factor of eight results in either quadrupling the flow and
velocity at the same differential head, or alternately, reducing
the frictional loss by a factor of 16 at the same flow and
velocity. This error cannot be considered to be negligible and
is the minimum error which can be expected if frictional effects
govern. In systems not controlled by frictional losses the
effect, of course, would be less. Additionally, this error
affects only rectangular conduits. Since these may be limited in
number or totally absent in some systems, not all prior studies
may be affected. Following correction of the computational error,
the entire data set for the city of New Orleans was rerun. The
results of the original and revised version were compared. In
nearly all cases the surcharging was reduced in the corrected
version. In the two subareas which showed minor flooding in the
original run, the flooding was reduced, with one exception, in
the corrected version. Overall, the differences were
sufficiently small that modification of the recommendations of
the original study was not .justified.
TESTING THE IMPROVED EXTRAN BLOCK
Following correction of the hydraulic radius computational
error in the original version as described above, the test system
in Figure 4 was rerun. The maximum difference between flows was
commonly 0.01 cfs and averaged 0.04 cfs. The maximum difference
between depths was commonly 0.01 feet and averaged 0.015 feet.
The maximum differences in both flows and depths occurred early
or late in the simulated events when flows and depths were low.
Shapes actually found in the New Orleans system were then
placed in the system in Figure 4. The results were compared with
those obtained for hydraulically equivalent rectangular and
circular shapes. Substantial variations in depth were
encountered between the two techniques at less than full flow
conditions. As the channels approached full flow, the
differences diminished and at full flow the differences were
negligible. Since the design basis for the system is full, the
original approximation method appeared to be feasible. Under
conditions other than full flow, the division of flow among
parallel conduits and the depth of flow are much better simulated
-------
with the modified version of EXTRAN.
Permitting user-defined conduits in the EXTRAN Block of SWMM
provides additional flexibility in the program. Although the
program has been adapted to permit the use of sewers of any
shape, data sets prepared for use in the original version can
still be run in the modified version without alteration.
The authors wish to acknowledge the assistance of Mr. Daniel
E. Rau in this project.The work described in this paper was not
funded by the U.S. Environmental Protection Agency and therefore
the contents do not necessarily reflect the views of the Agency
and no official endorsement should be inferred.
REFERENCES
1. Huber, Wayne C., et al. Storm Water Management Model User's
Manual Version III, EPA, Cincinnati, Ohio 1981.
2. McGhee, T.J. and Moira L. Yasenchak Applications of SWMM in
the New Orleans Area Proceedings, Stormwater and Water Qualitj'
Model Users Group Conference, Denver 1988.
-------
REVISED RUNOFF BLOCK OF SWMM
by: JAMES C. Y. GUO, Ph.D., P.E.
Department of Civil Engineering,
University of Colorado at Denver,
Denver, Colorado 80204.
BEN URBONAS, P.E.
Urban Drainage and Flood Control District
Denver, Colorado 80204.
BACKGROUND OF THE SOFTWARE
Storm Water Management Model (SWMM) was developed for the
Environmental Protection Agency under a joint effort of Metcalf &
Eddy, Inc., University of Florida, and Water Resources Engineers,
Inc.. SWMM was released in September 1970. The model consists
of hydrologic watershed simulation, water guality modeling,
hydraulic routing, contamination prediction, erosion estimation,
as well as other features to function as a complete water quality
and quantity model. The model was updated by the University of
Florida in June 1973. Further changes were made to the Runoff
Block of SWMM in 1974 by the Hydrologic Engineering Center,
Missouri River Division (MRD) of the Corps of Engineers, to
include the option of overflow section for pipes and channels,
and routing capability to model storage reservoirs such as
detention ponds.
In March 1985, the Boyle Engineering Corporation, Denver,
Colorado, in cooperation with the Urban Drainage and Flood
Control District (UD&FCD) converted the MRD version of the Runoff
Block of SWMM to a micro computer version and named it UDSWM2-
PC. UDSWM2-PC includes only the rainfall and runoff subroutines
required for stormwater drainage modeling. In their revision,
the software was modified to be capable of reading and routing
hydrographs previously generated by the UD&FCD software CUHP
which uses the synthetic unitgraph method to predict storm
hydrograph.
SWMM requires the transformation of an urban catchment to
its equivalent rectangular basin. This revision removes this
limitation and enables the engineer to use the unitgraph
convolution to predict storm hydrograph. However, UDSWMM2-PC
requires the user to provide the entire drainage network. Often,
10
-------
it causes inconvenience in designing a new drainage system.
In this study, the University of Colorado at Denver further
modified UDSWM2-PC to include pipe sizing capability which
enables the software to compute the required pipe diameter for
the given hydrographs. The new version, named UDSWM3-PC, is
capable of simulating flood propagations in a drainage system
with or without existing pipes in the network.
FLOOD ROUTING METHODS USED IN SOFTWARE
SWMM, as well as UDSWM3-PC, utilizes the kinematic wave
theory for both overland flow and gutter flow routing. The
kinematic wave equation is a simplified form of the dynamic wave
equation. For a one dimensional open channel flow, the dynamic
wave equation states:
5 u + u 6 u + g 6 y - g(S -S ) = 0 (1)
S t 6 x 6 x of
in which t = time, u = velocity, y = depth of flow, g =
gravitational acceleration, So = channel slope, and Sf = friction
slope
Assuming that friction loss is balanced only by the
gravitational effect, Eq.l is reduced to
SQ = Sf (2)
Eq.2 implies that Manning's equation is suitable to predict
flow motion in a conveyance element.
Overland Flow Modeling
A drainage basin of one foot wide is shown in Figure 1 in
which the infiltration loss is expressed by the Morton's formula.
-k t
I = k1 + (k2-k1) e J (3)
in which I = infiltration rate, t = time, k-^ = final infitration
rate, k2 = initial infiltration rate, and k3 = decay coefficient.
Using the wide open channel approximation, overland flow may
be described by the Manning's equation with flow depth equivalent
to hydraulic radius. Thus, we have
Q = S1/2 W (d + d - d ) 5/3 (4)
n o 1 s
11
-------
in which S = overland flow slope, W = width of flow, n =
Manning's roughness, d0 = flow depth at time t, d]_ = flow depth
at time t + At, and ds = depression loss determined by the basin
soil characteristics.
Storage variation in an overland flow is expressed by its
depth variation, d^, and modeled by the volume difference between
inflow and outflow for the given period of time, At.
dt= R - I + (Qi-Q)/ As (5)
in which R= rainfall intensity, As = overland flow area, Q-^ =
inflow rate at time t and Q = out-flow rate at time, t + At.
In computation, the two unknowns, outflow rate and its
corresponding depth can be obtained by simultaneously solving
Eq's 4 and 5.
Channel Flow Modeling
UDSWM3-PC requires the user to provide a downstream drainage
element for each sub-basin to collect overland flow generated
from the sub-basin. Drainage conveyance element can be either a
round pipe or a trapezoidal channel with or without an overflow
section which can be either a pipe or a trapezoidal channel.
When drainage element becomes full, the program will include its
overflow section to convey flow downstream. To model flow in a
conveyance element, Manning's equation states:
(6)
in which Q = outflow, So = conveyance element slope, A = flow
area, n = Manning's roughness, and Rn = hydraulic radius.
Storage change in the conveyance element is described by
AV = (Qi+ Qw- Q) At (7)
in which AV = volume change in the channel section, and Qw =
channel lateral inflow.
To solve the two unknowns, outflow rate and its depth, the
Newton-Raphson iterative method is used to expedite the solution
convergence .
Reservoir Routing
The reservoir routing in UDSWMM3-pc requires the user to
input a storage-outflow relationship for each detention pond.
Release rate is computed using the storage-outflow curve for the
average reservoir storage during the time increment.
Q0 = 2 S/At + Qi (8)
-------
fov •
•a**-
', //////////////A/// / (j.
*••*••• %M*M» —I-L-fcWw-^j"•^JU^a—»^T—^.1—. Jt»j__J -^B—-——— -,^-|j ^.j™
T>
-------
in which, Qo = outflow rate; S = storage at the beginning of the
time increment, Q^ = inflow rate during the time increment and *
= time increment.
SOFTWARE MODIFICATIONS
After analyzing the UDSWM2-PC source code, it was determined
that the efficient approach to enable the software to size
circular pipes was to introduce an additional option, type 6
gutter: Auto Pipe Sizing, to the existing five types of
conveyance elements. By doing this, we can take advantage of the
existing program structure and computational algorithm of the
UDSWM2-PC model. An auto-pipe sizing routine was introduced to
UDSWMM2-pc to size the pipe diameter required to maintain an open
channel flow. For the purpose of reducing iterations in
computation, a preliminary pipe sizing routine was also developed
to approximate the initial estimate of pipe diameter.
Revisions were made to the RHYDRG subroutine. The newly
introduced variable PIPEFG (Auto-Pipe Sizing Flag) will be set
to one when the type 6 gutter is selected. This variable
activates the auto-pipe sizing routine and controls printout
changes. The new input variable NPRELIM (preliminary pipe sizing
routine) is read from column 80 on card 3.1. When NPRELIM is not
equal to one, the default, 18 inch pipe, will be used as initial
guess for a type 6 gutter. When NPRELIM is set to one, the
preliminary pipe sizing routine in RHYDRG will perform a pipe
sizing for a type 6 gutter. This routine first traces the
drainage paths throughout the entire drainage network to identify
all of the sub-basins that contribute to this pipe to be sized
and then sets the variable, TRIBFG (Tributary Flag), to 1 for
those tributary sub-basins. The hydrographs for all tributary
sub-basins to the new pipe are then added together to approximate
the peak flow for the pipe to be sized. No routing time is
accounted for in the hydrograph summation because the hydraulic
lag times and gutter storage volumes are not known at this point
in computation.
When runoff flows through a detention pond or a diversion
device prior to the new pipe element, only 50 percent of its peak
inflow rate is added to the hydrograph summation. This reduction
factor is adopted based on hydrologic modeling observations,
engineering design experience, and common detention basin release
rate criteria. After the peak flow of the unrouted hydrograph
from all tributary sub-basins has been estimated, the preliminary
pipe sizing routine calculates the pipe size using Manning's
equation with 75 percent of the unrouted peak value and then the
program enters the auto-pipe sizing routine to refine the pipe
size until open channel flow is maintained in the pipe. The
reduction factor, 75 percent, is employed to account for the
hydrograph lagging and gutter storage factors. Although it is
14
-------
considered to be a conservatively low estimate of peak flow
reduction for urbanized basin, it works well with the auto-pipe
sizing routine which can only increase pipe size during its
iteration computations. The user should be aware of any possible
anomalies that may cause erroneous results. When the attenuation
of runoff is considered greater than 25 percent, the option of
using the default of 18-inch pipe as initial estimate should be
chosen.
Revisions were also made to the GUTTER subroutine for pipe
sizing. It takes the first estimate determined by the preliminary
pipe sizing routine or by the default of 18-inch pipe and then
routes hydrographs through the drainage network. Whenever any new
pipe becomes surcharged, its size is increased by 3 inches in
diameter if the pipe is less than 36 inches, or by 6 inches in
diameter if the pipe is greater than 36 inches. These
increments are in conformance with manufacture standard sizes.
After each change in the pipe diameter, hydrograph routing
computations are repeated from the beginning of rainfall. This
procedure is repeated until every new pipe maintains an open
channel flow.
CASE STUDY
A drainage basin of 1428 acres located in the city of
Pueblo, Colorado, is planned for its future developed conditions.
The drainage network is illustrated in Figure 3. The general
sub-basins characteristics are listed as follows:
HYDROLOGIC
INFORMATION
Size, acres
Length, miles 1
Centroid, miles 0
Average slope, % 2
Land use
Hydrologic soil group
Infiltration
initial, inches/hour
final, inches/hour
decay rate xlOE-4
Depression storage
pervious area, in
impervious area, in
DESIGN
104
83
.13
.42
.13
SF
C/D
3.0
0.5
18
0.5
0.1
Note: SF - single family
The CUHP software
203
484
0.97
0.98
3.62
SF/OS
B/C
3.5
0.6
16
0.5
0.1
and OS
was run
103
226
1.
0.
1.
10
46
47
SF
C/D
3
0
0
0
.0
.5
18
.5
.1
102
172
1.18
0.51
1.23
SF
C
3.0
0.5
18
0.5
0.1
POINT
202
238
1.48
0.68
3.83
SF/OS
B/C
3.4
0.6
17
0.5
0.1
201
84
0.93
.33
5.10
SF/OS
C/D
3.0
0.5
18
0.5
0.1
101
129
1.
0.
2.
03
47
13
SF
C/D
3
0
0
0
.0
.5
18
.5
.1
-open space.
for
determine the 10-year storm runoff
this
drainage
hydrographs .
basin
The
to
15
-------
precipitation of a 10-year, 1-hour rainfall is 2.00 inches for
this area. The calculated peak runoff flows for sub-basins were:
Sub-basin number 104 203 103 102 202 201 101
Peak flow, cfs 114 447 284 194 193 113 166
These hydrographs were then stored on a disk for flood
routing using UDSWM3-PC.
Overland flows were collected by the drainage network
depicted in Figure 3. Street gutters were combinations of pipes
with overflow channel (type 5), channels with overflow channel
(type 4), and new pipes (type 6). This combination should
demonstrate the versatility of UDSWM3-PC in handling a drainage
system with and without existing gutters. Gutters 101, 104, and
210 are pipes to be sized. The remaining gutters are pipes,
channels and detention ponds with known sizes. Gutters 55 and 56
are dummy gutters that combines hydrographs before detention
basin and the outlet. The option 3, direct routing method, was
used for these dummy gutters.
The following table lists the configurations of the existing
conveyance elements and their design flows or storage volumes:
GUTTER CHANNEL PIPE DESIGN STORAGE
ELEMENT WIDTH SIDE SLOPE DIAMETER FLOW
(ft.) (inch) (cfs) (acre-ft)
200
201
202
203
211
203
51
50
4
15
5
2.5:1
2.5:1
2:1
48
60
60
275
115
193
419
67
193
4.7
29.9
Results of pipe sizing are presented in the following table:
PIPE TO
BE SIZED
DESIGN
FLOW
cfs
PRELIMINARY
PIPE SIZE
inch
FINAL
PIPE SIZE
inch
ITERATION
101
100
201
165
111
748
36
24
84
16
54
42
102
3
5
3
-------
In this test case, new gutters 101, 104, and 210 were sized
to be 54-, 42-, and 102-inch pipes, respectively, in a single
computer run. This would have been a tedious process if UDSWM2-
PC were used because each new gutter would have had to be
determined individually by several trials before sizing the next
new gutter. This can be a very time-consuming process.
The second case study is to compare the predictions from
UDSWM3-PC and another UD&FCD software, UDSEWER, which sizes storm
sewer system. UDSEWER uses the rational method to predict runoff
for a given basin. The example basin is located in Denver and is
divided into three sub-basins with four conveyance elements to be
sized. The schematic layout of drainage network is presented in
Figure 4.
Design information are given as follows:
Sub-basin information are given as follows:
Designation number
Size acres
Width, feet
Average slope, %
Land use
Infiltration
initial, inches/hour
final, inches/hour
decay rate xlOE-4
Depression storage
pervious area, inch
impervious area, inch
Rational Method
runoff coefficient
95
5.7
500
2.0
3.0
0.5
18
0.5
0.1
0.6
96
7.3
400
1.0
Residential
3.0
0.5
18
0.5
0.1
0.6
97
22.9
1000
3.0
3.0
0.5
18
0.5
0.1
0.6
A 10-year, 1-hour storm was used in this case study. The
rainfall hyetograph for the UDSWM3-PC model was determined using
the Denver Urban Drainage and Design manual with a time increment
of five minutes. Peak flow rates predicted by both models are
shown below:
Sub-basin Number : 95 96 97
Manhole (for UDSEWER) : 11 12 13
UDSWM3-PC Peak Flow,cfs: 14 15 50
UDSEWER Peak Flow,cfs : 13 15 49
The following table summarizes the pipes sized by both
models.
17
-------
104) U03I 1031 102 (£08) 12011 (101
202 201 101
(5) (5) <6)
TTPC
a> PIPC
DIRECT ruiv
«> Pipt
<3) CHUJtCL •/ DVEHUV^
PIPE
' 2> cwmtn.
I 3> BIRECT F1.DW
4> PIPE w/OVERFLOV
, 5> CKAWrtX »/ DVERPUJV*
,' 6) PIPE - WTO SIZE
DITTLEI
Figure 4. Layout of Drainage Basin for Case Two,
13
-------
GUTTER 5678
Length in feet
Slope in percent
UDSWM3-PC
Peak flow in cfs
Pipe size in inches
UDSEWER
Peak flow in cfs
Pipe size in inches
800
1.0
14
21
13
21
800
3.6
16
18
15
18
200
3.9
53
30
49
27
1000
3.0
30
24
27
24
Results indicate only slight differences in peak runoff
rates and pipe sizes. These differences may be caused by the
different routing methods and sub-basin hydrologic factors
required by these two models.
CONCLUSIONS
The auto-pipe sizing routine greatly enhances the
versatility of the UDSWM2-PC model. UDSWM3-PC can now be used
for a drainage basin analysis with or without existing storm
sewers. When designing a new storm sewer pipe, it is now a one-
step process with UDSWM3-PC to size the pipe sizes rather than
the trial and error process required with UDSWM2-PC. The
preliminary pipe sizing routine helps reduce the iteration in
computing pipe sizes and enables the program to arrive at final
pipe sizes more efficiently.
The UDSWM2-PC model is inefficient with computer memory
space. Approximately 90 arrays of 399 points are reserved for
storing gutter and sub-basin data. Only a portion of these
arrays are utilized during computations depending on the total
numbers of gutters and sub-basins in the drainage network. The
remainder of the array space remains idle. For instance, the
case study only used 4.5 percent of the reserved array space.
The correction of this problem would require a complete
restructure of the software. Although this effort was not within
the scope of this study, it is suggested for the future
investigation.
Aknowledgment
The authors wish to acknowledge the assistance of Dr. Young Yoon
in this project. The work described in this paper was not funded
by the U.S. Environmental Protection Agency and therefore it does
not reflect the views of the Agency.
19
-------
REFERENCES
(1) "Urban Highway Storm Drainage Model," Volume 4, Federal
Highway Administration, December 1983.
(2) "Urban Storm Drainage Criteria Manual," Wright-McLaughlin
Engineers, March 1969. Distributed by the urban Drainage
and Flood Control District, Denver, Colorado.
(3) "A Short Course on Urban Storm Water Modeling Using Colorado
Urban Hydrograph Procedures," Department of Civil
Engineering, College of Engineering and Applied Science,
University of Colorado at Denver in cooperation with the
Denver Urban Drainage and Flood Control District, January 8-
10, 1986.
(4) "Urban Storm Drainage Management," John Sheaffer, Kenneth
Wright, William Taggart, and Ruth Wright; Marcel Dekker,
Inc., New York, New York; 1982.
(5) "Elements of Computational Hydraulics," Christopher
Koutitas, Chapman and Hall, New York, New York, 1983.
20
-------
by: Wayne C. Huber and Robert E. Dickinson
Department of Environmental Engineering Sciences
University of Florida
Gainesville, Florida 32611
ABSTRACT
Version 4 of the EPA Storm Water Management Model (SWMM) was released
during September 1988. Improvements and changes include: full adaptation for
microcomputer use, addition of natural channel geometry to the Extran and
Transport Blocks (using HEC-2 input formats), addition of subsurface quantity
routing to the Runoff Block, ability to access recent National Weather Ser-
vice precipitation and meteorological data and perform statistical analysis
on these data, variable time steps in the Runoff Block, metrification of the
Extran Block, simplification of input data, and many other changes. The
model is available from the EPA Center for Exposure Assessment Modeling in
Athens, Georgia. Support and documentation continues from the University of
Florida.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
21
-------
INTRODUCTION
The U.S. Environmental Protection Agency (EPA) Storm Water Management
Model (SWMM) is a comprehensive model for simulation of the quantity and
quality of runoff in urban areas, with extensions to non-urban areas as well.
It was originally developed as a single event model for analysis of combined
sewer overflows (Metcalf and Eddy et al., 1971), but later evolved into a
model for continuous as well as single event simulation and has been applied
to all hydrologic, hydraulic and water quality aspects of urban drainage,
including storm sewers, open channels, combined sewers and sanitary sewers.
In particular, the Extran Block of SWMM contains a solution of the complete
dynamic wave equations (St. Venant equations) for simulation of drainage sys-
tem hydraulics (not water quality) including effects of backwater, looped
connections, surcharging and pressure flow. A bibliography of SWMM usage and
case studies is available (Huber et al., 1985).
Version 4 (Huber and Dickinson, 1988; Roesner et al., 1988) includes and
supersedes Version 3 and its 1981 documentation. (These two user's manuals
are available from NTIS or from the University of Florida.) Although inclu-
sion of case studies within the two user's manuals still leaves something to
be desired, there are 79 example input data files contained on the distribu-
tion disks, all with annotated input. One of the most useful enhancements to
SWMM4 input is the ability to include comment lines within the input data
file. These are indicated by an asterisk in the first column and are
stripped from the file before passing the input data to the rest of the pro-
gram. As indicated in Figure 1, these comments may be used to identify input
variables and to annotate the input. Additional information about the pro-
gram and its implementation on microcomputers is provided in "README" files
on the distribution disks, as is a description of each sample input data
file.
The EPA Center for Exposure Assessment Modeling (CEAM) at Athens, Geor-
gia distributes the model from a bulletin board or on diskettes. They have
developed a standardized format for their programs, in which the program
(both the Fortran code and compiled, executable versions) and the example
data and other files are compressed in an archive format, and automatically
"uncompressed" and installed on a hard disk by installation programs con-
tained on the distribution diskettes. (The same distribution diskettes are
also available from the University of Florida.) Before release, all programs
are tested using three different compilers and all sample data sets run
through the model. Users wishing to run on a mainframe may upload the For-
tran code into their machine for compilation. SWMM4 has been tested on the
CEAM VAX computer as well as on IBM-compatible microcomputers. Information
about program availability and access is available from the authors of this
paper and from:
Mr. David Disney
Center for Exposure Assessment Modeling
Environmental Protection Agency
College Station Road
Athens, Georgia 30613
(404) 546-3123
22
-------
* SPECIAL CONTROL SYMBOLS NOW CONTROL THE SWMM INPUT FILE
* THE SPECIAL SYMBOLS ARE '*', '(§' , AND '$'.
* * ---> COMMENT LINE
* @ ---> SAVE A FILE PERMANENTLY OR USE A SAVED FILE
* $ ---> CALL A SWMM BLOCK
* THE FIRST LINE CONTAINS THE NUMBER OF BLOCKS TO BE RUN AND
* THE JIN AND JOUT INTERFACE FILE UNIT NUMBERS.
(NOTE THAT A SECOND BLOCK IS NOT RUN IN THIS EXAMPLE.)
* NBLOCK JIN(l) JOUT(l) JIN(2) JOUT(2)
SW 2 0 10 10 0
* THE SECOND LINE CONTAINS UP TO 6 SCRATCH FILE UNIT NUMBERS.
MM 6 11 12 13 14 15 16
* FILE 10 (OR ANY UNIT NUMBER) COULD BE PERMANENTLY SAVED.
@ 10 'SAVE10.0UT'
* CALL THE RUNOFF BLOCK USING $RUNOFF
$RUNOFF
* THERE ARE TWO Al OR TITLE CARDS IN EVERY BLOCK.
* ALL CHARACTER DATA MUST BE ENLOSED IN SINGLE QUOTES.
Al 'RUNOFF EXAMPLE 2, SIMPLE CONFIGURATION'
Al 'SINGLE CATCHMENT PLUS SINGLE PIPE, CONST. RAIN'
* COMMENT LINES CAN BE USED TO IDENTIFY INPUT VARIABLES.
* METRIC ISNOW NRGAG INFILM KWALTY IVAP NHR NMN NDAY MONTH IYRSTR
BIO 0 1 0 0 0003 17 88
B2 0 0 2
* 5-MIN TIME STEP, 2-HR SIMULATION
B3 300. 300. 300. 2 2.0
B4 0 0
Dl 0
* KTYPE KING KPRINT KTHIS KTIME KPREP NHISTO THISTO TZRAIN
El 2 1 0 0 0 0 3 60.0 0.0
* STEP-FUNCTION HYETOGRAPH
* TIME-REIN(1) RAIM=REIN(2)
E3 0.0 1.0
E3 60.0 0.0
E3 120.0 0.0
* 2-FT DIAMETER CIRCULAR PIPE
* NAMEG NGTO NPG GWIDTH GLEN G3 GS1 GS2 G6 DFULL GDEPTH
Gl 101 102 2 2.0 300.0 0.005 0.0 0.0 0.014 0.0 0.0
* 2-AC IMPERVIOUS CATCHMENT
* JK NAMEW NGTO WW1 WW2 WW3 WW4 WW5 WW6 WW7 WW8 WW9 WW10 WWII
HI 1 201 101 200.0 2.0 100.0 0.01 0.020 0.20 0.03 0.3 3.0 0.3 0.001
* PRINT CONTROL PARAMETERS
Ml 2 1
* DEFAULT STARTING AND STOPPING PRINT TIME IS DURATION OF SIMULATION.
M2 1 0 0
* PRINT HYDROGRAPHS FOR PIPE 101 AND INLET 102.
M3 101 102
* END THE SWMM SIMULATION BY USING $ENDPROGRAM.
* (ANOTHER BLOCK COULD FOLLOW INSTEAD.)
$ENDPROGRAM
Figure 1. Example SWMM-4 input data file.
23
-------
CHANGES FOR VERSION 4
Not all users will require Version 4 since many computational aspects
are identical to Version 3. Significant modifications are listed below.
1. Input/output has been enhanced. All input is free-format with line
(data group) identifiers (Figure 1). The line identifiers are now a require-
ment since the program uses them as the only means of separating one data
group from another. Program-generated error messages make it easier to lo-
cate problems causes by improper entry of data. Input strings of up to 230
characters are allowed in SWMM4. Strict column sequencing of input data is
still possible as long as at least one space separates the fields.
Comment lines are allowed in this version of SWMM. A comment line begins
with an asterisk in the first column. Sample input data files include iden-
tification of each input variable.
2. Errors have been corrected for all blocks as best they are known.
3. Extran is available in a metric format and uses data group identifi-
ers. Additional features include: a "hot start" capability (restart from end
of previous run); natural channel cross sections, with cross-sections input
as in HEC-2; improvements to surcharge and flow routing routines; and auto-
matic adjustment of small pipe lengths. The natural channel cross section
information is illustrated in Figure 2, based upon application work performed
by Camp, Dresser and McKee. The inclusion of natural channels essentially
allows Extran to be used for dynamic flood routing in any channel system.
4. SWMM output may be linked to the CEAM DYNHYD4 (water quantity) and
WASP4 (water quality) programs for receiving water quality simulation (Am-
brose et al., 1986). Runoff, Transport, Storage/Treatment, and Extran inter-
face files can be read by both DYNHYD4 and WASP4. DYNHYD reads only the
flows from the interface file. WASP4 reads water quality loading rates from
Runoff, Transport, and Storage/Treatment. A model of an estuary therefore
can include Runoff to generate surface pollutant loadings, Transport or
Extran for detailed simulation of surface routing network, DYNHYD4 for simu-
lating a link-node estuary model, and WASP4 for simulating the water quality
of the estuary under the stress of the Runoff or Transport pollutant load-
ings.
5. The microcomputer version permits greater manipulation of interface
files and other scratch and I/O files. The Combine Block may be used to con-
vert any interface file to formatted (ASCII/text) files capable of being read
by programs such as Lotus 1-2-3 or other software. All interface files can
be permanently saved and retrieved. Users can input their own interface
files.
6. A subsurface routing package (quantity only) has been added to the
Runoff Block (Cunningham et al., 1987). A separate accounting is made for
the unsatu'rated and saturated zones, and the water table elevation can fluc-
tuate (Figure 3). Baseflow to Runoff channel/pipes may be generated from the
saturated zone.
24
-------
EL(1), STA(l)
EL(NUMST), .STA(NUMST)
TOP OF BANK ELEVATION
NATURAL CROSS-SECTION
LEFT
OVERBANK
A
MAIN
CHANNEL
"BEST FIT" TRAPEZOIDAL
CROSS-SECTION
MANNING'S N -
STCHL STCHR
XNL XNCH
RIGHT
OVERBANK
XNR
DEEP(N)
Figure 2. Natural channel cross section used in Extran and Transport Blocks.
-------
Ox
'ETD
IMPERVIOUS
UPPER
ZONE
LOWER
ZONE
AREA 1
V/////////////A '
^Mv
fll ENFIL ]
DET/
t
Dl
«
J
MF t 1
III DWT1
if 1, „
W J GWF
^ PERC
u
-v
/—i -r?Rin FV
\ /
\ / ^
IW~*X ^ /*
LW \
V
B
DEPPRC
7 7 7
T*
0 T
/
-• — STG
DTOT
k
*^ BELE>
Figure 3. Conceptualization of Runoff Block subsurface quantity routine.
-------
7. Instead of processing continuous meteorological data in the Runoff
Block, two new blocks have been added: Rain and Temp. These include the cap-
abilities of the former Subroutine CTRAIN in Runoff with additional statisti-
cal analysis similar to the SYNOP program of Hydroscience (1976, 1979). It
is also possible to process rainfall data with the SWMM Statistics Block.
Both new and old National Weather Service (NWS) formats for precipita-
tion tapes may be read, as well as NWS rainfall data on floppy disks. In
general, continuous simulation is easier, with several options for input of
precipitation data and other time series. User-defined input time series may
also be used. Continuous simulation is capable of using up to 10 rain gages.
8. Numerical methods have been improved in the Runoff Block. A varia-
tion of the extrapolation method (Press et al., 1986) is used to couple the
nonlinear reservoir equations, evaporation, infiltration, and groundwater
flow. Subroutine Gutter no longer has convergence problems. There is no
distinction any more between single event and continuous simulation, elimina-
ting parameter ICRAIN.
Runoff uses a wet, dry and intermediate (wet/dry) time step defined by
the user. The wet time step is used whenever there is rainfall or snowmelt
occurring, the wet/dry time step is used whenever there is water remaining on
the surface, and the dry time step is used otherwise. For instance, typical
values could be 5 min, 1 hr and 1 day, respectively.
9. This version of SWMM tries to use more Fortran primitives. There is
one subroutine to read interface files, one subroutine to write interface
files, one clock subroutine (eliminating occasional timing disparities be-
tween interfacing blocks), one file opening routine etc. for all blocks. The
common functions of all blocks are exactly the same.
In general, the Fortran code has been completely revised to minimize GO
TO statements in favor of IF-THEN-ELSE statements and other code improve-
ments. The Fortran is compatible with Fortran 77 standards.
10. This version can be made more modular than the EPA Version 3 for the
microcomputer. It is possible to run files containing only the blocks of
interest, saving the interface file for use by the next block. This permits
file compression for ease of distribution and much faster execution times.
However, as noted earlier, the EPA distribution of SWMM-4 for the microcompu-
ter is in the form of one large executable file that uses overlays for execu-
tion in a 640 kb PC environment.
11. The Graph Block is no longer limited to 200 data points. An unlim-
ited number of points for both measured and predicted graphs can be plotted.
Graph may be used to plot both loadographs (mass/time versus time) and pollu-
tographs (concentration versus time). Graph Block output remains simply
line-printer graphics; the microcomputer version still lacks graphics de-
signed especially for the PC. However, information regarding the structure
of the interface file should be sufficient to easily access ancillary graph-
ics packages, especially with the capability mentioned earlier (item 5) to
27
-------
convert the interface file to an ASCII format.
12. The user has more control over printout in this version of SWMM.
Most printout can be bypassed at the user's discretion. Error messages are
summarized at the end of a run instead of being printed every time step.
13. Microcomputer users will see the current time or time step printed
on the screen during the simulation as well as other program messages. Exam-
ples are illustrated in Figures 4, 5, 6 and 7 for the Graph, Runoff, Trans-
port, and Extran Blocks, respectively. These "reassurance" messages are
especially useful during long runs to know the status of the computations.
The length of time required to execute a SWMM block on an IBM AT-compatible
microcomputer varies with the degree of discretization and number of time
steps, but most runs will require less than 5 minutes.
SUMMARY
Version 4 of SWMM attempts to update and correct errors found in earlier
versions, add new computational features (most notably the subsurface routing
portion of Runoff and the natural channel sections for Extran and Transport),
and make the program easier to run on a microcomputer without eliminating the
option for use on main-frames. The conceptualization of the rainfall-runoff-
quality processes remains the same, with attendant strengths and weaknesses.
SWMM is expected to remain a familiar and improving tool for analysis of
urban hydrologic and similar problems for the foreseeable future.
ACKNOWLEDGMENTS
This research was supported in part by EPA Cooperative Agreement CR-
811607. We gratefully acknowledge the support of personnel of the EPA Center
for Exposure Assessment Modeling in Athens, Georgia. The Extran Block was
was developed and has been supported by Camp, Dresser and McKee, Inc. We
appreciate the efforts of Larry A. Roesner and John A. Aldrich of CDM in ex-
tending its capabilities. Finally, many persons at the University of Florida
and elsewhere have contributed to SWMM components over the years, for which
we are very appreciative.
REFERENCES
1. Ambrose, R.B., Vandergrift, S.G. and Wool, T.A. WASP3, a hydrodynamic
and water quality model -- model theory, user's manual and programmer's
guide. EPA/600/3-86/034. Environmental Protection Agency, Athens, Geor-
gia, September 1986.
2. Cunningham, B.A., Huber, W.C. and Gagliardo, V.A. A description of a
new groundwater subroutine in the Storm Water Management Model. In
Proceedings of Stormwater and Water Quality Model User's group Meeting.
Denver, Colorado. EPA/600/9-87/016. Environmental Protection Agency,
Athens, Georgia, March 1987, pp. 70-104.
3. Huber, W.C. and Dickinson, R.E. Storm Water Management Model, version
4: user's manual. EPA/600/3-88/001a (NTIS PB88-236641/AS). Environmen-
28
-------
tal Protection Agency, Athens, Georgia, 1988.
4. Huber, W.C., Heaney, J.P. and Cunningham, B.A. Storm Water Management
Model (SWMM) bibliography. EPA/600/3-85-077 (NTIS PB86-136041/AS).
Environmental Protection Agency, Athens, Georgia, September 1985.
5. Hydroscience, Inc. Areawide assessment procedures manual. Three vol-
umes. EPA-600/9-76-014. Environmental Protection Agency, Cincinnati,
Ohio, 1976 et seq.
6. Hydroscience, Inc. A statistical method for the assessment of urban
stormwater. EPA-440/3-79-023. Environmental Protection Agency, Washing-
ton, DC, May 1979.
7. Metcalf and Eddy, Inc., University of Florida, Water Resources Engi-
neers, Inc. Storm Water Management Model, volume I - final report. EPA
Report 11024DOC07/71 (NTIS PB-203289). Environmental Protection Agency,
Washington, DC, July 1971.
8. Press, W.H., Flannery, B.P., Teukolsky, S.A. and Vetterling, W.T. Nu-
merical Recipes. Cambridge University Press, New York, 1986.
9. Roesner, L.A., Aldrich, J.A. and Dickinson, R.E. Storm Water Management
Model, version 4: user's manual. Addendum I, Extran. EPA/600/3-88/OOlb
(NTIS PB88-236658/AS). Environmental Protection Agency, Athens, Georgia,
1988.
*************************************************
* THIS IS AN IMPLEMENTATION OF EPA SWMM 4.0 *
* "NATURE IS FULL OF INFINITE CAUSES WHICH *
* HAVE NEVER OCCURED IN EXPERIENCE" da Vinci *
*************************************************
ENTER INPUT FILE NAME - GRAPH3.DAT
ENTER OUTPUT FILE NAME - BOB.OUT
READING THE INPUT FILE AND DELETING COMMENT LINES
*************************************************
* ENTRY TO GRAPH BLOCK. LAST UPDATED SEPT. 1988.*
* "All art is quite useless," *
* Oscar Wilde(1891) *
*************************************************
PLOTTING 1 LOCATIONS
PLOTTING GRAPH #
1
Figure 4. Output to CRT during Graph Block execution.
79
-------
READING THE INPUT FILE AND DELETING COMMENT LINES
***************************************************
* ENTRY MADE TO THE RUNOFF BLOCK, LAST UPDATED BY *
* THE UNIVERSITY OF FLORIDA DURING JUNE 1988. *
***************************************************
* "And wherever water goes, amoebae go along for *
* the ride" Tom Robbing *
***************************************************
ENTERING INPUT SUBROUTINE
READING RAINFALL INFORMATION.
READING CHANNEL/PIPE INFORMATION.
READING SUBCATCHMENT INFORMATION.
READING WATER QUALITY INFORMATION.
BEGINNING TIME STEP LOOP. END AT TIME 1440.000 HOURS. FINAL DATE IS 66061
CURRENT STEP/TIME =
STEP= 40 158.170 HOURS. JULIAN DATE = 66007
Figure 5. Output to CRT during Runoff Block execution.
30
-------
READING THE INPUT FILE AND DELETING COMMENT LINES
*****************************************************
* ENTRY MADE TO THE TRANSPORT BLOCK, LAST UPDATED BY*
* THE UNIVERSITY OF FLORIDA JUNE 1988. *
*****************************************************
* "The sewer is the conscience of the city." *
* Victor Hugo (1862)*
*****************************************************
READING ELEMENT DATA.
READING INFILTRATION DATA.
READING WATER QUALITY DATA.
CALCULATING INITIAL CONDITIONS.
BEGINNING LOOP THRU 60 TIME STEPS
TIME STEP #
5
Figure 6. Output to CRT during Transport Block execution.
31
-------
READING THE INPUT FILE AND DELETING COMMENT LINES
*******************************************************
* ENTRY MADE TO EXTENDED TRANSPORT MODEL (EXTRAN) *
* UPDATED BY THE UNIVERSITY OF FLORIDA (UF) AND *
* CAMP DRESSER AND MCKEE INC. (COM), SEPTEMBER, 1988. *
* *
* "Smooth runs the water where the brook is deep." *
* Shakespeare, Henry VI, II, III, 1 *
*******************************************************
READING CONDUIT DATA.
READING JUNCTION DATA.
READING REMAINING SIMULATION DATA.
SCC ==> SUPERCRITICAL CONDUITS. TOTAL # OF CONDUITS. 9
SJ ==> SURCHARGED JUNCTIONS. TOTAL # OF JUNCTIONS. 10
BEGINNING LOOP THRU 480 TIME STEPS
TIME STEP #. # OF ITERATIONS. # OF SCC. # OF SJ.
61 167 3 3
Figure 7. Output to CRT during Extran Block execution. Additional informa-
tion includes the current number of conduits flowing with supercritical flow,
number of surcharged junctions, and the number of iteration used by the
entire model. Two iterations per time step would be an absolute minimum for
the two-step explicit solution method used by Extran.
32
-------
IMPROVEMENTS TO SURCHARGE CALCULATIONS IN EXTRAN
by: Laura K. Belvin
Brown and Caldwell
Seattle, Washington 98119
ABSTRACT
Unsteady non-uniform surcharged flow is a condition frequently found in
combined sewer systems. Yet, to date there is no well-documented, publicly
available sewer model which can accurately simulate such flows. The Extended
Transport (EXTRAN) block of the Environmental Protection Agency's Stormwater
Management Model was designed to model such flows, but is has deficiencies.
Continuity is not necessarily preserved in the model, especially under sur-
charge condition, which causes errors in flow balance, peak flow time, and
peak heads. In this paper, modifications have been made to the model to im-
prove the non-uniform surcharge flow calculations. By accounting for fluid
volumes in the pipes, the calculations of continuity are improved, especially
under surcharge. These modifications were incorporated into the existing
model without changing the input or output parameters, and the improved model
does not require additional user training to run the model. The result is a
publicly available, well-documented hydraulic sewer model that more accurately
models unsteady, non-uniform surcharged flows.
The first section of this paper explains the problems with the continuity
formulation in EXTRAN. Next, the proposed improvements to the continuity
formulation of the model are explained, along with other minor modifications.
A sample application is included to compare the difference in the models. Fi-
nally, remaining model limitations and suggestions for further study are
outlined at the end. The EXTRAN documentation by Roesner,et al. (1984) is a
complete explanation of the remainder of the model. The additions made in
this paper should be used in conjunction with that document.
FORMULATION OF EXTRAN
CONCEPTUAL REPRESENTATION OF EXTRAN
The sewer systems conceptually represented as a series of links and nodes
in EXTRAN. The link-node representation facilitates the description of each
element (link, node, or device) in the system, and provides a convenient
33
-------
method to account for each element throughout the solution process. Links (or
conduits) transmit flow from node to node. Nodes are the connections between
the links and correspond to manholes or pipe junctions. The nodes also act as
storage elements within the program. The storage volume is equal to the water
volume in the half-pipe links connected to each node, as seen in Figure 1.
The change in volume at the node forms the basis of the head calculations.
0,(N-II
Figure 1. Conceptual Representation of EXTRAN
(Roesner, et al., 1984)
GOVERNING EQUATIONS
The flow in a sewer follows the physical principles of conservation of
mass, momentum and energy. Assuming hydrostatic pressure, uniform velocity
distribution, and negligible spatial gradient of internal stresses, flow
through a cross-section of pipe can be expressed mathematically as a pair of
first-order partial differential equations of continuity and momentum. The
combined continuity and momentum equations are customarily referred to as the
Saint Venant equations for unsteady open channel flow. Solving the combined
equations for the change in flow with respect to time, the equation used to
determine the flow through the links can be written as:
dQ/dt = -gASf + 2VdA/dt + V2dA/dx - gAaH/dx
where Q = discharge through the conduit
V = velocity in the conduit
A = cross-sectional area
(1)
H = total head
Sf = friction slope
A separate continuity equation based on storage routing is used at each
node to determine the heads at each junction. The continuity equation solved
for the change in head with respect to time is described as:
an/at = So.t/Ast
(2)
where As^. = area of the entire water surface extending into each half-link at
a node. The water surface area is assumed constant over the time period.
34
-------
During surcharge, the water surface area at the top of the pipe is re-
duced to zero. The continuity equation used in a surcharged manhole is thus:
2(Qt + dQt/dHj-dHj(t» = 0 (3)
A transition function is used to smooth out the change over from gravity to
surcharged flow, and an iteration process is used to boost the heads.
DISCUSSION OF EQUATIONS
Although the flow equation (1) maybe sufficiently accurate, the head
equation does not conserve continuity. A more appropriate continuity equation
might be of the form:
dS/dt = 2Qt where S = Storage (4)
Applied to equation (4), equation (2) would become:
ds = dH-Ast = dH-T-L where T = width (5)
For equations (2) and (5) to be equivalent the surface area must be assumed to
be constant over the time period. Since the length is also constant, this
assumption implies that the width is constant. The heads at adjacent nodes
are also assumed to be fixed for the time interval, and are not included in
the equation.
These assumptions are not applicable to most real drainage systems. Most
sewer systems have closed, circular pipes where the water width is constantly
changing, especially near the top and bottom of the pipe. The difference in
the width in these regions can be more than the change in height during a time
step. Unsteady flow conditions could imply varying depths and depth deriva-
tives in each pipe, effecting the storage volumes at the adjacent nodes.
The existing model does not keep track of storage volumes. This situ-
ation is especially critical in a surcharged manhole. The head and flow equa-
tions are not solved simultaneously, and the flow in and out of the manhole
does not always equal zero, resulting in a continuity error.
For a system of closed, circular pipes under surcharge, the governing
equations presented will not conserve continuity under all circumstances. An
improvement in the use of the continuity equation should improve the head
calculation at each junction, with or without surcharge.
IMPROVEMENTS TO FORMULATION OF EXTRAN
To improve the continuity formulation of the model, an account of fluid
volumes will be made at each time step and node. That way all volumes are
accounted for and continuity is preserved. The determination of heads from
these volumes is not easily determined, though, so that process is also ex-
plained, other minor elements of the program have been modified to enhance
the new solution technique and to improve model hydraulics.
35
-------
CONTINUITY FORMULATION IMPROVEMENTS
Previously, the head at each node was determined using an incomplete
continuity equation (Eq. 2), which can be rewritten as:
H2 = HI + !QtAt/A3t (6)
The change in head is based on a calculation of the water surface area at each
node. No volume computations are made, and continuity is not necessarily
preserved. Continuity can be preserved by computing the volume at each node
and solving the storage routing continuity equation (Eq. 4) which is rewritten
as:
82=8!+ 2QtAt (7)
The storage volumes are determined from a simple calculation of the known
differences in flow through the node. Given the volumes at every node in the
system, a head at each node can be computed from the volume-depth relationship
that will give the known volumes. Continuity is then preserved, with the
result of improved head and flow computations.
Relationship of Volume and Depth
Given the volume at a node, the depth is not easily determined. Several
pipes could be connected to the node, each with differing flows and volumes.
A system of volume equations must be solved to incorporate the effects from
each adjacent node. The volume in a circular pipe is a non-linear function of
the depth, which further complicates the solution. A unique formulation of
the volume-depth relationship of circular pipes will aid in the computation of
the depths at each node.
Volume-Depth in a circular Pipe— The volume of a partially filled con-
duit in this model is equal to the average cross sectional area of flow multi-
plied by the length. The volumes need to be expressed in terms of the depths
in the conduits, instead of the average areas. For a box shaped conduit, the
area is a constant width times the depth, and the depth is easy to separate
out. For circular conduits, the area is a non-linear function of the depth
and is not easily separated.
In order to separate out the depth in the volume equation a special func-
tion was used. Just as in a box shaped conduit the volume is equal to the
depth x width x length, an "effective width" could be developed for the circu-
lar section to separate out the depth. Defining a variable wb = effective
width = A/H, the volume can be expressed as:
V = A-l = A/H-H-1 = wb-H-1 (8)
Although wb is dependent on the head, its normalized values almost become
a constant once the pipe is greater than half full, as seen in Figure 2 below.
The smoothness of this improved wb function will later aid the stability and
accuracy of the solution.
36
-------
Application to Systems of Pipes—The storage in a system node is a func-
tion of the head at the immediate node and at those nodes adjoining connecting
links. The depth in an adjacent node will have an effect on the volume at an
immediate node. For a conduit system with closing tops, several cases of head
conditions between adjacent nodes need to be considered; they could be open
channel, partially submerged, and submerged, as shown in Figure 3.
In a completely open channel flow, the volume in the half link at the
immediate node is equal to the average end areas at the node and at midspan
times the half length. If the conduit is completely submerged, the volume is
equal to the area of the pipe multiplied by the half-link length. The compu-
tation of volume is more complex for the partially submerged condition. The
resulting volume at each node will equal the volume of water in each half-link
connected to it, plus the volume of water in the manhole.
0.9 1
Figure 2 Normalized Effective Width
The change in volume with respect to a change in the water surface eleva-
tion at the immediate node (dV/dH^) can be computed for each node. An example
for the open channel case is included in the figure. The effect of the chang-
ing head at the node on the volume would be the sum of all the dV/dH± terms
from the adjacent links.
Solution of Volume-Depth Equations
The equations for volumes at each node make up a system of volume equa-
tions that are non-linear. A separate solution technique must be used to
determine the heads. An iterative technique is used to solve the volume
equations and find the head at each node. The relaxation method was chosen by
the writer because it converges faster than the Guass-sidel method.
37
-------
An initial guess of the head based on the last known values will be used
to compute a trial volume V2. The difference between this volume and the
volume calculated by continuity, vp, is defined as dv. By definition dv =
QV/dH^-dH^ + dV/dH2-dH2. In this case the dV/dH^ term will dominate, as each
connecting conduit contributes a term to it. The dH2 is assumed to be small
for now. The head equation can then be approximated by:
or
Hl,n + (V2-Vp)/(3V/aH1)
(9)
(10)
VOLUME CASES
PARTIAL
SURCHARGE SURCHARGE
H1CDEFTH
PARTIAL
SURCHARGE
EXAMPLE OPEN CHANNEL
J V L
*H,~7
Figure 3. Volume conditions
This gives a system of head equations for each node. These equations are
solved one node at a time, updating the heads continually. This method of
solution can be as accurate as needed by setting a tolerance check on the
error term, and iterating until that tolerance is maintained. The equations,
although non-linear, should converge within a few iterations as the sum of the
terms dominates the equation.
Surcharge Condition
The continuity equation for a surcharge manhole is much simpler than that
of a partially filled pipe. Under surcharge the continuity equations can be
simplified and substituted into the Saint-Venant equation (1), making an
automatic hydraulic calculation of the required head. Knowing the surface
area of the manhole, AC, the continuity equation is:
38
-------
V2 = V1 + .5-(I.Qn + 2<2n+1)At = Vfull -f (Hn+1 - Hf)-AC (11)
Given the downstream boundary conditions and the previous flow values,
all term of the equations are known, and the combined equations can be solved
for both head and flow at a node. Solving the head and flow simultaneously
reduces the need for a trasition function, preserves continuity, and improves
the surcharge calculations.
OTHER MODIFICATIONS TO EXTRAN
In further investigation of the model, more hydraulic deficiencies were
discovered. These deficiencies are minor compared with the problems with the
continuity equation, but are important enough to make a difference in the
model output. The application of the boundary conditions and friction losses
can be improved and the code has been modified to include these improvements.
The surcharge calculations of head puts heavy emphasis on the most down-
stream boundary condition. An error in that last depth is propagated up
through all of the surcharged junctions. An instability at the end condition
is also passed along the pipeline. Modifications were made to subroutine
BOUND to account for volumes. Knowing the volumes, and assuming critical
depth at the outfall condition, the flow and head at the downstream node can
be solved simultaneously. This improves the calculation of the most down-
stream head.
The known change in Manning's n over the depth in a circular conduit
(Chow,1959) was not previously included in the model. It is incorporated into
this version by modifying the hydraulic radius calculations. This helps to
improve stability considerations near the top of the pipe, and proves to
smooth out flows in that region.
IMPLEMENTATION OF EXTRAN MODIFICATIONS
PROGRAM STRUCTURE
The modifications in the theory described above were implemented into
the original structure EXTRAN code. The original program can be found in the
existing EXTRAN documentation (Roesner,et al, 1984) This documentation also
describes data preparation, trouble shooting methods, formulation of the
program, and example problems. The modified program can be described by the
same documentation except for changes in the theory and formulation as de-
scribed previously, and some resulting modifications to the program structure.
The organization of the program is diagramed in the master flowchart in
Figure 4. The original EXTRAN block had 13 subroutines in addition to the
main program which controls execution. One additional subroutine has been
added to the code to calculate the heads, NHEAD (New-Head). Three of the
subroutines, TRANSX, BOUND, and INDATA, have been modified to incorporate the
changes in formulation.
39
-------
BOUND
OUTPUT
HYDRkD
Figure 4. Flow Chart - Master Program
SAMPLE APPLICATION
The modified code was tested by using a simple conduit network. A simple
application has the advantage of showing the differences between the models
more clearly, with an ability of comparing the output to hand calculated
values. The five pipe conduit system which was modeled is shown in Figure 5.
Each pipe has the same diameter of 2 feet, slope of .002 ft/ft, length of 400
feet, and roughness coefficient equal to 0.013. The boundary condition is a
free outfall
INFLOW
Q = NODE
= LINK
Figure 5 Schematic of Sample Pipe System
A triangle shaped input hydrograph with a peak flow of 15 cfs at 15 min-
utes was routed through the pipe system. The flow attenuation of this hydro-
graph can be observed easily, and the standard hand calculations of head and
flow are easily checked. According to steady flow conditions, the head with a
gravity flow of 8.8 cfs should equal 1.6 ft. The difference in heads between
manholes during a surcharge flow of 15 cfs should equal 1.2 ft.
40
-------
Comparison of Runs
As shown in Figure 6, the modified program performs better than the
original, and matches the hand calculated values. The difference between the
original and modified code is apparent, especially when the flow depth is near
the top 20% of the pipe. Here the effects of the friction slope modification
and improved continuity equation can be seen. Continuity is preserved during
surcharge also. A slight difference in the flows in and out of the surcharge
node is the result of the volume stored in the manhole due to a change in
head. Notice also that the flow conditions are stable and smooth throughout,
following the general shape of the inflow hydrograph, but lagged. The origi-
nal code hydrograph degenerates at conduit 45 and 56. Not only is continuity
preserved at the end of each run, but it is also preserved at each node for
each time period.
> a
Itm tr no
Figure 6 Comparison of Runs
SUMMARY AND CONCLUSIONS
As demonstrated in the sample application, the modified EXTRAN code pre-
serves continuity and matches computed head values better than the original
EXTRAN code. Because the improvements were made without changing the user
parameters of the model, this new version of EXTRAN, in conjunction with the
available EXTRAN documentation, is the only publicaly available, well-docu-
mented program that accurately models surcharge flow under non-uniform, un-
steady conditions.
MODEL LIMITATIONS
This new version of EXTRAN still has the same limitations that the old
version does. Again, these should be carefully noted and adjusted for accord-
ingly. The methods for dealing with the limitations are discussed in detail
in chapter 4 of the EXTRAN user's manual.
41
-------
Although this version of EXTRAN gives better results under surcharge
conditions, it still does not model all of the peculiarities of sewer flow.
Under some conditions the model still cannot give accurate results. Flow
instabilities at the transitions between gravity and surcharge flow, man-
hole losses, and supercritical and critical flow are not accounted for by the
model. "Equivalent length" pipes used to satisfy the Courant stability condi-
tion must now be used with caution as the pipes are being used as storage. In
addition, instabilities are created when a large peak is routed through the
system without initialization of flows in the pipe. This is partially due to
the nature of the solution scheme and is not controlled by the old transision
function, but it is a major fault of the new model.
SUGGESTIONS FOR FURTHER STUDY
Although improvements have been made to the theory and program results,
some items are still left to be accomplished. An expanded version of the
program could apply the 'effective width' idea to other shape pipes, modify
boundary conditions to account for volumes and continuity, and improve upon
the stability conditions. The model is only a tool in the examination of the
hydraulics of drainage systems. The approximations and limitations of the
model may limit the appropriateness of using this model in some situations.
The computational results should always be checked for suspicious behavior and
compared against measured values and rough hand calculations. The experienced
hydraulic engineer should be able to detect errors and make adjustments for
them.
By using the appropriate continuity relationship at the manholes and
boundary conditions, the results of the modified EXTRAN block of SWMM have
been improved. Hopefully municipalities and consultants will make an effort
to use these ideas to improve their analysis of drainage systems.
REFERENCES
1. Camp, T.R. Design of sewers to Facilitate Flow. Sewage Works Journal.
18:1-16, 1946.
2. Chow, V.T. Open Channel Hydraulics. McGraw Hill, New York, 1959.
3. Roesner, L.A., Shubinski, R.P. and Aldrich, J.A. Stormwater Management
Model User's Manual Version III, Addendum I EXTRAN. EPA-600/2-84-109b.
Municipal Environ. Res. Lab., US. EPA, Cincinnati, Ohio.
4. Yen, B.C. Hydraulics of Sewers. Advances in Hvdroscience. 14:1-157.
Academic Press, Orlando, Florida, 1986.
42
-------
URBAN RUNOFF MODELING FOR ADMINISTRATIVE PURPOSES
by: William P. Ruzzo, P.E.
Wright Water Engineers, Inc.
S490 W. S6th Avenue, Suite 100A
Denver, CO 80S 11
ABSTRACT
Computer modeling of urban runoff has advanced in the last decade or so,
particularly with the advent of the personal computer. The modeling
techniques for simulating watershed responses and real time rainfall have
increased our ability to more precisely predict the runoff peaks and
hydrographs. This advance in modeling capabilities has benefited our overall
understanding of the physical sciences.
Another aspect of modeling urban runoff has also advanced along with the
understanding of the physical processes. This aspect of modeling deals with
the input data as it relates to the actual end use of the model results. In
many cases the model results are used to develop a floodplain for regulatory
purposes, or to develop hydrographs to design flood mitigation facilities.
This aspect of modeling is more for administrative purposes and the decisions
on the input data are not necessarily based on simulating the actual physical
processes. Instead, decisions are made to facilitate the administration of
the floodplain regulations or to provide a sound engineering basis for
facilities which are designed for anticipated future development.
This paper presents the decisions made during urban runoff modeling for
several watersheds in the Denver area, which required the model to account for
anticipated future watershed conditions. The decisions are based on
administrative considerations such as (1) worst case scenarios, (S) limited
jurisdictional control of development, (3) ability to accurately predict
future conditions, and (4-) local policies regarding stormwater management.
Case studies are presented regarding the use of inadvertent detention that
occurs upstream of road or railroad embankments, flood flows which become
split from the main channel, projections of impervious land densities and
selection of watershed characteristics and routing parameters.
43
-------
INTRODUCTION
Stormwater runoff modeling is very complex. Mother Nature has chosen to
be very unpredictable when She decides how much, how long and under what
conditions She elects to make it rain. However, it is because of these
uncertainties that many of us have made a career of estimating what will be
the final outcome of a particular storm event.
Today, we are even being challenged further. Now we are also being
asked to predict the quality of urban runoff and of the receiving water body.
We are ready to take on this new challenge. We have our personal computer on
our desk and are armed with a complete arsenal of computer software and a new
understanding of the physical process. We now can more precisely account for
the various flow routing conditions in the EXTRAN block of SWUM. We can
account for rainfall variations by real-time modeling. We can even automate
the process using CAD. All of this technology has advanced the field of
Stormwater management tremendously over the last 5 to 10 years.
The reason we have undertaken the challenge, however, generally falls
into one or two categories: (1) to design a facility to solve a problem, or
(2) to define a condition such that we can regulate certain activity. To
develop a reasonable runoff model to design a facility or define a regulatory
condition does not always require us to model the actual physical process.
Instead, we can make decisions based on the actual end use of the information.
For this paper, I have called the process urban runoff modeling for
administrative purposes.
For example, when we have a small culvert through an embankment that
impounds large volumes of water, we can rather precisely define the actual
physical process of storage routing the runoff hydrograph atid developing an
outflow hydrograph. However, if we are interested in defining a floodplain
for regulatory purposes, we may not need to know exactly what happens at the
culvert. Instead, we want to provide information which protects the property
owners downstream under existing and future development conditions. For the
latter condition, we do not always include the routing effects on the peak
flows to define downstream floodplain.
Four conditions are discussed in this paper where the decisions on how
to model the watershed are based more on administrative needs rather than
trying to model actual physical processes. These conditions are: inadvertent
detention, flow splitting conditions, future land use conditions, and channel
watershed characteristics and channel routing conditions.
INADVERTENT DETENTION
Inadvertent detention is defined as an unplanned storage of storm runoff.
Good examples of inadvertent detention occur upstream of highway or railroad
embankments which have small culverts and large storage volumes upstream of
the embankments. The storm runoff cannot pass through the culvert without
44
-------
creating backwater and therefore storage. It is referred to as inadvertent
detention since the owner of the facility probably did not intend to create a
detention area, but was only trying to save money by under sizing the culvert.
The Colorado Water Conservation Board recently adopted rules and
regulations for the designation of regulatory floodplains (1). Under Rule 5,
Section B.5 regarding detention, the rule states:
"The hydrologic analysis shall consider the effects of on site detention
for ... highways, road fills, railroad embankments, diversion dams, ...
or other structures only if they have been designed and constructed with
the purpose of impounding water for flood detention and are owned, and
operated, and maintained by a governmental body."
From a physical modeling process, this means that the flood peaks
upstream and downstream of the inadvertent detention will not reflect the
routing effects of the embankment storage upstream. Since the embankment is
in place, the floodplain will reflect the head required to pass the 100-year
flood without reduction of the peak flow from routing (with embankment
overtopping if appropriate).
The reason for this rule is related to the ultimate purpose of the
modeling, which is to protect the property owners from flood hazards. Since
the inadvertent detention was not specifically planned, nor is it publicly
owned and maintained, then the regulatory agency cannot guarantee that the
facility will be in place in the future to protect the downstream property
owners from the increased flooding caused by the removal of the storage.
In addition, the owner of the embankment generally does not own the
property on which the storage occurs. Even if the embankment owner would
agree to having the culvert maintained in perpetuity in its present condition,
the upstream property owner still has the right to fill in the floodplain up
to the limits of the floodway. This filling would reduce the storage volume
and could significantly reduce the flood reduction benefits of the storage.
There are many instances in Colorado where roads and railroads cross
streams and creeks in rural areas and the culverts back up storm runoff
creating inadvertent detention areas. As urbanization takes place, the
storage volume behind the embankments can become important to the protection
of the downstream property and can amount to several hundred acre feet. For
this reason, when preparing stormwater master plans, the inadvertent detention
is almost always included as an alternative solution to the flooding problems,
in which case the models then reflect the actual physical process.
An example of inadvertent detention is shown on Figure 1, (S). Colorado
Boulevard crosses Grange Hall Creek in the City of Thorton, a north Denver
suburb. The culvert restricts the flood flows and creates a substantial
storage area. For the purposes of defining the floodplain, the storage
benefits were not accounted for in the downstream flood peaks. However,
because of the flood control benefits of the detention, the storage was
included as part of the recommended plan by the consultant.
45
-------
LEGEND
-X Limits of 100-year floodplain with existing facilities
•—*• Limits of 100-year floodplain with proposed facilities
• Cross section location
Figure 1. Example of inadvertent detention in a major drainageway
FLOW SPLITTING CONDITIONS
Flow splitting is defined as the hydraulic condition where a portion of
the flood flow leaves the main channel and becomes hydraulically disconnected
for a significant distance. Flow splitting can occur for any frequency flood
and not necessarily for all flood frequencies at a particular location.
Whereas the condition does occur naturally, the frequency of occurrence
increases as urbanization increases. This is primarily due to the
inadequately sized drainage facilities constructed prior to urbanization.
An example of flow splitting conditions is shown in Figure E, (3). This
area is the Henry's Lake drainageway located in south west Denver, which is a
right bank tributary to a major channel called Bear Creek. The total
46
-------
3 •<
o m
cr T
rr
cu
cr Q.
ID
3
(D cr
CU 3-
cr (0
rr
Ul
I
x •<
0- (D
3 DJ
TJ -(
Q.
ID
3 -S
^—'
O
1> O
< CL
3
C CU
ID •-
fD ID Cu Cu
Cu Cu CL 3
Q. 3T ui TJ
2 2: O-
CU O D ID
cr -*, -i 3
ID cr
in •- <
tu ID
m ui CD 3
T3 O C
— n c m
•— in ro
o cu
c -i
cr s: D.
3-
O fD i—
-*» 3 Ifl
cr CU
TO 3- TJ
ID ~S
n o
3- n x
cu cu »-•
3 T)
Q) ct-
3 3-
Q. ID
a>
o «c •<
3
cr O
O -*i -J
U)
[u cr n
CL 3- -h
U> fD I/I
2
o n
-i c: 2
3- < •-•
tD n
DO I 3"
51 Q cr
a. H- i— M.
in ID in in
Q. Q
O <
0) ID
in i
-h
3 «—
O O
rt- S
O
1 3
ID <^~
n- o
C
3- !-•
ID O
n
BJ in
•a TJ
Oi i--
n i—
o
o n
-h n
C
cr -I
3- Ifl
ID
ID Cb
x cr
a
0>
fU cr
=: 3-
o o
-1
cl-
ct- J
3"
m
m
o
x c
ID —
rs ro
-i <
•< Oi
- -i
in a.
3-
(— m
a> a>
x- Q.
m in
a> ID •—
-i x (D
a. n in
• m ui
m
a.
T1 (D cr
o a a-
cr
O cr
rt- 3-
3" (D
m
o
-t> c
-P »-•
= ID
<
n a>
3 -i
TD Q.
n
c o.
— c
< ut
ID cr
cr n ri)
"
3" >-•
ID Ul
ru 3
>Q I
in
cr o
m c
3 cr
ID 3"
Oi
cr O
"
03
C
-1
ID
ru
m
X
CU
-o
in
T)
n
o
a.
o
in
3
CU
3
cu
a.
i
tu
t-l.
Hi
lO
m
cu
n -- o
O O O
n o i
a Q. x;
in *~ m
ID 3 CU
Q. lQ 1
Q. -h -t\
ID T i—
cr O O
ID 3 o r~
3 Q. m
cr -t» -a en
•-•• K- i—i rn
o o cu z
3
•o m
o TJ »-
D. «-• 3
cr i—•
-1
m
CU
ifi
CU
CL
ID
XI
c
CU
cr
ID
n
c
1 — '
<
ID
~\
-+,
i — '
0
s;
in
T)
i — •
t-<
c-t-
cT
V— •
33
an
Cu
3
CL
0
3
ID
3"
C
3
CL
~^
ID
a.
<
cu
-^
xi
t^
3
id
— fj
— ^
O
3
H- '
•-0
5;
£U
d-
ro
~i
in
3"
ID
CL
Cu
1
m
CU
in ~-
in
rr
TJ ID
CU
in ->
in fD
o cu
ID n
cu -HI
-i in
cr C
m
ID
~1 c-r
0) O
n n
c o
-' Ul 3
i— ID O
Cu O
1 -»»
*-<•
O cr
C 2
01 O
o cu
O <-r
a. nj
03
3T
ID
3 c
n n no
ID -h 3
in n
(-• fD
cr Q. t-"
ID ID m
-i a.
< c ^
cu n •
t— cr -(>
i— UI
-»\ 3
<—• id 01
O XI
O cr C
a 3- a.
tfi ID ~I
ID
C -*i
CL O 3
ID 2 •-•
-\ i—
in ro
i TJ Ifl
3- c
v-' cr *j<
n C cr £
3- T in —
(D " cr
3- - J
cu CL
< ID ~*i cr
ro < o 3"
*D 1 ID
t—'
3 O cr -«i
O TJ 3- t—
cr 3 ID O
ID O
T) 3 cr CL
-\ cr 2
O . O
< T)
t- -< ID
a. —i ro cu
(0 3" 0) X-
Q. ID 1 UI
-------
drainageway. For storms of greater intensity, a portion of the split flow
actually returns to the main channel area. The magnitude of the flow split to
the north under Wadsworth Boulevard varies from approximately 80 cfs for the
S-year flood to 180 cfs for the 100-year flood with existing drainage
facilities. The S-year developed flow is 130 cfs and the 100-year developed
flow is 620 cfs.
The second flow split occurs at Hampden Avenue (called the Academy Park
flow split) at the outfall further downstream. This flow split was also due
to an inadequate culvert with subsequent overflow away from the main channel.
The flow overtopping the frontage road travels east out of the basin to Pierce
Street and combines with local runoff. A portion of this total runoff will
then flow north into Bear Creek. The magnitude of the flow split varies from
100 cfs for the 10-year flood to 620 cfs for the 100-year flood with existing
drainage facilities. The 10-year developed flow is 260 cfs and the 100-year
developed flow is 630 cfs.
The basis for the flood routing decisions depend on the specific
location and detailed estimations of the physical process at each location.
WADSWORTH BOULEVARD FLOW SPLIT
Estimates were made of the flood peaks and the volumes that would split
from the main channel. This was accomplished by using the weir equation at
the point of overtopping to calculate the portion of flow that would be
divided. Using the maximum elevation obtained from this calculation and the
corresponding discharge, the hydrograph of the split flow was also obtained by
extracting the residual hydrograph above the portion of the flood peak which
would remain in the main channel.
Since a majority of the flood peaks for the more intense storms and most
of the flood volume for all storms will cross Wadsworth Boulevard and stay
with the main channel, the total flood peaks and volumes were considered to be
routed downstream and were used to define the flood hazards and evaluate the
improvement alternatives. This decision was due to the high probability that
the Wadsworth Boulevard culvert would be replaced, redirecting the flow split
back to the main channel. Therefore the downstream floodplain and any future
downstream improvements would be based on the most likely flood condition. If
the decisions were made to model the actual physical process, any future
improvements to the Wadsworth Boulevard culvert would have to account for the
increased flooding from the total flows in the drainageway, at least from the
regulatory standpoint.
However, because the potential for the flow split was significant and
because of the corresponding flood hazards from the flow split, a flood plain
for the flow split area was defined for administrative purposes and will be
regulated accordingly, until the flow split condition is corrected.
48
-------
ACADEMY PARK FLOW SPLIT
The hydraulic analysis at Hampden Avenue was similar to the Wadsworth
Boulevard analysis, except the decisions were different. Since culvert
improvements under Hampden Avenue at the flow split location were not
considered viable due to costs, the flow split was not combined with the local
flows in Henry's Lake drainageway downstream of Hampden Avenue. Therefore the
floodplains and any drainage improvements downstream of Hampden Avenue were
based on the actual physical modeling process. As with the situation at
Wadsworth Boulevard, the split flow area was also floodplain zoned and
regulated.
FUTURE LAND USES
The projection of future land uses to estimate impervious land densities
is very common in stormwater master planning. However, since we are only
interested in impervious surfaces, the actual land use becomes less important
and the percent impervious of the area becomes the important parameter.
A recent hydrological investigation for First Creek and Irondale Gulch
was performed by Wright Water Engineers, Inc. (^). These two watersheds are
located in northeast Denver and are comprised of 73.9 square miles. Both
watersheds are right bank tributaries to the South Platte River, which is the
major drainageway for the Denver metropolitan area.
Because of the close proximity to the new Denver International Airport
site, the development pressures in the area are very high. As a result, the
future land use projections, in some cases, were very detailed, with land
areas as small a ^0 acres being defined. However, the actual land uses were
also under constant revision since the land use is part of negotiations for
annexation. Because of the volatile and politically sensitive nature of the
land use designation, Wright Water Engineers, Inc. was extremely careful not
to develop any maps which would be in conflict with the ever changing land use
plans of the various cities.
To overcome the sensitive nature of the land use designations and to
provide a reasonable basis for estimating future runoff peaks and volumes,
Wright Water Engineers, Inc. elected to prepare a projected impervious land
density map instead of a projected land use map. This map simply defines the
boundaries for various impervious land density percentages (ie: 30'/., 35%, *tO'/.,
etc.). The map was developed by combining the various land uses with similar
impervious percentages (ie: within 2 to 3 percent), which were calculated
based on the information in the comprehensive plans.
By using the impervious land density map, the type of land use becomes
less important to the projection of future runoff peaks and volumes. In
addition, future land use changes from the master plan need only to identify
if there are any differences in percent impervious values, and not if the land
use changes from residential to commercial. This provides a greater
flexibility in the use of the master plan, even though the model developed for
49
-------
the watershed does not exactly follow the actual physical process. These
decisions were based primarily on local policies and the ability to predict
future development.
SUB-WATERSHED ROUTING
The process of defining and routing the watershed hydrographs involves
many assumptions regarding the physical characteristics of the watershed and
channels. Information is required to model the characteristic routing slope,
the type and condition of routing elements, the length of the routing elements
and the maintenance of these elements. The assumptions required for these
parameters to estimate the future development condition runoff peaks and
volumes are not necessarily based on actual conditions. In many cases, they
are based more on the jur isdictional ability to control future development.
Examples of such decisions made during the First Creek hydrological
investigation (*t) and the basis are discussed below.
To define the characteristic basin slope under future development, the
assumption was made that the existing jur isdictional regulations for drainage
channels would result in an overall slope in which the 100-year runoff
velocity would not exceed 7 feet per second (or 5 feet per second in sandy
soils). In general, this means that the slope of the future channels would
not exceed approximately one-half percent, even though the existing channel
slopes exceed this value in many cases.
This assumption generally would result in a projection of lower flood
peaks for future development. In the past the assumption was based more on the
actual physical process, since future development could occur without any
improvements to the channels. Therefore the routing velocities would be higher
as well as the resulting flood peaks.
More recently, however, we have found that natural channels must be
protected from increased runoff due to urbanization, regardless if development
avoids the channel floodplain; otherwise extensive erosion can occur.
Therefore, as a minimum, the channel must be stabilized in the future in
accordance with local requirements.
The type of the routing element assumed for the future was also based on
existing regulations and experience. For channels, the assumption was made
that the regulations would limit the channel configuration to a depth of 5
feet, a velocity of 7 feet per second and a channel with 4 to 1 side slopes.
This design is in accordance with the requirements of the Urban Drainage &
Flood Control District and takes advantage of the channel storage and lower
flow velocities inherent in the design, which results in lower flood peaks.
However, the flow resistance was based on a channel which was assumed
not to be maintained, because there are insufficient institutional mechanisms
to assure that the channels will be maintained. In both of these cases, the
assumptions are based on realistic possibilities rather than present physical
conditions.
For smaller watersheds, however (less than about 130 acres), the
50
-------
characteristic flow path for future development would probably consist of a
combination of storm sewers and street flow. In addition, urbanization would
increase the actual flow path over existing condition by the addition of
streets which parallel the existing watershed contours. Both the anticipated
future channel maintenance and the characteristic watershed length were
modeled based on anticipated conditions, rather than the current physical
condi t ions.
SUMMARY
Hydrological modelling has advanced considerable over the last five to
ten years, due in part to the personal computer. Our understanding of the
physical processes involved in urban runoff has also increased. However, in
many cases the assumptions required for the model are based on administrative
reasons, rather than on the actual physical process. Various examples of
these assumptions were presented and decisions discussed regarding inadvertent
detention, flow splitting conditions, selection of future impervious land
densities, and typical watershed routing parameters. The basis for these
assumptions can be summarized as follows:
1. Worst case scenarios - In several cases, regulators must assume the
worst case scenario will take place in order to protect the property
owners from flood damage. Inadvertent detention assumptions and flow
splitting decisions generally fall into this category.
£. Limited control over future development - In spite of all the stormwater
management regulations, future development projects, in many instances,
need only to avoid the major drainageways to avoid the regulations.
Under this condition, the assumption that culverts will be replaced and
future problems solved may not be entirely accurate.
3. Ability to predict future conditions - Even if the best planners could
all agree on a future land use plan, future development will probable
not follow the plan to closely. Therefore, modeling the future land use
plan too precisely may not be appropriate.
<+. Local jurisdicticnal policies - Because local jurisdictions look toward
the Denver Urban Drainage & Flood Control District for guidance and
generally enforce their rules and regulations, certain drainage and
flood control improvements will probable be constructed in the future in
accordance with the standards. These standards can then be modeled for
future conditions with reasonable certainty.
The work described in this paper was not funded by the U. S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the
views of the agency and no official endorsement should be inferred.
51
-------
MODELING STUDIES FOR THE CITY OF AUSTIN STORMWATER MONITORING PROGRAMS
by: George C. Chang, John H. Parrish, and Channy Soeur
City of Austin Environmental Protection Department
Austin, Texas 78701
ABSTRACT
This paper presents statistical modeling studies for the City of
Austin's two stormwater quality monitoring programs. One program monitors
creeks of various large multiple-land use urban watersheds. The other
program monitors flow and water quality of small single-land use suburban
watersheds and control structures.
The stormwater quality and rainfall-runoff data generally follow
log-normal probability distributions. Based on the assumptions of
normality or log-transformed normality, the data were analyzed using SAS
computer programs. Regression equations relating runoff and rainfall
variables were successfully developed for each watershed. Total and
incremental pollutant loads for storms were regressed on runoff variables
and antecedent rainfall conditions. The validation of a regression
equation depends on statistical tests and specific precision standards. In
some cases the storm pollutant load was simply estimated as the product of
storm runoff volume and mean EMC's.
The amount of impervious cover in a watershed was chosen to represent
the degree of urbanization for the watershed. The pollutant load per storm
linearly increases with the increase of watershed impervious cover. On the
other hand, the pollutant concentration depends on various factors. Many
of these factors are also related to the amount of impervious cover. For
the large watersheds, the concentrations of many parameters generally
increase with impervious cover. For small suburban watersheds these
relationships do not exist as the concentration mainly depends on land use
and maintenance. Two filtration basins and one wet pond were studied. The
removal efficiencies from filtration or sedimentation are estimated by
comparing the EMC's between inflows and outflows.
These results support the City of Austin's watershed ordinance which
specifies impervious cover limitations and requires sedimentation and/or
filtration basins for controlling stormwater quality for developing areas.
52
-------
INTRODUCTION
OBJECTIVES
The City of Austin has experienced rapid urban development in the past
decade. This growth generated concerns about possible water quality
degradation. The objectives of this study are to document local stormwater
runoff pollutant loading data and the effect of urban development on
stormwater quality, and to evaluate the effectiveness of water quality
control basins.
STORMWATER MONITORING PROGRAMS
Data in this study were obtained from two stormwater-quality
monitoring programs. One of them is creek monitoring conducted by the U.S.
Geological Survey (USGS) under the City of Austin/USGS cooperative
program(l). The USGS performs streamflow measurements and water quality
sampling for various creeks and lakes. The City implemented a separate
monitoring program in 1984. This program(2) monitors flow and water
quality of single-land-use suburban watersheds and control structures. A
total of 18 monitoring stations have been successively installed.
WATERSHED DESCRIPTION
The watersheds are listed in Table 1. All five creeks are tributaries
of the Colorado River which bisects the City of Austin. The river is
impounded in two riverine lakes, and it provides the City's drinking water
and recreation resources. Bull, Barton, and Williamson watersheds have
been undergoing rapid development. The effects of construction activities
on Bull and Williamson water quality may be significant. Boggy and Shoal
creeks receive substantial quantities of urban stormwater runoff from high
density residential and commercial areas. The small watersheds are all
single-land-use suburban areas. Bear creek is primarily an undeveloped
ranch land area. Rollingwood, Hart Lane, and Highwood Apartments are all
better maintained residential areas. Barton Creek Square Mall (BCSM)
drainage basin consists of a shopping mall and its parking lot.
DEGREE OF URBANIZATION
The runoff pollutant load (mass of wash-off per unit area per storm)
is an increasing function of impervious cover as evidenced by previous
studies(3-4). On the other hand, the pollutant concentrations of
stormwater generally depend on various factors such as land use,
maintenance, population density, traffic volume, atmosphere deposition,
channel erosion, construction activity, and the number of connections of
waste discharges to the storm sewer system. This effect is evidenced by a
combination of several studies(5-9). But in many cases, the above
mentioned factors are also related to the impervious cover, thus the
pollutant concentrations are also indirectly related to the impervious
cover. Therefore, the amount of impervious cover in a watershed is chosen
to represent the degree of urbanization.
53
-------
TABLE 1. WATERSHED CHARACTERISTICS AND DEVELOPMENT CONDITIONS
D.A.
Watershed (Acres)
Large
Watershed
1976-86
Barton 74,240
Creek @
Loop 360
Bull Creek 14,272
@ Loop 360
Williamson 4,032
Creek @ Oak
Hill
Boggy Creek @ 8,384
US Hwy. 183
Shoal Creek @ 7,872
12th Street
Small
Watershed
1984-87
Bear Creek 301
(BC)
Rollingwood 63
(RO)
Hart Lane* 371
(HL)
Highwood* 3
(HI)
BCSM* 47
Imp. Location &
Cover Land Use
7(2) Suburb
Multiple
12 Suburb
Multiple
15 Suburb
Multiple
41 City
Multiple
47 City
Multiple
3 Suburb
Undeveloped
21 Suburb
Single Family
39 Suburb
Single Family
50 Suburb
Multi-Family
86 Suburb
Commercial
Development
Condition
Developing
1978-86
Developing
1978-88
Developing
1978-85
Saturated
Saturated
Undeveloped
Saturated
Saturated
Saturated
Saturated
Degree of
Maintenance
Good
Fairly GooH
Fairly Good
Poorer
Fair$
Natural
Good
Good
Good
Good
* Indicates period of data.
$ Channel improvements along
# Hart Lane, Highwood, and
three control structures,
BCSM filtration basins.
Shoal Creek occurred between 1982 anH 19R6.
BCSM Watersheds are drainage basins of the
i.e., Woodhollow wet pond, and Highwood and
54
-------
STATISTICAL MODELS
PREVIOUS STUDY
The City of Austin(lO) previously presented a stormwater quality
modeling study. This paper is a continuation and improvement of the
previous work. The data were analyzed using SAS computer programs.
UNIVARIATE MODELS
Most of the stormwater quality and rainfall-runoff data, such as
pollutant concentrations, pollutant loads, and storm rainfall and runoff
parameters, follow log-normal probability distributions. In some cases,
these data values can be fitted to both normal and log-normal
distributions. The fittings were adequate as tested by Shapiro-Wilk
statistic and normal probability plots. The variables of storm duration
and time between storms are stochastic independent and can be estimated
from univariate probability distributions. The marginal sample
distributions of all other variables were also obtained. Based on a
statistical plan such as the completely randomized design, the
distributions of various variables can be compared. For example, the
variances and means of EMC's among various sampling sites were compared.
REGRESSION MODELS
Normal error linear regression models(ll) were used to correlate
stormwater pollutant load-runoff-rainfall relationships. The models
require that the dependent variable be normally distributed and the
independent variables be mutually independent. The calibration of the
models can be tested using statistical parameters such as the coefficient
of determination (R ) and the coefficient of variation(C ). C is defined
as the sample mean of the dependent variable divided by the regression's
standard error of estimate. A model was considered adequate if either
R > 0.85 or Cv < 0.50.
The regression models were successfully developed for various rainfall
and rainfall to runoff relationships. The accumulated pollutants load
within a rainfall storm were regressed on accumulated runoff, number of dry
days before the storm, and other runoff variables. Other than the
accumulated runoff, the rest of the independent variables are generally not
significant in formulating the equations. In many cases, the regression
itself cannot precisely represent the load-runoff relationship, i.e., the
values of R or C are not within the pre-specified range. Under these
conditions the loads were simply estimated as the product of runoff volume
and mean EMC's.
SIMULATION AND VERIFICATION
The pollutant loads washed off from watershed surface during a storm
were simulated using runoff and loading regression equations. The
pollutant loads for any time period can be obtained by summing the loads
55
-------
from individual storms occurring during this period. The simulation and
its verification have not been completed and are not a part of this
presentation.
RESULTS AND CONCLUSIONS
COMPARISONS OF MEDIAN EMC'S
The median event mean concentrations for large and small watershed are
listed in Tables 2 and 4. For large watersheds the concentrations of TP,
BOD, TSS, and fecal coliform generally increase with the imperviousness.
The higher TSS concentration for Bull creek is likely due to construction
activities. In reviewing the Williamson Creek data, the TSS concentrations
values for the major developing period (1983-85) are also significantly
higher although the overall median value is lower. The concentration
values of TP, TKN, and TOC are significantly related to that of TSS as
shown in table 3. For small watersheds, the concentration values depend
mainly on land uses and maintenance. Except for the nitrogen parameters,
the concentrations for all residential watersheds are about the same.
These watersheds are all better maintained subdivisions. The higher
concentration values of the nitrogen parameters for single family areas are
probably due to higher rate of fertilizer applications. The concentration
values for the undeveloped watershed are generally lower than those of the
developed areas. The fecal coliform levels for the high density commercial
area, BCSM, are significantly higher than those of other watersheds. The
effect of the daily street sweeping practice in BCSM is significant and
improves the runoff quality in every case. As compared with the large
watersheds, the concentration values for TSS, TOC, TKN, phosphorous, and
fecal coliform for the smaller watersheds are significantly lower. The
higher TSS and TKN values for large watersheds are likely due to greater
channel erosion. The higher fecal coliform and phosphorous values are
probably due to higher population densities, and greater channel erosion
and traffic volumes, respectively.
POLLUTANT LOADING
Storm pollutant loads linearly increase with the increases of
impervious cover and the amount of storm rainfall. Table 5 is a list of
unit pollutant loads (mass per acre) per 1-inch rainfall for all
watersheds. As with the comparisons of concentrations, for the same
parameters (TSS, TKN, TOC, phosphorous, and fecal coliform) the unit loads
for large watersheds are also significantly higher than those of small
watersheds.
EFFTCTENCIES OF CONTROL STRUCTURES
The City of Austin Comprehensive Watershed Ordinance and various
design criterial manuals(12) provide documentations for water quality
management. The City requires sedimentation and/or filtration basins for
most developing areas. This study analyzed data from two filtration basins
and one wet pond. The characteristics of these control basins are shown in
56
-------
Table 6. According to the City's new standards, these structures are not
ideal water quality control basins. The main deficiency is that the water
captured for treatment is not isolated off-line from the balance of the
runoff. Nevertheless, the pollutant removals achieved by the filtration
and/or sedimentation are generally significant. Table 7 is a list of
percent reductions in EMC's between inflow and outflow. The removal
efficiencies were not computed pending the completion of flow accounting.
Based on the past studies(13), however, the values of removal efficiency
are generally about 10-20 percent higher than those of EMC reductions.
TABLE 2. COMPARISONS OF EMC'S AMONG LARGE MULTIPLE LAND USE WATERSHED
Barton Bull Williamson Boggy Shoal
D.A. (sq. mi.)
Imp. Cover(X)
No. of Storms
TSS
BOD
TOC
NO™
NH3
TKN
TP
Fe. Col. 28,
116
7
12
740
4.0
23
.22
.08
1.5
.18
500
mwow^vnin
22.
12
14
2,060*
5.
46
•
3.
48,100
_^_^-., __ ., - „
3 6.3
15
13
1,000
3 9.0
34
48 .35
09 .08
5 3.9
39 .69
103,000
, ,.» JT ,,-, -^
13.1
41
21
2,340
9.5
40
.35
.14
2.8
1.30
172,000
. „.„ .^ ,
12.3
47
15
2,010
11.3
33
.50
.12
3.4
1.10
130,000
* Values shown are medians of the EMC's. The unit of fecal coliform is
colonies per 100 milliliter. The unit of other parameters is milligram
per liter. Values grouped by the same number of underscores are not
significantly different from each other.
TABLE 3. CORRELATION OF EMC'S BETWEEN TSS AND OTHER PARAMETERS.
Mean Correlation BOD TOC N00 NH., TKN TP TPO, Pb Zn FeCnl .
— ; .:— . _• ' r—rrrr—• - rrr—-"-- r—-,.r J 3 £L _-_^.v-
Large Watersheds TSS .38 .75* .32 .28 .50* .76* -
Small Watersheds TSS .17 .22 .28 .24 .34 - .36 .51* .61* .13
* Indicates significant correlation. Otherwise the correlation is not
significant.
57
-------
TABLE 4. COMPARISONS OF EMC'S AMONG SMALL SUBURBAN WATERSHEDS
BC
RO
HL
HI
BCSM
BCSMS
D.A. (Acres)
Imp. Cover(%)
No. of Storms
Land Use
TSS
BOD
COD
TOC
N02+N03
TKN
NH3
TN
P04
Cu
Fe
Pb
Zn
Fe. Col.
Fe. Stp.
301 68 371
3 21 39
18 16 17
Undev. S.F. S.F.
88* 189 105
7. 8. 6.
13. 21. 19.
6. 10. 6.
.13 .66 .92
.34 .76 .62
.07 .11 .12
.44 1.83 1.64
.04 .10 .14
.004 .006 .01
.33 .29 .35
.003 .02 .02
.006 .03 .04
6,000 10,000 14,400 19
3,500 17,000 13,840 12
3 47 47
50 86 86
26 24 23
M.F. Comm. Comm.
70 48 160
7. 9. 9.
23. 32. 64.
8. 9. 16.
.25 .35 .40
.56 .78 1.50
.18 .15
.87 1.15 1.80
.19 .10 .21(TP)
.01 .005
.26 .24
.01 .02
.03 .12
,000 16,000 46,000
^000 6^300 45,000
* Data collected during 1985-87 when the site was maintained by sweeping
the entire parking lot everyday.
$ Data collected during 1982-84 when the site maintenance was at minimum.
# Values shown are medians of the EMC's. The unit of fecal coliform and
fecal streptococci is colonies per 100 milliliter. The unit of other
parameters is milligram per liter. Values grouped by the same number of
underscores are not significantly different from each other.
58
-------
TABLE 5. POLLUTANT LOAD GENERATED FROM 1-INCH RAINFALL
Watershed TSS BOD N03 TKN TP Fe. Col. Pb
Zn
Barton
Bull
Williamson
Boggy
Shoal
Bear Creek
Rollingwood
Hart Lane
Highwood
BCSM
5
22
12
96
83
0.1
2
4
11
16
0.02
0.07
0.11
0.39
0.45
0.01
0.06
0.38
0.92
1.17
0.002
0.005
0.005
0.015
0.020
0.0002$
0.006
0.04
0.05
0.08
0.01
0.05
0.04
0.11
0.14
0.0004
0.001
0.024
0.090
0.292
0.005
0.015
0.034
0.159
0.120
0.0001$
0.001
0.005
0.029
0.041
430
2914
5771
32,021
24,047
32
411
2517
13282
40668
-
-
-
_ _
0.0000 0.0000
0.0002 0.0003
0.0008 0.0020
0.0020 0.0050
-
* The unit of fecal coliform load is million colonies per acre. All other
units are pound per acre.
$ For small watersheds, values are NO^+NO,. and PO, loads, respectively.
TABLE 6. CHARACTERISTICS OF WATER QUALITY CONTROL BASINS
Contri.
Control D.A.
Basin (Acres)
Highwood 3
Filt.
BCSM Filt. 79
Woodhollow* 371
Water Quality
Volume
(Inch-Runoff)
0.5
0.5
0.5
Surf. Area
of Sand Bed
(Acres)
§
0.003 An
0.005 AD
N.A.
Ave. Time
of Outflow
(Hrs.)
19
32
13
Est. Deten.
Time
(Hrs.)
4
6
3
* Detention time is estimated as the time between the centroids of the
inflow-outflow hydrographs.
$ A~ is contributing drainage area of the structures.
# A wet pond was formulated by closing the lower gate of a floodwater
detention basin.
59
-------
TABLE 7. MEDIAN VALUES OF REDUCTIONS OF EMC'S BETWEEN
INFLOW AND OUTFLOW IN PERCENT*
TSS BOD COD TOC N02+N03 TKN P04 TN Cu Pb Zn FeCol.
Woodhollow
Wet Pond 63
BCSM Filtr. 73
Highwood
Filtr. 81
-2 -9 10
32 20 39
8 21 9
37 28 60 23 33 63 46
-4 54 43 33 38 82 82
-72
5 -16 9 41 56
39
31
18
* The basin removal efficiency is about equal or greater than the EMC
reduction if the basin outflow is equal or less than the basin inflow.
CONCLUSIONS
(1) Statistical models can be applied to stormwater runoff loading
process without complexity. For better precision in predicting,
one should probably refer to small scale
distributive/mathematical models.
(2) This study confirms previous finding that stormwater runoff
pollutant loads linearly increase with watershed imperviousness.
(3) The pollutant concentration depends on various factors. Under
certain conditions, many of these factors are related to
watershed imperviousness.
(4) As compared to the small single-land use suburban watersheds of
same imperviousness, the concentrations of TSS, TKN, TOC,
phosphorous, and fecal coliform for the large multiple-land use
watersheds are significantly higher. This is likely due to the
increase (per unit area) of channel erosion, traffic volume, and
population.
(5) Both sedimentation and filtration basins of average design can
significantly improve the quality of stormwater runoff.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents Ho not
necessarily reflect the views of the Agency and no official
endorsement should be inferred.
60
-------
REFERENCES
1. U.S. Geological Survey. Hydrologic data for urban studies in the
Austin, Texas Metropolitan Area. Prepared in cooperation with the City
of Austin, 1975-1986.
2. City of Austin. City of Austin stormwater monitoring program
description. 1986.
3. Engineering Science and the City of Austin. Final report of nationwide
runoff program, 1983.
4. Chang, G.C. and Hartigan, P. Statistical models for urban stormwater
quality studies. Paper presented at the 1985 Joint Statistical
Meeting, Las Vegas, Nevada. August 5-8, 1985.
5. Griffin, D.M., Grizzard, T.J., Randall, C.W., Helsel, D.R., and
Hartigan, J.P. Analysis of nonpoint pollution export from small
catchment. Journal WPCF, Vol. 52, No.4, April 1980.
6. Glenne, B. Simulation of Water pollution generation and abatement on
suburban watersheds. Water Resources bulletin, Vol. 20, No.2, April
1984.
7. Schmidt, S.T. and Spencer, D.R. The magnitude of improper waste
discharge in an urban stormwater system. Journal WPCF, Vol. 58, No. 7,
July 1986.
8. Schueler, T.R. Controlling urban runoff: a practical manual for
planning and designing urban BMPs. Metropolitan Washington Council of
Governments Publication No. 87703, Washington, D.C., 1987.
9. U.S. Environmental Protection Agency. Results of the nationwide urban
runoff program. Vol. 1 - final report, Washington, D.C., 1983.
10. City of Austin. Stormwater quality modeling for Austin Creeks.
Watershed Management Division, November 1984.
11. Neter, J., Wasserman, W., and Kutner, M. Applied linear regression
models. Richard D. Irwin Inc., Homework, Illinois, 1983.
12. City of Austin. Environmental criteria manual. Interim draft, sec.
2.36, stormwater filtration criteria, June 1988.
13. Welborn, C.T. and Veerhuis, J.E. Effects of runoff controls on the
quantity and quality of urban runoff at two locations in Austin, Texas.
Prepared in cooperation with the City of Austin, USGS Report No.
87-4004, Austin, Texas, 1987.
61
-------
APPLICATION OF SWMM IN THE_.NEW_ORLEANS .AREA
by: Terence J. McGhee and Moira L. Yasenchak
Tulane University
New Orleans, LA 70118
ABSTRACT
This paper reviews recent applications of SWMM in the New
Orleans Metropolitan area - a region which offers an unusually
interesting array of drainage problems.
The city and adjoining areas of Jefferson Parish are
entirely enclosed by levees, have very little surface relief, and
have drainage systems which are thoroughly interconnected and
subject to reversals of flow. Rainfall amounts are heavy. The
normal annual precipitation exceeds sixty inches and individual
storms may produce totals of ten to twelve inches in as many
hours. In recent years extensive property damage resulting from
flooding has been the impetus for studies intended to improve the
capacity of systems which, in at least some cases, were built to
dewater marsh and swamp land and are now used to drain developed
urban areas.
In the studies reported herein the standard SWMM blocks
RUNOFF, TRANSPORT, and EXTRAN have all been used depending upon
the particular circumstance. In addition, certain modifications
have been made which make the model more useful in this region.
Among these are the use of "Standard Streets" in a manner
analogous to that employed in the Chicago Drainage Model(1),
inclusion of user-defined conduits in EXTRAN, and improvement of
the pumping calculations in EXTRAN to better simulate a system
with multiple pumps and variable suction and discharge bay
elevations.
The results of calibration studies and values selected for
those parameters affecting the timing and quantity of flow in
this metropolitan are also presented.
62
-------
BACKGROUND
The surface geography and physiography of southeastern
Louisiana reflect the different courses of the Mississippi River
during the Holocene. The river presently deposits some 500
million tons of sediment per year into the Gulf Coast Geosyncline
and similar or greater amounts in the past have produced vast
delta complexes across the region. The sediments consist
primarily of unconsolidated sand, silt and clay in a fining
upward sequence. The entire region is gradually subsiding as a
result of overburden pressures on over 10,000 feet of clastic
sediments.
Locally, the land surface slopes downward from natural
levees along the river and Lake Pontchartrain to inter-levee
basins. Soils are coarsest near the river, consisting of fine
sands and silts. The low-lying areas consist of clays and organic
deposits resulting from decay of swamp and marsh vegetation.
Development was originally confined to the higher relatively
stable natural levee, but gradually spread to lower areas which
had to be drained by pumping even for agricultural purposes.
Later construction of man-made levees along the lakefront and
river prevented the regular deposition of sediment by flooding.
This, coupled with consolidation and decay of organic material
which resulted from lowering the ground water table by drainage,
has produced continual subsidence within the metropolitan area.
Ground elevations range from 10 to 15 feet above MSL along the
river to 5 to 10 feet below MSL in the interior.
The original drainage system - which included most of the
canals now in use - was completed prior to 1900. Flow drained to
the central portion of the city from which it was pumped by a
series of paddle-type drainage machines into Bayou Bienvenue.
During the last 90 years the system has been continually
modified. Canals have been enlarged, lined and covered; the flow
pattern has been redirected to Lake Pontchartrain; and the
original paddle pumps have been replaced by vertical axial flow
or horizontal screw pumps up to 14 feet in diameter. Development
in Jefferson Parish occurred considerably later. The area was
largely rural until after WW II. The New Orleans pattern of
drainage, development, flooding, more drainage, more development,
more flooding has been repeated there and the entire metropolitan
area has suffered considerable flood damage in recent years.
On May 3, 1978 a particularly intense storm deposited
approximately 10 inches of rain in about six hours, 6 inches of
which fell within two hours. This storm produced street flooding
even in the highest areas near the river and flood waters
completely covered cars and nearly filled the ground level of
buildings in low-lying zones. This event and a number of lesser
63
-------
but still severe storms (8 inches in 4 hours) which occurred
within the next two years stimulated a series of studies of the
various drainage systems which, for the first time in New
Orleans, employed techniques truly capable of analyzing such
complicated networks. Among these have been several applications
of SWMM.
THE WEST BANK STUDY
In 1981 URS Engineers completed a master plan for drainage
of most of the West Bank of Jefferson Parish Louisiana(2). The
study area (Figure 1) covered some 36,000 acres of which about
10,000 acres had been developed for commercial, industrial or
residential use.
LAKE
PONTCHARTRAIN
J
A
A
17th STREET CANAL
DRAINAGE BASIN
JEFFERSON PARISH
WEST BANK
\)
Figure 1 - New Orleans Drainage Study Areas
This region is subdivided into eight drainage basins, ranging
from 600 to nearly 9000 acres in size, which are not completely
isolated hydraulically.
In the application of SWMM to the West Bank, the eight
drainage basins were assumed to be hydraulically independent,
with interbasin transfers occurring only through pumping. The
procedure employed the blocks RUNOFF, TRANSPORT and EXTRAN. Each
drainage basin was subdivided into subareas tributary to
64
-------
particular nodes of the distribution system. The average subarea
contained about 37 acres.
The then existing drainage system was never modeled since it
was clearly inadequate. Rather, a preliminary design was laid
out, generally following existing ditches and canals, and
analyzed using TRANSPORT in design mode. The resulting circular
conduits were converted into hydraulically equivalent open
channel sections and reanalyzed using EXTRAN with the design
being adjusted until it was capable of handling the design storm.
No calibration was attempted in this study. The design storm
was a ten-year 12-hour event. The default values which were
provided in SWMM Version II were used for Manning's n and
depression storage in RUNOFF. The final design was checked by
calculating a backwater profile with HEC-2(3) using the maximum
flow found in each element of the system. This procedure
confirmed that the sections chosen were capable of containing the
flows produced by the design rainfall provided adequate pumping
capacity was provided.
THE SEVENTEENTH STREET CANAL STUDY
The Seventeenth Street Canal is the outfall from New Orleans
Pumping Station No. 6 to Lake Pontchartrain. The canal lies along
the border between Orleans and Jefferson Parishes and the pumping
station, while primarily serving uptown New Orleans, also
receives some flow from Jefferson Parish {Figure 1). The New
Orleans Sewerage and Water Board had proposed increasing the
capacity of the pumping station and canal - an action which would
displace some residents of Jefferson Parish. In order to resolve
the issue of need for the project and evaluate alternative
solutions, an independent study was provided by the firm of
Linfield, Hunter and Gibbons, Inc. of New Orleans.
The Seventeenth Street Canal Study!4), completed at the end
of 1982, utilized the SWMM blocks RUNOFF and TRANSPORT and HEC-2.
In addition, "Standard Streets" similar to those used in the
Chicago Drainage Model(l) were employed to simplify and reduce
the calculations required in RUNOFF.
STANDARD STREETS
The total drainage area tributary to Pump Station No. 6 is
approximately 10,000 acres - all of which is developed. Streets
are generally parallel or perpendicular to the river. Those
streets which are parallel have negligible slope, while those
which are perpendicular slope away from the river at an average
of about 0.15%. Major drainage canals lead away from the river
and are fed by smaller conduits along the streets parallel to the
river. Development is relatively uniform, consisting chiefly of
single and two-family residences on small lots. There is little
65
-------
to distinguish one such street from another, hence standardized
"typical" streets were developed which permitted each individual
street in the area to be included in the analysis.
The Orleans Parish portion of the system was described using
six different standard streets, while seven such streets were
needed for Jefferson Parish. The several streets vary in length,
area, conduit sizes and hydraulic slope. Pipe sizes in each
Parish were based upon design criteria presently in use.
Hydraulic slopes were calculated based upon the assumption that
the water surface would be at street level at the most remote
inlet and six inches below the top of the intercepting canal.
The Standard Streets were developed to match existing
drainage patterns and tributary areas as closely as possible.
They represent very closely, but not exactly, the actual streets
within the basin. The total area of each sub-basin was matched
exactly by the total area of the Standard Streets which it
contained.
ANALYSIS OF THE SYSTEM
The Seventeenth Street Canal Drainage Basin was analyzed by
generating outflow hydrographs for each of the Standard Streets
and calling these hydrographs as often as necessary to completely
represent the various subbasins. Because of the limited number of
elements which can be represented in SWMM, the area was
subdivided into six subbasins ranging from 150 to 5500 acres. The
subbasins are not significantly different in development but do
have differing discharge conditions. Only the largest of the
basins, that tributary to Pumping Station No. 1, offered the
opportunity of calibrating the model.
Ca1 i b r a t i on
The model was calibrated by selecting storms which were
reasonably uniform throughout the city and which were not so
intense that surcharging or street flooding occurred. The storms
selected had net precipitation of about 2 inches in 5 hours.
The pumped hydrograph at Pumping Station No. 1 was developed
from stage-discharge relationships for the individual pumps, and
operational records and stage recordings at the station. Pumping
times were determined for each pump for the pump logs. For each
time period the average suction and discharge heads were
determined and the differential head calculated as the difference
between the two plus 1 foot. The additional foot accounts for
water surface depression between the location of the suction
basin recording gage and the pumps. This loss was measured during
actual pumping events during the study.
From the differential head and the pump curves the discharge
66
-------
of each pump during each time period was determined. The total
pumped hydrograph consisted of the sum of the individual pump
rates. Such pumped hydrographs were developed for the storms of
November 26, 1980 and April 22, 1979.
The model was calibrated by adjusting the various parameters
until the inflow hydrograph produced by TRANSPORT closely matched
the pumped hydrograph. For the storm of November 26, 1980, the
pumped volume differed from the calculated volume by less than 3
percent. The pumped hydrograph lagged the calculated inflow
hydrograph by 15 to 20 minutes. The calibrated model was then run
for the storm of April 22, 1979. The inflow and outflow
hydrograph volumes differed by about 1 percent and a lag similar
to that, in the calibration run was obtained. The model was judged
to be calibrated arid the values obtained (Table 1) were applied
throughout the basin.
TABLE 1. CALIBRATION VALUES FOR SUBCATCHMENTS - 17th STREET CANAL
STUDY
Parameter Calibrated Value
Percent Imperviousness 76 percent.
Ground Slope 0.015 ft/ft
Depression Storage
Impervious Area 0.01 inches
Pervious Area 0.10 inches
Infiltration Coefficients
Maximum Rate 1.50 in/hr
Minimum Rate 0.25 in/hr
Decay Rate 0.0011 /sec
Manning's n
Impervious Area 0.018
Pervious Area 0.200
D e s i g n and A n a1y s i s
Following calibration the model was run using a design storm
selected by the client. This storm delivers five inches of
rainfall in five hours and is nearly identical to a 10 year
event. The model showed many inadequacies in the system and was
then run in design mode. The circular conduits selected by SWMM
were replaced by hydraulically equivalent rectangular or
trapezoidal sections and the system was reanalyzed to insure that
no flooding occurred. A number of new canals were added to the
system during the design process. These were generally parallel
to the existing major structures. The final design was shown to
be capable of transporting the flows generated by the design
storm to the pumping stations without producing major flooding.
67
-------
MAJOR CANALS
The Seventeenth Street Canal Drainage Basin contains two
major canals - the Palmetto Canal and the 17th Street Canal. The
first carries the pumped flow from Pumping Station No. 1 to
Pumping Station No. 6 while intercepting gravity flow from most
of the rest of the study area. The second carries the pumped flow
from Pumping Station No. 6 into Lake Pontchartrain. Neither of
these canals could be satisfactorily modeled by SWMM since both
were crossed by numerous bridges and pipe lines which offered
obstructions to the flow and neither had a cross-section
corresponding to those available in the model. For these reasons
the major canals were modeled using HEC-2. The flows used were
the peak flows generated by SWMM although it was recognized that
the peaks from the several subbasins did not coincide in time.
The simulation with HEC-2 showed that the canals were
inadequate for the design storm. The Palmetto Canal was able to
be improved sufficiently to carry the peak flows by removal of
most of the minor bridges and other structures. The 17th Street
Canal was not so easily improved since the major obstacles to
flow consisted of a railroad bridge, three interstate highway
bridges, and two bridges on the major commercial highway in
Jefferson Parish. A further problem area existed within the last
1200 feet of the outfall canal where what had once been docks had
gradually become houses with fill placed around them and the
width of the canal had been reduced substantially. The analysis
with HEC-2 showed that the least expensive solution involved
deepening and widening the canal throughout its length except
immediately adjacent to the highway and railway bridges. This
alternative also required that the encroachment and structures at
the end of the outfall canal be removed.
THE ORLEANS PARISH MASTER PLAN FOR DRAINAGE IMPROVEMENTS
In late 1984 the firm of Daniel, Mann, Johnson and
Mendenhall completed a study(5) of the entire drainage system of
Orleans Parish - including the Cities of New Orleans and Algiers
(Figure l!. The total area of 98,000 acres was divided into 14
subareas ranging from 1450 to 29,800 acres. The two largest
subareas are presently undeveloped. The largest developed subarea
contained 9500 acres. The system contained 191 miles of major
canals, 89 miles of which were covered, and 16 pumping stations
with capacities ranging from 100 to 6350 cfs.
Developed areas of the Parish were generally modeled using
RUNOFF and EXTRAN. The procedure used in calibrating the model in
this instance varied somewhat from that in the 17th Street Canal
Study. Values similar, but not identical, to those in Table 1
were selected based upon a review of other studies. The model's
output was then adjusted to match pump station hydrographs by
varying characteristic width and percent imperviousness.
68
-------
Calculated peaks, total volumes, and time of peak flow generally
corresponded reasonably well to pumped hydrographs.
Simulation of storms in excess of the drainage system
capacity illustrated the deficiencies by showing discharges at
internal points. The inadequate sections were increased in
dimension and other modifications were made in the modeled system
until it was capable of conveying the flow to the pump stations
without flooding.
SOURCES OF ERROR
SWMM is a very useful tool for simulation of complex storm
drainage systems. For the system in New Orleans, however, it has
certain features which may lead to error. These include the need
to have relatively uniform conduit lengths in order to insure
stability of the solution at reasonable time step length, the
limited number of cross-sections available, and the relatively
simple simulation of pumping stations.
CondujLt Length
The time step in EXTRAN is generally controlled by the wave
celerity in the system. In order to insure a stable solution the
time step should not exceed the time required for a surface wave
to travel from one end to the other of a conduit. Since the
actual system consists of both long and short segments, the
celerity condition was satisfied by substituting for the short
conduits a longer section of lower frictional resistance which
was hydraulicaliy equivalent.
In a typical situation a 10 foot by 20 foot canal 300 feet
long with n=0.015 was replaced by a 10 foot by 20 foot section
900 feet long with n=0.0087. The two sections will carry the same
flow at the same head loss (assuming equal depths) but do not
have the same storage capacity. The false storage capacity
included in the model by this technique is bound to influence the
results - reducing peaks and depth of flow under some
circumst ances.
Cqnclu 11 Cross-Sect ions
EXTRAN supports six conduit shapes - including those most
commonly encountered in storm drainage systems. Unfortunately,
the New Orleans System includes a large number of cross-sections
which are neither trapezoidal nor rectangular nor, in places,
even symmetrical. These sections were simulated in the model by
hydraulicaliy equivalent rectangular or trapezoidal sections -
that is - by sections which had the same conveyance when full.
This procedure is satisfactory when the conduits are all full,
but results in different depths at intermediate flows and, in a
complex system, can result in incorrect distribution of the flow
69
-------
amongst the elements of the system.
Pump Simulation
EXTRAN permits simulation of on-line and off-line pumping
with a maximum of three pumping rates controlled by either wet
well volume or junction depth. The actual pump stations in the
New Orleans system contain multiple pumps and these pumps are
operated under varying conditions of suction and discharge head.
The "step" type pumping simulated by EXTRAN does not correspond
to the reality and introduces a computational instability which
does not exist in the real system.
MODIFICATIONS TO SWMM
Since 1985 the data describing the drainage system of the
city of New Orleans has been maintained on the IBM 3081 at Tulane
University. In an effort to improve the accuracy with which the
model can simulate the actual conditions in the city a number of
modifications have been devised and tested.
USER-DEFINED CONDUITS
A recent study(6)'has been reported in which the capability
of employing user-defined conduits was added to EXTRAN. The
modified version of EXTRAN will accept and run data sets prepared
for the standard version of SWMM but will also permit simulation
of cross sections of any shape whatsoever, open or closed. The
differences between the standard and modified versions are
particularly significant under less than full conditions. This
appears to be a useful improvement in the model.
PUMPING STATIONS WITH MULTIPLE PUMPS AND VARYING HEAD
The actual pumping stations in New Orleans may contain as
many as ten separate pumps. These commonly include large
horizontal Wood screw pumps, smaller vertical turbine pumps, and
one or more small constant duty pumps.
In operation during a runoff event the Wood screw pumps are
free-wheeled when it is clear that a substantial flow may occur.
As the depth in the suction basin begins to rise the vertical
pumps are started and vacuum priming of the horizontal pumps
begins. The large pumps are loaded sequentially by the operator
when he judges, based upon the rate of rise in the suction basin,
that the flow is sufficient to support their operation. The
vertical pumps are used primarily to recover pumping capacity
when a screw pump loses prime and between loading cycles on the
large pumps. The water elevation in the suction basin may vary by
10 to 15 feet during a major storm. Additionally, the elevation
on the discharge side may vary by as much as 10 feet on interior
canals and by up to 5 feet on outfall canals.
70
-------
In the last year a modification has been made in EXTRAN
which permits more accurate simulation of the actual pumping
operation. At present the version being tested is dimensioned for
up to six pumps per station. Three points are entered from the
characteristic curve for each pump. These are used to fit an
equation relating head and discharge. Starting suction basin
elevations are also specified for each pump. During a simulation
the pumps turn on sequentially as the depth in the suction basin
rises. Additionally, the individual pump discharges vary with the
differential head between the suction and discharge basins. The
effect of this improvement upon the calculated discharge
hydrograph may be seen in Figure 2. The impact of this smoothing
of the discharge hydrograph upon the tributary conduits has not
yet been fully evaluated for the New Orleans System.
1600 ..
jg 1200 ..
800 ..
400 .
0.0
MULTIPLE UNIT PUMP
STATION HYDROGRAPH
STANDARD PUMP
STATION HYDROGRAPH
0.0 0.8 1.6 2.4 3.2 4.0 4.8 5.6
CLOCK TIME (HOURS)
Figure 2 - Comparison of Pumped Hydrographs
The authors wish to acknowledge the assistance of Mr. Daniel
E. Rau in this project. The work described in this paper was not
funded by the U.S. Environmental Protection Agency and therefore
the contents do not necessarily reflect the views of the Agency
and no official endorsement should be inferred.
71
-------
REFERENCES
1. Tholin, A.L. and C.J. Kiefer Hydrology of Urban Runoff.
Journal Sanitary Engineering Division, Proceedings American
Society of Civil Engineers 85:SA-2, 1959.
2. URS Engineers West Bank Master Drainage Study, Volume 1
1981.
3. U.S. Army Corps of Engineers HEC-2 Water Surface Profiles,
User's Manual 1982.
4. Linfield, Hunter and Gibbons, Inc. Seventeenth Street Canal
Drainage Basin Study 1983.
5. Daniel, Mann, Johnson & Mendenhall Master Plan for Orleans
Parish Drainage Improvements 1984.
6. Yasenchak, Moira L. and Terence J. McGhee User-Defined
Conduits in the EXTRAN Block of SWMM Proceedings, Stormwater
and Water Quality Modelling Users Group Meeting 1988.
72
-------
USE OF SWMM/EXTRAN AND TR-20 TO DEVELOP REGIONAL STORMWATER
DETENTION PLANS IN THE WASHINGTON, D.C. REGION
by: Brian W. Mack, Thomas S. George, and John P. Hartigan
Camp Dresser & McKee
7535 Little River Turnpike
Suite 200
Annandale, Virginia 22003
ABSTRACT
The regional approach to stormwater detention is the current trend for
stormwater roaster plan development in the Washington metropolitan region.
This paper discusses the development of criteria for locating and designing
regional detention basins, and the modeling approaches used to maximize
regional detention benefits on a watershedwide basis.
Applications of stormwater models to develop a regional detention basin
master plan for Fairfax County, Virginia and a preliminary stormwater
management investigation for Montgomery County, Maryland are described.
Following the selection of regional detention basin sites and the
completion of conceptual designs the SWMM/EXTRAN model and the Soil
Conservation Service TR-20 model were used to determine the watershedwide
impacts of alternative detention systems. To assess regional benefits,
various locational schemes were analyzed for both county plans. The
Fairfax County plan included the design of maximum efficiency basins which
utilize lower maximum release rates to compensate for areas not controlled
by regional facilities.
The regional detention basin network, recommended in the Montgomery
County investigation, demonstrated the use of extended detention on top of
a permanent pool for water quality benefits. In several cases, in addition
to water quality benefits, this type of design reduced the post-development
2-year peak flows to levels less than pre-development conditions. The
TR-20 model was used to evaluate the watershedwide impacts of this type of
design. In addition, a PC graphics package was developed to illustrate the
watershedwide interactions of the routed TR-20 hydrographs.
INTRODUCTION
The regional approach to stormwater detention has many advantages over
the traditional onsite detention approach, including: increased
effectiveness; reduction in capital and maintenance costs; opportunities to
73
-------
manage existing as well as projected stormwater problems; opportunities to
provide water quality management as well as erosion and flood control
protection. This paper describes the use of stormwater models to develop a
regional detention basin master plan for Fairfax County, Virginia and a
preliminary stormwater management investigation for Montgomery County,
Maryland.
The plan for Fairfax County, Virginia involved identifying and
providing conceptual designs for up to 200 detention facilities serving
approximately 100 sq mi (259 sq km) in the rapidly urbanizing sections of
the County. The stormwater management investigation for Montgomery County,
Maryland included siting and developing conceptual designs for a maximum of
seven regional detention basins for each of the three study areas that
range between 4 sq mi (10.5 sq km) and 13 sq mi (34 sq km). The study
areas are growing regions of the County and are of interest to the
Department of Environmental Planning. The regional systems will provide
streambank erosion protection, flood control, and water quality benefits.
DELINIATION OF REGIONAL DETENTION SYSTEMS
The key to a successful regional stormwater detention system was a
comprehensive analysis of the watershed environment. Sites for regional
detention basins were selected based on a review of County maps and
reports. Maps include topographic, flood plain, wet land, property ID,
zoning, aerials, comprehensive plans, sanitary sewer maps, and hydrologic
factors within a watershed. Each of these factors governs the need for
stormwater controls and defines the physical constraints for siting and
designing stormwater detention facilities. The development of criteria for
locating and designing regional detention basins was the first step for
successful stormwater management.
Topography, drainage area (200 - 400 acres), soils, land development,
and critical environmental areas were of prime importance in locational
criteria; however, property access and adjoining land use are items that
were also be addressed. Consideration was also given to the size of the
detention basin. For Fairfax County, detention basins were initially
chosen with maximum dam depths less than 25 feet and maximum storage less
than 50 ac-ft, thus allowing them to be exempt from the permitting
requirements of the Virginia Dam Safety Program. Additional checks were
made to prevent detention basins from being located in the floodplain of
the main stem and in wetland areas. Because Fairfax County, Virginia and
Montgomery County, Maryland are rapidly developing areas in the Washington,
D.C. region, it was imperative in the development of the regional detention
system that County agencies provide their imput into the site selection
process. County agencies such as the Department of Public Works, the
Department of Environmental Management, and the Park Authority for Fairfax
County, and the Department of Environmental Protection, the Montgomery
County Soil Conservation Service, and the Maryland-National Capital Park
and Planning Commission for Montgomery County, met with Camp Dresser &
McKee in work sessions to evaluate each of the proposed sites for the
regional detention basins.
74
-------
The design criteria focus was on the particular type of detention
facility to be recommended at a site. Fairfax County facilities included
wet detention and extended dry detention with 10-year flood control and/or
2-year erosion control while Montgomery County used designs that had
extended detention in addition to a wet detention volume with 10-year flood
control and/or 2-year erosion control. The design depths used in Fairfax
County for the 2-year and 10-year storm events were 3.2 inches and 5.0
inches respectively. Montgomery County used 3.2 inches and 5.1 inches for
the same design storms. The SCS Type II 24 hour distribution was used for
both studies. Montgomery County was also interested in the storm events
less than the 2-year storm. Their philosophy is that these are the storms
that cause most of the channel erosion problems. The smaller storms
reviewed were the 2.6-, 2.0-, 1.5-, and 1.0 inch rainfall events.
The Fairfax County master plan developed wet detention basins in the
water supply Occoquan watershed and extended dry detention basins in the
remaining portions of the County. The wet and extended detention storage
requirements for Fairfax County were based on the land use upstream of the
detention basin. For wet detention the inches of storage range from 0.5
(undeveloped land) to 1.3 (high commercial) and for extended dry detention
the inches of storage range from 0.0 (undeveloped land) to 0.8 (highly
commercial). THe extended detention volume was designed to be released
over a 24-hour period. The current Montgomery County stormwater management
policy requires a permanent pool to have 0.5 inches of drainage area
storage and extended detention must have an additional 0.5 inches of
drainage area storage that is released over a 40-hour period.
A storage capacity check was performed to determine if the candidate
site was adequate to control the desired water quality and flooding for the
upstream drainage area under projected land use conditions. Based on the
best location of the dam for a regional detention basin, the available
storage was calculated by developing an elevation-storage relationship for
the site.
During the storage check for the Montgomery County, it was found that
in several instances the 2-year runoff volume was almost entirely
controlled by the required extended detention volume. Because of this, the
extended detention volume limited the design of the 2-year outlet and
provided release rates as low as 26 percent of the pre-developed flows. At
each site in Fairfax County, the maximum level of protection was checked
first to see if the available storage was sufficient for the required
storage, if not, then the next level of protection was tested. For
example, if an extended dry detention basin, which achieved both 2- and
10-year protection, could not fit at a particular site, then an extended
dry detention basin with just 2-year protection was evaluated. The amount
of storage required for wet detention, extended dry detention, and 2-year
and 10-year peak flow protection was determined from the drainage area to
the site and the percent imperviousness based on future land use.
The evaluation of required storage not only included the storage for
the types of detention basins as described above, but also included the
storage required for the passage of the emergency spillway design storm.
75
-------
The design storm, as given in the Fairfax County Public Facilities Manual,
is a function of the storage and height of the dam. Larger detention
basins are required to have emergency spillways which will pass larger
storms. For Montgomery County, the spillway design stormand regional
storage was based on soil conservation service procedures for a given pond
size.
Once the Fairfax County regional detention basin sites were selected
and conceptual designs were formulated, a SCS Hydrology model was linked
with the SWMM/EXTRAN model to simulate watershedwide hydrology and screen
the benefits of alternative regional detention basin sites. In addition,
the stormwater model package was used to evaluate the watershedwide benefit
of oversized detention basins.
The conceptual designs for Montgomery County were implemented into the
SCS TR-20 hydrologic/hydraulic model to maximize erosion protection, flood
protection, and water quality benefits. To help visualize the watershed
wide effects of the detention basins a graphics program was developed to
interactively display reach widths which represent the magnitude of the
simulated flows. Snapshots of the channel reaches, at selected time
intervals can be manually or automatically advanced on the computer screen
through a selected time duration. The graphics package has been set up and
run successfully on a microcomputer.
APPLICATION OF STORMWATER MODEL
EXTRAN is a link-node type hydraulic model used to simulate the Fairfax
County watershedwide benefits of conventional and maximum efficiency
regional detention basin designs. USGS channel cross-section data provided
the channel geometry used in the EXTRAN model. A SCS hydrologic model was
used to simulate the runoff hydrograph from each subbasin and also to route
through existing or proposed detention basins. The hydrologic model can
route flows through a facility by specifying a storage-discharge
relationship or by providing direct representation of outflow structure
geometry.
Traditional design criteria require that the post-development peak
discharge at a development site be reduced to pre-development levels for a
specified design storm. These performance standards typically result in
40-70 percent reductions in 2-year and 10-year peak flows for
post-development conditions below the regional detention basin. However,
applying these criteria on a regional scale in Fairfax County resulted in
insignificant watershedwide benefits due to runoff from existing and future
development in subwatersheds where regional detention systems were not
feasible. Because of this, the model was used to evaluate the benefits of
maximum efficiency regional detention basins designed to achieve lower
release rates by maximizing the use of available storage and thus
compensating for areas not controlled by regional detention facilities.
Recent evaluations of erosion control criteria in other areas (e.g.,
State of Maryland) have concluded that a peak release rate, based upon a
2-year predevelopment peak flow may not maintain post-development stream
76
-------
channel erosion at predevelopment levels It is being suggested that
release rates considerably less than the 2-year predevelopment peak flow
are required to prevent post-development increases in erosion. The peak
release rate (33% of the predevelopment peak flow), used for maximum
efficiency detention basins (2-year control) in the Fairfax County study,
is equivalent to 0.05-0.1 in/hr or less. This release rate is consistent
with some of the preliminary results of erosion control standard
evaluations carried out in other areas. Of course, it is not feasible to
achieve a peak release rate of 33% of the predevelopment peak flow for all
regional detention basin sites due to storage constraints. However, the
reduced release rates achieved at most maximum efficiency sites are still
preferable to conventional release rates from an erosion control
standpoint, and they offer the added advantage of compensating for flood
flows of uncontrolled areas.
Time of travel studies were also performed to evaluate the most
effective detention basin locations by analyzing the impacts of the
regional basins at various key locations within the watershed. The benefit
of an upstream detention basin on the peak flow at a downstream location is
a function of the timing of the detention basin outflow hydrograph, the '
timing of the downstream hydrograph peak at the location of interest, and
the time of travel associated with the distance from the detention basin to
the downstream location of interest.
Table 1 summarizes the watershedwide benefits of the maximum efficiency
regional detention basin system for one subwatershed within Fairfax County.
in the upper half of Difficult Run watershed (35 sq mi), 2-year storm peak
flows are summarized for future land use conditions without regional
detention basins, with 40 regional detention basins that have
pre-development peak release rates, and with about 20 regional detention
basins (maximum efficiency detention basins) that have less than
pre-development release rates (33 percent of pre-developed peak flows).
Table 1 shows the peak flows and percent reductions for these three cases
at six locations (nodes) within the watershed. The maximum peak flow
reductions occur for node 50140 on the Little Difficult Run tributary and
the minimum reduction in peak flow occurs further downstream at nodes 35000
and 40000.
Figure 1 shows the increase in 2-year peak-shaving benefits achieved by
smaller detention basin release rates which are 33 percent of the
pre-development peak flows. For example, at location 50140, the 2-year
post-development peak is reduced by 30 percent with pre-development peak
flow release rates and by 54 percent with less than pre-development release
rates. This provides an 80 percent increase in 2-year peak-shaving
benefits with the lower maximum release rates from the maximum efficiency
regional detention basins. The increase in peak-shaving benefits,
summarized in Figure 1, required only a 40% increase in capital costs to
oversize about 20 regional detention basins.
In each watershed in Fairfax County, the maximum number of detention
basin sites were selected based on the available storage and other site
constraints. The evaluation has shown that each detention basin provides
77
-------
Table 1. PEAK FLOW COMPARISON AT KEY NODES
DIFFICULT RUN. FAIRFAX COUNTY. VIRGINIA
Node
Difficult Run
35000
40000
50000
80000
74000
LKBa Difficult
50140
Without
Detention
Baavu
1.170
1,300
930
644
393
Run
4SS
Two- Year Storm
Peek Flow (da) lor Future Land Use
With Detention Baaina
Predevatopment
Relaaaa Ratea
1,090
1.220
747
560
296
319
H Lea
Reduction
7
6
20
13
25
30
a than Predev
Releaaa Rataa*
981
1,142
582
456
290
210
Reduction
18
12
37
29
26
54
'Apptod to ddwilion basui which could b* ovwtuid to handl* reduced rdMM ralet
128%
80%
10O%
Figure 1. INCREASE IN 2-YEAR PEAK-SHAVING BENEFITS ACHIEVED
BY SMALLER DETENTION BASIN RELEASE RATES
(33% of Pr«d«v«lopment Peak Flow)
78
-------
localized benefits immediately downstream. However, the watershedwide
impacts are largely a function of the total area controlled by the
detention basin, the number of maximum efficiency basins and the
distribution of the detention basins. The watershed evaluations have
demonstrated that the greatest benefits for downstream areas are produced
when several detention basins are clustered together in the upstream area
thus controlling more of the drainage area tributary to downstream
locations. Although they may control immediate downstream areas, a more
scattered detention basin network, especially along the main stem, cannot
produce the same level of benefits in the downstream areas as do detention
basins which are clustered together in the upstream areas.
CONCLUSIONS
The result of the Fairfax County master plan is a well tuned, regional
detention system designed to control runoff under existing and projected
land use conditions. The stormwater detention system includes a mixture of
conventional and maximum efficiency basins to provide the maximum benefits
countywide. By reducing maximum release rates to 33 percent of the 2-year
pre-development peak plow, post-development peak discharges (2-year and
10-year) are reduced to as much as 90 percent of the post-development peak
flows immediately below the detention facilities, and up to 65 percent at
downstream main stem locations. Therefore, a relatively small increase in
capital costs produces a very significant increase in watershedwide
benefits.
The Montgomery County preliminary stormwater investigation provided a
site location and conceptual design for the proposed regional detention
basins in each study area. The detention basins provide erosion
protection, water quality benefits, and flood protection benefits. The
extended detention storage requirement for the regional detention basins
provided, in most cases, 2-year post-development release rates below the
flows generated from the 2-year storm under pre-development land use
conditions. This resulted in velocity reductions immediately below the
detention basins to a lower level than what would be obtained by setting
the 2-year post-development release rate equal to the 2-year pre-
development flow. This supports the County philosophy that storms below
the 2-year event are the greatest source of erosion problems and that the
regional detention basins should protect the downstream reaches from the
smaller storm events.
REFERENCES
1. Camp Dresser & McKee, 1988. "Regional Stormwater Management Plan."
Prepared for Department of Public Works, Fairfax County, Virginia.
2. Camp Dresser & McKee, 1988. "Preliminary Stormwater Management
Investigation for Clarksburg Study Area." Prepared for Montgomery
County Government Department of Environmental Protection, Rockville,
Maryland.
79
-------
3. Hartigan, J.P. and George, T.S., 1988. "Use of Stormwater Models to
Optimize the Performance of a Regional Stormwater Detention System."
Paper presented at the 15th Annual Water Resources Conference, Norfolk,
Virginia.
80
-------
II TO EXPIDRE WASTEIDAD AT-TQCATION
by:
Angelo S. Liberti, Rhode Island Department of Environmental
Management, 291 Promenade St., Providence, RI 02908
Raymond M. Wright, Ph.D., P.E. Department of Civil and Environmental
Engineering, University of Rhode Island, Kingston, RI
Kevin Scott, Metcalf and Eddy, Inc., 10 Harvard Mill Square,
Wakefield, MA 01880
ABSTRACT
The Pawtuxet River flows through heavily urbanized areas, receives
effluent from 3 municipal WWTFs and one industrial WWTF, and has summer DO
concentrations well below the 5.0 rog/1 standard for much of its length.
Data collected during three 48 hour sampling surveys was used to calibrate
and validate QUAIr-II by Scott and Wright (1). Ihe Water Quality Branch
staff at EPA Region I and the modeling group at EPA, Atlanta, Georgia,
reviewed the model calibration and validation and noted that although BOD
decay and nitrification rates were calculated from field data, they varied
greatly among surveys for a given reach, and between adjacent reaches.
To provide a more defendable wasteload allocation (WIA), the model
was recalibrated and revalidated using one set of decay rates which
successfully predicted the field data. The data used to revalidate the
model was collected during a flow profile very close to the 7Q10 flow and
alternative WIA strategies were explored using this model.
Flow augmentation, instream aeration, increasing the number of
outfalls, and advanced treatment (AT) simulations indicated that discharge
limits of BOD 10, NH-j 2 mg/1 are required to attain the instream DO
criteria. Seasonal limits were developed using monthly USGS flow and
temperature data. When simulating AT, effluent DO concentrations were set
to 6.0 mg/1 and instream BOD decay and SOD rates were reduced.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official
endorsement should be inferred.
81
-------
INTRODUCTION
The Pawtuxet River basin is located in west - central Rhode Island and
encompasses 230 square miles of forest, open and urban land. The Pawtuxet
River is formed by the junction of the 6.8 mile North Branch and the 9.0
mile South Branch. The North Branch originates at Gainer's Dam which
itrpounds the Scituate Reservoir (a major drinking water supply) and the
South Branch originates at the Flat River Reservoir. The main stem of the
river flows 10.9 miles and discharges into the Pawtuxet Cove on
Narragansett Bay. The river receives discharges from five major point
sources (Hoechst Celanese Corporation on the south branch, Original
Bradford Soap, West Warwick, Warwick, and Cranston publicly owned
treatment works on the main stem), as well as runoff from roadways, urban
areas, and a landfill. The West Warwick, Warwick and Cranston treatment
facilities use the conventional activated sludge process and have average
daily flows of 4.2, 3.5, 12 M3D. Historical data indicates that summer
dissolved oxygen (DO) concentrations are well below the 5.0 mg/1 standard
for much of the main stem's length. The close proximity of the 3
municipal treatment facilities is illustrated in Figure 1.
Figure 1. The Pawtuxet River basin
The purpose of this investigation is to develop the best technically
sound and legally defendable scenario to distribute the waste assimilative
capacity of the Pawtuxet River among the municipal and industrial
dischargers in such a manner as to attain the minimum DO criteria of 5.0
mg/1 during the 7Q10 flow. Computer modeling of dissolved oxygen dynamics
provides the opportunity to evaluate the effect of various pollution
control strategies on instream DO and therefore, is an integral part of
waste load allocation and discharge permit development.
82
-------
The SEMCOG version of the Qual II model validated by Scott and Wright
(1), was based on three 48 hour sampling surveys conducted in the summer
of 1985. In addition, the sampling survey on July 30 - 31, 1985 was
conducted when the river flow was very close to the 7Q10 flow and is used
as the basis for the waste load allocation process. Numerous preliminary
simulations have shown that the discharge of wastewater from Hoechst
Celanese and Original Bradford Soap have an insignificant impact on the DO
concentration of the main stem of the Pawtuxet River and for this reason
are not included as part of this analysis.
The objectives of this investigation include:
1) Recalibration and revalidation of the steady - state dissolved oxygen
model, Qual II developed by Scott and Wright (1).
2) Development of permit limits for the West Warwick, Warwick and Cranston
municipal wastewater treatment plants which will allow the Pawtuxet River
to attain the instream dissolved oxygen concentration standard of 5.0
rag/1.
MONITORING DATA
As mentioned earlier, field surveys were conducted to gather the data
necessary to define model input parameters and to calibrate and validate
the models. Velocity-flow and depth-flow relationships were developed
from previous dye studies as well as those conducted by Wright and
McCarthy (2), and Scott and Wright (l). Water quality data was collected
from the five point sources and from instream water quality stations,
(WQS). The flows for the point sources on the survey dates were taken
from discharge monitoring reports (DMR) submitted to DEM by the
facilities. The river flow profile was determined from the USGS gages in
Coventry and Cranston. Groundwater inflow rates were estimated from the
river flow and point source flow data.
Field surveys were conducted on June 5-6, July 10-11, and July 30-31,
1985. The first and third surveys ran less than 48 hours due to rain.
The 16 WQS were sampled every 4 hours for; DO, temperature, conductivity,
pH, BOD5/ NH3, NO2, NO-j, TDS, Ortho-Phosphorus, and Cl~. Ortho
Phosphorus, pH, and Cl were not modeled in this analysis. The water
quality parameters required for the point sources were the same as for the
WQSs. This data, however, was obtained from DMR reports.
RECALIBRATION AND REVALIDATION SUMMARY
The Water Quality Branch staff at EPA region I and the modeling group
at EPA, Atlanta Georgia reviewed the model calibration and validation and
noted that although BOD decay and nitrification rates were calculated from
field data, they varied greatly among surveys for a given reach, and
between adjacent reaches. In an effort to provide a more defendable
83
-------
wasteload allocation (WIA), the model was recalibrated and revalidated
using one set of BOD decay, nitrification, and SOD rates which
successfully predicted the field data. The key components contributing to
the ability to defend the selection of input parameters and the model's
success are outlined below.
1) One set of consistent decay rates (K^, K~) was used for all 3
field surveys. Decay of BOD and NH^ is the result of bacterial
activity and should not vary significantly between surveys or
locations, without justification.
2) SOD rates were the last parameter input to the first model and were
used to tune the DO predictions. Model inputs of SOD were adjusted
within the range of the field data and their magnitudes and
locations are supported by the fact that when input to the other 2
applications DO concentrations were successfully predicted.
3) Model predictions of water quality parameters track the observed
trends.
4) The predicted DO profiles fell within the observed range with only
minor exceptions. These were in the range of 0.0 to 0.7 mg/1.
Figure 2 illustrates the observed and predicted DO concentrations.
45678 123 9 10 II 12 13 14 15 16
10 0
80
6.0
4.0
2.0
a o
*- 10.0
z
g 80
X 60
0
O 40
LU
>
^ 2.0
O
OT
CO 0
0
10.0
8.0
60
4.0
2.0
n
1 — i — n — T
SOUTH BRANCH
- MODEL PREDICTION
r MAXIMUM
4 AVERAGE
I MINIMUM
-
JUNE S, 1985
• 1 — r— n — r
-
SI
* i JT *
*
_
_
JULY 10, 1985
1 i n — r
-
t-? i
-
i
r—
_
JULY 30, 1985
ill
— • 1 r T
NORTH BRANCH
V
1 1 !
T
i
I"" — ^\
$
1 — n —
^A
I
1 1 i
II 1 1 lit
PAWTUXET RIVER
jT^^r
1 TV y T
•• — »—
i
1 1 i i it
T
1 T
^Nv
i\.
Tj
•i i
1 T^M^^I~J
-^- -^- T J
J-
-n — r- -n i M
fev
i\
ji
^i i
i i i i j.
6420
DISTANCE FROM CONFLUENCE (mi) DISTANCE FROM
PAWTUXET COVE (mi)
Figure 2. Recalibrated and revalidated model dissolved oxygen
predictions.
84
-------
WASTE IDAD ALIOCATION
The objective of the Pawtuxet River WIA is to maximize the vise of the
river's waste assimilative capacity, while protecting the river's
designated use. Hie main stem of the Pawtuxet River is classified as a
warm water fishery (Class C), and as a result is subject to an instream DO
criteria of 5.0 mg/1 and an Ammonia (NH-j), criteria of 1.7 mg/1. By
comparing the predicted concentrations to water quality criteria, an
appropriate pollution abatement strategy and subsequent water quality
based discharge permit limits can be developed. Ihe results of computer
simulations used to assess the impact of a variety of pollution control
strategies are outlined below.
The MA must be performed under the following conditions as mandated
by, RI Water Quality Regulations, RI Pollution Discharge Elimination
Regulations (RIPDES), and accepted practices of RIDEM and USEPA:
1) the 7Q10 flow of the receiving water.
2) discharge facilities design flow.
3) the groundwater inflow was recalculated using the 7Q10 flows at the
Coventry and Cranston USGS gaging stations and average point source
flows.
4) the Scituate Reservoir Release was set to 9 MGD (as specified in a
1920' s Riparian rights agreement), and the Flat River Reservoir Release
was calculated using flows measured at the Coventry gage and the
incremental inflow.
POLLUTION CONTROL STRATEGIES
Instream aeration
Instream aeration was simulated with WWTF limits of BOD = 30 mg/1 and
NH3 = 10 mg/1 by increasing the reaeration rate at the first point where
the DO dropped below 5.0 rog/1 and then the model was re-run to determine
the new location where the River violated the DO criteria. These steps
were repeated until the entire river attained a DO of 5.0 mg/1. As a
result of the above theoretical analysis it was concluded that 6 instream
aerators would be required to meet the DO criteria.
Although instream aeration appears to be a theoretical solution to
low DO levels in the main stem of the Pawtuxet River, the Clean Water Act
of 1987 mandates that best available technologies must be employed before
alternative technologies (such as instream aeration), may be used. For
this reason instream aeration is not an acceptable pollution control
strategy and the results of the model simulations are not included in this
paper.
85
-------
Flow augmentation
Flow augmentation was modeled at effluent concentrations of BOD 30
mg/1 and NH3 20 mg/1, (limits which officials from the 3 WWTF have
indicated can be met by their present facilities), BOD 30 and NH3 10
mg/1 and also at BOD 20 and NH3 10 mg/1. Flow augmentation may Be
available from the Scituate Reservoir or from the proposed Big River
Reservoir (which would encompass Flat River Reservoir (FRR)). For the
purpose of this analysis the flow from FRR was increased until the DO
level throughout the Pawtuxet River reached 5.0 mg/1. At treatment limits
of BOD 30, NH3 20 mg/1, FRR release must be increased from 11.18 to 240
cfs raise the DO to 5.0 mg/1. Treatment limits of BOD 30, NH3 10 mg/1,
require a FRR release of 190 cfs to raise the minimum DO from 1.1 to 5.0
mg/1. The flow from FRR must be increased to 175 cfs to achieve a minimum
DO of 5.0 mg/1 with treatment limits of BOD 20, NH3 10 mg/1. As noted
above, unrealistically large volumes of flow augmentation would be
required to meet the minimum DO criteria.
Although flow augmentation is not a realistic alternative to advanced
treatment, it does have a positive impact on DO concentrations in the Main
Stem during low flow periods. The possible construction of the proposed
Big River Reservoir provides the unique opportunity of modifying the
reservoir design to incorporate flow augmentation capabilities and avoid
competition with water supply needs. It should also be noted that
construction of Big River Reservoir could also reduce flow to the Pawtuxet
River and decrease water quality if its effect on the River is not
addressed.
Advanced Treatment
The steps taken to assess the impact of various point source loadings
on the DO profile are presented below. The issues are common to all
regulatory agencies performing WIAs. For these simulations the effluent
DO concentration was increased to 6 mg/1 at all WWTF and the BOD decay
rate (Kd) was set to 0.23 day'1 in all reaches.
The effluent DO level was set to 6.0 mg/1 since preliminary
simulations indicated that simply mixing low DO effluents with higher DO
river water caused a marked decrease in DO (41% of the total river flow,
at the Cranston WWTF outfall, is effluent, based on WIA flow rates.
How should BOD decay rates change in response to AT?
The BOD decay rate was decreased from 0.35 to 0.23 day"1 in all
reaches downstream of the West Warwick WWTF discharge to correspond to
advanced treatment. Thomann (3), and Leo et. al. (4), have reported that
high levels of sewage treatment leave only refractory materials in the
effluent which are difficult to degrade, and result in lower stream
86
-------
oxidation rates. A post audit of 6 rivers where WWTF were upgraded to
advanced treatment revealed that BOD decay rates were reduced in 3 of the
rivers and remained the same in the other 3, (4). The average decline in
Kd, in the above post audit analysis, was 60%, (4). In this analysis Kd
was decreased 34%, from 0.35 to 0.23 day"1.
The first set of simulations was run to assess the impact of various
levels of WWTF effluent concentrations. The 4 levels of effluent
concentrations which were simulated are listed below.
(mg/I) (mg/1)
30.0 10.0
20.0 10.0
15.0 3.0
10.0 2.0
How will SOD rates chancre after implementation of AT?
Simulations were performed to evaluate the effect of completely
removing (dredging) and reducing SOD to background levels (the rate
measured in clean streams by Butts and Evans (5)). Since dredging would
be difficult to carry out and would suspend toxics present in the
sediment, it doesn't appear to be feasible. In addition, advanced
treatment with limits lower than BOD 20, NH3 10 would be required to
meet the DO criteria even with the complete removal of SOD. If the
organic matter present in the sediment is gradually decomposed and
additional inputs of settleable organics from the WWTFs are eliminated,
SOD rates could return to those measured in clean streams. If SOD is
reduced to background rates, then WWTF limits of BOD 15, NH3 3 would be
required to meet the DO standard.
After further consideration, it was determined that if advanced
treatment was necessary it would only be required seasonally. If seasonal
advanced treatment limits are imposed it is no longer logical to expect
that SOD will return completely to background rates. For this reason the
decrease in SOD from the recalibrated (present) values to background rates
was weighted for seasonal treatment using the formula below:
= Background SOD + (Present SOD - Background SOD) X (0.75)
0.75 - is the percentage of the year that SOD reduction is not
anticipated (when advanced treatment is not required)
To assess the impact of seasonal advanced treatment, simulations were
run with the weighted background SOD at WWTF limits of BOD 15, NH3 3 and
BOD 10, NH3 2. Figures 3 and 4 show the predicted DO profiles with SOD
at recalibrated (present), background, and weighted background rates, at
WWTF limts of BOD 15, NH3 3 and BOD 10, NH3 2, respectively.
87
-------
SOD RATES
Background
--•- Weighted Background
— Recalibrated
SOD RATES
Background
Weighted Background
— Recalibrated
DO Standard
Distance from Pawtuxet Cove (miles)
10 987654321
Distance from Pawtuxet Cove (miles)
Figure 3. Predicted DO Profiles at
WWTF limits of BOD 15,
NH3-N 3 mg/1.
Figure 4. Predicted DO profiles at
WWTF limits of BOD 10,
NH3-N 2 mg/1.
Hie rest of the simulations presented in this section were run to
test the conclusion that WWTF effluent limits of; BOD 10, NH3 2, and DO
6.0 mg/1 would enable the Main Stem to reach the DO criteria of 5.0 mg/1
using weighted background SOD rates and K^ 0.23 day'1 under 7Q10
conditions.
Are effluent DO limits necessary?
It was noted earlier that all advanced treatment simulations were run
with WWTF effluent DO concentrations of 6.0 mg/1. Increasing the WWTF
effluent DO concentration from 3.0 to 6.0 mg/1 provides an additional 0.8
mg/1 and is necessary in order to meet the instream DO criteria.
Is maintenance of a minimum release from Scituate Reservoir crucial?
Throughout this analysis the Scituate Reservoir release was set to a
historical minimum of 9 MGD, however, since termination of hydropower
generation at Gainer's Dam in 1984, this rate has occasionally been
reduced to 0.0 MGD. At this high level of treatment, reducing Scituate's
release to 0.0 MGD causes a maximum decrease in DO of 0.3 mg/1. It should
be noted that the negative impact of no release from Scituate Reservoir on
the Main Stem DO concentrations would be magnified at lower levels of
treatment and higher SOD and K^ rates (present conditions).
88
-------
Can higher limits be set through the use of additional outfalls?
To make better vise of the River's waste assimilative capacity, the
discharge of each WWTF was equally divided into 3 separate outfalls. The
9 outfalls were all equally spaced along the main stem of the River and
model simulations of WWTF limits of BOD 30, NH3 10, and BOD 20, NH-,
10, predicted that the instream DO concentration violated the standard.
Simulations indicate that with this outfall design and treatment levels
of BOD 20, NH-, 10 mg/1, a FRR release of 90.0 cfs would be required to
raise the minimum DO from 3.2 to 5.0 mg/1.
Is the discharge of effluent from one regional AT facility acceptable?
In addition, a simulation was run to explore the water quality impact
of piping secondary effluent from the West Warwick and Warwick WWTFs to a
single advanced treatment facility at the current Cranston WWTF site.
Using a single outfall pipe at this location would require effluent limits
of BOD 10, NH3 2 mg/1, to reach the minimum DO criteria.
How will the seasonal limits be developed?
After exploring several pollution control strategies it is apparent
that advanced treatment is required to enable the Pawtuxet River to attain
the DO standard of 5.0 mg/1. To effectively utilize the waste
assimilative capacity of the Pawtuxet River, seasonal advanced treatment
permits were developed for the West Warwick, Warwick, and Cranston WWTFs.
To evaluate the water quality impact of seasonal permit limits, a low
flow must be determined for each month, and can be calculated using data
collected by USGS at the Cranston Gage. Two methods were used to
determine the monthly low flow rates. For the first method the lower 90%
confidence limit of the average flow for each month was calculated. For
the second method, a 7Q10 flow was determined for each month using the
average daily flows. This method was analyzed in order to be consistent
with the concept of the annual 7Q10 flow. The monthly 7Q10 is the minimum
average 7 consecutive day flow for a given month with a return frequency
of once in 10 years. The monthly low flow rates calculated by these 2
methods were very similar and resulted in identical monthly permit limits.
A second factor which varies seasonally and affects a stream's waste
assimilative capacity of BOD and NH3, is instream temperature. As
opposed to flow, the high monthly temperature is of concern since it
promotes biological decay and results in lower instream DO levels.
Therefore, to simulate the worst case monthly conditions, the upper 90%
confidence limit for temperature was input along with the monthly low
flow.
89
-------
Can total oxygen demand limits allow for variations in BOD & NH3?
At the request of the cxmrnunities the use of total oxygen demand
(TOD) limits was explored. TOD limits allow the concentration of BOD and
NH-j to vary while maintaining a constant instream DO concentration. The
use of TOD limits allows only slight variations in effluent BOD / NH3
concentrations (BOD from 10-1 mg/1 with NH3 2.0-4.8 mg/1). There is
little advantage to using TOD limits since a 1 mg/1 increase in the NH3
limit requires a 3 mg/1 decrease in the BOD limit.
The monthly RIPDES permit limits developed for the West Warwick,
Warwick, and Cranston WWTFs as a result of the conventional pollutants WIA
are listed in Table 5.
TABLE 1. Final Daily Maximum, Average Weekly and Average Monthly
BOD, NH3 and DO Limits (mg/1)
July
June
Oct.
Nov.
Date
1 - Sept. 30
1 - June 30,
1 - Oct. 31
1 - May 3
Parameter
BOD5
NH3-N
DO
BOD5
NH3-N
DO
BOD5
NH3-N
DO
Permit
Average
Monthly
10.0
2.0
6.0
15.0
3.0
6.0
30.0
NR
NR
limits
Average
Weekly
10.0
2.0
(minimum)
15.0
3.0
(minimum)
45.0
NR
NR
Daily
Maximum
15.0
3.0
20.0
5.0
5O.O
NR
NR
NR - effluent limits are not required
RECOMMENDATIONS AND CONSIDERATIONS
Based on the simulations made in this analysis the only realistic
option to ensure that the main stem of the Pawtuxet River reaches its low
flow DO and NH3 criteria is to require seasonal, effluent limits of BOD
10 mg/1 and NH3 2 mg/1 at the 3 WWTFs.
As with any water quality modeling exercise, this analysis could
benefit from additional monitoring data to further support the selection
of input parameters. Decay rates could be better defined if river reaches
were divided around the WWTF discharges. These changes in the sampling
90
-------
and modeling procedures may serve to reduce the original variation in K^
and Kn rates that were calculated by Scott and Wright (1), and to
support the revalidated model. As part of this analysis BOD decay and SOD
rates were reduced to levels measured in clean streams and the simulations
revealed that even when using these rates advanced treatment was required.
Therefore, it is not likely that additional sampling will result in new
input parameters which predict that advanced treatment is not required.
Ihe cost of upgrading one existing WWTF to a regional advanced
treatment facility which would accept effluent from the other 2 secondary
WWTFs, should be compared with the cost of 3 separate advanced treatment
facilities. Model simulations should be run to ensure that instream water
quality criteria will not be violated if a single discharge point is used.
If nitrification is used to reduce the effluent ammonia level the
nitrate concentration in the effluents will rise. Nitrates are considered
to be the limiting nutrient in salt water bodies and the impact of nitrate
loading to Pawtuxet Cove should be considered. If little nitrification
takes place in the River then denitrification of WWTF effluents may be
necessary to avoid algae blooms in Pawtuxet Cove.
It may also be prudent to determine if phosphorous removal should be
required at this time. Although alga productivity in the River is
currently low, industrial pretreatment, coupled with advanced treatment at
the WWTF may remove an unknown parameter which is currently suppressing
macrophyte growth.
REFERENCES
1. Scott, K. and Wright, R.M. Modeling Dissolved Oxygen in Transient
Flow Conditions. Unpublished Draft Report, 1987.
2. Wright, R.M. and McCarthy, B.J. A study of the water quality of the
Pawtuxet River: Chemical monitoring and computer modeling of
pollutants, Volume 2: Computer modeling of toxic pollutants in the
Pawtuxet River, 1985. 173 pp.
3. Thomann, R.V. and Mueller, J.A. Dissolved oxygen sources and sinks of
DO-Kinetic relationships. IN; Principles of Surface Water Quality
Modeling and Control. Harper and Row, Publishers, Inc., Cambridge,
1987. p. 261.
4. Leo, W.M., Thomann R.V. and Gallagher T.W. , Before and after case
studies: comparisons of water quality following municipal treatment
plant improvements. EPA 430/9-007, U.S. Environmental Protection
Agency, Washington, D.C., 1984. 183 pp.
5. Butts, T. and Evans R. Sediment oxygen demand studies of selected
Illinois streams, Circular 129, Illinois State Water Survey, Urbana,
Illinois, 1978. 177 pp.
91
-------
FREQUENCY ANALYSIS
OF
TRACE LEVEL WATER QUALITY DATA
WITH A
TIME VARYING CENSORING LEVEL
by: S. Rocky Durrans, P.E.
Project Engineer, Merrick & Company, Denver,
Colorado 80222; and, Graduate Student, Dept.
of Civil Engineering, University of Colorado,
Boulder, Colorado 80309
ABSTRACT
Databases representing measurements of trace level contaminant concen-
trations are often censored by a constraint on the range over which measure-
ments may be made. Technology is such that there is often a detection limit
below which contaminant concentrations can not be measured and, resultingly,
databases can be proliferated with entries such as "undetectable" or "less
than detection limit." A number of recent studies have addressed the fre-
quency analysis of censored data sets but have all been limited to the spe-
cial case where the detection limit is constant over time. One can expect,
however, that technological advancements will result in decreasing censoring
levels over time. Therefore, techniques should be investigated for the case
of a time varying censoring level. This paper addresses this more general
problem and compares two methods which may be employed for parameter estima-
tion when there is a number of discrete, well defined censoring levels. It
is argued that the method of maximum likelihood should be the method of
choice.
92
-------
INTRODUCTION
Censored data. What is it? Where does it come from? An how may it be
treated? These are not trivial questions, nor are they of merely academic
interest. Indeed, the establishment of water quality standards by regulatory
bodies such as the U.S. Environmental Protection Agency is often based upon
statistical analyses using data which might be highly censored. Society's
increased awareness of the effects of various contaminants has manifested
itself in demands for cleaner and cleaner water and there is therefore a need
to be able to address databases which represent contaminant concentration
levels which may be technologically difficult to measure.
A problem which is becoming increasingly encountered with water quality
data involves the measurement of extremely low contaminant concentrations
where there is a detection limit (censoring level) that is governed by tech-
nological measurement constraints. Previous studies have presented analytical
techniques for the case of a time invariant censoring level; however, as time
progresses and technological advancements are made, the problem will be com-
plicated in that detection limits will tend to move downward. Databases which
are maintained over significant periods of time then will be subject to a
number of discrete, well defined censoring levels and thus the potential util-
ity of a method for the treatment of time varying censoring levels should be
fairly obvious. The proper (or improper) statistical treatment of such data-
bases for decision-making purposes might have significant long term social,
economic and/or environmental impacts.
This paper presents an examination of the third question raised above
and compares two alternative estimation techniques which may be utilized in
frequency analyses of data exhibiting a time varying censoring level. To the
writer's knowledge, this paper represents the first consideration of this
generalized case. Before proceeding however with the more technical details,
types of problems which might be encountered in practice are briefly dis-
cussed. In so doing the first two questions posed may also be answered, at
least to the extent necessary to support one's use of the methods evaluated
later.
Techniques which are presented in the following pages are intended to be
applied to the lognormal distribution; but, because of its similarity, may
also be applied to the normal distribution. Alterations to the equations
must be made if one wishes to address distributions other than he normal or
lognormal. While any one of a number of distributions could ha^e been con-
sidered in the following discussions, the lognormal was selected for three
reasons. First, it has a lower bound of zero and is thus consistent with the
range of possible contaminant concentrations; second, it has been found by
McCarty and Reinhard (1), Hashimoto and Trussell (2), and Gilliom and Helsel
(3,4) to be an acceptable model in many instances; and third, its genesis
lies in the central limit theorem, coupled with the hype.nesis of multiplica-
tive effects, and it thus has an undeniably sound theoretical basis.
93
-------
PROBLEM TYPES
As implied earlier, there are a number of types of problems that one may
actually encounter in practice. It seems appropriate that at least a brief
discussion of the various types should be undertaken before proceeding. While
such will serve the immediate purpose of answering the first two questions
posed, it will also clarify the distinctions between truncated data and vari-
ous types of censored data. The following few paragraphs define these data
types and should assist the reader in deciding whether the techniques presen-
ted later are applicable to a problem at hand.
As an example of a truncated data set, consider the following problem
which has been adapted from Kendall and Stuart (5): Arrows are shot at a cir-
cular target and it is desired to estimate the distribution of distances from
the center of the target to the points where the arrows strike. Some arrows
however miss the target entirely. Furthermore, it is not even known how many
arrows were shot. While the true distribution of distances must include the
possibility of distances greater than the target radius, only those distances
which are less than the target radius can be measured and utilized for para-
meter estimation purposes. Also, since there is no knowledge of how many
arrows missed the target, not even relative weights may be assigned to the
available, non-truncated measurements. Problems of this type occur rather
frequently but are seldomly recognized as truncated data problems.
In contrast with truncated data is what is known as censored data. This
latter categorization is also subdivided into two types of censoring: Type I
and Type II. Returning to the arrow and target example, in Type I censoring
one knows how many arrows missed the target but does not know by how much.
The fact however that the number of misses is known clearly differentiates
this data type from that of truncation and also provides valuable information
for parameter estimation purposes. The censoring level is also known and, in
this example, is equal to the target radius. Type II censoring occurs fre-
quently in reliability testing such as might be performed to define the dis-
tribution of say the lifetime of an electronic component. One might take, for
example, a sample of 100 components and decide to test them simultaneously
until 80 have failed. In this case, like that of Type I censoring, the number
of non-quantifiable data entries is known (20 in this example); but, unlike
Type I censoring, the censoring limit is not known a priori. In the case of
Type I censoring then the censoring level is known and the number of non-
quantifiable data entries is a random variable; in the case of Type II censor-
ing, the number of non-quantifiable data entries is known and the censoring
level is a random variable.
The most commonly encountered problem type in the context of water qual-
ity data involves Type I censoring where the censoring is from below. What
is meant by this is that quantifiable values are all above some well defined
censoring level (detection limit) while the non-quantifiable, censored data
entries are all below the censoring level. The example given earlier of the
arrows and target represents a case of censoring (or truncation) from above.
Either censoring from above or censoring from below, or both, may occur in a
94
-------
given problem.
The preceeding paragraphs have answered the first two questions posed by
defining censored data and explaining how and why it occurs. The remainder of
this paper may now be devoted to a discussion of methods which may be utilized
in the statistical treatment of censored data sets.
STATISTICAL METHODS
A study of the literature reveals that there are two methods that are
particularly attractive for parameter estimation purposes for the lognormal
distribution when censored data must be used. The first of these involves the
use of a plotting position formula and, subsequently, a least squares linear
regression procedure. The second technique is known as the method of maximum
likelihood and is theoretically much more rigorous in nature. Previous stu-
dies have indicated that these two methods are essentially comparable to one
another in terms of their results; but, because of its relative simplicity,
the regression technique is probably more frequently applied. It should be
noted however that previous works have all addressed the special case of a
time invariant censoring level. As will be shown shortly, the method of maxi-
mum likelihood exhibits a distinct advantage in the more general case of a
time varying censoring level.
Figure 1 provides a graphical portrayal of a probability density function
curve and may help to clarify the following discussions. The area under the
curve has been subdivided into the three regions A, B and C by the two censor-
1
Figure 1. Subdivision of population distribution
into regions by discrete censoring levels.
95
-------
ing levels DLi and DL~. Only one of these censoring levels is in effect at
any given time and is replaced by the other when an external effect such as a
technological advancement occurs. The fact that only one censoring level
change (from DLj to DL.2) is illustrated should not be construed to reflect
that the following discussions are limited to this particular case. Indeed,
equations which are presented later are general enough to permit the consider-
ation of any number of changes. It should also be noted that the following
procedures are applicable whether the detection limit decreases in time, in-
creases in time or, in a series of changes, randomly oscillates between in-
creases and decreases.
The situation that will be most frequently encountered with water qual-
ity data will be that of a censoring level that decreases in time. Thus,
prior to some technological advancement, and referring to Figure 1, the detec-
tion limit will be equal to DL^ and observed concentrations in Region C only
will be able to be quantified. Concentration levels falling within Regions A
and B will be censored. Now, supposing that a technological advancement
occurs, the detection limit will move to DL.2 and Region B, which was previous-
ly within the censored range, will move into the non-censored range. Concen-
trations observed in Region B prior to the change remain censored of course
but those observed in that range after the change can be quantified.
The following subsections present discussions of each of the two estima-
tion techniques mentioned above. The discussion related to the regression
approach is intentionally brief and presents only enough information to illus-
trate its shortcomings. The reader is referred to the literature for more
thorough treatisas of this method.
REGRESSION TECHNIQUE
The regression approach to parameter estimation for the lognormal distri-
bution takes advantage of the fact that the cumulative distribution function
curve will plot as a straight line on lognormal probability paper. Recogni-
tion of this provides one with a valuable bit of information for if one can
accurately assign plotting positions to observed data values a least squares
linear regression technique may be applied to estimate the parameters of the
distribution best fitting the plotted data points. Note the qualification
here of "accurately" assigning plotting positions. While any one of a number
of plotting position formulae may be utilized to accomplish this task, no
single one can be claimed to be better, or more accurate, than the others.
Regardless of the plotting position formula selected for use, application
of that formula requires that the data set be ranked. It is here that the
regression approach exhibits its greatest weakness. When a time varying cen-
soring level must be considered, not even all of the non-censored data entries
may be utilized in the analysis. To illustrate this point, suppose that data
observations are made with a certain detection limit DL} and that k censored
and m non-censored observations are obtained. Suppose also that a technologi-
cal advancement occurs which results in a detection limit reduction to DL.2
and that, subsequent to the change, an additional q censored and r non-
censored observations are made. Since the regression approach relies on plot-
96
-------
ting positions, and since these in turn are a function of data value ranks,
one should easily be able to see that only the initial m non-censored observa-
tions, plus that portion of the subsequent r non-censored observations which
are greater than DLi, can be utilized. This of course is because some or all
of the remaining portion of the subsequently obtained r non-censored values
could be lower in magnitude than the largest of the k initially observed cen-
sored data values. In the case where DL2>DLj, which is not discussed at
length here, similar reasoning may be applied.
METHOD OF MAXIMUM LIKELIHOOD
A fundamental concept of probability theory states that the joint proba-
bility of observing say the two outcomes A and B in some experiment is equal
to the probability of observing A times the probability of observing B given
that A has already occurred, or
Pr(A and B) = Pr(A)»Pr(B|A). (1)
If A and B are independent of one anther then Pr(B|A) = Pr(B) and Eq. (1) be-
comes
Pr(A and B) = Pr(A)»Pr(B). (2)
Similarly, given n outcomes x^, i = 1, 2...n, the joint probability of ob-
serving all of those n outcomes is
Pr(x, and x« and ... and x ) = J[Pr(x.). (3)
n i=l 1
This concept of joint probability, coupled with the premise that the pro-
bability of observing a certain value of the variate X, say x, is directly
proportional to the density function ordinate f(x), forms the basis upon which
the method of maximum likelihood is founded. Indeed, the likelihood function
L itself is cast as a joint probability and takes the form
-ft
f(x ; a, p, ...). (4)
1-1 X
Here, f(x^; a, (3, ...) denotes the density function ordinate f(xi) correspon-
ding to the abscissa x^ where the density function contains the parameters a,
P which are to be estimated. It is wished to maximize the probability of
having observed the available data set and thus one wishes to maximize L.
This may be performed by setting partial derivatives of L with respect to each
of the parameters equal to zero and solving the resulting set of equations
simultaneously. Since many probability density functions contain exponential
terms, this procedure may often be simplified by maximizing the log-likelihood
function in L where
in L = in[|f(x ; a, p, ...) = £ in f(x ; a, p, ...). (5)
97
-------
This of course is valid because the logarithmic function is monotonic in
nature.
When censored data must be considered, the joint probability concept un-
derlying the method of maximum likelihood permits one to write modified forms
of the likelihood and log-likelihood functions as
m
and
m
In L = kin[Pr(X
-------
and
(13)
DL. - M.1
rDL - ,n
H—
L * J
where f[(DLj - i^)/(r] and F[(DLj - |JL)/(T] respectively denote values of the den-
sity and cumulative distribution functions of the standardized variates
(DL-: - |j.)/(r. When individually set equal to zero and solved simultaneously,
Eqs. (13) and (14) yield estimates of the parameters JJL and 1 but such are
beyond the scope of this paper and are not discussed here.
DISCUSSION
There are a few attributes of each of the two estimation techniques that
become apparent from the previous discussions. This section provides a brief
presentation of some of these and illuminates some basic principles that must
be satisfied for an analysis to be valid. A qualitative comparison of the
two techniques in terms of their statistical efficiencies is also presented
and forms the basis for a sound argument related to a preference for the
method of maximum likelihood.
It was noted in the introduction that the procedures presented here are
intended to be applied to the lognormal probability distribution, but that
they may also be applied to the normal distribution. Reality is such however
that the opposite is true; i.e., techniques discussed in the previous section
are presented for the normal distribution but may be used for the lognormal
distribution if the available data entries are log-transformed prior to analy-
sis. In other words, variates x^ used in equations presented earlier repre-
sent actually measured data values if the normal distribution is applied and
represent logarithms of actually measured values if the lognormal distribution
is applied. The same is true for the detection limit DL.
99
-------
An additional and very important point related to basic data utilized
pertains to the independence of data values. This is an assumption underlying
the development of the method of maximum likelihood but must be true regard-
less of the estimation technique employed. If data entries are not indepen-
dent of one another, as might be the case if measurement intervals are too
short, time series analysis techniques should be invoked to properly account
for the internal dependence structure in the data.
The final issue addressed here pertains to the statistical efficiencies
of each of the two parameter estimation techniques. Since the variances of
parameter estimates are inversely proportional to the sample size n, and since
utilization of the regression technique constrains the effective sample size
to a value less than n, standard errors in estimates obtained using that
technique can be expected to be higher on average than those obtained using
the method of maximum likelihood. Earlier discussions imply that this effect
becomes particularly pronounced when the detection limit varies in time.
SUMMARY
This paper has presented a comparison of two different techniques that
may be utilized to estimate the parameters of a lognormal distribution with a
censored data set. It has particularly examined the case where the censoring
level varies in time and has formulated a generalized maximum likelihood esti-
mation technique which has not been heretofore presented.
The implications of the comparison presented here appear to be signifi-
cant. The regression approach, because of its relative simplicity, is proba-
bly the most frequently applied of the two estimation techniques. This
method has previously been shown to be comparable to the maximum likelihood
technique and its use has thus been justified. A qualification must now be
imposed on the statement of comparability however. Previous studies have all
addressed the special case of a time invariant censoring level rather than
the more general situation presented here. While the regression technique
might perform well in the special case, it fails in the general case since not
even all of the non-censored data values can be utilized. The method of maxi-
mum likelihood, on the other hand, makes use of all observed data, including
that which is censored, and thus serves to maximize the benefit that is
afforded by a reduced censoring level.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the
views of the Agency and no official endorsement should be inferred.
REFERENCES
1. McCarty, P.L. and Reinhard, M. Trace organics removal by advanced waste-
water treatment. Journal, Water Pollution Control Federation. 52: 1907,
1980.
100
-------
2. Hashimoto, L.K. and Trussell, R.R. Evaluating water quality data near the
detection limit. In; Proceedings of the Advanced Technology Conference.
American Water Works Association, Las Vegas, Nevada, 1983. p. 1021.
3. Gilliom, R.J. and Helsel, D.R. Estimation of distributional parameters
for censored trace level water quality data 1. Estimation techniques.
Water Resources Research. 22: 135, 1986.
4. Gilliom, R.J. and Helsel, D.R. Estimation of distributional parameters
for censored trace level water quality data 2. Verification and applica-
tions. Water Resources Research. 22: 147, 1986.
5. Kendall, M.G. and Stuart, A. The Advanced Theory of Statistics. Vol. 2.
Fourth Edition. Charles Griffin & Co., London, 1979.
6. Gupta, A.K. Estimation of the mean and standard deviation of a normal
population from a censored sample. Biometrika. 39: 260, 1952.
7. Cohen, A.C., Jr. On the solution of estimating equations for truncated
and censored samples from normal populations. Biometrika. 44: 225, 1957.
101
-------
APPLICATION OF THE HSPF MODEL TO WATER MANAGEMENT IN AFRICA
by: Robert C. Johanson,
School of Engineering,
University of the Pacific,
Stockton, CA 95211
ABSTRACT
The Hydrological Simulation Program - Fortran (HSPF)
performs deterministic simulation of hydrology and water
quality, for watersheds of arbitrary complexity.
Like any such model, HSPF requires good calibration data.
Thus, most applications (especially for water quality) have
been in North America and Europe. However, HSPF was recently
applied to two catchments in S. Africa. One was small (90 ha)
and highly urbanized; the other was much larger (300 sq. km.)
and rural.
The results of simulations involving hydrology, sediment
and phosphorus were very satisfactory, and this type of
modelling is being continued there. The goal is to use it to
help manage water and constituent cycling in the larger Mgeni
basin (approx 4000 sq. km.), which serves as the water supply
and effluent conduit for several rapidly expanding urban
centers, as well as numerous rural villages.
102
-------
APPLICATION OF THE HSPF MODEL TO WATER MANAGEMENT IN AFRICA
BACKGROUND
During my recent sabbatical leave at the University of
Natal (S. Africa) I discovered that the Council for Scientific
and Industrial Research (CSIR) was involved in a project to
develop a comprehensive computer model for simulating the
hydrology and water quality of the Mgeni river system, which
has a catchment of about 4000 sq. km. (Figure 1) It contains
two major cities and numerous villages, and is the water supply
and effluent conduit for almost all of this area.
Previously, I was involved in designing the Hydrological
Simulation Program - Fortran (HSPF), which was developed for
the U.S. EPA for similar purposes. It can simulate relatively
complex catchments, both the hydrological behavior and a wide
range of chemical and biochemical parameters. It is in the
public domain and is written in Fortran 77. Thus, it is easy
to acquire and to instal on most machines. It has been
documented by Johanson, et. al. (1984) and Donigian, et. al.
(1984) .
APPROACH
We decided that the best way for local researchers to
benefit from my experience would be for them to apply the HSPF
model to parts of the Mgeni system. They would then be able to
assess its strengths and weaknesses and, by learning the
structure of the software, gain insights valuable to their
future work. It was agreed that the CSIR would apply HSPF to
their research catchment in Pinetown (mainly urban) and the
Dept. of Ag. Engineering at Natal Univ. would apply it to part
of the Midmar catchment (mainly rural).
SOFTWARE MODIFICATION
Although this study principally involved the application
of HSPF, it was necessary to make one major modification. HSPF
runs with a constant time step. I foresaw problems with this
feature for the small urban Pinetown catchment. During storm
103
-------
events a time step of a few minutes is required for good
hydrograph reproduction. However, if this time step is
maintained between storm events, excessive computer time is
used. I therefore set about implementing the RESUME mode in
HSPF, a feature that had been designed into it but never
implemented because of budget constraints. The idea is that a
simulation period is broken up into many consecutive periods,
typically forming an event, inter-event, event, etc., sequence.
Then, the simulation is done by making many runs, each one
covering one of these periods. Each run can have a different
time step. The final conditions from one run become the
initial conditions for the next run. All this is done with
minimal repetition of input from the user. This work involved
changes to 23 of the 400 HSPF subroutines, and took about two
weeks.
APPLICATION TO PINETOWN CATCHMENT
INTRODUCTION
Staff from the CSIR had instrumented a 90 ha catchment in
the central business district in Pinetown, Natal (Figure 1).
From 1982 to 1987 they collected rainfall and streamflow data,
at a 2-minute resolution, and took flow-weighted composite
samples during each storm event, which were later analyzed for
a variety of constituents. The results of this work have been
detailed by Simpson (1986).
Continuous simulation models, like HSPF, usually require
and produce large quantities of "time series" data, such as
rainfall, observed streamflow, simulated streamflow and
pollutant loads, etc. Preparing and manipulating these time
series is a substantial part of the modelling project. it can,
in fact, "bog down" the entire project. The HSPF system was
designed to minimize these problems. All time series are
placed in a single, internally-partitioned dataset, called the
Time Series Store (TSS). After the TSS has been created, input
time series are read into it. Then, simulation runs can be
done and output time series written into the TSS for further
analysis. The data sets in the TSS each have their own
specified time step, ranging from 1 minute to 1 day, and the
HSPF software automatically converts data between the time
steps used in the simulation and in the TSS, as they are read
from, or written to, the TSS.
For the Pinetown project, three years of data were used.
Data from the two autographic raingages and the flow recorder
were read in (at a 2-minute resolution). The "zero
compression" option in HSPF was useful here; by compressing out
repeating zero values, disk storage requirements were reduced
about 95%. Daily evaporation data were also read in.
104
-------
The User's Control Input (UCI) is the other major input to
HSPF. In it, the user specifies the simulation period and
timestep, the manner in which the catchment has been
subdivided, the range of constituents being modelled and their
associated parameters, and the connectivity between the
simulated units. Much of the required input could be obtained
from published sources such as topographical maps. For this
study, we also had the benefit of previous modelling work by
Simpson (1986). Figure 2 shows how the catchment was
subdivided. Parameters that were not already available were
estimated, using published guidelines, such as those given by
Donigian, et al (1978) and (1984). These values were refined
by calibration.
HYDROLOGICAL CALIBRATION
This must come first, because simulation of all water
quality parameters depends on it. The systematic method
described by Donigian, et al (1984) was followed and, after
several iterations, a good fit between observed and simulated
flows was obtained. Figure 3, which shows the accumulation of
the flows over time, indicates that the simulation was about 7%
low overall. This is probably due to some of the observed
baseflow being non-natural in origin, e.g. leaking water
mains, wash water from industries, etc. We were not
attempting to include these components. Figures 4 and 5 show
typical simulated and observed storm hydrographs. The
agreement is considered good, because only two raingages were
available to supply input to the model. This does cause
significant errors in individual events. However, it has been
shown by Johanson (1971) that, even with relatively few
raingages, it is possible to obtain simulated streamflows which
exhibit statistical properties very close to those of the
observed flows.
SEDIMENT CALIBRATION
This was the logical next step because it depends on the
hydrological simulation, and sediment-associated quality
constituents, such as adsorbed phosphorus, depend on the
sediment simulation.
In HSPF, sediment (in this study defined as TSS) is
simulated separately for pervious and impervious areas.
However, in both cases the principle is similar. The model
simulates sediment generation or deposition, removal by
operations such as street sweeping, and washoff by overland
flow. For pervious areas, the effects of sediment detachment
by rainfall and the protection provided by ground cover are
also considered. For each land-segment simulated, there are
105
-------
approximately 4 parameters that need to be adjusted through
calibration.
The observed data for sediment and other quality
constituents were in the form of event loads. There were about
240 rainfall events in the three year simulation period. The
approach used for calibration was similar to that for the
hydrology; that is, parameters were adjusted (following
guidelines) until the simulated and observed sediment outflows
were as similar as possible (Figs. 6 and 7). The simulation
indicated that, except in the case of severe events, almost all
the sediment came from the impervious areas. Considering the
fact that the pervious areas are generally well covered by
vegetation, this was expected.
PHOSPHORUS CALIBRATION
It was decided to simulate sediment-associated and
dissolved P separately, because the mechanisms involved are
different. Because the samples had been analyzed for both
forms, they could be individually calibrated. However, we did
not attempt to separate organic and inorganic P in the
simulation.
The sediment-associated P was handled by specifying its
concentration on the sediment washed off the land. Different
values were used for the pervious and impervious areas, because
the sediment that comes from the impervious areas has a higher
"fines" content, and this is the sediment fraction that is most
important for adsorption. These concentrations are called
"potency factors". The dissolved component was handled as
follows. Surface accumulation and removal rates were
specified, and the washoff of this material depended on the
rate of overland flow. This component was similar to sediment
simulation. The subsurface component was simulated by
specifying a (constant) concentration. The results of this
work are shown in Figure 8. Although the simulated accumulated
quantities of soluble and total P agree quite closely with
observed values, the agreement for individual events is not
good. Time did not permit further refinement of the
calibration.
APPLICATION TO MIDMAR CATCHMENT
INTRODUCTION
The Midmar catchment was selected as the rural test case
because it is an important sub-catchment of the Mgeni system.
Also, Midmar reservoir is important as a water source and is
eutrophically sensitive.
106
-------
Because of time constraints it was decided to confine this
exploratory study to just part of the system. The Midmar
catchment consists of two major sub-catchments, drained by the
Mgeni river in the south and the Lions river in the north
(Figure 9). The Mgeni sub-catchment (Weir U2M13) of area 299
sq. km. was selected and subdivided into 6 pervious
subcatchments. The subcatchments were delimited by taking
catchment physiographic and land use characteristics into
account as well as regional variations in mean annual
precipitation.
It was decided to model the period 1980 through 1985.
Daily rainfall data for 5 gauges (Figure 9) were selected for
the study. Since it was thought that daily totals would not
suffice for simulation of hydrograph shape, the data were
disaggregated into plausible hourly values using data from 3
autographic gauges at the Cedara research station, which is
reasonably close by. Daily evaporation data and observed
breakpoint runoff data were also obtained
HYDROLOGICAL CALIBRATION
Preparation of the User's Control Input and hydrological
calibration was done in a similar manner to that for the
Pinetown catchment. However, because no previous modelling
work had been done, all the data had to be prepared from
scratch. Despite the fact that the staff at the Department of
Agricultural Engineering had no prior experience with the
model, they were able to calibrate it with surprisingly little
outside help. Typical daily flows are shown in Figure 10 and
accumulated flows in Figure 11. Note that the period 1980
through 1982 was used for calibration, and the period 1983
through 1985 for verification. Thus, the data in the first
half of Figure 11 were used for parameter fitting, but the data
in the second half represent "prediction" by the model without
further parameter adjustment. Considering the quality of the
basic data, the agreement for long term water yield as
illustrated in Figure 11 is considered very good.
WATER QUALITY SIMULATION
Comprehensive, regular water quality data were only
available for a point below the confluence of the Mgeni and
Lions rivers at the inlet to Midmar dam (Figure 9). However,
only the Mgeni component of the system was simulated. Because
of time limitations and because the objective of the study was
mainly to become familiar with the capabilities and features of
the model, it was decided to treat the quality concentration
data as if they applied to the Mgeni. Although we knew this
would involve some error, the land use mix in both the Mgeni
and Lions sub-catchments is similar, so the error involved in
this assumption should not be severe. Water quality
107
-------
information from daily grab samples collected between April
1980 and March 1985 were obtained, processed and input to the
TSS as mean daily concentrations. The quality determinands
input to the TSS included suspended, dissolved and total
solids, and particulate and soluble nitrogen and
phosphorous. However, in this phase of the study only TSS was
calibrated.
We obtained good simulations of accumulated suspended
solids except for two periods; 27 March 1982 and 14 February
1985, when both suspended solids and runoff was over-simulated
(Figures 12 and 13). During these periods the poor runoff
simulations were associated with large recorded rainfall totals
at rain gages 0238662 and 0239002 respectively (which were
both heavily weighted in determining rainfall input to the
subcatchments). Rainfall records from the other stations
indicated, however, a marked spatial variation in storm
distribution, which is typical in this area. This, together
with somewhat inadequate representation of temporal rainfall
distributions, necessitated by having to disaggregate daily
totals, is regarded as the major cause of inaccuracies in
modelling individual events.
CONCLUSIONS
The use of the HSPF model in sub-catchments of the Mgeni
catchment gave local practitioners some valuable experience in
using a comprehensive model. They appreciated the ease with
which time series could be stored and processed and the
versatility of the model. However, they found that efficient
use of the model requires considerable effort in becoming
acquainted with the modelling approach, data management
techniques, etc.
ACKNOWLEDGMENTS
Many people were involved in the project. The following
were particularly helpful:
CSIR: Dr.A. Twinch, Mr.B. Gardner and Mr.D. Simpson.
Dept. of Ag. Eng., Natal University: Prof.R. Schulze, Mr.E.
Schmidt, Mr.S. Lynch.
Computer Center for Water Research, Natal University: Mr.M.
Dent, Mr.A. Kure, Mr.R. de Vos.
Mgeni Water Board: Dr.H. Furness, Ms.M. Pillay.
108
-------
The work described in this paper was not funded by the U.S. EPA
and therefore the contents do not necessarily reflect the views
of the Agency and no official endorsement should be inferred.
REFERENCES
Donigian, A.S., and Davis, H.H., 1978. User's Manual for
Agricultural Runoff Management (ARM) Model. U.S. EPA, Athens,
Georgia, USA. Report EPA-600/3-78-080.
Donigian, A.S., Imhoff, J.C., Bicknell, B.R., and Kittle, J.L.,
1984. Application Guide for Hydrological Simulation Program -
Fortran (HSPF). U.S. EPA, Athens, Georgia, USA. Report EPA-
600/3-84-065.
Johanson, R.C., 1971. Precipitation Network Requirements for
Streamflow Estimation. Ph.D thesis, Dept. of Civil Eng.,
Stanford Univ., Stanford, Calif., USA.
Johanson, R.C., Imhoff, J.C., Kittle, J.L., and Donigian, A.S.,
1984. Hydrological Simulation Program - Fortran (HSPF): User's
Manual for Release 8. U.S. EPA, Athens, Georgia, USA. Report
EPA-600/3-84-066.
Simpson, D.E., 1986. A Study of Runoff from an Urban
Catchment. M.Sc thesis, Dept. of Civil Eng., Univ. of Natal,
Durban, Natal.
109
-------
THE CATCHMENT
v> UMGENI
POWER
STATION/
SW
NE
RESIDENTIAL AREAS
WESTVILLE
QUEENSBURGH
CHATSWORTH
2km
FIG. 1. LOCATION OF PINETOWN STUDY AREA
-------
SUB CATCHMENT BOUNDARIES
131 SUB CATCHMENT NUMBER
j) INLET NUMBER
/ CONDUIT NUMBER
— 37!— CONTOUR IN METRES
FIG. 2. DISCRETIZATION OF PINETOWN CATCHMENT INTO 7 SUBCATCHMENTS
-------
1500
1200
Fig 3. Cumulative Flow (1000 m**3)
Pinetoun Catchment Run 6
— Observed
Simulated
01/10/83 07/10/83
01/07/84 07/06/84 01/03/85
Date (rcn/dd/yy)
07/03/85 12/31/85
-------
rig 4. Event on 8 March 1983
Pinetoun Catchment
T
-jf- Observed
O Simulated
2000
2100
2200
2300
Time (hhrom)
Tig 5. Event on 12 October 1982
P-inetown Catchment
•f- Observed
•O Simulated
400
500
600
Time (hhmm)
113
-------
(X 100000)
•3
Fig 6. Accijtviu 1 at»d Sediment Load
Finetoijjn Run 9
i i i i T n i i i i i i ( i i i T i i i r n i i i i i r i TTLT
I I I I I I I I I I I III I....1.. II I I I I I I II ..I. II I I I. I t I I I I
40 60 80 100 120 140 160 180 200. 220 240
Event Number
(X 10000)
Regression of Simulated on Observed
Fig 7. Pinetoun Suspended Sediment Run 9
114
-------
Fig 8. Accumulated Phosphorus Loads
Pinetown Run 4
-fit- Total Obsd
•O Total Si rod
L_ I
400
300
k3
200
100 I—
I I j i i I i i I I i i I I Mill IM
t t i i i I
Qted
Disvd Simd
•~j-
20 40 60 80 100 120 140 160 180 200 220 240
Event No.
-------
.0239
tNORTH
:EDARA
FIG. 9. THE MIDMAR CATCHMENT
-------
7ig. 10.Daily Flows for UZM13 Catchment
— Observed
Simulated
09/01/83
06/30/1
(X 100000)
4
h
o
u
s
a
n
d
c
u
b
i
c
Fig 11. Accumulated Flows
U2M13 Catchment
— Observed
Simulated
01/10/80 01/08/81 01/07/82 01/06/83 01/05/84
Date (rni/dd/yy)
01/03/85 01/02/86
117
-------
(X 10000)
2
1.6
1.2
0.8
0.4
Fig 12. flccumulated Suspended Solids
U2ml3 Catchment Run 1
— Observed
• •• Simulated
I
04/10/80 04/09/81 04/08/82 04/07/83 04/05/84 04/04/35
Date (rni/dd/yy)
(X 1000)
10
13. Accumulated Suspended Solids
U2M13 Catchment Sun 2
04/10/80 04/09/81
— Observed
- Simulated
04/08/82 04/07/83 04/05/84
Date
-------
MULTI-MODEL MICRO-COMPUTER BASED WET DETENTION BASIN
DESIGN METHODOLOGY
by: Sidney L. Harrell
Environmental Engineer
Water Quality Section
Division of Environmental Management
North Carolina Department of Natural Resources
and Community Development
Raleigh, NC 27611.
ABSTRACT
A regulatory driven technical guidance manual, including a
compendium of model series, is being synthesized from widely
available manuals and models to assist developers in planning
stormwater control mechanisms and to aid municipal officials in
reviewing these plans. Design of wet detention basins, the
subject of the first volume of this manual, usually consists of
determining in four steps: 1) the minimum surface area of the
permanent pool, 2) the storage volume that will detain a
specified runoff, 3) principal spillway size and additional
storage volume for flood control and sediment accumulation, and
4) the dam and emergency spillway design parameters. Tables of
required surface area for a given drainage area, imperviousness,
and watershed characteristics are used for the first step. These
tables were developed previously using Driscoll's Model (1) which
runs in Basic@. For completing the remaining steps, a LOTUS123@
spreadsheet model is pulled up, appropriate values entered, and
the remaining design parameters computed. This process, the
model, and an example of its application are presented.
INTRODUCTION
Control of nonpoint source (NFS) pollution is a stated goal
of the 1987 Federal Water Quality Act. A primary source of these
pollutants is stormwater runoff from urban areas. The approach
of the North Carolina Department of Natural Resources and
Community Development, Division of Environmental Management (DEM)
to control stormwater quality is based first on mi nimi ?.ing
i mpervi mis surfaces through land use controls and secondly
on treating stormwater runoff using engineered stormwater
controls.
119
-------
Dissemination of technical information to both engineers and
local officials on the design and maintenance of engineered
solutions is essential for adequate pollutant removal . The
planning design of wet detention basins for stormwater control is
the subject of this paper.
The design of wet detention basins is based on retaining
storm runoff for an extended length of time in order to settle
out suspended solids and pollutants such as heavy metals and
nutrients. Eugene Driscoll's model (1) was chosen by DEM for the
permanent pool component of the design. The model uses as input a
long-term average storm statistically calculated from the
historical rainfall record. By using this storm and the
appropriate watershed characteristics (e.g., impervious cover), a
permanent pool is sized to detain the storm runoff long enough to
attain the target Total Suspended Solids (TSS) removal. The
model incorporates settling that occurs during the storm
(dynamic) and between storms (quiescent). The movement of the
storm runoff through the basin is assumed to be via plug flow.
In addition to the permanent pool, the basin should have a
temporary water quality pool. This storage volume, located above
the permanent pool, is necessary for periods when runoff entering
the basin is significantly warmer than the permanent pool. Under
these conditions runoff could flow across the top of the
permanent pool and exit the basin without being detained long
enough to achieve maximum settling. To counteract this, the
runoff from the designated storm is detained and then slowly
released through a negatively sloped pipe (Figure 1).
Additional storage volume and a principal outlet and
emergency spillway may be added for flood and/or erosion control.
DEM has adopted a conservative policy in which the storage
allocated to flood control is located on top of both water
quality pools. The storage for erosion control may occupy the
same storage as the temporary water quality pool (Figure 1).
WET DETENTION BASIN DESIGN
Wet detention basins for water quality control usually
consist of four components: 1) a permanent pool, 2) a temporary
water quality volume, 3) a flood control volume and outlet
device, and 4) a dam with emergency spillway (Figure 1). Because
of physical and performance constraints, planning design should
begin at the bottom by determining the permanent pool surface
area and then proceed upward in a step-by-step approach.
120
-------
10 year
flood storm
1/2 inch or
v 1 inch storm
Orifice for
drawdown
of 1/2 inch or
1 inch storm
Emergency
.spillway •,
Permanent
.pool elevation
Non-erosive at
10 year storm
Emergency drain pipe with valve
FIGURE 1. WET DETENTION BASIN
PERMANENT POOL
The permanent' pool provides a high pollutant removal rate
through gravity settling, chemical flocculation, and biological
uptake. The settling removal rate is related to pond geometry,
surface area, depth, volume, residence time, and the size,
density, and shape of the particles in the runoff volume.
Charts 1 or 2 below are used to determine the minimum
surface area (SA) that will satisfy DEM's TSS removal criteria
(stated below) depending on the type of location, and the fully
developed site characteristics:
85% TSS Removal from Runoff from first 1" of rainfall:
Water Supply Watershed Critical Area with greater than 6%
Imperviousness - Chart 1
Water Supply Watershed Non-Critical Area with greater than
30% Imperviousness - Chart 2
65% TSS Removal from Runoff from first 1/2" of rainfall:
Water Supply Watershed Non-Critical Area with greater than
12% but less than 30% Imperviousness - Chart 2
The critical area is the area within 1/2 to 1 mile of the
reservoir or intake point depending on the watershed size.
A sediment storage pool underlies the permanent pool. It's
depth depends on the permanent pool area, the expected sediment
121
-------
CHART 1
FOR NORTH CAROLINA
PIEDMONT AREAS
INSIDE CRITICAL AREA
Impervious %
< 6% ?
YES
ONLY TO BE IN UNAVOIDABLE SITUATIONS
Wet Detention Basin
SA/DA % for basin depths
Impervious (%) 3.0ft 3.5ft 4.0ft 5.0ft 6.0ft
7-29
30
50
70 (max)
1.8 1.5
2.4 2.1
4.0 3.5
5.7 4.8
1.4
1.8
3.0
4.3
1.1
1.5
2.5
3.5
1.0
1.3
2.1
2.9
(controls 1-inch rainfall)
Interpolate intermediate values
* Surface area basin/drainage area * 100
PREFERRED METHOD
No structural controls needed
yield and the planned maintenance interval. A method outlined in
Schueler (2) is used in the example design exercise (Tables 1 and
l.(a)) for estimating sediment yield.
TEMPORARY WATER QUALITY POOL VOLUME
This is the volume of runoff from the developed watershed,
resulting from the applicable rainfall depth (1/2 inch or 1 inch)
that will flow into the wet pond. If the basin drains only
impervious surfaces, this volume can be calculated as the
rainfall depth times the area drained. Otherwise, the volume
would be that computed by some defensible method.
In order to ensure plug flow thru the pond, this volume of
runoff, stored above the permanent pool, is to be detained for a
minimum of 2 days and a maximum of 5 days. It is to be slowly
released through a negatively sloped pipe or other non-clogging
device. The diameter of this pipe or orifice is determined by an
iterative reservoir routing procedure such as that shown in Table
2.
122
-------
If flood control is not incorporated into the pond design,
all flow in excess of the temporary water quality pool volume
must be routed around rather than through the settling basin.
This is to prevent such excess flow from stirring up already
settled material and flushing it out of the basin. A device such
as a diversion box may be used for this purpose.
FLOOD CONTROL VOLUME
This is the volume of excess rainfall running off the pond
watershed that must be detained in order to reduce the peak flood
flow to an acceptable level, usually the predevelopment peak flow
for the designated design storm. Erosion control regulations may
limit the peak outflow even further because of the increased
duration (impulse) of elevated outflows. Local, state, and/or
federal governments or agencies having jurisdiction for a site
will designate the return period and duration of the design
storm.
The flood control storage volume is usually less than the
design storm runoff volume since water is released during the
storm. Because outflow is dependent on the depth of water in the
basin and the depth is dependent on the inflow, outflow, and the
CHART 2
FOR NORTH CAROLINA
PIEDMONT REGIONS
OUTSIDE CRITICAL AREA
.YES
Impervious %
< 12%?
•YES
Special
case analysis
PREFERED METHOD
No structural
controls needed
.YES
ONLY TO BE USED IN
UNAVOIDABLE SITUATIONS
Wet Detention Basins
SA/DA % for basin depths
Impervious % 3.0ft 3.5ft 4.0ft 5.0ft 6.0ft
31 2.4 2.1 1.8 1.5 1.3
50 4.0 3.5 3.0 2.5 2.1
70 5.7 4.8 4.3 3.5 2.9
(controls 1 .inch rainfall
Interpolate intermediate values
* Surface area basin/drainage area * 100
PREFERRED METHOD
No structural
controls needed
ACCEPTABLE METHOD
Wet Detention Basins
SA/DA for basin depths
Impervious % 3.0ft 3.5ft 4.0ft 5.0ft 6.0ft
13-30 1.0 0.9 0.8 0.7 0.5
(controls 1/2-inch rainfall)
* Surface area/drainage area * 100
123
-------
basin size and shape, a time-by-time routing is usually required.
The example design exercise (Tables 2 and 3) uses the HRM (H.R.
Malcom) method of routing (3), which is easy to execute and
gives results similar to the Storage-Indication Method.
A riser-barrel is the most efficient outlet device in terms
of minimizing storage required. Usually it is designed with the
barrel as the peak flow limiter. Aesthetic concerns may however
dictate a weir with a slightly larger basin.
Particular attention must be paid to how routing is set up
if reservoirs will occur in series or if any outlet device will
experience high tailwater. Care must also be taken to ensure that
the peak of the release of a pond off the main stem will not
coincide with that from upstream. A bottom withdrawal riser design
and re-aeration may be necessary to reduce downstream thermal and
low DO impacts during the summer.
EMBANKMENT AND EMERGENCY SPILLWAY
The emergency spillway should be designed in accordance with
SCS methods (4) for at least the 100-year storm. The NC Dam
Safety Act (5) gives guidance on design storms for spillways in
larger basins. Storms of other durations should be checked for
overtopping of the dam (3). Calculations for wave height and wind
setup should be included in a detailed freeboard analysis (6).
The minimum is one foot of freeboard above the emergency spillway
(2). The water level rise above the emergency spillway during an
extreme event can be found by simply adding a weir to the
flood control routing (Table 3), with the weir crest set just
above the maximum stage occurring during the flood control event.
The work described in this paper was not funded
by the U.S. Environmental Protection Agency and
therefore the contents do not necessarily reflect
the views of the Agency and no official endorsement
should be inferred.
REFERENCES
1. Driscoll, E. Methodology for Analysis of Detention Basins
for Control of Urban Runoff Quality. EPA 440/5-87-001.
U. S. Environmental Protection Agency, Office of Water,
Nonpoint Source Branch, Washington, D.C., 1986.
2. Schueler, T.R. Controlling Urban Runoff: A Practical Manual
for Planning and Designing Urban BMPs. Department of
Environmental Programs, Metropolitan Washington Council of
Governments, Washington, D.C., 1987.
124
-------
3. Malcom, H.R. and New, V.E. Design Approaches For Stormwater
Management in Urban Areas, prepared for CE383 at NCSU,
Raleigh, NC, 1975.
4. United States Department of Agriculture, Soil Conservation
Service. Engineering Field Manual for Conservation
Practices. 1986.
5. North Carolina Department of Natural Resources and Community
Development. Dam Safety, Title 15, Subchapter 2K. Raleigh,
NC, November 1, 1985.
6. Lindsley, R.K. and Franzini, J.B. Water-Resources
Engineering. McGraw-Hill, USA, 1972.
125
-------
APPENDIX
EXAMPLE WET DETENTION BASIN DESIGN EXERCISE
Site specifications: 10 ac., 25% imp., non-crit., sewer
TABLE 1. PERMANENT POOL SURFACE AREA
With drainage area (DA) = 10 acres, for 85% TSS removal;
Required surface areas (SA) using Chart 2 are:
Depth(ft)
SA/DA
SA (acre)
SA (sqft)
3
1.0
0.1
4356
3.5
0.9
0.09
3920
4
0.8
0.08
3485
5
0.7
0.07
3049
6
0.5
0.05
2178
TABLE l.(a) SEDIMENT STORAGE POOL
Using the method described in Schueler (2), page 1.19:
L = [(P)(Pj)(Rv)/12](C)(A)(2.72)
where: L = sediment export load (Ibs)
P = rainfall depth (in) / duration (yr)
Pj = non runoff producing storm corr. factor
Rv = runoff coef. = r/P = 0.05 + 0.009 * I
I = site imperviousness (%) = 25 %
C = sediment concentration (mg/1)
A = area of site (acres) = DA
12,2.72= unit conversion factors
Assuming: P = 40 , Pj = 0.9 ,
C = 280 , Rv = 0.275
L = 6283 Ibs/yr = 3.1 tons/yr
Assuming 85% removal for 20 years;
L = 4.2 tons/year * .85 * 20 years = 53.4 tons
Assuming 1 ton = 1 cuyd, Sediment Volume = 1442 cuft
Assuming Sediment Pool Area = Permanent WQ Pool SA,
and thus Sediment Depth (Sed D) = Sediment Volume / SA,
the respective pond depths are:
WQDep(ft) 3 3.5 4 5 6
Sed D(ft) 0.33 0.37 0.41 0.47 0.66
Total(ft) 3.33 3.87 4.41 5.47 6.66
126
-------
TABLE 2. TEMPORARY WATER QUALITY POOL
Using the method described in Schueler (2) page 1.11:
P =
r = Rv * p
0.14 in
Runoff Volume = r * A =
4991 cuft
0.5 inch
(Chart 2)
Assuming a 6 foot deep permanent WQ pool;
Compute diameter of negatively sloped pipe (D orf):
BASIN DATA:
STORM DATA:
0.5 inch
Ks =
b =
P orf =
D orf =
Cd =
Inv Z =
ROUTING:
TIME
[hr]
0
0
10
20
30
40
50
60
70
80
16.4
3
6.7
0.5
0.59
0
f (area)
f (side
ft
in
ft
slopes)
Criteria:
QP =
Tp =
dT =
2-5 day
drawdown
n/a
n/a
10
or 48-120
cfs
min
hr
hour
HRM Method
INFLOW
[cfs]
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
STORAGE
[Cf]
4837.
9828
9154
8488
7830
7181
6542
5912
5293
4685
STAGE
[ft]
6.7
8.4
8.2
8.0
7.8
7.6
7.4
7.1
6.9
6.6
Outflow
[cfs]
0.000
0.019
0.019
0.018
0.018
0.018
0.017
0.017
0.017
ERR
Surf Area
[sqft]
2178
3494
3332
3169
3003
2835
2664
2490
2313
2132
*
#
* Permanent Pool
# with runoff volume
127
-------
TABLE 3. FLOOD CONTROL VOLUME
10 - Year Storm Peak Discharge (QplO) Control
Rv = r/p assume duration (Tc)= 5 min and
or r = Rv * p p= 0.62 in (depth-dur.-freq. table)
r = 0.17 in r(Vol) = 6189 cuft (watershed)
assuming a triangular hydrograph;
Vol = 1/2 * (2*dur) * Qp Rational check: Qp = CIA
Qp = Vol / Duration if C = Rv and I = 7.28 ,
QplO = 20.6 cfs QplO = 20.0 cfs
Criteria: Predevelopment QplQ; assume Rv = 0.1
r = 0.06 in r(Vol) = 2251 cuft (watershed)
QplO = 7.5 cfs
BASIN DATA:
D
P
Ks =
b =
riser =
Cw =
riser =
D bar =
Cd =
Tnv Z =
ROUTING:
TIME
[min]
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16.4
3
10
3.33
8.4
10
0.59
0
Results
INFLOW
[cfs]
0.0
2.0
7.1
13.5
18.7
20.6
18.7
14.5
11.2
8.6
6.7
5.1
4.0
3.0
2.4
1.8
in
ft (temp
in
ft
: Max Z =
STORAGE
[cf]
9828
9828
9946
10371
11150
12155
13138
13842
14247
14450
14498
14428
14267
14036
13752
13428
STORM DATA:
WQ pool)
9.61
STAGE
[ft]
8.4
8.4
8.5
8.6
8.8
9.1
9.3
9.5
9.5
9.6
9.6
9.6
9.6
9.5
9.4
9.4
Qp =
Tp =
dT =
Max Q =
Outflow
[cfs]
0.0
0.0
0.1
0.5
1.9
4.3
6.9
7.8
7.8
7.8
7.8
7.8
7.8
7.8
7.8
7.7
10
20.6
5
1
7.8
-yr
cfs
min
min
cfs
Surf Area BarrelQ
[sqft]
3494
3494
3522
3622
3801
4026
4240
4390
4476
4518
4528
4513
4480
4431
4371
4302
[cfs]
0.0
7.3
7.3
7.4
7.5
7.6
7.7
7.8
7.8
7.8
7.8
7.8
7.8
7.8
7.8
7.7
RiserQ
[cfs]
0.0
0.0
0.1
0.5
1.9
4.3
6.9
9.0
10.2
10.8
11.0
10.8
10.3
9.6
8.7
7.8
128
-------
MODELING AND FIELD EVALUATIONS OF URBAN WET DETENTION PONDS
Jy S, Wu
Associate Professor
Department of Civil Engineering
University of North Carolina at Charlotte
Charlotte, North Carolina 28223
ABSTRACT
An extensive stormwater sampling program was conducted
on three existing urban wet 'detention ponds in the Piedmont
region of North Carolina. By analyzing the pollutant removal
data collected from eleven runoff events, a performance
relationship was observed, permitting incorporating the water-
quality improvement requirements into proper sizing of wet
detention ponds. To achieve a minimum level of urban runoff
pollution control, the surface area ratio of detention ponds
must be greater than 0.5%. About 1.0% to 2.0% of the
watershed area is needed for developing detention ponds to
control 70% or more of the sediment load. In addition, an
EPA model was examined and verified for its usefulness in
analyzing the water-quality improvement performance of urban
wet detention ponds.
INTRODUCTION
Stormwater detention facilities of proper design can
serve not only for flood control but also for retention of
sediment and other pollutants associated with settleable
particulates (1,2,3). Wet detention ponds, in particular, may
also provide aesthetic amenities and recreational
opportunities to the community. Installing wet ponds at
strategic locations within a watershed can eliminate a major
portion of nonpoint pollutant loadings for the entire area
(4). In contrast, dry ponds were found inefficient in
removing suspended solids and other pollutants (5).
Methods are available to predict sediment trapping
efficiencies in detention ponds (6,7,8,9,10). In general, wet
ponds constructed with length-to-width ratio of 2-to-l or
129
-------
greater provide a better trapping efficiency of sediment.
shallow and long ponds are more ideal than deep and short
ponds; a minimum of 3-ft depth is recommended. The performance
of wet ponds depends greatly on influent particle size
distributions (11). The American Society of Civil Engineering
has conducted a survey on issues related to the hydraulic
function, public safety, maintenance, water quality and
aesthetic aspects of outlet controls for detention ponds (12).
Information pertinent to the beneficial use of
detention ponds is limited in North Carolina. Consequently, a
field sampling program was initiated in 1986 to establish a
data base for examining the performance of existing urban wet
detention ponds. Based on results of field monitoring and
computer modeling, a performance relationship is obtained for
evaluation of wet detention ponds in controlling urban runoff
pollution.
FIELD MONITORING PROGRAM
Three urban wet detention ponds were selected for
study, including Lakeside (LS), Waterford (WF) and Runaway Bay
(RB). They are located in the Piedmont region of North
Carolina, in the city of Charlotte. The entire watershed has
a drainage area of 437 acres and comprises of three subareas
of Lakeside (65 acres), Waterford (302 acres), and Runaway Bay
(70 acres). The watershed layout and information pertinent
to each detention pond are given in Table 1 and Figure 1,
respectively.
The upper portions of Lakeside (30 acres) and Waterford
(302 acres) subareas consist primarily of single family
residential land use. Adjacent areas of LS (35 acres) and RB
(70 acres) ponds are characterized by intensive development of
multifamily housing such as condominiums and apartments.
Storm runoff originating from the adjacent impervious areas
discharges into the detention ponds by overland flow or
through a number of storm pipes.
The percentage of watershed area devoted for detention
pond development is defined as a surface area ratio, SAR.
Thus, the SAR's is 7.5% for LS pond, 0.6% for WF pond, and
0.75% for RB pond. The overall SAR for all three ponds and
the entire watershed is 2.27%.
Flow measurements were made at the upstream inflow and
downstream outflow of each detention pond using recording
gaging stations. A NITRON paper tape reader transmits the
records to the U.S.G.S. Prime computer system for data
processing. For large storms, recorded stream stages might
130
-------
TABLE 1. WATERSHED AND DETENTION POND CHARACTERISTICS
Lakeside Waterford Runaway Bay
Watershed Characteristics
Land Use Single Family Single Family Apartment
Apartment Apartment Wooded
Condominium Wooded
Acreage, acres *
Upstream 30 302 367
Local 35 - 70
Total 65 302 437
Detention Pond Characteristics
Acreage, acres
Volume, acre-ft
Mean Depth, ft
SAR, %
4.88
38.84
7.97
7.51
1.79
5.08
2.84
0.59
3.25
12.27
3.78**
2.27
* IncludingLakeside and Waterford subareas.
** For all three ponds and the watershed.
exceed the highest stage for which a current-meter measurement
was made. In these cases, rating curves were extended using
the conveyance-slope method (13,14). Local runoff originating
from adjacent areas of a detention pond was estimated with the
TR-20 hydrologic model (15).
Stormwater samples were collected, using ISCO
automatic samplers, from outflows of each detention pond and
at the inflow of LS pond. The sampling frequency varied from
half an hour to two hours depending on whether samples were
taking during or after a storm event. Normally, the sampling
of outflow from a detention pond continues for about two days
after the end of a storm event. Runoff samples were also
manually collected at 10- to 20-minute intervals from the
inflow of LS pond and from the two selected storm pipes in the
R8 subarea. Non-storm samples were collected once every two
to three weeks at pond outlets to establish a background
condition of water quality.
The analyses of TSS, ammonia nitrogen (NH,-N), total
Kjeldahl nitrogen (TKN), ortho-phosphorus (OP), and total
phosphorus (TP) were in accordance with procedures outlined in
Standard Methods (16). Metals were analyzed by flame atomic
absorption spectrophotometry method. Dissolved metals were
analyzed by filtering the storm samples through a 0.45 micron
131
-------
U)
NJ
9) Stream Gaging Station
[9] Sampling Station
• Rain Gage
Figure 1. Watershed and Detention Ponds
-------
Nalgene syringe filter (cellulose acetate membrane). Total
metals were analyzed by digestion of storm samples at 95
degrees centigrade overnight, using 2 ml of concentrated
nitric acid per 50 ml of storm sample. The procedure for
determining particle settling velocity was similar to a
settling column test with certain modifications (14).
The pollutant mass exported from local drainage areas
is obtained by multiplying the areal runoff by an event mean
concentration (EMC), defined as the constituent mass per unit
runoff volume (mg/1) or a weighted average pollutant loading
rate (Ibs/acre/in runoff). The EMCs were developed based on
runoff quality and flow information obtained from the RB
subarea.
The pollutant mass entering a detention pond is the
summation of pollutant mass from upstream inflow and local
drainage areas. The removal or trapping efficiency of
pollutants is computed as the percent difference of the total
pollutant mass entering and leaving the detention pond.
DATA PRESENTATION
A total of eleven runoff events have been monitored, as
summarized in Table 2. The average magnitude, duration,
intensity of the monitored storms are 1.19 inches, 37 hours,
and 0.08 in/hr, respectively. The average runoff coefficients
determined for LS and WF subareas, and the watershed are 0.7,
0.43 and 0.4, respectively.
The quality of storm runoff expressed in terms of EMCs
is presented in Table 3. The EMCs are based on data collected
from storms 2, 3 and 4, and are considered representative of
the runoff quality in the study area.
The performance of a detention pond is influenced by
the particle size of incoming sediment, which may vary among
storms, within a storm, and from one area to another.
Consequently, ten sets of storm samples were collected from
different storms and at various stages of a storm event. The
average settling velocity is reported in Table 4.
Sampling of runoff samples for storms 1, 2, 3, 4, 5, 6,
7 and 10 was focused on LS and RB subareas. This data base
was employed to calculate the pollutant removal efficiencies
of LS pond (SAR=7.51%), and RB pond (SAR=0.74%). Runoff
samples obtained for storms 8, 9 and 11 were primarily within
the WF subarea, permitting an evaluation of the removal
efficiencies of WF pond (SAR=0.59%). Based on the average
performance of each detention pond, the overall performance
was calculated for the detention pond system (SAR=2.27%).
133
-------
TABLE 2. RAINFALL STATISTICS OF MONITORED STORMS
Storm
No.
Date
Volume
(in)
Duration
(hrs)
Intensity
(in/hr)
Time Since
last Storm
(hrs)
1
2
3
4
5
6
7
8
9
10
11
01/01/87
02/26/87
03/25/87
04/15/87
06/04/87
09/05/87
10/27/87
11/10/87
12/10/87
02/02/88
02/19/88
Mean
Cv
1.39
3.60
1.73
1.48
0.64
1.06
0.52
1.08
0.56
0.80
0.26
17.8
66.7
146.8
80.0
2.8
23.6
8.3
6.8
3.7
42.8
11.4
0
0
0
0
0
0
0
0
0
0
078
054
012
019
226
045
062
160
153
019
0.023
1.19 37.3 0.077
0.77 1.2 0.919
Continuous Rain Gage Records
(11/18/86-12/01/87)
184
89
137
383
9
16
647
329
273
149
352
233
0.8
Mean
Cv
Mean
Cv
0.50
1.19
N.C.
0.36
1.45
6.6
1.4
Annual Stati
5.9
1.1
0.222
1.840
sties*
0.066
1.320
112
1.9
77
1.1
* see reference 1 o. Cv = coefficient of variation
Water quality response and removal efficiencies of
detention ponds, are summarized in Tables 5 and 6,
respectively. It can be seen from Table 5 that the average
concentrations of WF pond outflow during storm events are
generally higher than those of LS and RB ponds, particularly
for TSS, Zn and Fe. The average of storm peak concentrations
exceeds the background level. The average of mean event
concentrations is not significantly higher than the background
level, with the exception of TSS.
LS pond accounts for a large SAR; it performs well for
the removal of TSS, Zn and Fe. However, its removal of TP and
TKN is less efficient due to the additional source of nutrient
input from waterfowl droppings. In some cases, LS pond is
capable of retaining the total runoff pollutant loads produced
by small storms; e.g. a 100% removal for all pollutant
parameters is reported for storms 5,6 and 7. The removal
134
-------
TABLE 3. EVENT MEAN CONCENTRATIONS OF STORM RUNOFF
Water Quality Runaway Bay NURP Values
Parameters Local RUnoff Residential Overall
TSS, rag/1
TKN, mg/1
NH,-N, mg/1
TP; mg/1
OP, mg/1
T-Fe, mg/1
S-Fe, mg/1
T-Cu, ug/1
T-Pb, ug/1
T-Zn, ug/1
S-Zn, ug/1
135.00
0.88
0.22
0.14
0.10
6.11
0.24
*
*
66.00
20.00
228.00
2.85
0.62
56.00
293.00
254.00
1.50
0.33
0.12
34.00
144.00
160.00
*Concentration below detection limit of flame atomic
absorption spectrophotometric analysis
T-Fe = total iron S-Fe = soluble iron
TABLE 4. PARTICLE SETTLING VELOCITIES
Size
Category
1
2
3
4
5
Percent Particle
in Each Category, %
0-20
20 - 40
40 - 60
60 - 80
80 - 100
Average Settli
Velocity, ft/
0.01
0.08
0.40
1.80
6.00
ng
hr
efficiencies for R8 pond, on the average, are 62% for TSS, 21
% for TKN, 36% for TP, 32% for Zn and 52% for Fe. The average
TSS removal efficiency for WF pond is 41%. TSS is the only
water quality parameter measured for WF pond.
LS pond accounts for a large SAR; it performs well for
the removal of TSS, Zn and Fe. However, its removal of TP and
TKN is less efficient due to the additional source of nutrient
input from waterfowl droppings. In some cases, LS pond is
capable of retaining the total runoff pollutant loads produced
by small storms; e.g. a 100% removal for all pollutant
parameters is reported for storms 5,6 and 7. The removal
135
-------
TABLE 5. WATER QUALITY RESPONSES OF DETENTION PONDS
Mean Constituent Concentration
TSS TP TKN Zn Fe
mg/1 mg/1 mg/1 ug/1
LS
WF
RB
Pond
(a)
(b)
Pond
(a)
(b)
Pond
(a)
(b)
Outflow
Outflow
Outflow
17
11
108
51
47
26
0
0
0
0
0
0
.25
.14
.25
.14
.27
.12
1.5
1.0
1.3
0.8
1.1
0.7
Non-storm
LS
WF
RB
Pond
(c)
Pond
(c)
Pond
(c)
(a
(b
(c
Outflow
Outflow
Outflow
) Average
) Average
) Average
6
9
13
of
of
of
0
0
0
storm
mean s
.15
.15
.18
peak
torm
n on -s form
1.
1.
0.
concent
concent
1
Condi
2
4
8
ration
ration
70
18
20
32
83
24
t ion
45
36
28
s .
s .
2
1
8
3
3
2
0
1
1
.2
.4
.0
.5
.6
.1
.9
.5
.7
concentrations.
efficiencies for RB pond, on the average, are 62%
% for TKN, 36% for TP, 32% for Zn and 52% for Fe.
TSS removal efficiency for WF pond is 41%. TSS
water quality parameter measured for WF pond.
for
The
is the
TSS, 21
average
only
MODELING OF DETENTION POND PERFORMANCE
The model for computing the long term removals of
pollutants by detention ponds was developed by U.S. EPA (17).
The analysis is based on the assumption that the removal of
sediment is due to the combined effect of dynamic settling
during runoff period and of quiescent settling during the
interval between successive storms. The variable nature of
storm runoff is treated by specifying the rainfall and runoff
in probabilistic terms, established by an appropriate analysis
of a long term precipitation record. The following information
is required to perform the computations.
136
-------
TABLE 6. PERFORMANCE OF LS, RB AND WF PONDS
(Field Data)
Percent Removal, %
Storm No. TSS TP TKN Zn Fe
1.
2.
3.
4.
1
2
3
4
5 1
6 1
7 1
10
Avg .
1
2
3
4
5
6
7
10
Avg.
8
9
11
Avg.
Rainfall stat
for rainfal
Watershed cha
area of the
Runoff coeffi
Short circuit
82
94
95
85
00
00
00
91
93
56
7
62
74
87
78
87
57
62
67
39
18
41
i s t i c s
-20
10
82
-55
100
100
100
45
45
55
-10
62
19
- 4
36
93
40
36
i ncl
1 volume, i
racter i
basin,
cient .
st i c
and
LS Pond
-58
- 7
- 9
4
100
100
100
22
32
RB Pond
2
27
- 4
37
35
31
15
21
21
WF Pond
uding coeff
ntensity du
s including
drainage a
72
82
69
71
100
100
100
48
80
46
-29
85
67
40
-15
22
40
32
ici ents of
ration and
depth and
rea .
_
78
78
71
100
100
100
81
87
_
2
46
55
79
67
85
30
52
variation
interval .
surface
parameter.
5. Particle size distribution.
The EPA model was applied to the study area. Model
input included rainfall records (11/18/86-12/01/88), a short
137
-------
circuit coefficient of "3", and the particle settling velocity
distribution presented in Table 4. Model computations were
performed for LS and WF ponds to determine the removal
efficiencies and the particle size distributions of their
outflows. The particle size distribution for RB pond inflow
was obtained by combining the computed particle size
distribution of LS and WF pond outflows. Model calculation
was then performed for RB pond using the combined distribution
as an input and other pertinent information.
Results of simulation for TSS removal are presented in
Table 7. The computed average long term TSS removal for LS,
WF and RB ponds are 99%, 47% and 49%, respectively. The
overall removal is calculated as 74%. A comparison of field
data and model predictions is included in Table 8.
TABLE 7. MODELING RESULTS OF TSS REMOVAL
Size*
Category
1
2
3
4
5
Pe-rcent
in Each
0
20
40
60
80
Part i c 1 e
Category
- 20
- 40
- 60
- 80
-100
Removal
LS
99
99
99
100
100
Efficiency,
WF
9
38
44
58
85
%
RB
11
48
51
58
80
Overal 1
99
47
49
Refer to Table 4 for size category
TABLE 8. RELATIONSHIP BETWEEN DETENTION POND
PERFORMANCE AND SURFACE AREA RATIO
s
0
0
2
7
AR,%
.59
.74
.27
.51
Pond
System
WF
RB
LS+WF+RB
LS
TSS
41
(47)
62
(49)
79
(74)
93
(99)
Percent Remov
TP TKN
29*
(31)
36
(32)
53
(52)
45
(65)
22*
(24)
21
(25)
37
(38)
32
(50)
al , %
Zn
22*
(24)
32
(25)
51
(38)
80
(50)
Fe
22*
(24)
52
(25)
66
(38)
87
(50)
Numbers i n p a r en t he s so ba n e dby moeng .
* Estimated as fractions of TSS removal.
138
-------
SUMMARY AND CONCLUSIONS
The extent of urban runoff pollution is highly variable
and site specific, depending on the characteristics of the
drainage ares as well as rainfall intensity and duration.
The results of the investigation suggest that the water
quality improvement performance of Piedmont North Carolina wet
ponds can be correlated to the surface area ratio. To achieve
a minimum level of water quality improvement, the SAR must
maintain at 0.5% or higher. About 1.0%-2.0% of the watershed
area is needed for developing detention ponds to control 70%
or more of the sediment load; the corresponding levels of
control for nitrogen, phosphorus, zinc and iron would be 25%,
40%, 35% and 55%, respectively. The removal of nutrients is
less than satisfactory because of the presence of urban geese
at LS pond. The daily contribution of nutrients from geese
was estimated at 0.12 Ibs of nitrogen and 0.03 Ibs of
phosphorus per 100 geese. The rate of sediment accumulation
in wet ponds has not been investigated; however, research
conducted elsewhere indicates a 13% reduction of the storage
capacity may be expected over a 10-year period (2). Lead and
copper were not detectable either in storm samples or in pond
effluent by flame atomic absorption spectrophotometry
technique. The pollutant removal efficiency is highly
influenced by individual storms; therefore, the reported
performance should be regarded as the long-term average
performance of the Piedmont North Carolina wet detention
ponds.
The quality of storm runoff from the study area is
characterized by EMCs and a particle size distribution. The
EMCs are 135 mg/1 for total suspended solids, 0.88 mg/1 for
total Kjeldal nitrogen, 0.22 mg/1 for ammonia nitrogen, 0.14
mg/1 for total phosphorus, 0.10 mg/1 for ortho-phosphorus,
6.11 mg/1 for total iron, 0.24 mg/1 for soluble iron, 66 ug/1
for total zinc, and 20 ug/1 for soluble zinc. The average
concentrations of total suspended solids from the detention
pond effluent range from 10 to 50 mg/1, corresponding to
63%-93% reduction of the influent sediment concentration.
The average concentrations of other pollutant parameters from
the detention pond effluent approach the EMCs of storm runoff.
The removal of particulate pollutants is achieved by
sedimentation. The settling of fine particles requires a
longer detention time. The mean particle size (50 percentile)
in storm runoff is 6 micron (0.3 ft/hr) corresponding to the
very fine silt and clay soils in the Piedmont. Because of the
fact that sediment wash-off is affected by storm intensity,
the reported particle size of sediment should be considered as
a long-term average distribution.
139
-------
The degree to which Piedmont wet detention ponds reduce
hydrographic peaks was observed to be less than satisfactory
because local runoff enters near the downstream portion of the
detention ponds. A coordinated plan of housing development
near a detention pond is necessary to minimize the discharge
of local runoff into the lower portion of a detention pond.
In summary, the following conclusions can be stated.
(l)Although existing urban wet detention ponds in the
Piedmont region of North Carolina were not originally
built for the purpose of water quality improvement, they
were found effective for controlling urban runoff
pollution.
(2)The quality of storm runoff (EMCs) from the study area is
better than that reported by the National Urban Runoff
Program. The study site is representative of the Piedmont
urban setting with a good management for community
development.
(3)A performance relationship was observed, correlating
pollutant removal effectiveness and surface area ratio of
detention ponds. This permits incorporation of water
quality improvement requirements into design of wet
detention ponds in the Piedmont region of North Carolina.
(4)An EPA computer model was found to be reasonably useful for
initial sizing of urban wet detention ponds to achieve
targeted levels of pollution control. If local
meteorological, hydrological and soil properties are
available, the model could provide an estimation of the
long-term efficiency of sediment control using wet
detention ponds.
ACKNOWLEDGEMENTS
This project was funded in part by The Water Resources
Research Institute of The University of North Carolina.
Additional funding was made available through the Engineering
Major Grants program from The University of North Carolina at
Charlotte. The work described in this paper was not funded
by the U.S. Environmental Protection Agency and therefore the
contents do not necessarily reflect the views of the Agency
and no official endorsement should be inferred.
140
-------
LITERATURE CITED
1 Schueler, T.R., 1987. Controlling Urban Runoff: A Practical
manual for Planning and Designing Urban BMPs. U.S.
Metropolitan Washington Council of Governments.
2. Striegl, R.G. 1987. Suspended Sediment and Metals Removal
From Urban Runoff by a Small a Lake. Water Resources
Bulletin. Vol.23, No.4, pp.985-996.
3 . U.S. EPA, 1983. Results of Nationwide Urban Runoff
Program-Executive Summary. US Environmental Protection
Agency.
4. DEM. 1985. Toxic Substances in Surface Waters of the
Falls and the Neuse Lake Watershed. Report no. 85-08.
Division of Environmental Management, N.C. Department of
Natural Resources and Community Development.
5. Dally, L.K. 1983. Operation of Detention Facilities
for Urban Stream Quality Enhancement. M.S. Thesis,
University of Washington.
6. Camp, T.R. 1945. Sedimentation and the Design of
Settling Tanks. Trans. ASCE, pp. 895-936.
7. Rausch, H.G. and Heinemann, H.G. 1975. Controlling
Reservoir Trap Efficiency. Trans. ASCE, pp.1105-1113.
8. Griffin, D.M. Randall, C. and Grizzard, T.J. 1980.
Efficient Design of Stormwater Holding Basins Used for
Water Quality Protection. Water Research. Vol.14,
pp.1549-1554.
9. McCuen, R.H. 1980. Water Quality Trap Efficiency of
Stormwater Management Basins. Water Research. Vol.1,
pp.15-21
10. Wu, J.S. and R.C. Ahlert, 1985. A Trajectory Model for
Analyzing Sediment Trapping Efficiencies in Storm
Water Detention Basins. In: Proceedings Conference on
Stormwater and Water Quality Management Modeling. W.James
(ed.) CHI Report R149, McMaster University, pp.257-263.
11. Wu, J.S. and R.C. Ahlert, 1986. Modeling Methodology for
Dual-Function Stormwater Detention Basins. PB87-159711,
National Technical Information Service.
12. ASCE. 1985. Stormwater Detention Outlet Control
Structures. Task Committee on the Design of Outlet
Control Structures, ASCE.
13. USGS. 1977. National Handbook of Recommended Methods for
Water-Data Acquisition. US Geological Survey.
14. Wu, C.J. 1988. Performance of Urban Wet Detention Ponds,
M.S. Thesis, Univ. of NC at Charlotte.
15 SCS. 1983. Computer Program for Project Formulation:
Hydrology. Technical Release 20, US Soil Conservation
Servi ce.
16. APHA. 1985. Standard Methods for the Examination of Water
and Wastewater. APHA, AWWA, WPCF.
17. U.S. EPA, 1986. Methodology for Analysis of Detention
Basins for Control of Urban Runoff Quality-. EPA 440/5/87-
001, US Environmental Protection Agency.
141
-------
HYDROLOGIC DATA AUTOMATION USING AUTOCAD
by: Jim Y. Chang
Kiowa Engineering Corporation
Vice President
Denver, Colorado 80210-3809
and: James C. Y. Guo
University of Colorado at Denver
Assistant Professor
Denver, Colorado 80236
ABSTRACT
CUHPCAD is a computer program which serves as a control program to
link between AutoCAD and the CUHP program. The program was developed for
the purpose of hydrologic data automation. The program consists of two
parts. The first part is the CUHP.LSP program which analyzes and stores
the data generated by using AutoCAD, the second part is the CUHPCAD.BAS
program which can abstract the data from CUHP.LSP and perform data calcula-
tion and preparation of CUHP data input file. The CUHPCAD was tested in a
major basin study and proved to be efficient, accurate and very flexible.
INTRODUCTION
The concept of hydrologic data automation by utilizing personal com-
puter systems and AutoCAD^ was formulated by Kiowa Engineering Corporation
(KEC) in November 1987 to seek an efficient way to prepare data input to
the Colorado Urban Hydrograph Program^ (CUHP). In general, CUHP user need
to prepare storm and basin data as outlined in the user's manual. Storm
data can be either a storm distribution or total rainfall depth. Basin
data consists of basin area, weighted slope along flow path, flow length,
flow length from centroid to the outfall point, percent of imperviousness
for specific land use, storage loss for pervious area and impervious area,
soil infiltration rate for the initial and final conditions and it's decay
rate. Storm distribution data for the metropolitan Denver area can be
obtained from the Drainage Criteria Manual and rainfall depth data can be
obtained from the National Oceanic and Atmospheric Administration (NOAA)
Atlas. Basin data can be obtained by using appropriate basin maps, soil
maps, and land use maps. Basin maps can be studied for obtaining subbasin
area, flow path, and slope. Soil and land use maps can be used for
defining soil types and land use types for the determination of soil
infiltration rate and percent of imperviousness. A typical process for
preparing a completed data input for the CUHP includes the following
142
-------
steps: (1) basin delineation and assignment of basin identification
number, (2) determination of representative flow path, (3) identify flow
path segment for apparent slope conditions, (4) determine the centroid
point in the subbasin, (5) measurement of the area and length for each sub-
basin, (6) calculating the weighted slope, percent of impervlousness and
soil infiltration rate for each subbasin, (7) preparing storm data, and (8)
coding, typing, running, and checking for correct data input and reasonable
output. With the advancement of measuring tools such as digitized plani-
meter and the powerful spread sheet and word processing programs for the
personal computer, the preparation of data input to the CUHP is becoming
easier. If the study area contains only a few basins, the preparation of
data input from measuring to data recording is just a matter of a few hours
of work. Efficiency becomes a major concern only when dealing with a big
drainage basin which may consist of hundreds of subbasins. For such a
major basin hydrologic study, a good data management system is essential.
How to make the work easier to handle and still flexible enough for data
preparation and input is the main reason for the hydrologic data automation
using AutoCAD.
DEVELOPMENT OF CUHPCAD
CUHPCAD was developed as a result of the hydrologic data automation
concept. The main purpose of CUHPCAD is to provide a control program to
link between AutoCAD and CUHP. AutoCAD can be utilized to digitize the
basin map, soil map, land use map, and to record the basin data after the
measurement of area and distance by using a digitizer board. CUHPCAD con-
sists of two parts: (1) CUHP.LSP and (2) CUHPCAD.HAS (Reference 1).
CUHP.LSP is a computer program written in AutoLisp language which provides
a simple' way for the user to prepare the basin information. By following
the program prompt, the user can digitize all the areas related to basin
size, soil distribution, and land use. Also, the user can digitize all the
distance related to flow path length, flow path length from centroid to the
outfall point, and channel length for each slope condition. Table 1 is a
list of program prompts for the user to respond. The data file created by
CUHP.LSP can be linked in a later stage to the CUHPCAD.HAS program for
further data analysis and conversion to a CUHP input file. CUHPCAD.BAS is
a computer program written in BASIC language which can function as a data
process program to allow the user to input: (1) storm depth, duration, and
frequency, (2) storage loss for pervious and impervious area, (3) CUHP out-
put format, and (4) rational method option for basin area less than 90
acres. The user can also input project title, specify drawing scales and
edit data files as generated by the CUHP.LSP program. Table 2 shows the
data input steps for the CUHPCAD.BAS program. After completion of data
input and analysis, CUHPCAD.BAS will generate three summary tables which
^AutoCAD is a trademark of AUTODESK, Inc.
^CUHP is a PC version of the hydrolgic computer program developed by Ben
Urbonas with the Urban Drainage and Flood Control District, Denver,
Colorado.
143
-------
TABLE 1. GUHP.LSP DATA INPUT STEPS
Step 1: Enter a subbasin ID number.
Step 2: Enter the Elevation Drop or flowline segment 1 (or 2) (or 3).
Step 3: Measure Length of flowline segment 1 (or 2) (or 3).
Step 4: Measure Length to Centroid.
Step 5: Measure Basin Area.
Step 6: Enter Soil Type 1 (or 2) (or 3).
Step 7: Measure Subarea of Soil Type 1 (or 2) (or 3).
Step 8: Give an Imperviousness Percent for Land Use Type 1 (or 2) (or 3).
Step 9: Measure Subarea of Land Use Type 1 (or 2) (or 3).
include: (1) summary of data as measured by using CUHP.LSP, (2) summary of
data for all the CUHP parameters, and (3) completed CUHP data input ready
for process by the CUHP program. A sample of each of the output from
CUHPCAD.BAS is shown in Table 3. In order to access the program "CUHPCAD",
the user needs to have a personal computer system which includes a PC "XT"
or "AT", a digitizer, a color monitor, and AutoCAD software.
CASE STUDY
The CUHPCAD program has been tested and applied to a major drainage
basin planning project for the study of Second, Third, and Box Elder Creeks
in Adams County, Colorado. The study area is approximately 70 square
miles. Total subbasin number is 390. Soil types include A, B, and C
groups based on Soil Conservation Service's soil survey report. Land use
for existing conditions are mostly agricultural. Land use for future basin
conditions include the proposed Mew Denver Airport, the E-470 Beltway,
residential, commercial, and business developments (References 2 and 3).
As part of the project requirement, three baseline hydrological conditions
for the frequencies of 2-, 5-, 10-, and 100-year need to be modelled by
using CUHP which includes existing basin conditions, and future basin con-
ditions with or without the proposed New Denver Airport. All subbasin
delineation needs to reflect the existing and future basin conditions
because of land use differences such as the E-470 Beltway and the airport
runways. How to manage the enormous amount of data for all the various
basin conditions becomes a major concern. With the use of CUHPCAD, K.EC was
able to complete the hydrology portion of the above-mentioned project
144
-------
within allowable schedule and budget. Figure 1 shows a portion of the
study basin map generated by using AutoCAD. After the manual preparation
of the basin map, soil map, and the land use map, KEC was able to complete
the input of all maps to the computer and conduct the CUHPCAD operation for
20 basins within three to four hours. With the aid of CUHPCAD, KEC felt
that both accuracy and efficiency were significantly improved and revision
of basin conditions was more flexible.
CONCLUSION
CUHPCAD was developed for the promotion of the hydrological data auto-
mation concept. With the popularity of personal computer systems and
PC-CAD software, the CUHPCAD can be best utilized for either small or large
basin studies. This program has been tested for a major basin study by KEC
and concluded that the program can improve work efficiency and accuracy and
provide flexibility for basin changes.
TABLE 2. CUHPCAD.BAS DATA INPUT STEPS
Step 1: Give the Project Title.
Step 2: Give the starting Subbasin ID number.
Step 3: Give the last Subbasin ID number.
Step 4: AUTOCAD Screen Scale: 1 Unit = Z feet Z = ?
Step 5: Edit existing Data Files.
Step 6: Delete Data Files from Data Process.
Step 7: Summarize Subbasin Hydrologic Data Measurements.
Step 8: Create an Input Data File for CUHP Program.
Step 9: Summarize and Tabulate Subbasin hydrologic Data.
Step 10: Exit to DOS.
145
-------
ruu 3. SDNNUI or CIIKU.UI OITPIT
uiu lion UOK uci now.ui:
UIII
ID
K
1
2
3
run uiu n miis UD mil iioru u t
1ST
ruun
IEKTI
1.57
LI)
1.91
1ST
IIOPI
LI3
l.ll
I.S7
UD
rWlIH
IUGTI
l.ll
1.3)
l.ll
210
lion
l.ll
1.41
l.ll
31D
ruim
IUGTI
l.ll
I.2S
l.ll
3ID
lion
Lit
l.ll
l.ll
CltO
1C
IIKTI
l.3«
1.31
Lit
BUII Mil TIPIS UD DIITIHITIOIS:
UIII
ID
M
1
2
3
MU mu UD mu ructmci
in
Mil
nri
c
c
c
1ST
Utt
t
111.11
111.11
111.11
210
Mil
Tin
.
.
-
2ID
uu
»
l.ll
Lit
l.ll
3ID
Mil
nri
.
.
-
31D
UU
t
l.ll
l.ll
l.ll
torn
uu
14 Kill
LS)
I.4S
l.3t
USIl UIOISIS UD TIIII IMPUVIOIJIISI:
USIl
ID
M
1
2
3
uioisis uo mu Mpnviouisj t
in
IW
t
2.11
2.11
2.11
1ST
UU
t
111.11
111.11
111.11
2ID
UP
V
l.ll
l.ll
l.ll
210
UU
I
l.ll
l.ll
l.ll
31D
IMP
\
l.ll
l.ll
l.ll
310
UU
t
l.ll
l.ll
l.ll
n'o
ur
\
2.11
l.ll
2.11
IUKUI or usii noioiocic ricTou
UIII USIl UUl VIICniD IIOKQUCIC riCTOU
ID Tom roTiiuicn uiii BUII in riui urn mis i*m
M uu riot TO n'D n'o ro re i toss loss
UGTicmiD tun IMP
SQH Kill Kill R/R \ II/IR II/U 1/HC 1ICI IICI
1
2
3
LS)
1.45
1.34
LS;
1.73
I.M
1.14
1.31
l.fl
l.lll
1.113
l.llf
2.111
2.111
2.111
CIIP IIPOT UT1
3.141
3.441
3.111
nu
4.544
4.544
1.544
1
1
1
.1111
.1111
.1111
L)S
I.3S
1.35
l.ll
l.ll
l.ll
2 n PIOJICT
1
2D)
71 1 15.1
LS) I.S7
71 1 1S.I
I.4S 1.73
11 1 1S.I
L3i L)l
1
1
.34 2.IILII!
2
.31 2.III.II3
3
.11 2.ULIK
l.ll 1.35 3.44.44U LSI
l.ll 1.35 3.44.1414 LSI
l.ll L3S 3.11.1111 LSI
146
-------
Figure 1. Box Elder Basin - Future Condition,
147
-------
ACKJMUWLEDGMENT
As part of KEC's in-house research and development team, Dr. James Quo
was responsible for the conduct of computer programming for CUHPCAD.
Appreciation was extended to Urban Drainage and Flood Control District,
Adams County, the City's of Aurora, Brighton, Commerce City, and Denver,
and the office of the New Denver Airport. Because of their award of the
Second, Third, and Box Elder Creeks project to KEC, KEG was able to develop
and test the CUHPCAD program with enthusiasm.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
REFERENCES
1. Hydrologic Data Automation between AutoCAD and CUHP Computer Software.
Dr. James Guo and Kiowa Engineering Corporation. 1988.
2. Hydrology Report for the Second Creek, Third Creek, DFA0053 and Barr
Lake Drainage Basin Planning Study. Kiowa Engineering Corporation.
1988.
3. Hydrology Report for the Box Elder Creek Watershed in the Proposed New
Denver Airport. Kiowa Engineering Corporation. 1988.
143
-------
DISTRIBUTED RAINFALL-RUNOFF MODELING
BASED ON DIGITAL MAP DATABASE
Lynn E. Johnson, Charles Huffman
Department of Civil Engineering
University of Colorado at Denver
1200 Larimer St.
Denver, CO 80204
ABSTRACT
A cascade-of-reservoirs rainfall-runoff model, called MAPHYD, has been
developed which incorporates automatic generation of model input data using
geographical information system functions integrated into a user-friendly
workstation. Digital terrain modeling is accomplished to obtain slope and the
direction of slope for a user-selectable grid cell resolution. Associated data
on soils and infiltration characteristics, land use and percent
imperviousness, antecedent moisture conditions, and rainfall distribution are
also mapped as digital color images accessable directly by the rainfall-runoff
simulation routines. Output results of the spatial distribution of runoff at
each time step are displayed using an interval color scale. Input, retrieval,
and editing of the digital database is accomplished by interactive computer
graphics techniques.
INTRODUCTION
Hydrologic modeling efforts have been hampered by limitations on data and
processing for input to hydrologic models. Recently developed methods for
digital data capture, image processing and interactive computer graphics, and
geographic information system software provide the tools to incorporate
greater spatial and temporal detail into deterministic rainfall-runoff models.
Hydrologic model parameters are predominantly derived from mapped
attributes of the land. Watershed topography, soils, land use and rainfall
patterns are the primary data sets of interest. From these can be derived
model parameters of drainage pattern, slope, infiltration and other
abstractions, soil moisture, and rainfall inputs.
The MAPHYD model has been developed as an integrated geographic
information systems' software and hardware system for watershed data
management and simulation. MAPHYD has automated functions for digital terrain
modeling as a preprocessor of input data directly linked to distributed
rainfall-runoff model bassed on a cascade-of-reservoirs algorithm.
149
-------
MAPHYD is demonstrated using the Lena Gulch watershed, an
urbanizing drainage system located on the west side of Denver,
Colorado. Lena Gulch has an area of 25 sq. km. The watershed is
instrumented with radio signal reporting rain and stream gages as
part of a flood warning program developed by the Denver Urban
Drainage and Flood Control District.
GEOGRAPHIC INFORMATION SYSTEMS
Geographic informations systems (GIS) are computer software
and hardware systems dedicated to development, processing,
archival, and retrieval of spatially distributed data. The x, y, z
coordinate data are displayable on computer graphic screens and
hardcopy outputs, a capability which aids understanding, error
detection, and operator control of geoprocessing and hydrologic
prediction software. The data coordinates also provide a primitive
attribute key to other alphanumeric data archived in a database
(e.g. streamflow records).
GIS software and hardware tools can be integrated into a
flexible, adaptive interactive computer graphics based data
manager supportive to automatic modeling of watersheds. Since
there are many subjective aspects of watershed modeling and flood
forecasting, the interactive and user-friendly character of the
MAPHYD workstation permits the operator to incorporate their
judgements into the database development, and to control the
rainfall-runoff analysis process.
MAPHYD has integrated geographic information systems'
functions for watershed data management. Digital maps of watershed
characteristics are obtained using direct entry imaging, manual
digitizing, and remote sensing methods. Source maps typically
include maps of topography, soils, and land use/zoning. The maps
may be obtained in digital formats for some attributes (e.g.
digital terrain model) or as paper stock obtained from a local
planning agency. Radar-rainfall imagery is obtained by networking
to a central mainframe computer.
Once in digital format it is possible to archive the digital
data sets, and retrieve selected data to support rainfal1-runoff
computations. The maps are preprocessed for scale resolution,
drainage system definition, and parameter estimations. Soil
moisture accounts are updated periodically between rainfall
events. Map overlay methods are used to determine composite runoff
characteristics and to input rainfall distribution.
MAPHYD is implemented on a desktop personal microcomputer,
having 1.6 mb RAM (random access memory), math coprocessor chip,
and MS-DOS 3.1 operating system. Graphic displays are executed by
a Vectrix VX/PC graphics card set with high resolution color
monitor providing 672 x 480 pixel (picture element) resolution and
150
-------
nine bitplanes of graphic memory. Enhanced configuration allows
512 displayable colors selectable from 16.8 million. A digitizing
tablet with four-button puck enables a user-display interface that
functions as a command pick device and allows the user to trace
watershed characteristics onto a digital base map. Keyboard input,
displayed on a separate monochrome monitor, and output data
selections are minimized using the menu command approach. Total
cost of the workstation is less than $10,000.
The workstation functionality is representative of a new
class of water resource decision support system (Johnson, 1986).
Graphics software developed for computer-aided planning (French
and Taylor, 1986) was used to provide utility routines for menu
definition and selection, image storage and retrieval, map
mosaicing and color overlays.
Tablet digitizing and database interaction capabilities are
most helppful in hydrologic modeling. Data for large and small
areas can be edited for model parameter modifications as part of
calibration and forecasting operations.
Image digitizing is a means of providing a representation of
an image (e.g. photograph, map, or remote scene) in a computer
compatible form. The image, or raster, format is in contrast to
the vector format noted above. Image digitizing methods require
special-purpose computer hardware which converts light intensity
observed at a particular location within a scene to a digital
representation with the attributes color, hue, and intensity. The
digital imagery can be interpreted as data, or as a location
framework for display of other manually digitized, imported or
computer-generated data.
DEFINING HYDROLOGIC SYSTEM CHARACTERISTICS
Base maps of the Lena Gulch watershed were input using video
digitized topographic maps and mosaiced to form the watershed base
map (Figure 1). The interactive computer graphics system provides
an easy means for inputting, editing and displaying data', and
integrating computational algorithms directly with the digital
database.
There are a spectrum of digital maps which portray the
watershed and storm data to be processed: composite soil/landuse,
elevation, groundwater, upslope distance, and precipitation. In
MAPHYD, many data can be represented as colors, and input/editing
of watershed data can be accomplished using "painting" functions.
Colors are overlaid as transparencies on the gray scale base map,
a technique which provides immediate visual feedback on location.
The figures show pictures of the digital datasets used for
the Lena Gulch watershed simulation modeling. Figure 1 is the
151
-------
digital base map obtained using a video image capture or direct
entry camera device. Separate maps can be mosaiced to form
a watershed base map upon which (color-coded) land attribute data
of various types can be registered and applied.
Topography of the Lena Gulch watershed is defined by color
coding of the elevation contour bands. The highly refined digital
terrain model shown was developed in order to obtain as much
detail on elevations as possible - this to support experiments in
which the spatial resolution of the terrain data was degraded and
the impact on the predicted runoff hydrograph observed.
The elevation map is processed using digital terrain modeling
algorithms to obtain slope parameters (Figure 3) and runoff
directions and distances (Figure 4). Slope has a strong influence
on the hydraulics of overland and channel flow, and is an
essential parameter for rainfall-runoff modeling. Aspect map
displays the facing compass direction and is derived from the
digital terrain model. Aspect determines the direction of flow,
one cell to another, and aspect data can be used to define the
drainage pattern.
A basic digital dataset to support infiltration and other
abstractions accounting is the soil map (Figure 5). Here the SCS
hydrologic soils group data has been hand-digitized on the
watershed base map. There are two different soil parameters -
porosity and saturated hydraulic conductivity - for each soil
group.
Landuse data (Figure 6) play a very important role in
computing the excess precipitation from a watershed. Tn areas of
high density land development, the calculated percent of
imperviousness will reflect areas of little or no infiltration
rapacity. The MAPHYD procedure involves digital mapping of
commercial, high density residential, low density residential, and
open space landuse types. Four levels of land imperviousness were
used to describe the watershed's landuse. The program's default
settings are 90, 70, 50, and 2 percent impervious for commercial,
high density residual, low density residual, and open space,
respectively (Denver Regional Council of Governments, 1975).
The soils and landuse maps are logically combined to form a
composite runoff potential map (Figure 7). The runoff potential
map derived reflects a matrix of combinations of soil and landuse
categories having distinct pervious-impervious characteristics.
If only point rainfall data are available at sites of rain
gages it is standard hydrologic analysis procedure to determine
the area assignments associated with each gage. The so-called
"Thiessen polygon" procedure is readily implemented as a G1S
function (Figure 8) as are other geometry-based algorithms.
152
-------
Rainfall distribution can also be defined using
radar-rainfall imagery (Figure 9). The overlay operation shown
provides great detail on the spatial distribution of rainfall.
Storm duration and time distribution of rainfall can also be
defined by moving the image using the interactive graphic data
manipulation techniques.
Runoff magnitude (Figure 10) and other model-generated data
can be displayed on the base map and provide a quick visual key of
the location of high flows. The runoff distribution pattern shown
was generated by the MAPHYD cascade-of-reservoirs model, discussed
below,
MAPHYD CASCADE-OF-RESERVOIRS MODEL
A spectrum of hydrologic models have been integrated into
MAPHYD, including: (a) time-area, (b) unit hydrograph, (c)
partial-area method, and (d) cascade-of-reservoirs. MAPHYD's
cascade-of-reservoirs model is of primary interest here. The other
modeling activities are described elsewhere (Johnson and Dallmann,
1987); Toms and Johnson, 1988). MAPHYD hydrologic simulation
models perform calculations of contributing areas, infiltration
rates, soil moisture, evapotranspiration, water table positions,
and runoff routing using digitized maps. The models use parametric
equations for areas of infiltration, groundwater fJow, rainfali
distribution, and initial abstraction losses.
MAPHYD's cascade-of-reservoirs model involves complete
integration of the digital terrain model and preprocessing
products of slope and aspect with rainfall, infiltration, and
overland and channel flow hydraulics. Here, each cell is treated
as a reservoir with the basin treated as a collected of
reservoirs, each cascading downslope one to another (Chow, 1964).
Although computationally intensive, this distributed parameter
model is physically based and permits simulation of the runoff
hydrograph at any location in the basin.
All models are implemented using the interactive command
programming approach. That is, geographic data processing for
model parameter estimations, as well as model procedure selection
arid activation, are controlled using menu commands. If new
geographic data are to be entered, the command program accesses
the geographic database program which in turn provides the means
of displaying, editing, and storing the digital spatial data. The
command program also provides access to any of the available model
functions for analyzing and using the data. Built-in "Help"
commands display user instructions and serve as an interactive aid
for teaching watershed modeling methods.
The MAPHYD cascade-of-reservoirs model has been checked
against an accepted rainfall/runoff model. Storm hydrographs
153
-------
obtained from the CUHP model for Lena Gulch watershed compared
veil with the results generated by the MAPHYD models in terms of
the shape of the hydrograph and peak discharge. Calibration and
verification of the model will require further testing using
monitored storm data. Extensive simulation runs were conducted to
test the model's sensitivity to changes in resolution, friction
and storage parameters on the resultant peak flow, volumes, and
timing. These results are described by Huffman (1988).
CONCLUSIONS
Development and use of MAPHYD demonstrates the 1 inked
capabilities for digital data management, simulation and graphic
communication of flood runoff conditions. MAPHYD provides useful
and powerful tools for watershed modeling. Higher productivity is
believed realized for watershed database development, modeling
research, and its use for hydrologic design and real-time flood
monitoring and prediction applications.
MAPHYD has the capability to quickly update hydrologic
parameters either by interactive user control or through
computation. Soil and land use parameters can be initialized once
and updated multiple times in response to development and
development proposals. Real-time data can be accessed and
displayed to provide decision support functions useful for flood
forecasting.
MAPHYD is more flexible than the unit hydrograph procedure
and accounts for more variables. In particular, spatial variations
in impervious lands and spatial and temporal variations in
rainfall are shown to have a significant effect on the magnitude
and timing of flood runoff. Basin averaging and lumped parameter
modeling approaches can introduce substantial error in the runoff
hydrographs. Also, it has been demonstrated that distributed
parameter modeling can be accompJished rapidly on a microcomputer,
a factor contributing to use of MAPHYD for real-time flood
forecasting operations.
ACKNOWIJDGEMENT
Partial support provided by National Science Foundation Grant No.
ECE-8513122. The work described in this paper was not funded by
the U.S. Environmental Protection Agency and therefore the
contents do not necessarily reflect the views of the agency and
no official endorsement should be inferred.
REFERENCES
Chow, V.T. 1964. Handbook of Applied Hydrology. McGraw-Hill pubs.
New York, pp 14-1 to 14-54.
154
-------
Denver Regional Council of Governments. 1975. Urban Storm Drainage
Criteria Manual. March.
French, P.N. and M.R. Taylor, 1986, Computer-Aided Planning Libray
Programming Guide (CAPLIB). Resources Planning Assoc., Ithaca,N.Y.
Huffman, C. 1988. Digital Terrain Modeling for Distributed
Parameter Watershed Modeling. M.S. Thesis. University of Colorado.
Johnson,L.E. 1986. Water Resource Management Decision Support
Systems. ASCE J. Water Resources Planning and Mana. Vol.112, No.3.
pp 308-325. July.
Johnson, L.E. and J. Dallman. 1987. Flood Flow Forecasting Using
Microcompuer Graphics and Radar Imagery, in MICROCOMPUTERS IN
CIVIL ENGINEERING, An International Journal. Elsevier Pubs., vol.
2. Number 2. June.
Toms, E.A. and L.E. Johnson. 1988. Parital-Area, VAriable-Source
Rainfall-Runoff Model. ASCE Hydraulics Specialty Conference.
Colorado Springs, Colorado. August 8.
155
-------
Figure 1. Video image capture techniques are used to obtain a
digital base map of the Lena Gulch watershed area.
Figure 2. Digital terrain model of Lena Gulch watershed
defines elevations for each picture element.
156
-------
Figure 3. Slope classification map is generated from digital
terrain model.
Figure 4. Aspect map displays the facing compass direction
and is derived from the digital terrain model.
157
-------
Figure 5. Soils data are obtained from SCS soils maps
categorized into four hydrologic soils groupings per their
infiltration capacity and rate.
Figure 6. Land use data is obtained from local land use and
zoning maps and is digitized onto the watershed base map.
158
-------
Figure 7. Overlay of the land use attribute map onto the
soils map generates a composite map defining runoff potential.
Figure 8. Theissen polygon procedure is implemented to
determine the watershed area associated with each rain gage.
159 .
-------
Figure 9. Overlay of radar-rainfall image onto the watershed
provides an efficient means for defining the spatial
distribution of rainfall and its movement.
Figure 10. Spatial distribution of runoff resultant from
rainfall event is displayed (color coded) at each computation
time step.
160
-------
PC-SYNOP. A Rainfall Analysis Tool
By: Eric W. Strecker, Eugene D. Driscoll, and Gary E. Palhegyi
Woodard-Clyde Consultants
500 12th Street, Suite 500
Oakland, California 94607
ABSTRACT
This paper reports on the porting of the EPA Synoptic Rainfall Analysis Program (S YNOP)
to the Personal Computer and describes and illustrates some useful features that have been added
to the program. S YNOP analyzes an hourly rainfall record that can be obtained on floppy disk from
the National Climatic Data Center for rain gages located throughout the country. It generates a
variety of storm event based summary statistics. By specifying a minimum dry period that separates
successive storms, S YNOP groups the hourly rainfall record into storm events and analyzes the
resulting record. A user friendly interface has been provided for the selection of computational and
printout options. Some of the added features include the ability to (1) select either a water year or
calendar year organization, (2) to conduct an analysis on a seasonal basis (useful for dealing with
the pronounced wet and dry seasons in some areas of the country), (3) to exclude from the statistical
analyses all storm event volumes that are less than some user specified minimum (used to determine
the characteristics of only those storm events that will produce runoff), and (4) the ability to write
out a separate file containing storm event information on all storm events in the record for down
loading to a spreadsheet or statistical analysis program. The paper also presents some comparative
analyses that examine the effect and some of the implications of specifying a minimum storm event
volume that generates runoff.
INTRODUCTION
SYNOP is a computer program that reads and analyzes a long term rainfall record,
segregates the data into discrete storm events, and outputs statistical properties of the storm event
parameters, volume, intensity, duration, and time between storm event midpoints (delta). It was
developed by Hydroscience, about 10 years ago, as an element of an EPA contract (1), and was
subsequently maintained at different times on the mainframe computers of EPA and the USGS, and
161
-------
at the University of Florida. The program was relatively inconvenient to access via this route
particularly by parties not actively involved in programs with these agencies or having some type
of working arrangement with the University.
The growth of microcomputer applications of many programs formerly developed for
mainframes naturally extended to the SYNOP program. The most notable example was SYNOPs
incorporation into the PC version of the SWMM model (2). The availability of hourly rainfall
records from the National Climatic Data Center on floppy disk increased the practicality of having
a PC version of SYNOP.
A study that we conducted for the Federal Highway Administration where we have
developed a methodology for evaluating the water quality impacts of stormwater runoff from
highways called for the use of SYNOP-type rainfall event statistics in some of the procedures.
Typical values for the rainfall parameters of interest were available from summaries prepared using
SYNOP results obtained from a large number of analyses performed at various times over a number
of years (3). However, results of many of these analyses are now out of date, as many of the records
analyzed did not include data from the past 10 years. In addition, the results obtained from the few
west coast and southwest stations included in these summaries were questionable because of the
inability of the original program to deal effectively with a record exhibiting pronounced wet and
dry seasons. Finally, while the "typical" regional parameter values presented in the available
general summaries might be suitable for broad scale screening type analyses, more accurate
parameter values obtained from the analysis of a local gauge are preferred.
The FHWA recognized the importance of the latter consideration, as well as the practical
value of making the program readily available to users. They included in our contract the task of
providing a PC version of the SYNOP program. This paper describes the program, including a
number of features that were incorporated to make it user-friendly and to expand its capabilities in
several areas. Tables are presented that illustrate examples of the program outputs and examine the
information that certain of the new features make it possible to evaluate.
DESCRIPTION OF PROGRAM
Rainfall (and the runoff it generates) may be viewed as a series of independent, randomly
occurring events as shown in Figure l(a). This representation can be further simplified by
schematizing each event as a uniform, rectangular hydrograph as shown in Figure 1 (b). Each event
is characterized by its duration, volume, average intensity, and the time elapsed since the last event
(interevent time). The interevent time is the time between event midpoints. The rainfall event
statistics that are desired are summarized in Table 1. Here, the coefficient of variation (ratio of
standard deviation to the mean) is used in place of the standard deviation in order to have a
convenient dimensionless parameter representing variance. The SYNOP program computes
statistics of the four parameters given in Table 1. from long-term hourly rainfall records. For
162
-------
fe
CO
LLJ
Z
Mini mum dry period for
seperating events
Event
>l
0
TIME
(a) Actual Record and Event Deliniation
•z.
LLJ
DC
0
M-Duration-*
Average
intensity
Time between
event midpoints
Volume
0
TIME
(b) Simplified Representation Used in SYNOP
Figure 1. Actual and simplified representation of independant rainfall events.
163
-------
Table 1. Rainfall/runoff event parameters and statistics.
PARAMETER
INTENSITY
DURATION
VOLUME
UNITS
(L/T)
(T)
SYMBOL FOR
EACH EVENT
q
d
TIME BETWEEN
EVENT MIDPOINTS (T)
EVENT
MEAN
QP
DP
VP
TP
COEFFICIENT
OF VARIATION
CVQP
CVDP
CVVP
CVTP
Table 2. Regional differences in typical rainfall statistics.
REGION
VOLUME
(inches)
MEAN CV
NORTHWEST
ROCKY MT
NORTHEAST
SOUTHEAST
0.45
0.20
0.40
0.45
1.5
1.6
1.5
1.6
INTENSITY
(in/hr)
MEAN CV
0.02 0.9
0.04 1.0
0.08 1.1
0.12 1.3
DURATION
(hours)
MEAN CV
20 1.3
4 1.2
6 1.0
5 1.3
INTERVAL
(hours)
MEAN CV
100 1.0
100 1.0
80 1.0
72 1.0
example, representative values showing regional differences for these rainfall statistics are shown
in Table 2.
An assumption of the SYNOP analysis is that the hourly record will be aggregated into a
series of independent events. The usual approach used for convenience in processing hourly
rainfall data is to choose a minimum dry period, MDP, such that rainfall values separated by less
than MDP are considered part of the same storm. Storms separated by times greater than or equal
to MDP are considered to be independent events. Figure l(a) demonstrates this. Several methods
for choosing MDP exist (4), but the most common is to assume that interevent times are gamma
distributed (4; 5; 6). When the interevent times have a coefficient of variation equal to one, they
are exponentially distributed, which is a special case of the gamma distribution. Thus, trial values
of MDP are chosen until the coefficient of variation of time between event midpoints (interevent
time, Table 1) equals 1.0. When the resulting interevent times are exponentially distributed, they
are considered to be independent events. Resulting values of MDP are usually in the range of 3 to
24 hours (1). Although for western gauges, we have noted MDPs as high as 90 hours for the summer
dry period.
SYNOP sorts through the rainfall record with user inputted bounds on the MDP and
interpolates to an estimated MDP that would produce an exponential distribution of interevent
164
-------
times. The selection of the MDP bounds is a trial and error process where before results are
considered to be acceptable, the interpolated MDP should be between the user specified bounds.
Once the proper MDP is found, the SYNOP program generates the statistics shown in table 1, (a)
for all events in the record, (b) for each individual year, and (c) for the record stratified by month
(i.e., all Januaries, all Julys, etc.).
The SYNOP program is essentially the same program that was available on the mainframe,
but some enhancements have been added and minor errors have been corrected. The program is
available on 1.2 Megabyte Floppy Disks. It requires 640K of RAM and a hard disk is recom-
mended.
SPECIAL COMPUTATIONAL FEATURES ADDED
Some of the special computational features that have been added to the SYNOP program
are described briefly and their purpose discussed, in this section.
Wet/Dry Season Analysis. The program permits the user to specify the starting and ending
months in a year that will be considered in the analysis. This option can be used to perform
completely separate analyses for wet seasons and dry seasons such as exist in the west and
southwest. The number of storms and the dry intervals between events are radically different in
these two seasons, and the statistics so produced provide a much more reliable characterization of
the rainfall. Previously, the substantial seasonal differences resulted in a distorted overall
characterization of storm event properties.
Calendar Year or Water Year Breakdown. This option allows the user to organize outputs
on either a water year or calendar year basis. This feature is more a convenience than a substantive
computational modification. It has proved to be helpful in situations where it was desired to make
comparisons of long-term monthly rainfall and streamflow data for use in NPS assessments. It
makes it much simpler to assemble water year stream data and calendar year rainfall data on a
common basis. The potential value of such comparisons is illustrated by Figure 2 developed for
an NPS project we are presently working on. The data plotted are the annual volumes of rainfall
and runoff based on 36 years of record for the Guadalupe River. The varying pattern for the amount
of the rainfall that shows up as stream flow in this largely urbanized watershed, is attributed to a
combination of upstream reservoir storage/release practices, to changes in soil moisture over the
years, to water table changes over the years, and to the average intensity of rainfall in any one year.
We are presently analyzing data to identify the significance of each of these factors. The results
will be used to assist in the calibration of a SWMM model of the area.
Minimum Event Size. One of the added features is the ability to filter out all storms smaller
than a user-specified minimum volume, and compute the statistical parameters of the larger storm
events expected to produce runoff. This feature responds to the concern that applying a runoff
coefficient to the complete rainfall record provides a biased estimate of the statistics for runoff
165
-------
40
30-
0)
CD
"P 20 H
10
[J Rainfall (in)
HjH Runoff (in)
i O) O> O) O) O O) i
Year
Figure 2. Annual rainfall vs. runoff volume in inches for the Guadalupe River, San Jose.
flows and volumes. Most of the very small storms (e.g., those that produce total depths less than
about 0.06 to 0.12 inch) are completely absorbed and generate no runoff at all. Schueler has made
this observation for the Washington D.C. area (7), as has Urbonas for the Denver area (8).
The general effect of the removal of a group of small values from a data set is that the value
of the mean will increase, and the coefficient of variation will decrease. The higher the cutoff level
selected, the greater will be the percentage change in the statistic. Another effect is that the number
of storm events will decrease. This can have a significant impact on the prediction of the frequency
of exceedance of water quality standards due to stormwater runoff.
Based on some preliminary results from the SYNOP program, some analyses made made
using artificial data sets, and suggestions offered by Urbonas (8), an initial guess was made that the
mean values of the rainfall statistics will increase by about 30 percent, and the coefficients of
variation will decrease by 15 percent. These conditions were employed in a sensitivity analysis to
assess the degree to which elimination of the small storms that produce no runoff might affect
predictions of detention basin performance using the wet pond analysis methodology developed by
Driscoll (9). It was found that the change in rainfall statistics had little impact on the overall
performance of the basin. Table 3 presents an example detention pond analysis using LAX rain
gauge wet weather statistics.
Inspection of the general performance relationships shown by Table 3 indicates that the
changes in statistical characteristics tend to compensate. The higher mean values reduce efficiency,
166
-------
Table 3. Effect of elimination of small storms on performance predictions for the
LAX rain gauge (wet season)
RAINFALL EVENT STATISTICS
ALL STORMS
MEAN COV
W/O SMALL STORMS
MEAN COV
VOLUME inch
INTENSITY in/hr
DURATION hours
INTERVAL hours
0.71
0.040
20.0
237
1.51
0.99
1.27
1.00
0.90
0.048
24.4
234
1.24
0.87
1.07
1.00
WET POND PERFORMANCE ESTIMATES
BASIN %REMOV %REMOV %REMOV
SIZE RATIO DYNAMIC QUIESCENT COMBINED
% REMOV % REMOV % REMOV
DYNAMIC QUIESCENT COMBINED
0.10%
0.20%
0.33%
0.50%
1.00%
2.00%
13
16
18
19
22
25
4
7
12
18
33
53
17
22
27
34
47
65
14
16
18
20
23
26
3
6
10
15
29
51
16
21
26
32
45
64
NOTES-
Value shown is the ratio of the basin surface area to the area of the
catchment that contributes runoff (expressed as a percentage).
Computations assume an average basin depth of 3 feet, and
but the lower coefficients of variation operate to increase efficiency. The analysis compared arange
of basin size ratios with completely impervious catchments through a 3 foot deep basin. Table 3
compares the performance predictions for the two sets of rainfall/runoff statistics. It is seen that for
the assumed changes, there is no significant effect on long term performance efficiency. Additional
testing, using better estimates of the actual changes in the rainfall/runoff statistics for different
regions of the country will be necessary before concluding that these results are generally true.
The effect of removal of the small storms may have a much greater effect on the statistics
of reported rainfall volumes if another distribution, other than the normal, is used to describe the
population of rainfall volumes. As an example, Figure 3 presents probability plots of the rainfall
167
-------
Storm Volume (In.)
1
LRX Ulet Season No Minimum Uolume.
MEDIRN= 0.28
COU = 3.22
MERN = 0.94
PPCC = 0.983 ' '
99.9
Storm Volume (in.)
10
PERCENT EQUAL OR GREATER
LRX UJet Season .10 inch Minimum Uolume.
MEDIRN= 0.55
COU = 1.25
MERN = 0.89
PPCC= 0.989
99.9 99 95 90 80 50 20 10 5
PERCENT EQUAL OR GREATER
Figure 3. Comparison of the effects of specifying a minimum storm volume for analyses
168
-------
event volumes for wet season storms (October through March) at the LAX rain gauge over a 38 year
period. A lognormal distribution was assumed for the storm volumes. The data where obtained
from a SYNOP evaluation of the gauge, where storm event data were written to a separate file
(another new option). The upper plot assumes no minimum volume, while the lower plot has a 0.10
inch minimum volume. Note that the median volume storm doubles, but the coefficient of variation
decreases by a factor of 2.5. Note also that the lognormal distribution seems to fit the data better
where the minimum storm volume has been applied.
OTHER PROGRAM OPTIONS
The program still retains a number of other features which will be briefly described here.
The program can transform the event volumes and intensities to lognormal. So that the lognormal
distribution can be assumed for these parameters. There are a number of printing options, including
the ability to print hourly rainfall data, storm event summaries for each storm, and line printer
statistic and probability plots. The user may also have output printed for all runs of SYNOP or for
only the last run where the interpolated MDP is applied.
EXAMPLE PROGRAM OUTPUTS
This section presents some examples of the type of output that SYNOP provides. An
analysis of the LAX Rain gauge (#045114) for 39 years of hourly data. The analysis was com-
pleted by evaluating a separate analysis of the wet and dry seasons. For purposes of illustra-
tion, wet season results are presented here.
Presented in Table 4, 5, and 6 are illustrations of the some of the SYNOP outputs.
These talbes are based on an analysis of the LAX rain gauge data for a 38 year record. Table 4
presents a summary of the rainfall statistics by month for the period of record from the same
analysis. Presented in Table 5 is a abbreviated summary of rainfall statistics for each year.
Table 6 presents the overall summary of rainfall statistics of storms for the period of record. To
illustrate the type of information that can be assembled from SYNOP analyses, a summary
table indicating the rainfall statistics for the National Weather Service Portage Bay, Seattle
Gauge (#457458) is presented in Table 7.
Table 4. Summary of rainfall statitics by month (for period of record) for LAX rain
gauge for wet season months
MONTH DURATION INTENSITY VOLUME DELTA
AVERAGE COEF VAR AVERAGE COEF VAR AVERAGE COEF VAR AVERAGE COEF-VAR
1.10 257.94 1.06
1.37 237.16 .94
1.04 203.66 .89
.91 306.32 1.00
1.07 161.29 .54
1.00 218.03 1.00
169
1.
2.
3.
4.
11.
12.
27.97
33.72
19.65
18.91
22.28
21.70
1.05
1.13
1.05
1.03
.83
.86
.0484
.0521
.0484
.0434
.0578
.0412
.6997
1.1715
.8590
.7776
.8087
.7169
1.08
1.28
.65
.55
1.03
.77
-------
Table 5. Summary of rainfall statistics by year (for the period of record) for the
LAX rain gauge.
YEAR DURATION
AVERAGE COEF VAR
1.36
.56
.75
1.08
.85
.78
.66
1.20
1.31
.95
.93
49.
50.
51.
52.
53.
81.
82.
83.
84.
85.
86.
20.20
21.83
16.33
27.89
17.79
23.80
24.67
28.35
10.64
13.86
28.07
INTENSITY
VERAGE
.0770
.0400
.0501
.0430
.0445
.0448
.0409
.0460
.0720
.0636
.0470
COEF VAR
1.2761
.6983
.8538
.6605
.6944
.4706
.8023
.6217
.4952
.7973
.8046
VOLUME D;
AVERAGE
.49
.75
.51
1.04
.59
.82
.84
1.07
.55
.61
1.20
COEF VAR
.79
.63
.58
1.08
.82
.63
.83
1.36
.84
.91
1.09
AVERAG;
141.64
263.50
222.27
203.35
216.50
195.28
160.71
158.50
246.82
189.62
235.71
DELTA
.73
.14
.94
.96
.89
.69
.63
.89
.66
.76
.91
Table 6. Rainfall statistics by storm (for the period of record) for the LAX rain-
gauge.
NUMBER
TOTAL MINIMUM MAXIMUM AVERAGE STD DEV VARIANCE COEF-VAR
DURATION 476. 11597.000 1.000 201.000 24.3634 26.145 683.579 1.073
INTENSITY 476. 23.056 .003 .340 .0484 .042 .002 .874
VOLUME 476. 430.283 .110 9.130 .9040 1.118 1.250 1.237
DELTA 438. 102521.00 30.500 1523.500 234.0662 232.921 54252.240 .995
The work described in this paper was not funded by
the U.S. Environmental Protection Agency and
therefore the contents do not necessarily reflect the
views of the Agency and no official endorsement
should be inferred.
170
-------
Table 7. Rainfall event statistics for NWS Portage Bay, Seattle Gauge #457458
Month
Wet Season,
October
November
December
January
February
March
Wet Season
Statistical
Summary
Dry Season,
April
May
June
July
August
September
Dry Season
Statistical
Summary
Monthly Event Volume
No. of Average Volume Average C.V.*
Storms (in.) (in.)
Minimum Dry Period Between Storms =
6.29
10.21
9.36
8.14
9.43
9.00
52.43
Minimum
5.79
5.21
3.00
2.57
2.36
3.57
22.50
3.21
5.56
5.69
4.43
4.19
3.21
26.30
0.51
0.54
0.68
0.54
0.44
0.36
0.50
Dry Period Between Storms =
2.32
1.88
1.18
0.94
1.46
1.64
9.42
0.40
0.36
0.39
0.36
0.62
0.46
0.42
3hrs
1.06
0.97
0.96
1.13
0.87
0.47
1.02
13hrs
0.74
0.68
1.05
0.65
1.32
0.84
0.96
Duration
Average C.V.
(hr.)
12.81
13.55
15.25
15.00
14.57
12.10
13.93
19.40
19.18
15.17
11.83
20.03
17.78
17.73
0.79
0.75
0.72
0.68
0.77
0.76
0.74
0.69
0.64
0.66
0.78
0.74
0.75
0.71
Intensity
Average C.V.
(in./hr.)
0.0438
0.0413
0.0389
0.0397
0.0311
0.0313
0.0374
0.0271
0.0249
0.0337
0.0472
0.0343
0.0300
0.0310
0.66
0.59
0.49
1.00
0.43
0.47
0.66
0.78
0.73
0.79
1.10
0.70
0.48
0.87
Delta!**
Average ' C.V.
(hr.)
61.60
34.50
46.20
58.59
48.02
49.10
50.42
70.49
98.84
114.63
126.00
182.30
105.92
108.31
1.14
0.83
0.75
1.19
0.97
0.85
0.99
0.68
0.78
0.92
0.93
1.19
0.75
1.01
Analysis includes data from October 1973 to September 1987 for NWS gage # 457458 at Portage Bay, Seattle, Washington.
Minimum Storm Event Volume = .10 inches.
* C.V. = Coefficient of Variation (Average/Standard Deviation)
'* Delta = Time between storm event midpoints
-------
REFERENCES
1. Hydroscience. A Statistical Method for the Assessment of Urban Stormwater Loads -
Impacts - Controls, EPA 440/3-79-023, U.S. Environmental Protection Agency, 1979.
2. Huber, W.C., Dickinson, R.E., Cunningham, B.A. and Heany, J.P. Storm Water
Management Model, Version 4: User's Manual", U.S. Environmental Protection Agency,
Environmental Research Laboratory, Athens, Georgia, 1988.
3. Driscoll, E.D., Shelley, P.E. and Strecker, E.W. Evaluation of Pollutant Loadings From
Highway Stormwater Runoff: Design Procedures, FHWA-RD-88-006, U.S. Federal
Highway Administration, 1988.
4. Heany, J.P., Huber, W.C., Medina, Jr., M.A., Murphy, M.P., Nix, S.J. and Hasan, S.M.
Nationwide Evaluation of Combined Sewer Overflows and Urban Stormwater Discharges,
EPA-600/2-77-064, U.S. Environmental Protection Agency, 1977.
5. Restrepo-Posada, PJ. and Eagleson, P.S. Identification of Independent Rainstorms, Journal
of Hydrology, Vol. 55, pp. 303-319, 1982.
6. Haan, C.T. Statistical Methods in Hydrolgy, Iowa State Press, Ames, Iowa, 1977. 378pp.
7. Schueler, T.R. Controlling Urban Runoff: A Praticle Manual for Planning and Designing
Urban BMPs , Washington Metropoliton Council of Governments, 1987.
8. Urbonas, B. Personal Communication, 1988.
9. Driscoll, E.D. Methodology for Analysis of Detention Basins for Control of Urban Runoff
Quality, U.S. Environmental Protection Agency, Office of Water, Nonpoint Source Divi-
sion, Washington, D.C. 1986.
172
-------
COMPUTER AIDED PLANNING OF DRAINAGEWAY IMPROVEMENTS MADE EASY
WITH LOTUS 1-2-3
by: Michael B. Cooke, P.E. and R. Penn Gildersleeve, Jr., P.E.
Greenhorne & O'Mara, Inc.
3131 South Vaughn Way, Suite 228
Aurora, Colorado 80014
ABSTRACT
As an aid to engineers charged with the task of planning drainage
and flood control improvements, a computer program was developed by
Greenhorne & O'Mara, Inc. engineers which combines a hydraulics package
with a cost estimating program to allow quick planning level comparisons
of alternative planning scenarios.
The Stormwater Master Plan Model (SMPM) is a menu-driven program for
those involved in the master planning of drainageways. The program
computes hydraulic characteristics of existing and planned drainageway
improvements and. estimates their cost.
SMPM models four types of drainageway elements: channels, culverts,
bridges and detention ponds. Channels can be grass lined, riprap or
concrete and can include concrete or riprap drop structures. The
program accommodates four types of culverts, namely, corrugated metal
pipe, corrugated metal arch, concrete pipe and reinforced concrete box
culverts.
SMPM utilizes normal depth and inlet control for computing the
hydraulic characteristics of channels, culverts and bridges.
One of the most powerful features of SMPM is its cost routine. The
program computes a "planning level" cost estimate for each drainageway
element included in a model. The cost is based on user defined unit
costs for excavation, riprap, pipe right-of-way and other common
capital cost items.
The user builds a drainageway model by stringing together channel,
culvert and bridge elements. Each element is added to the model via a
menu which prompts the user for all needed input. Extensive error
checking protects the user from common input errors.
173
-------
Once a model is built, the program is run to determine hydraulic
characteristics and costs for each drainageway element and the overall
system. Again, this is performed via simple menu commands. Hydraulic
and cost results can be viewed on screen and printed on paper.
Modifications to design flows and any drainageway element can be
made quickly and easily via the menu. Each time a change is made, the
user can quickly see its impact on hydraulic characteristics and costs.
Thus, many "what if" scenarios can be studied more quickly and accurately
than previously possible.
SMPM is designed for those familiar with drainageway planning. For
such users, the program is a valuable tool that allows quicker and more
accurate analysis of planned improvements than is possible using only
engineering judgement and rules of thumb. Also, more numerous
improvement alternatives can be analyzed leading to better master
plans. Furthermore, the program provides documented results of any
improvement scenario studied by the planner.
SMPM works on any computer running Lotus 1-2-3 Release 2. Each
model can contain up to 35 channel, culvert or bridge elements.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.
INTRODUCTION
Greenhorne & O'Mara, Inc. has developed a computer model to aid
engineers who are involved with the planning of drainageway improvement.
The program combines a hydraulics package with a cost estimating module
to allow quick, planning level comparisons of alternative improvement
plans. The model is a menu-driven spreadsheet that will run on any
computer that can run Lotus 1-2-3, Release 2.
THE NEED FOR COMPUTER MODELLING TOOLS
Drainageway master planning is ideally suited to the use of computer
models because it involves assumptions about several variables; and
people often want to know how changing assumptions about one of the
variables might affect the end results.
Thus, drainageway planning is a highly iterative process. The
drainageway planner makes assumptions about the numerous variables
involved in drainageway planning and develops improvements that will
handle the anticipated drainage flows. The planner usually develops
174
-------
several improvement schemes before coming to a "best" or recommended
improvement alternative. Technical reviewers usually question
assumptions made by the planner and generally want to see how changing
some of the assumptions will affect the proposed improvements.
Subsequently, policy makers usually have questions and concerns that
can only be addressed by again varying some of the assumptions and
making new calculations. This "what if" questioning can go on
interminably and is generally limited only by the time and money
available to complete the project.
Since time and money are two commodities which are invariably in short
supply, planners are forced to rely upon rules of thumb and engineering
judgement when they are developing drainage improvement plans.
Unfortunately, the planner can develop blind spots and tunnel vision by
relying too much on past experience. In addition, policy makers are
often uneasy relying upon the planners "judgement" and are more
comfortable seeing actual calculations that support the planner's
contention that certain improvement alternatives are too expensive or
technically unfeasible.
Therefore, drainageway planning can benefit from any computer
modeling tools that automate the drainageway planning process.
CURRENT PRACTICES IN DRAINAGEWAY PLANNING
Drainageway planning involves a fairly standard procedure. First, a
planner develops design flood flows for the planning area using any one
of a number of commonly accepted hydrologic models. Then, the planner
tries to determine what improvements are necessary to handle the
projected flows. The planner works with four basic types of
improvements! channels, culverts, bridges and detention ponds. The
improvement options facing the planner are almost endless. For example,
can a grass-lined channel handle the flow through a certain area given
the fact that there is an existing sixty foot right-of-way constraint?
If not, is it more economical to acquire the additional right-of-way
needed for a grass-lined channel or would it be better to build a
riprap lined channel at a steeper slope that could handle the flow
without violating the sixty foot right-of-way constraint? The
drainageway planner faces enumerable questions of this type.
THE STORM WATER MASTER PLAN MODEL (SMPM)
SMPM computes hydraulic characteristics of existing and planned
drainageway improvements ajui determines their cost.
The SMPM program models four kinds of drainageway improvements:
channels; culverts; bridges; and detention ponds.
175
-------
In the model, each section of channel, each culvert, and each bridge
is called an element. Drainageway improvement models are built by the
user by stringing together elements from upstream to downstream along
the drainageway for which improvements are proposed.
Channel elements in the program can be grass, riprap or concrete
lined and can include drop structures. The program handles four types
of culverts, namely: corrugated metal pipe (CMP); corrugated metal
arch pipe (CMA); concrete pipe (CONG); or box culverts (BOX). Detention
ponds are modelled by decreasing flood flows at appropriate points in
the model.
Each element is added to the model via a menu system. For example,
if the user wants to add a channel, he selects "Edit", "Add" and
"Channel" from the menu system. An input screen then appears which
prompts the user for all the needed input. An example of the channel
input screen is shown in Figure 1.
This is a new CHANNEL element
Enter the following information on the new element...
Existing?
NO
Input
Input
10OO
5O
2
3
3
0.035
O.O2
GRASS
3
2
0.67
2
20
iLength/No of Drops
[Flow/Drop Type
iB-Width/Drop Width
iM-Left/Drop Keydown
!M-Right/Drop Keybac
! "n"/Drop Height
[Slope/Road Prot?
[Lining/Extra $
J Riprap Depth
', Riprap Keydown
iFilter Depth
|D/S Element No
[Extra ROW Width
CONCRETE
1
6
12
3
YES
$5,000
...When input
is completed:
Press ESC to
continue with
this operation
Figure 1. Input screen for channels as it appears on the
computer monitor.
176
-------
Because the program is menu-driven and prompts the user for all needed
input, the program is easy to use without constant reference to manuals.
Also, extensive error checking makes the program even easier. Whenever
the user provides any input, the program automatically checks the input
for common input errors. For example, the user is prompted to indicate
the type of channel lining. The program expects one of three answers:
grass, riprap or concrete. If the user enters something else or
misspells a word, the program "beeps" and prompts the user to try again.
Once a model is built, the program is run to determine hydraulic
characteristics and cost for each drainageway element and for the
overall system. This is done simply by selecting the appropriate
commands from the menu.
The program uses a normal depth routine to calculate the depth of
flow and velocity in channel elements. The program makes an initial
guess at depth of flow and computes capacity of the channel given the
input parameters defined by the user. It then compares the computed
channel capacity to the required capacity. Next, the program makes a
new guess at the depth of flow based on the magnitude of difference
between the required flow and the calculated flow in the channel. This
process is repeated until the difference between required flow and
calculated flow is less than one percent. At that point, the normal
depth routine for that channel element terminates.
The program calculates culvert capacity assuming inlet control.
Nomographs which relate culvert capacity to headwater depth are built
into the program. The user specifies culvert size and construction
material along with other input parameters and the program determines
capacity of the culvert. The program compares culvert capacity to
required capacity and alerts the user to any culverts that are
undersized. Where the culvert is too small to pass all of the required
flow, the program completes a weir flow calculation over the culvert
(if the user has indicated a weir length during the input routine).
Since the actual execution of the model can take as long as five
minutes depending on model size, the user is provided with a status
screen that provides continual information on the status of the program
execution. The status screen shows how many drainageway elements there
are in the model, how many have been hydraulically balanced and estimates
the time required for the program to finish execution. The screen is
updated every few seconds so that the user knows that the program is
running and knows when the program will finish.
177
-------
A sample of the hydraulic output of the program is shown in Figure 2.
CALCULATION
RESULTS
Browse as desired using the cursor Keys...
Name: SN1
Printed: O9/30/8B
...Press ESC to return to Study Sub-menu.
ELEMENT
LENGTH/
: CHANNEL
LINING
ELMT «/ (Type
TYPE for ovts)
1 50
CULVERT CMP
2 1000
CHANNEL GRASS
3 50
CULVERT BOX
4 1000
CHANNEL GRASS
5 50
BRIDGE NA
0 O
0 NA
0 0
0 NA
1
I
-------
For channels, the last column contains the critical output items,
namely, normal depth and normal velocity. In Figure No. 2 Element No.
2 is a channel where the normal depth and velocity is 4.80 feet and 6.8
feet per second respectively.
An examination of Figure 2 shows that a lot of the output table is
devoted to restating the input provided by the user. This is done so
that the user can see what input assumptions he might want to change if
the output results are not satisfactory. For example, Element No. 3 is
a culvert. The last column shows that this culvert is at 119Z of
capacity. Flow is coming up over the road with a depth of 0.90 feet.
If the planner is working a constraint of 0.5 feet maximum overflow
depth, he can look at the input and decide what to change. He might go
from a 6' x 7' box culvert (as indicated by data in Columns 3 and 4) to
a 6' x 8' box culvert. Or, he might see that he has used Inlet Type
No. 2 (Column 4, 2nd line) and decide to try another inlet type.
One of the most powerful features of SMPM is its cost routine. The
program computes a "planning level" cost estimate for each drainageway
element included in the model. The cost is based on user defined unit
costs for excavation, riprap, pipe and other common construction
elements. During the input routine, the user can call up a screen for
inputting unit cost.
A sample of the cost results table for a model is shown in Figure 3.
COST
RESULTS
(in $1,000)
Browse as desired using the cursor keys...
Name: SNl
Printed: 09/30/88
..Press ESC to return to Study Sub-menu.
ELMT NO TYPE
1 CULVERT
2 CHANNEL
3 CULVERT
4 CHANNEL
5 BRIDGE
6 0
7 0
>-fN
34 0
Sub-Totals
Elements Deleted
Sub-Totals
Contingency!?
Sub-Totals
Engr/Legal/Fiscal@
Totals
LENGTH
50
1000
SO
1000
50
0
0
0
25.00,
0.00%
s=s=sss==rs
CURRENT
$4
$79
$26
$106
$120
$0
$0
$0
$334
$0
$334
$84
$418
$0
$418
PREVIOUS
$4
$79
$26
$124
$120
$0
$0
$0
$353
$0
$353
$88
$441
$0
$441
ssssssszs:
CHANGE
$0
$0
$0
($18)
$0
$0
$0
$0
($18)
$0
($18)
($5)
($23)
$0
($23)
DEFAULT
$4
$79
$26
$115
S120
$0
$0
•x.
^v
$0
$344
$O
$344
$86
$430
$0
$430
N
IN
Figure 3. Cost table as it appears on the computer monitor.
179
-------
The table has four columns of costs: Current, Previous, Change,
Default. The "Current" column shows costs for improvements based on
the most recent input assumptions defined by the user. The "Previous"
column shows costs for improvements based on the input assumptions
prior to the most recent changes. The "Change" column is the difference
between "Current" and "Previous" costs. By looking down this column,
the user can easily see if the new input assumptions are improving or
worsening the cost of the proposed improvements.
The "Default" column is used for record keeping. Once the user
settles on an improvement scheme that will be the best or recommended
improvement plan, he can set the default costs equal to current costs.
Then, if he later uses the model, he can compare the cost of any new
improvement schemes to his preferred scheme. No matter how many new
scenarios he tries, he can always see if the "Current" improvements are
cheaper than his preferred alternative.
The ability to see both hydraulic and cost data for any given
drainageway improvement scheme is a great advantage to planners. They
can easily try out different drainageway improvement schemes, determined
that they are hydraulically adequate and then see the projected costs
for those improvements. Because the program does the tedious, time-
consuming calculations in a short time, the planner can try many more
improvement scenarios than previously possible.
Modifications to any drainageway model can be made quickly and
easily via the menu. This makes it easy to play the "what if" game
which is so necessary to drainageway planning. When the user indicates
he wants to modify a particular drainageway element, the existing
assumptions are displayed on the screen so that the user can easily see
what assumptions were previously made. After making modifications, the
user can again run the program to see the impact of the changes on
hydraulic characteristics and costs.
The program also allows the user to delete drainageway elements or
to insert new drainageway elements within a model.
USES FOR SMPM
First, as described above, the model can be used as an aid to
drainageway planners in developing drainage improvement plans.
Secondly, the model can be used by those responsible for implementing
drainage improvement plans to maintain a continually updated master
plan for a given area. For example, suppose that a city contracts with
a consulting engineering firm to develop a drainage master plan for a
certain drainage basin. As an end product, the city receives not only
a written copy of a drainage master plan, but also receives an SMPM
180
-------
program which models the proposed drainageway improvements. Now, each
time a parcel of ground within that drainage basin is developed, part
of the city's review process will be to check the developer's projected
flows against the flows used in the master plan. If the flows are
significantly above or below those used in the master plan, the city
personnel can change the flows in the SMPM program and see the hydraulic
and cost impacts on proposed downstream facilities. Cost assessments
or credits can then be granted to the developer. Or, the developer may
be proposing changes to the improvements shown in the master plan. It
would be easy for city personnel to modify the SMPM program to evaluate
the developer's proposed changes. Once the city accepts the developer's
drainage plan for a given undeveloped parcel in the basin, it can
incorporate any changes to the master plan into the SMPM program. In
this way, the city can maintain a continuously updated master plan. As
a result, the city can provide a higher degree of planning, insure that
costs are shared more equitably and reduce the need for frequent and
costly drainage plan revisions by the consultant.
THE FIRST USE OF SMPM
Greenhorne & O'Mara was retained by the City of Woodland Park in
January 1986 to develop a stormwater master plan for the city. In
addition to a traditional master plan report document, the contract
with the city called for providing a computer model of the planned
improvements for use by city personnel.
While conducting a traditional master planning study for the city,
Greenhorne & O'Mara developed the SMPM program. The development of the
program itself was conducted outside the scope of the work for the City
of Woodland Park. Then the program was used to study proposed
improvement schemes while traditional hand calculations were being used
to study the same improvement schemes. This allowed Greenhorne &
O'Mara engineers to debug the program. At the same time, the program
was used to provide analysis of more numerous improvement alternatives
than would have been possible using manual methods. Also, the program
uncovered some errors in the calculations that were done manually.
LIMITATIONS TO THE PROGRAM
SMPM is designed for those familiar with drainageway planning. It is
important to note that the program assumes the user is familiar with the
drainageway planning process and that it is intended as a tool for these
knowledgeable experts.
181
-------
A second limitation is that the SMPM program is suitable for planning
purposes only. It is not intended to be used as a design tool. The
hydraulics of the program are based on normal depth and inlet control
and cost estimates are derived by utilizing a number of simplifying
assumptions that have traditionally been used in master planning
projects. Because of these limitations, the program is not suitable
for use in the design phase of a project.
CONCLUSIONS
The SMPM program is a valuable tool for those involved in the master
planning of drainage improvements. It allows quicker and more accurate
analysis of planned improvements than is possible using only engineering
judgement and rules of thumb. Also, more numerous improvement
alternatives operating under multiple flow assumptions can be analyzed
leading to better master plans. The program also enhances the users
credibility with non-technical people because it allows more thorough
planning and because it provides documented backup to the user's
engineering judgement. Finally, the program makes it easier for
regulators to administer and implement a drainage master plan concept.
182
-------
HYETOGRAPH COMPOSITING EFFECTS ON URBAN RUNOFF MODELLING
by: Michael P. Jansekok, B. S.
Civil Engineer
Kiowa Engineering Corporation
Denver, Colorado 80210
Ben R. Urbonas, M.S., P.E.
Chief, Master Planning Program
Urban Drainage and Flood Control District
Denver, Colorado 80211
ABSTRACT
Rainfall and runoff data from a 3.08 square mile urban watershed in
Denver, Colorado was used to investigate the effects of compositing several
recorded rain storm hyetographs on urban stormwater runoff modelling results.
The watershed in this semi-arid region had data at five rain gages and two
flow gages. This data provided the basis for calibrating an Urban Drainage
and Flood Control District version of SWMM model. The calibrated model was
then used to examine the effects on runoff calculations using a single
composite hyetograph for each storm.
Compositing of hyetographs was performed using two types of area
weighted techniques. The five hyetographs were composited directly usiwj the
recorded rainfall depth at each clock time interval (i.e., "across
compositing"). In addition, the hyetographs were composited using a
technique that first shifted the five gage records so the peak rainfall time
increments of each hyetograph were aligned (i.e., "peak preservation
compositing"). The findings are described in this paper and their
implications for urban stormwater runoff modelling are discussed.
183'
-------
INTRODUCTION
Very little research has been done in compositing simultaneous rainfall
records into single model input hyetographs and their effect on calculated
stormwater discharges. This paper will explore two facets to hyetograph
compositing. First, the effects on calculated runoff from hyetograph
compositing of a storm event will be compared with runoff results calculated
using individual/multiple rain gage records. Second, two popular compositing
techniques will be compared and evaluated. The authors were fortunate enough
to have access to eight years of rainfall/runoff data from a stable urban
watershed with a relatively high gaging density.
The Harvard Gulch basin is located in Denver, Colorado and was used for
this investigation. An investigation of hyetograph compositions by Avon,
Collins and Kibler (1974) was performed for a smaller watersheds using two
hypothetical, four time increment, hyetographs. They recommended adopting
the hyetograph pattern from one gage and compositing hyetographs using a peak
pattern preservation technique. The results from the author's study using
rainfall and runoff records appear not to support this conclusion.
RAINFALL AND RUNOFF GAGES
The Urban Drainage and Flood Control District was a local cooperator
with the U. S. Geological survey in collecting rainfall/runoff data between
1979 and 1987 for the Harvard Gulch drainage basin. This cooperative effort
continues and data is still being collected. Data from two flow gages and
five rain gages were used in the investigations. Detailed records of
rainfall and flow stages are collected at each station by the operation of
two digital recorders which punch the data on a 16-channel paper tape at
5-minute intervals.
The Harvard Gulch basin was divided into two basins for modelling. The
upper basin had a flow gage located at Colorado Boulevard (1.12 mi Kwhich
contains two rain gages and one flow gage. The total basin (3.08 mi )
contain five rain gages and two flow gages. The locations of all gages are
shown in Figure 1.
SELECTION AND PREPARATION OF ANALYSIS DATA
MINIMUM RAINFALL AND RUNOFF CRITERIA FOR ANALYSIS
Seventeen storms were selected for this analysis based on the following
criteria:
1. Five rain gages and two flow gages must be reporting during the
storm.
2. Minimum recorded rainfall at any one rain gage must equal or exceed
0.08 inches during at least one 5-minute period within a storm.
184
-------
FIGURE 1, HARVARD GULCH BASIN
CO
Ul
8000
-------
FIGURE 2
HARVARD GULCH SWMM MODEL CALIBRATION
RUNOFF INCHES AT HARVARD PARK
04
035
03
025
02
015
01
005
0
DATA PONT
BEST FIT
"0 0.05 01 015 02 025 03 035
OBSERVED RUNOFF INCHES
FIGURE 3
HARVARD GULCH SWMM MODEL CALIBRATION
PEAK FLOW AT HARVARD PARK
800
700
600
500
400
300
200
100
0
0 100 200 300 400 500 600 700 800
OBSERVED Q PEAK (CFS)
186
-------
3. The recorded peak flow at one of the two flow gages must equal or
exceed 50 CFS.
PREPARATION OF UDSWM2PC MODEL
Stormwater drainage system maps from Denver Wastewater Control Division
provided the initial basis for both basin division and drainage system
networks. Denver's contribution towards this study is greatly appreciated..
Major basin boundaries and the degree of imperviousness were field verified
by UDFCD personnel in 1979 and 1980.
The Harvard Gulch Basin was divided into 59 subbasins, 23 of which
comprised the upper basin at Colorado Boulevard. Average values for
imperviousness, perviousness, slope, tributary width, Manning's n, etc. were
estimated for each subbasin area.
The model for the drainage system contained 78 conveyance elements, 35
providing drainage for the upper basin. Conveyance elements were divided
into five types as either pipe, pipe with overflow, channel, channel with
overflow, or non-routing. One detention element was incorporated into the
model to reflect field verified conditions.
The Thiessen Polygon Method was used to assign each subbasin to a
specific rain gage.
CALIBRATION OF SWMM MODEL
For convenience sake, two separate SWMM models were created and
calibrated for the Harvard Gulch Basin, namely an upper basin model and a
total basin model. The upper basin model consisted of the basin area east of
Colorado Boulevard. Each one of the 17 selected storms was processed through
both SWMM models. Calculated runoff volumes and peak discharges were then
plotted against observed values. Data regression was used to determine a
best-fit line through plotted points. Adjustments were made to rainfall loss
parameters, Manning's n, subbasin tributary widths, etc. and the model was
rerun for each of the 17 storms until each best-fit line approximated 45
degrees for both peak flows and volumes. The SWMM model calibration data are
shown in Figures 2 through 5. Examples of hydrograph comparisons are shown
in Figures 6 and 7.
PRESENTATION OF DATA AND DISCUSSION OF RESULTS
COMPOSITE TYPE COMPARISONS
Two types of multiple hyetograph compositing were studied. The first
type and the one most commonly used by engineers is to area weight-average
the rainfall depths recorded each time increment. The second technique first
shifts the individual hyetographs so that the time increments containing the
peak rainfall depth (i.e. intensity) are lined up. The shifted hyetographs
are then composited using area weighing technique.
187
-------
o
FIGURE 4
HARVARD GULCH SWMM MODEL CALIBRATION
RUNOFF INCHES AT COLORADO BOULEVARD
0.1 0.2 03 0.4
OBSERVED RUNOFF NCHES
0.5
FIGURE 5
HARVARD GULCH SWMM MODEL CALIBRATION
PEAK FLOW AT COLORADO BOULEVARD
400
300
200
100
50 100 150 200 250 300 350
OBSERVED Q PEAK (CFS)
188
-------
s
2
FIGURE 6
HARVARD GULCH AT COLORADO BOULEVARD
Calibration Run. Storm Date: 05/14/87
TME N MNUTES
FIGURE 7
HARVARD GULCH AT HARVARD PARK
Calibration Run. Storm Date: 05/28/81
50
100 150
TIME IN MNUTES
200
250
189
-------
EFFECT OF COMPOSITE TYPE
Examples of hydrograph comparisons between SWMM multi-gage runs and the
two composite type hyetograph runs are shown in Figures 8 through 11. The
most notable trend in both composite types is that they tend to under
estimate peak flows and runoff volumes. This is summarized in Tables 1 and 2
and in Figures 12 through 15. Both composite types resulted in peak flows
that were, at times, as much as 65% less than obtained using the five rain
gage runs. The divergence in volumes was up to 20% less from the muHi-gage
runs. Upper basin results were similar and ranged as much as 30% less for
peak flow and 10% less for runoff volumes. At the same time very little
difference was found in calculated peak flow and volume results between the
two compositing methods.
TABLE 1
HARVARD GULCH AT H. PARK - PEAK FLOW (COMPOSITE TYPES)
PERCENT DEVIATION FROM FIVE GAGE CALIBRATED RUN
Composite Range Mean Standard
Type Deviation
Pk. Pres. -65.1 to 4.9 -17.4 18.1
Across -60.5 to 9.7 -16.7 18.0
TABLE 2
HARVARD GULCH AT H. PARK - RUNOFF VOLUME (COMPOSITE TYPES)
PERCENT DEVIATION FROM FIVE GAGE CALIBRATED RUNS
Composite Range Mean Standard
Type Deviation
Pk. Pres. -18.8 to 9.0 -3.3 7.7
Across -20.1 to 9.0 -2.8 7.6
TABLE 3
HARVARD GULCH AT CO. BLVD. - PEAK FLOW (COMPOSITE TYPES)
PERCENT DEVIATION FROM TWO GAGE CALIBRATED RUN
Composite Range Mean Standard
Type Deviation
Pk. Pres. -27.9 to 9.3 -2.7 10.1
Across -23.3 to 9.1 -1.1 7.4
TABLE 4
HARVARD GULCH AT CO. BLVD. - RUNOFF VOLUME (COMPOSITE TYPES)
PERCENT DEVIATION FROM TWO GAGE CALIBRATED RUNS
Composite Range Mean Standard
Type Deviation
Pk. Pres. -10.3 to 2.2 -2.2 3.8
Across -10.3 to 2.2 -1.6 3.4
190
-------
FIGURE 8
COMPOSITE RAIN VS. 5 RAINGAGE RESULTS
Storm Date: 05/28/81 (At Harvard Park)
50
100 150
"PME N MNUTES
FIGURE 9
:50
COMPOSITE RAIN VS. 2 RAINGAGE RESULTS
Storm Date: 06/05/83 (At Colo. Blvd.)
6
250
200
100 150 200 250
TME N MMUTES
300
191
-------
FIGURE 10
COMPOSITE RAIN VS. 5 RAINGAGE RESULTS
Storm Date: 06/09/87 (At Harvard Park)
50
100 150
TIME IN MINUTES
FIGURE 11
200
250
COMPOSITE RAIN VS. 5 RAINGAGE RESULTS
Storm Date: 09/11/85 (At Harvard Park)
160
PEAK PRES.
1
ACROSS
50
100 150 200 250 300
TIME IN MINUTES
192
-------
FIGURE 12
COMPOSITE TYPE COMPARISONS AT H. PARK
VARIATION FROM CALIBRATED 5 GAGE RUN
80
GO
40
20
0
-20
-40
-60
-80
8
o-
COMPOSITE TYPE
FIGURE 13
COMPOSITE TYPE COMPARISONS AT H. PARK
VARIATION FROM CALIBRATED 5 GAGE RUN
3U
20
10
0
1U
-«iO
.f.r\
_. _ .
! o
t -Q
4i v7
1 ©
# A
w u - ~ - —,._-.
#= PK PRES
0= ACROSS
X= 5 GAGE
COMPOSTE
L93
-------
I
FIGURE 14
COMPOSITE TYPE COMPARISONS AT CO. BLVD.
VARIATION FROM CALIBRATED 2 GAGE RUN
.'1 1
10
0
-10
-20
if °
i ! x"^
# o
# 0
#
#= PK PRES
0= ACROSS
X= 2 GAGE
COMPOSITE TYPE
FIGURE 15
COMPOSITE TYPE COMPARISONS AT CO. BLVD.
VARIATION FROM CALIBRATED 2 GAGE RUN
1
1
g
£
i
i—
£
10
0
—10
1 e;
M §
K H X"
I o
I 8
#
I a
#= Pk PRES
0= ACROSS
X= 2 GAGE
COMPOSTE TYPE
194
-------
CONCLUSIONS
In this study very little difference was found in peak flow and runoff
volume simulations between the two types of hyetograph compositing
techniques, namely compositing straight across or compositing using peak
preservation. However, the authors believe it is premature to accept this
finding as a general finding applicable under all conditions. Both methods
tended to underestimate peak flows and volumes when compared against the
calibrated multi-rain gage hyetograph runs using a calibrated SWMM model.
Some hydrologic models require incremental rainfall depths composited
into a single input hyetograph. These numerical method models are then
calibrated by modifying runoff coefficients and other parameters in order to
increase calculated volumes and peaks (during initial calibration with
observed data) to bring calculated values in line with observed data.
One possible problem in calibrating a model using composited rainfall
data is that if other recorded point rainfall or long term non-composited
rainfall data are used later with the model to generate continuous
simulation, the calculated volumes and peak flows could be over estimated.
Because the model was calibrated using composite hyetographs which appear to
underestimate peak flows and volumes, the percentage by which calibration
parameters are adjusted to increase calculated peaks and volumes will be the
percentage by which the use of non-composited rainfall data later will
overestimate the peak flows and volumes. It is this possibility of
overestimating during long term simulations that should be considered by
modellers when calibrating models using composited hyetographs, particularly
when studying larger urban watersheds.
The work described in this paper was not funded by the U. S.
Environmental Protection Agency and therefore the contents do not necessarily
reflect the views of the Agency and no official endorsement should be
inferred.
REFERENCES
Avon, G., Collins, J. G., and Kibler, D.F. "Problems in Weighing of
Hyetographs," Water Resources Bulletin, Vol. 15, No. 6, American Water
Resources Association, December, 1979.
195
-------
FLOOD HYDROGRAPH FOR UNPAGED WATERSHED
by: Wolney Carstens Cunha, P.E.
Project Manager
Stewart Environmental Consultants, Inc.
214 N. Howes Street
Fort Collins, CO 80522
ABSTRACT
The flood hydrograph for ungaged watershed can be calculated
utilizing the software discussed in this paper. It uses the Soil
Conservation Service (SCS) data and physically-based soil
infiltration equations. Large heterogeneous watersheds can be
partitioned into several smaller homogeneous subbasins. Routing
of excess rainfall is performed with the SCS dimensionless unit
hydrograph to produce a runoff hydrograph for each individual
subbasin. The final flood hydrograph is obtained by routing and
summing the runoff hydrographs of subbasins according to the
travel time associated with them. Rain data can be supplied by
simply providing a depth and duration of rain or the user may
input variable rain depths.
The work described in this paper was not funded by the U.S.
Environmental Protection Agency, and, therefore, the contents do
not necessarily reflect the views of the Agency and no official
endorsement should be inferred.
1-96'
-------
FLOOD HYDROGRAPH FOR UNGAGED WATERSHED
INTRODUCTION
The first problem faced by the engineer when designing a
storm drainage system is to select an easy and flexible yet
accurate tool for calculating the peak design flow. There are
powerful programs available on the market for this purpose.
Nevertheless, they require, as a rule, extensive and costly data,
the learning process of the software operation is demanding, and
sometimes the output is not accurate.
In small-scale projects, appraisals or bids, money and infor-
mation are scarce. Therefore, the ideal software should be one
based on simple but sound theory, easy to operate, flexible
enough to accommodate the lack of data, and with the capability
to produce accurate analyses when required.
The documentation and program listing for this software is
already available to the public through the National Technical
Information Service, Springfield, Virginia 22161, under number
FHWA/RD-81/061. The title of the documentation is "User's Manual
For XSRAIN - A Fortran IV Program For Calculation of Flood Hydro-
graphs for Ungaged Watershed." The authors are Messrs. James P.
Verdin and Hubert J. Morel-Seytoux.
This software utilizes physically-based infiltration equa-
tions to calculate the abstraction of rainfall in basins for
which there is a minimum of available hydrologic information.
The hydraulic soil parameters can be easily calculated from a
relationship between the Soil Conservation Service's curve number
(CN) and the mentioned parameters.
OBJECTIVE
The objective of this paper is to present a modified version
of XSRAIN written in Fortran 77. This updated version permits
the user to divide a large basin into several smaller subbasins.
The purpose is to better model the watershed and, consequently,
to achieve more accurate results.
METHODOLOGY
From experience, when the rainfall is of low intensity and
short duration, very little runoff occurs. Even with high
intensity but short duration rain, runoff may not be present for
some soil compositions. Since there was no runoff, all of the
water from the rain was infiltrated into the soil or intercepted.
Also, it is clear that there is a limit for the infiltration
197
-------
ratio because if rainfall continues long enough, overland flow
will occur eventually.
There are several equations available for the calculation of
the infiltration capacity of the soil. Among them is Horton's
equation:
Ic - I. + (I0 - Ia)e-Kt (1)
Where Ic is the infiltration capacity at time t, I0 is the
initial infiltration capacity, Ia is the time asymptotic limit
infiltration capacity, K is a constant, and t is time.
Infiltration capacity is in inches per hour. This equation
assumes that there is a large volume of ponded water on the
surface in such a way that the soil becomes saturated very
quickly. This condition is usually satisfied in a furrow
irrigation. Actually, Horton's equation is used quite
successfully in agricultural projects. However, Horton's
equation should not be used to calculate the infiltration
capacity unless heavy rainfall occurs at the beginning of the
storm.
From the analysis of Horton's equation, it is clear that the
infiltration capacity is an exponentially decaying process. The
infiltration process is a function of soil characteristics and
also depends on the antecedent conditions of soil saturation as
the "Ia" parameter indicates. At this point, it is apparent that
if Horton's equation is used, it must be calibrated. The cali-
bration procedure involves varying the mentioned parameters in
such a way that the equation is compatible with the soil under
consideration. Horton's equation does not take into account the
rainfall intensity, and the calibration process is time
consuming. One way to circumvent these problems is to use
physically-based infiltration equations.
PONDING TIME
At the early stages of a storm, all of the rainfall
infiltrates into the ground. When the soil has completely
saturated and the rainfall rate is greater than the infiltration
rate, water starts to accumulate on top of the soil. The elapsed
time from the beginning of the storm until the time when water
starts to accumulate is called ponding time, tp. There are two
different situations in regard to ponding time calculation: 1)
Constant Rainfall Rate and 2) Variable Rainfall Rate.
For case number one, Constant Rainfall Rate, Mein and Larson
presented the following formula for the calculation of ponding
time:
tp = [(e - 9i)hc]/r[(r/K) - 1] = Sf/[r(r* - 1)] (2)
198
-------
Where e is the water content of the soil at natural
saturation; e; is the initial moisture content of dry soil (water
content of the soil is dimensionless) ; hc is the effective
papillary drive of the soil in inches; r is the constant rainfall
rate in inches per hour; K is hydraulic conductivity at natural
saturation in inches per hour, and r* is the normalized rainfall
rate.
For case number two, Variable Rainfall Rate, the following
formula is utilized iteratively:
tj.! + (1/rj) [(6 - ei)hc/(rj/K - 1) - rv(tv.!)]
V-l (3)
Starting with j=l, we calculate tp until tp is smaller than
For j=l, i.e. the first time step:
tp - 0 + (i/rj [(9 - e^hc/CiVK - 1)] - sf/[ri(r* - 1) ]
If tp is smaller or equal to the duration of the first
rainfall rate, ponding occurs during the first time interval. If
tp is greater than tlf set j=2 and recalculate tp. For j=2, the
following expression is obtained:
tp = t, + l/ra [(9 - 6i)hc/(r2/K - 1) - r^J
Again, if tp is smaller than t2, then ponding occurs in the
second time period. If t is greater than t2, set j=3 and
iterate again. If the rainfall rate is equal to or smaller than
hydraulic conductivity (K) , then ponding cannot occur at this
period of time because the rainfall rate has to exceed the
hydraulic conductivity to produce overflow.
INFILTRATION RATE AFTER PONDING
The cumulative infiltration rate at a given time step, Wjf
after ponding can be calculated by the following formula:
Wj = Wp + S(Wp,ei)[(tj - tp + B)1/2 - B1/2] + K(t - tp) (4)
Where SfWp,^) is obtained from the following expression:
S(Wp,9!) = [2K(Sf + Wp)Vs,31/a (in/hr)1/2 (5)
199
-------
and B, as follows:
B = (Sf + Wp)V[2KS{(rp/K - I)2] (6)
Wp is cumulative infiltration up to ponding time tp. Wp and
Wj are in inches.
The instantaneous infiltration ratio is in inches per hour
and is given by:
I = l/2S(Wp,ei) (t - tp + B)-1/2 + K (7)
THE SCS METHOD
The runoff calculation using the SCS method is based on a
curve number (CN) which is a characteristic of the watershed.
This number was derived by calculating actual rainfall and runoff
data in experimental watersheds with specific soil types and land
covers.
According to SCS, the cumulative excess rainfall depth, Pe,
can be calculated by the following formula:
Pe = t(P - Ia)V(P - Ia + S)] (8)
Where P is the cumulative depth of rainfall in inches, S is
the maximum watershed storage in inches, and Ia is the initial
abstraction in inches. Interception, depression storage, and
infiltration occurring prior to runoff are included in the Ia
index.
Curve number CN is linked to the maximum watershed storage S
by the following expression:
CN = 1000/(S + 10) (9)
Also, the initial abstraction can be estimated by the
formula:
Ia = 0.2S (10)
The cumulative infiltration depth W in inches, is calculated
as follows:
W = P - Pe + We - Ia (11)
200
-------
Where We is the cumulative infiltration depth at time tc,
which is the end of initial abstraction. The factor P-Pe can be
easily derived from Equation 8 and is expressed as follows:
P - Pe = [P(S + IJ - Ia2]/(P - Ia+s)
Replacing P - Pe in Equation 11 yields:
W = [P(S + IJ - Ia2]/(P - Ia + S) + We + Ia (12)
The above expression can be further reduced to:
W = [S(P - IJ/(P - Ia + S)] + We (13)
The first derivative of Equation 13 is taken in relation to
time. The expression for instantaneous infiltration rate I is:
I = (S2r)/(P - Ia + S)2 (14)
where r is the rainfall rate. It is seen from the above
equation, that the infiltration rate, using the SCS method, is a
function of rainfall rate r. This contradicts the field
experience and laboratory evidence, as well as the theory of
groundwater hydrology.
A formula for the excess rainfall rate, re, can be derived by
obtaining the first derivative of Equation 8 in relation to time.
The final expression can be written as follows:
re = [(P - IJ r(P + 2S - IJ]/(P - Ia + S)2 (15)
The above expression states that after the ponding time,
there will be excess rainfall rate, no matter how small the value
of r. Nevertheless, if r is equal to or smaller than the
hydraulic conductivity K, all of the rainfall will infiltrate
into the soil. In this case, re is equal to zero.
The flood hydrograph based on SCS's method can be misleading
depending on the rainfall pattern. Equation 14 states that the
infiltration rate varies with rainfall. Equation 15 indicates
that after ponding, runoff will occur. Both cases are erroneous.
201
-------
DERIVATION OF PHYSICAL SOIL PARAMETERS FROM CURVE NUMBER
The best way to circumvent the rough approximations of SCS's
method is to use Equations 2 through 7 based on physical soil
parameters. From Equation 2, it is concluded that there are only
two parameters to be estimated: storage suction factor (Sf) and
hydraulic conductivity (K) . One way to establish the
relationship between CN and (Sf,K) is to calculate the cumulative
infiltration depth by both methods and set them equal to each
other. Let R equal the retention due to interception and
depression storage. If R is added to Equation 13, the equation
for total abstraction from SCS's method is:
R + W = ia + S(P - IJ/(P - Ia + S) (16)
The equivalent expression for the physical infiltration
equation will be:
Wj + R = R + Wp + S(ei)[(tj - tp + B)1/2 - B1/2] + K(t - tp) (17)
or:
Ia + S(P - Ia)/(P - I. + S)
= R + Wp + S(Q-1) [ 75 (20)
K = 1.236 - 0.0154CN, 36 < CN < 75 (21)
K = 1.853 - 0.0324CN, CN < 36 (22)
S(6j) = (100 - CNJ/42.252, CN > 65 (23)
S(e;) = 1.191 - 0.00575CN, CN < 65 (24)
202
-------
The storage suction factor at field capacity, (Sf)fc/
calculated from the following equation:
can be
(Sf)fc = [S(0fc)]
(2K)
(25)
XSRAIN calculates this relationship automatically.
PROGRAM APPLICATION
The objective is to calculate a runoff hydrograph for the
entire catchment shown in Exhibit 1. The required period of
return is assumed to be 50 years for the rainfall under
consideration. The rainfall depth is 2.1 inches and time
concentration is 1.77 hours. The watershed under consideration
is an urbanized area divided in nine subcatchments. Each
subcatchment is a single family residential area with 38 percent
impervious surfaces. The exceptions are the subcatchments No. 8
and No. 9 which are considered to be a reserved green area.
As seen from the Exhibit l, the collection of the runoff
water is accomplished by a central channel designed specifically
for this purpose. All of the necessary data required by the
calculations are shown in the Table 1.
Table 1
Watershed Characteristics
Sub
NO.
Area
(mi.2)
Surface, CN
Description
Time Lag
(Hr.)
Time of
Concen-
tration
(Hr.)
Travel
Time
(Hr.)
1
2
3
4
5
6
7
8
9
0,
0,
1562
1562
0.2344
0.1875
0.2344
0.3125
0.0781
0.1562
0.1562
Forest 83 0.15
Forest 83 0.15
Fallow 83 0.18
Meadow 83 0.12
Fallow 83 0.18
Lawn 83 0.20
Meadow 83 0.12
Pasture 86 0.57
Pasture 86 0.57
0.25
0.25
0.30
0.22
0.30
0.33
0.20
0.95
0.95
0.11
0.22
0.30
0.41
0.49
0.58
0.70
0.79
0.82
From the analysis of Table 1, it is evident that subbasins
No. 8 and No. 9 have both a large time of concentration and
relatively small areas. This implies that the mentioned
subbasins are increasing the final time of concentration of the
basin, but their contribution to the superficial runoff is
relatively small. The reader should be aware that the peak
runoff calculated from the above data is not the maximum. A
203
-------
trial and error approach in regard to rainfall depth should be
used to calculate peak runoff. The results of the calculation of
peak flow runoff of the mentioned watershed are shown in Exhibits
No. 2 and No. 3.
* . *
* . 8.9 . *
* ... *
* | *
* 7 16 *
*
*
*
*
* 3
*
*
*
*
* 2
*
*
*
*
*
*
*
5 *
*
*
1 *
4 *
1 *
1 *
| ......... *
1 *
| 1 *
*
*
EXHIBIT 1
WATERSHED CONFIGURATION
CONCLUSIONS
The calculated peak flow for a watershed can vary
substantially depending on the way that the analysis is
performed. This is evident from Exhibits No. 2 and No. 3.
XSRAIN is a flexible computer model that has the capability
to use data from the SCS method but without its shortcomings. If
data from the field is available, XSRAIN can incorporate the
information into the calculation and, consequently, provide very
good results.
204
-------
w
on
3
o
o
Oi
EXHIBIT NO. 3
FLOOD HYDROGRAPH
Rain Depth = 2.1 inc. Tc = 1.77 hr.
9 SUB BASINS
Tlrae
-------
NS
O
OS
9GO
BOO -
70O -
60O -
50O -
40Q -
300 -
20O -
100 -
EXHIBIT NO. 2
FLOOD HYDROGRAPH
Rain Depth = 2.1 in. Tc = 1.77 hr.
O.5
l.O 1.5 1.8 2.3 2.8 3.3 3.8
Time (lor.}
4.3
4.8
5.3
5.8
6.3
-------
REFERENCES
1 Pinto, Holtz, Martins, Gomidi, "Hidrologia Basica," Edgar
Blucher, Ltd., 1976.
2 Linsley and Franzi, "Water Resources Engineering," McGraw-
Hill, 1979.
3 Verdin and Morel-Seytoux, "User's Manual for XSRAIN - A
Fortran IV Program for Calculation of Flood Hydrographs
for Ungaged Watersheds," FHWA, 1981.
4 McCuen, Richard H., "A Guide to Hydrologic Analysis Using
SCS Methods," Prentice-Hall, Inc., 1982.
5 Smedema and Rycroft, "Land Drainage: Planning and Design
of Agricultural Drainage Systems," Cornell University
Press, 1983.
6 "CE-577: Urban Water Management," Class Notes, Colorado
State University, 1985.
7 McWhorter and Sunada, "Ground-Water Hydrology and
Hydraulics," Water Resources Publications, 1985.
8 "CE-522: Engineering Hydrology," Class Notes, Professor
Morel-Seytoux, H.J., Colorado State University, 1986.
207
-------
UNIT-HYDROGRAPH PROCEDURES FOR ARID LANDS
by: George V. Sabol, Ph.D., P.E.
Consulting Engineer
Brighton, Colorado
and
Joe M. Rumann, Hydrologist
Davar Khali Ii, Ph.D., Hydrologist
Teresa A. Dominguez, Hydrologist
Flood Control District of Maricopa County
Phoenix, Arizona
ABSTRACT
Unit-hydrographs are used in most hydrometeorolog!cat flood analyses in
the arid west and throughout the United States. In many areas of the United
States there is adequate rainfall-runoff data to develop site-specific or
regional unit-hydrographs for use in a flood analysis; however, appropriate
unit-hydrographs or adequate rainfalI-runoff data for unit-hydrograph
development are seldom available in the arid west. Synthetic unit-
hydrographs are usually used for flood analyses in the arid west and numerous
synthetic hydrograph procedures are available; however, the applicability of
these procedures for use in the arid west is questionable. Furthermore,
there has been a renewed interest in S-graphs, a form of synthetic unit-
hydrograph, with the incorporation of this methodology in the Flood Hydrology
Chapter of the 3rd Edition of Design of SmalI Dams by the U.S. Bureau of
Reclamation. However, appropriate S-graphs may be difficult to obtain for
many applications since they were often developed for federal projects and
may have never been published.
Recently, two studies were conducted for the Flood Control District of
Maricopa County for the purpose of selecting or developing synthetic unit-
hydrograph procedures for use in Maricopa County, Arizona. A study was
conducted to compile S-graphs from the southwest and to select S-graphs for
use in the various physiographic land forms in Maricopa County, a second
study was conducted to collect rainfall-runoff data from the southwest and to
analyze this data to develop a synthetic hydrograph procedure for Maricopa
County. Rainfa I I-runoff data was compiled from the Walnut Gulch Experimental
Watershed in Tombstone, Arizona, Tucson Experimental Watersheds, urban
hydrology programs of the U.S. Geological Survey in Denver and Albuquerque,
and a U.S. Geological Survey data collection program in Wyoming. The first
study is briefly described, and the development and results of the second
study are presented.
203
-------
GENERAL
The Flood Control District of Maricopa County (FCDMC) is presently
preparing a Hydrology Manual and a Drainage Design Manual for use in planning
and designing flood control facilities in Maricopa County, Arizona. The
purpose of the manuals is to present recommended methods and design practices
so as to standardize design discharges and facilities across jurisdictional
boundaries in Maricopa County. Two volumes are being prepared; volume 1 will
be the Hydrology Manual which will contain criteria and recommended
procedures for performing flood hydrology, and volume 2 will be the Drainage
Design Manual which will contain engineering procedures and design guidelines
for the design of flood control facilities. The research, development, and
testing of the criteria and procedures that are to be incorporated into the
Hydrology Manual are being directed by the staff of the FCDMC. Certain
topics of the Hydrology Manual have been performed under contract to the
FCDMC by the senior author. The Drainage Design Manual is being prepared
under contract to the FCDMC by NBS/Lowry and Associates, Inc. of Phoenix,
Arizona, and Mclaughlin Water Engineers, Ltd of Denver, Colorado. The two
volumes are scheduled to be completed by the end of 1988.
The Hydrology Manual will contain criteria for the determination of the
design rainfall, including rainfall depth-duration-frequency information,
rainfall temporal and spatial distributions, procedures for estimating
rainfall losses, procedures for developing synthetic unit-hydrographs, and
procedures for routing runoff from the watershed or sub-basins. The
research, development, and procedures that are being considered for synthetic
unit-hydrographs for Maricopa County are presented.
HYDROLOGIC SETTING
Maricopa County has an area of 9,226 square miles which is about the
same size as the state of New Hampshire. The county lies in the Gila River
basin, a tributary of the Colorado River, and the area comprises a wide
diversity of physiographic and topographic conditions. Approximately 70
percent of the area is mountainous and the remaining 30 percent is valley.
The mountain areas above 3,000 feet in elevation are characterized by rugged
terrain and steep slopes. The valleys consist of alluvial fans, flat basin
floors, and alluvial floodplains. Much of the area is agricultural land that
has been leveled for irrigation applications. Urbanization has been, and
will continue to be, a major impact on the runoff potential in 'Maricopa
County.
Vegetation varies according to physiographic and climatic factors. In
general, the vegetation is sparse and cacti grow throughout the area. The
valley basin has sparse grass and shrub cover in its natural condition,
although much of the area in the Phoenix metropolitan area is irrigated turf,
particularly large areas of golf courses. Hi I[slopes are populated by cacti
and shrubs, and the higher mountains have stands of trees with underbrush.
The diverse physiographic, topographic, and land-use conditions within
209
-------
Maricopa County requires synthetic unit-hydrograph procedures for all
conditions. That is, procedures must be available for major watercourses
with large drainage areas, small urban watersheds, natural and lightly
urbanizing hi I(slopes, Iasar-leveled agricultural land, alluvial fans, and
Sonoran desert. To compound the situation, very little research or data are
available for these conditions.
TYPES OF SYNTHETIC UNIT-HYDROGRAPHS
Two types of synthetic unit-hydrographs are being considered for use in
Maricopa County; S-graphs and Clark unit-hydrographs. S-graphs will be
recommended for major watercourses for which existing S-graphs are available
and applicable. The Clark unit-hydrograph will be recommended for smaller
watersheds and urbanized basins.
A study has been conducted to compile existing S-graphs for Arizona and
the southwestern United States. This study has resulted in the compilation
of 53 S-graphs, documentation of some of the watershed characteristics, and
an investigation into the development of an empirical relation for the sole S-
graph parameter, lag. A second study has been conducted to develop a
procedure for developing synthetic Clark unit-hydrographs. This study has
resulted in the selection of an equation for estimating the time of
concentration (TO, the development of an equation for estimating the storage
coefficient (R), and the development of two time-area relations, one for
urban and one for natural watersheds.
An S-graph is a form of unit-hydrograph and is often used in performing
flood studies. S-graphs are usually defined by the reconstitution of
recorded flood events and numerous S-graphs are available from such
reconstitutions. Existing S-graphs for the southwestern United States have
been compiled and reviewed (1). These and other S-graphs can be used
(transposed) to other watersheds for the purpose of defining a unit-
hydrograph under certain limiting conditions. The concept of the S-graph
dates back to the development of the unit-hydrograph itself, although the
application of the S-graph has not been as widely practiced as that of the
unit-hydrograph. The use of S-graphs has been practiced mainly by the U.S.
Army Corps of Engineers, particularly the Los Angeles District, and the U.S.
Bureau of Reclamation (USER). Recently the S-graph has been adopted as the
unit-hydrograph procedure by several counties in southern California and
selected S-graphs have been presented in their hydrology manuals. The S-
graphs in those hydrology manuals have been selected primarily from S-graphs
that had previously been defined by the Los Angeles District of the Corps of
Engineers from a rather long and extensive history of analyses of floods in
California. Other areas may not have the advantage of such an extensive data
base.
S-graphs may gain more popularity and usuage due to the release of the
Third Edition of Design of Small Dams by the USSR (2). This engineering
reference book is widely used as a guideline in flood hydrology, particularly
in the western United States, and the revised Flood Hydrology Studies chapter
discussses and presents the application of S-graphs. Six regionalized S-
210
-------
graphs are presented in the revised chapter of that book.
A discussion of the S-graph study for Maricopa County was recently
published (3) and this information will not be reproduced herein. Rather,
the data, analysis, and results for the Clark unit-hydrograph are emphasized,
CLARK UNIT-HYDROGRAPH
DATA
A preliminary study of rainfall-runoff data sources identified the
following instrumented watersheds and recorded events:
1. Walnut Gulch Experimental Watershed, nine watersheds and nine events,
2. Tucson Experimental Watershed, four watersheds and forty-two events,
3. Denver urban hydrology program, fourteen watersheds and twenty-eight
events,
4. Albuquerque urban hydrology program, seven watersheds and twelve events,
and
5. Wyoming study, seven watersheds and twenty-one events.
ANALYSIS
A final data selection resulted in the analysis of 51 storm events from
28 watersheds. The final selection of the 51 storm events was made after the
rainfall hyetographs and runoff hydrographs were plotted, and after the data
was screened to assess whether the rainfall and runoff data appeared
representative of each other. Numerous problems are associated with rainfall-
runoff data, and a common problem that is difficult to assess is whether the
measured point rainfalls are representative of the temporal and spatial
rainfall over the watershed. The selection process is rather subjective
because of the uncertainties in the data.
Flood Reconstitutions
Prior to execution of the flood reconstitutions using HEC-1 two
preliminary analyses had to be performed; the effective Impervious area had
to be determined, and a representative rainfall distribution had to be
selected for watersheds with more than one recording raingages. The
inability of HEC-1 to distinguish between total impervious area and effective
impervious area is considered to be a serious deficiency of the HEC-1 model.
Effective impervious area is the impervious area of the watershed that
would drain to the outlet without passing over pervious area. This is also
called directly connected impervious area. For each urban watershed the
effective impervious area was estimated by selecting all the storms that
appeared to be of .low- to mediurn-intensity and uniform distribution over the
watershed. Most of these rainfalls were less than 1.0 inch and greater than
about 0.3 inch. Using these events, the effective impervious area was
calculated by dividing the recorded volume of runoff by the average depth of
rainfall (assumed equivalent uniform depth of rainfall and runoff).
Exceptionally high or low values of effective impervious area were eliminated
211
-------
and the average of all calculations was taken as the effective impervious
area.
The time distribution used in flood reconstitution was either the
recorded distribution if only one recording raingage was available or a
composite of all of the recording raingage data. The composite time
distribution was determined by plotting all of the rainfall mass diagrams for
each raingage on a single sheet of graph paper. A single representative mass
diagram was drawn by considering individual raingage location and also the
timing of the runoff hydrograph. This method was preferred to using the
option in HEC-1 of weighting rainfall depths and rainfall distributions from
individual raingages.
Flood reconstitut ions of the 51 selected events were performed to
determine the "best fit" Clark unit-hydrograph parameters that would
reproduce the storm hydrograph from the recorded rainfall. The flood
reconstitutions were performed by using the parameter optimization option of
HEC-1. The resulting unit-hydrograph is listed as output of the HEC-1
optimization runs. The Clark unit-hydrograph has three parameters therefore
numerous combinations of the three parameters could result in equally good
reproductions of the storm hydrograph; however the individual parameter
values could be in error. The error in estimating TC and R may be
particularly significant if the third parameter, the time-area relation, is
fixed a priori such as by using the HEC-1 default time-area relation and then
determining the optimum values of TC and R. It is therefore desirable to
estimate one of the three parameters before using the HEC-1 optimization
option.
An estimate of R was obtained through a recession analysis of each of
the runoff hydrographs (4). Parameter optimization runs were then made using
various time-area relations in a trial-and-error procedure until the
optimized value of R was reasonably close to the previously estimated value
of R. Where more than one storm event had been selected for the same
watershed it was determined that the same time-area relation provided the
best fit for optimization of all events for the watershed. This was
encouraging and provided confidence in the optimization process.
The computation interval could be a controlling factor within HEC-1 when
optimizing on TC. The computation interval had to be selected such that it
was less than or equal to 1/3 TC. However, the shortest computation interval
that can be used is one minute. It must be realized however, that the
resolution of the rainfall data was not such that the rainfall distribution
could always be accurately reproduced at this small of a computation
interval. There probably was an artificial "smoothing" of the rainfall
distribution by this process and this can be expected to result in error in
the optimized values of the parameters, particularly TC.
The reconstitutions were evaluated and 13 of the reconstitutions were
rejected, leaving 38 "valid" reconstitutions. Often this rejection was based
on the belief that the recorded rainfall was not truly representative of the
rainfall over the watershed. The data base was again critically reviewed and
19 control events were selected as being "most accurate."
212
-------
Time of Concentration, TC
Two approaches were used to investigate the development of a TC
prediction equation. First, a literature search was conducted to determine
methods that are presently available for estimating TC, and second, an
attempt was made to develop a TC prediction equation from the data base of
flood reconstitutions. The best TC equation was judged to be that of
Papadakis and Kazan (5).
The selected TC equation Is
TC= 11.4 I'50 n'52 S-31 i"38 (1)
where TC is in hours,
L is length of flow path, In miles,
n Is the Mannings coefficient,
S is watersourse slope, in feet/mile, and
I is average rainfall excess Intensity, in Inches/hour.
Papadakis and Kazan (5) present data and a discussion of the development of a
graphical relation for the estimation of "n". Use of this figure eliminates
the uncertainty in estimating "n" and at least makes the slection of "n" for
a watershed a reproducible process from one individual to the next. The
figure presented by Papadakis and Kazan has been modified, and equations for
selecting "n" for use in Maricopa County are presented in Table 1.
Verification of the equation resulted in redefining "i" as the rainfall
excess intensity during a time-interval, TC.
Table 1. Estimation of "n" in the Maricopa County time of concentration
equation (TC) for the Clark unit-hydrograph
n = m log DA + b
where DA is drainage area, in acres
Equation Parameters
Land Classification
(1 )
Urban
Bare or nearly bare ground
m
(2)
-.0025
-.00625
b
(3)
.02
.04
(alluvial fan, agricultural land,
desert rangeland)
Rough and/or moderately vegetated -.01375 .08
(hill slopes)
Very rough and/or dense vegetation -.025 .15
(mountains)
213
-------
Storage Coefficient, R
Methods to estimate R for ungaged watersheds were investigated in a similar
manner as used for TC, however there is much less literature available. An R
prediction equation was developed from a stepwise multiple regression of the
data for the 19 control events. The R equation is
R = .37 TC1'11 A"'57 I'80 (2)
where R is in hours,
TC is in hours as calculated by Equation 1,
A is drainage area in square miles, and
L is length of flow path, in miles.
A comparison of the estimated R to the R from the flood reconstitutions was
conducted, and the equation is unbiased and appears to be equally as valid
for the non-control events as for the control events for which the equation
was developed.
Time-Area Relation
The best fit time-area relation was determined using a trial-and-error
process of 1) selecting a dimensionless form of a time-area relation, 2)
performing a HEC-1 optimization with that relation, and 3) evaluating the
results. Different time-area relations were tried until the following
criteria were met:
1. The peak discharge and time to peak of the reconstituted hydrograph
were the best fit to the recorded hydrograph.
2. The general shape of the reconstituted and recorded hydrographs were
si mi Iar.
3. The reconstituted value of R was as close as possible to the R value
from the recession analysis.
The development of time-area relations from maps and watershed
information is a tenuous procedure, and unreliable and inconsistant resul\ts
will be achieved. This is especially true of urban watersheds because of the
complex and convoluted drainage patterns that usually result from
development. It is desirable to have dimensionless time-area relations that
can be used for various types of watersheds. The Hydrologic Engineering
Center has developed a default time-area relation for use with HEC-1 (6).
This relation between travel time and contributing drainage area is very
nearly linear. This HEC-1 default relation is listed in Table 2. All flood
reconstitutions started with the default relation and were modified on a
trial-and-error basis until the final relation was developed.
For urban watersheds the time-area relation is generally advanced
indicating a rapid runoff response. A composite of the time-area relations
that evolved for urban watersheds is shown in Table 2, and is designated as
the urban default time-area relation. For natural watersheds the time-area
relation is generally delayed indicating a retarded runoff response. A
composite of the time-area relations that evolved for natural watersheds is
shown in Table 2, and is designated as the natural default time-area
re I at ion.
214
-------
Clark Unit-Hydrograph Procedure for Ungaged Watersheds
A procedure to synthesize a Clark unit-hydrograph is as follows:
1. Estimate TC using Equation 1.
of flow path (L), watercourse
excess (i), and resistance to
This equation is a function of length
slope (S), intensity of rainfall
flow (n). Use Table 2 to estimate
This equation
1,
is a function of time of
length of the flow path
Estimate R using Equation 2.
concentration (TC) calculated by Equation
(L), and drainage area (A).
Develop the appropriate time-area relation which would be expected
to fall within appropriate envelopes or select the default time-area
relations.
Discussion of Procedure
This procedure should account for the physical processes that are
occurring in a watershed; both watershed and rainfall characteristics are
incorporated into the procedure. The effect of urbanization is reflected in
"n" and in the time-area relation. Since R is a function of TC, R wi I I also
vary due to urbanization. The selection of "n" is simplified by the use of
equations in Table 2 and use of these equations will remove much of the
subjectivity and uncertainty in the selection of "n".
The procedure can be used over a wide range of watershed characteristics.
The shape of the resulting unit-hydrograph should vary in a predictable
manner based on the watershed characteristics. Impervious area has not been
a significant variable in the development of these procedures. However
impervious area will be incorporated in the calculation of rainfall excess
which is a very sensitive parameter for rainfall-runoff modeling, therefore
the absense of impervious area as a variable in the unit-hydrograph procedure
is not a concern. Impervious area is indirectly incorporated into the
procedure through the variable "i" of the equation for TC.
Table 2. Time-Area relation in the Maricopa County Clark unit-hydrograph.
Dimensionless
time as percent of
time of concentration
(1)
Dimensionless area as percent of total area
HEC-1
default
(2)
Urban
default
(3)
Natural
default
(4)
0
10
20
30
40
50
60
70
80
90
100
0
4.5
12.6
23.2
35.8
50.0
64.2
76.8
87.4
95.5
100.
0
5
16
30
65
77
84
90
94
97
100
0
3
5
8
12
20
43
75
90
96
100
215
-------
ACKNOWLEDGEMENT
The research and development that was performed in the undertaking of
this study was conducted by the senior author of this paper while under
contract to the Flood Control District of Maricopa County, Arizona. The
studies were conducted in close collaboration and review by the staff of the
Flood Control District, and the senior author acknowledges the support and
cooperation of the Flood Control District and its staff, and without their
support this study could not have been completed. Results that are presented,
herein, are preliminary and are not necessarily the final results that will be
adopted in the Hydrology Manual. Conclusions that are presented, herein, are
not necessarily those of the Flood Control District of Maricopa County.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the
views of the Agency and no official endorsement should be inferred.
REFERENCES
1. Sabol, G.V., S-Graph Study. Report for the Flood Control District of
Maricopa County, Phoenix, Arizona, 29 pgs. plus appendices, 1987.
2. U.S. Bureau of Reclamation, Design of SmalI Dams. Third Edition, 1988.
3. Sabol, G.V., Development, use, and synthesis of S-graphs. In; Proceed-
ings of the Engineering Hydrology Symposium, Amer. Soc. of Civil Engrs.,
WiI Iiamsburg, Virginia, pp. 627-632, 1987.
4. Sabol, G.V., Clark unit-hydrograph and R-parameter estimation. Jour, of
Hyd. Div., Amer. Soc. of Civil Engrs., V. 114, No. 1, pp. 103-111, 1988.
5. Papadakis, C.N., and Kazan, M.N., Time of concentration in saml1 rural
watersheds. In; Proceedings of the Engineering Hydrology Symposium,
Amer. Soc. of Civil Engrs., WiI Iiamsburg, Virginia, pp. 633-638, 1987.
6. U.S. Army Corps of Engineers, HEC-1 Flood Hydrograph Package. 1987.
216
-------
DETERMINATION OF DESIGNATED FLOODWAY BOUNDARIES
AROUND LONG ISLANDS IN STREAM CHANNELS
by: J.F. Harp, Ph.D.
Professor, Civil Engineering Dept.
The University of Oklahoma
Norman, Oklahoma 73019
ABSTRACT
Whenever river islands are encountered in hydraulic analysis, special
problems exist with respect to the proper application of most backwater pack-
age programs. There is a problem because the flow division in the side
channels must remain constant, unless there is provision for short circuiting
of the flows along the riverine pathway. The correct solution is achieved
whenever the flows divide such that the head loss around each side-channel is
the same.
The correct flow division, and resulting backwater profile is achievable
using the famous HEC-2 Computer Program (1). However even with computer as-
sist, a trial and error procedure is required. This procedure is a short and
simple process and requires only a small effort.
Once the proper flow division is achieved around the islands, a further
ominous problem exists in the computation of the designated floodway
stations. The problem arises because there are two separate side-channels,
and there is only one floodway, and one set of encroachment stations at the
outer streambank area. Uniqueness is not possible, even with present tech-
nology. However, reasonableness is available.
The methods described in this paper strive for a solution and there are
three procedures that are briefly described. One method simply utilizes a
wide cross-section containing all the GR points, and a usual application of
the HEC-2 methodology. Another method assumes that the original flow ratio
division remains constant, and two runs are made around the island using the
tributary option and the preserved energy gradient. A final and preferred
procedure is set out whereby smooth encroachment stations are set at outer
reasonable delineations, and the side-channel flows are rebalanced until
equal head losses again occur around each side-channel, while the incremented
water surface elevation is achieved. Any other procedures are likely to
217
-------
produce inaccuracies.
The vehicle utilized in the solution techniques described in this paper
is the famous HEC-2 Corps of Engineers Package Computer Program (1). It is
assumed that interested readers have familiarity with this vital tool. The
procedures described here should be workable with other programs as long as
the basic criteria is met.
INTRODUCTION
In recent times, accurate backwater analysis, reasonable encroachment
stations, erudite bridge routines, and non-reproachable computer models,
generally, are absolutely required at City, County, State, and Federal
levels.
Few computer applications are as simple as merely writing down a field
data scheme into a package program. Every engineering problem has its own
difficulties and unusual configuration aspects. It is a rare problem, or
engineering job, that entails only rote substitution and application of data
directly into a satisfactory model. As an unknown philosopher once said,
"all the easy jobs are already done."
Some bridges have awkward shapes, some channels have poorly defined
cross-sections, extended valley conveyance, or variable coefficients, and it
is always expected that the encroachment routines will produce floodways that
do not result in a smooth delineation. Some reaches of stream will run crit-
ical at some flow rates. Islands sometimes exist in inconvenient places.
Whatever exists wherever a computer model is applied must be modelled with
dispatch and propriety. Engineers have long been famous for solving the
unsolvable.
DIVIDED RIVERINE CHANNELS
BACKWATER COMPUTATIONS
When a channel divides and flows around an island, the backwater model
around the island must be performed properly and with uncompromising accura-
cy. A proper solution lies in the fact that the head loss must be the same
around each side-channel. This is not a direct solution result from HEC-2,
or any other known backwater program.
Most islands sustain divided flows of different ratios for each recur-
rent flood event, seasonal roughness coefficients, or other modeling diffi-
culties. Sometimes there is a zero flow in one branch at low frequency
events. All these problems have been encountered, but rarely solved proper-
ly. In these days of litigation over even small gliches, we must be ready to
defend our work whenever a challenge is made.
A direct and immediate, but cursory, solution around islands or ob-
21!
-------
section 2 using the tributary option, and the appropriate minus sign, paying
attention to the flow values. Simply proceed upstream around this side chan-
nel using QTWO until cross section 5 is again reached. Stop the program and
compare the water surface elevation from the original side run with the water
surface elevation from this side run. Whenever the two values are the same,
a proper solution has now been achieved, if not, one must go back to the
original side channel run, change the flow component division, using engi-
neering judgement, and repeat the two runs around the island until the water
surface elevations are the same at the upstream end, section 5. Once this
has been accomplished, the rest of the model can be continued on upstream as
usual.
ENCROACHMENT STATIONS
A similar, but much more difficult problem exists with floodway deter-
mination around islands in stream channels. The divided flow procedure in
conjunction with one of the methods of encroachment will obviously provide
two sets of encroachment stations when only one is desired, see Figure 2.
Prudent users will note that an expedient delineation of the overbank
stations might eliminate this problem, but even then a proper interpretation
of the model results must be performed.
NATURAL FLOODWAY
DESIGNATED FLOODWAY
UNWANTED
/UINWAL1
Vf
Figure 2. The Encroachment Dilemma
The usual floodway will have only a left and right encroachment station.
Most usually, the island will be within the floodway, and properly so.
Therefore, how can a proper single valued floodway be achieved? One applica-
tion, known to this author, was to simply omit the island mound points, and
proceed with an erroneous model. This could be done by considering long
reaches and just not using cross sections at the island itself. In Figure 1,
this could be accomplished by skipping from cross section No. 1 to cross sec-
tion No. 6. The error of this procedure is obvious. A similar erroneous
219
-------
structions can be achieved by application of the famous HEC-2 package program
by simply coding the cross sections across the island as a single series of
ground points. Divided flow will surely exist, and, in successive upstream
sections, the flow division will not remain constant. Therefore, an errone-
ous solution will result. Errors of unsatisfactory magnitude are common
using this naive approach. The correctness of any solution lies in the fact
that the flow division remains constant in successive side channel cross
sections as the solution computations proceed upstream. The flow division
component must also be maintained around the other side-channel.
QTOTAL
\
V
\
QONE J.
\ f.
Figure 1. A Typical Island Encounterment
According to the Hydrologic Enginering Center (2) , a proper solution is
achieved when a correct flow division is assumed, and a backwater run is made
around one side of the island to a point just upstream from the flow
division. Then using the tributary option, picking up the energy line at the
downstream confluence and proceeding upstream around the other side of the
island until the water surface elevation is the same at the upstream selected
cross section where the flow divides initially. This is best explained by
the use of a simple example. Consider Figure I, with downstream
cross-sections Number 1 and 2, then cross section Numbers 3, 4 and 5 past the
island to cross section Number 6 where the usual backwater run will continue
on upstream.
To further elucidate the procedure, assume that the total flow in the
River is, say QTOTAL, and that the backwater computations have already pro-
ceeded up to cross section 2. Now, assume a. flow division, say QONE, and
proceed up one side channel using only that part of the cross section 5 per-
taining to that side. Stop the codeup at cross section and return to cross
220
-------
solution was published, accepted by local, state, and federal agencies, and
was not noticed for a decade, since the site was not near anything of concern
at that early date.
To accomplish the solution to the encroachment problem around an island,
procedures are suggested whereby the tributary option is utilized and the en-
croachment stations are set using one of the "manual" options, such as Method
One in HEC-2, along with one of the suggested procedures set out here.
SOLUTION ONE
The solution proposed first applies whenever the island is short, the
flow line elevations are roughly the same, vegetation and roughness are well
defined, and a near constant flow division exists around the side channels.
Simply code up the cross sections across one set of GR records, make a rea-
sonable decision about the overbank selection, and simply make a single run
up the channel, in the case of subcritical flow, or down the channel in the
case of supercritical flow. In order to evaluate the flow distribution in
the side channels, either the flow distribution option can be employed, or
the QLOB, QROB, and QCH values can be examined for satisfactory results.
SOLUTION TWO
To utilize this procedure, assume that the original flow division will
remain the same, or approximately so. This is important in order to achieve
the associated solution. This procedure assumes that the flows are the same
both before and after encroachment, or else the problem has too many vari-
ables, posing some ponderance about just what must be done. This solution
may or may not be reasonable generally.
To accomplish solution two, using HEC-2, the floodway delineation run
can proceed upstream to cross section 2 using the method of your choice or
requirement. Next, switch the ET record to Method One whereby manual setting
of the encroachment stations is appropriate. Then using engineering judge-
ment select the extreme, or outboard encroachment station as the run proceeds
up one branch, while setting the island encroachment station at a point "high
and dry," effectively eliminating the island from encroachment delineation.
Continue this procedure up to cross section 5. Then call up the tributary
option, and repeat the procedure until the other side channel encroachment
run is made and the water surface elevations balance. This will require some
trial and error, but with expeditious engineering judgement it will be a
minimal exercise.
Even after the run is completed, the results must be interpreted and the
island encroachment stations must be ignored, since they are high and dry.
Note now that the floodway has been completely determined and any adjustments
can now be made to satisfy the requirements of prudent and reasonable
floodway determination.
221
-------
SOLUTION THREE
If side channel roughness is highly variable, or a large disparity in
the channel flow division is anticipated, an approach must be considered so
that the flow division ratio is allowed to be different from the initial
backwater run upstream past the island. One procedure is to determine the
floodway encroachments upstream and downstream from the island, and using en-
gineering judgement delineate reasonable encroachment stations on the right
and left, and achieve a proper flow division using the method already de-
scribed. Floodways can be delineated using the guidelines set out by Harp
and Hayes (3), insofar as is proper and reasonable. These guidelines are
listed below for the convenience of interested readers.
(1) That the hydrology and hydraulics be based upon existing con-
ditions .
(2) That the discharges be based upon a one percent exceedance
frequency.
(3) That the flood plain will be divided into a central designated
floodway and a floodway fringe area on each side of the designated
floodway.
(4) The designated floodway will pass the flood discharge without caus-
ing the water surface to rise more than one foot above the natural
water surface elevation.
(5) The floodway fringes are assumed filled solid for purposes of hy-
draulic computation.
(6) There should not be a significant increase in stream velocity.
(7) That there should not be unreasonable depths in the floodway
fringes.
(8) There should not be undulating top widths.
(9) That the floodway should be consistent with local needs.
(10) That the floodway should be consistent with engineering judgement.
(11) That in improved channels where the capacity of that channel will
be sufficient to carry the one percent exceedance discharge, the
encroachment stations can be set at the channel overbanks where
they will be high and dry, and meet all agency rules and
regulations.
222
-------
CLOSURE
The most challengeable backwater models used today deal with the flood-
way delineation problem in general. The HEC-2 package program provides no
less than six encroachment algorithms. Uniqueness is virtually unobtainable,
even though some of the methods, or algorithms, are so-called automatic de-
lineations. In the case of floodway delineation around islands, where only
the outer floodway definition is wanted, special consideration must be used
in achieving a satisfactory result.
As with any computer program utilization, careful interpretation must be
performed. Sometimes, the novice will stumble upon an acceptable solution
through naivety, but day in and day out, the prudent engineering practitioner
will strive for the solution that satisfies all the data and legal require-
ments .
Undoubtedly, other programs besides HEC-2 can be used to accomplish the
same goals, and many readers will think of ways to interpolate, or enhance
some of the procedures presented here as an assist to the profession.
The advantages, and disadvantages of the three, and possibly four, so-
lutions are presented here will be intuitive to qualified readers and a de-
tailed discussion is deemed unnecessary. The author welcomes comments from
others regarding their own experience, or other solution.
The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessarily re-
flect the views of the Agency and no official endorsement should be inferred.
REFERENCES
1. Water surface profiles. HEC-2 Users Manual, Hydrologic Engineering
Center, Davis, California, September 1982.
2. Accuracy of computed water surface profiles. Research Document No. 26.
Prepared for the Federal Highway Administration, The Hydrologic Engi-
neering Center, Davis, California, December 1986.
3. Harp, J.F., and Hayes, R.J. The variability of floodway encroachment
determination. Flood Hazard Management in Government and Private Sec-
tor. In; Proceedings of the Ninth Conference of the Association of
State Floodplain Managers, New Orleans, Louisiana, May 1985.
223
-------
GULF COAST FLOOD ROUTING
by: Ronald L. Rossmiller, Ph.D, P.E., Project Manager
Kenneth R. Wright, P.E., Chief Engineer
Wright Water Engineers, Inc.
2490 West 26th Avenue, Suite 100-A
Denver, CO 80211
ABSTRACT
Low elevation urban centers along the coast of the Gulf of Mexico
require special flood control measures to provide adequate drainage. Gulf
of Mexico climate and hydrological characteristics create speHal
challenges. Normal tidal variations of minus 2 to plus 1.5 feet mean sea
level (msl), typical storm surges ranging to 5 feet above msl, rainfall
rates of 8 to 11 inches per hour, and annual rainfall amounts of 55 inches
are characteristic.
Draining a city using Federal Emergency Management Agency (FEMA)
criteria for base floods requires special evaluation of concurrent
tailwater elevations, potential wind friction on flowing water, and careful
conservation of energy for channel design.
Of particular concern to the drainage engineer are the flat slopes of
the land surface and channels with usual grades of 0.0002 to 0.0005 feet
per foot. Flood flow modeling and routing of stormwater where bayou
thalwegs may range from minus 5 to 10 feet below msl are described at
Beaumont, Texas.
A plan for a mid-city interceptor channel is presented which provides
downstream bayou flood relief. Another potential southern interceptor
channel would provide not only additional downstream bayou flood relief,
but water quality improvement to the stormwater runoff as well.
The intent of this paper is to provide its readers with information
gained from an investigation of runoff in a southern coastal city. The
techniques and alternatives investigated may prove useful in other
locations as well.
224
-------
GULF COAST FLOOD ROUTING
INTRODUCTION
The design of flood control and stormwater management measures in
coastal Texas areas is different than, for example, those in the Denver,
Colorado area where this conference is being held. Engineers must play the
hand that is dealt to them. In the case of Beaumont, Texas, the hand
includes:
1. an average of 55 inches of rain each year
2. land slopes ranging from 0.02 percent to almost 1.0 percent
3. channel slopes ranging from 0.02 percent to 0.1 percent
4. clayey soils
5. tidal effects
6. hurricanes
7. large available storage volumes in the floodplains
8. an emerging stormwater management policy
9. usual social, political, administrative and funding problems
Playing this hand requires knowledge and experience so that recommended
solutions to current flooding problems reflect what actually occurs during
a runoff event of a particular magnitude and which build on the drainage
system, both natural and manmade, currently existing in the Beaumont area.
Tools, such as HEC-1 and HEC-2 of the U. S. Army Corps of Engineers
(USAGE), are available to help develop these solutions. However, the input
data to these computer programs and analysis of their output must be made
to insure that the results realistically portray what occurs during some
rainfall event. This paper describes these inputs and analysis of the
results.
Several alternatives were investigated to alleviate current flooding
problems. The objective of each of the alternatives was to produce a
system of detention facilities and/or channels which would contain the
100-year discharge within the channel banks and also to obtain a water
surface elevation for the 10-year event which would allow the design of
economical storm sewers flowing into the open channels.
The paper also describes two interceptor channels within the city which
would alleviate current flooding problems. One of them will also have a
beneficial impact on forthcoming EPA stormwater runoff quality rules and
regulations.
DESCRIPTION OF AREA
Beaumont, Texas is located about 30 miles north of the Gulf of Mexico.
It is located in Jefferson County, which is located on the coast
immediately west of Louisiana. Beaumont is home to about 120,000 people in
an area of about 80 square miles. The city has a mayor, council, city
manager form of government.
As noted above, annual rainfall averages about 55 inches. Table 1
lists rainfall amounts for various storm durations and recurrence
intervals. As shown in Table 1, the 100-year, 24-hour event has 13.2
225
-------
inches of rainfall. These values were obtained from Technical Paper No. 40
(1). Because of the flat slopes, a local saying is that when it rains 12
inches in Beaumont, the water gets a foot deep. This is almost literally
true in some sections of the city.
TABLE 1. RAINFALL DEPTHS IN BEAUMONT, TEXAS FOR VARIOUS DURATION STORMS AND
RECURRENCE INTERVALS (INCHES)*
Storm
Duration
Hours
Recurrence Interval (Years)
5 10 2B~
TO"
TOO"
0.5
1.0
2.0
3.0
6.0
12.0
24.0
1.9
2.5
3.1
3.4
4.0
4.6
5.5
2.4
3.1
3.8
4.3
5.2
6.2
7.5
2.7
3.5
4.4
4.9
6.1
7.4
8.8
3.1
3.9
5.0
5.6
7.1
8.6
10.2
3.4
4.3
5.7
6.3
7.9
9.8
12.0
3.7
4.7
6.3
7.0
8.8
11.0
13.2
* From Technical Paper No. 40 (1)
The principal products of the area are oil and rice. Spindletop, the
original oil field, is located in Beaumont. Refineries and storage
facilities are also located in the city. The Port of Beaumont is located
along the Neches River. Products are loaded at the port and shipped down
the river to Sabine Lake, which is connected to the Intercoastal Waterway
and the Gulf of Mexico.
As described above, both the land and channel slopes in Jefferson
County are very flat. Land elevations range from about elevation 15 at the
southern end of the city to elevation 40 at the northern end. Tidal surge
from a force 5 hurricane is estimated to reach elevation 14.
Most of the original bayous have been widened and straightened. They
have trapezoidal shapes and are lined with either concrete or grass. Their
invert elevations are several feet below msl downstream of Beaumont.
Invert elevations of the major channels, such as Hillebrandt Bayou, Willow
Marsh Bayou, and Hillebrandt Bayou Oxbow, are also still below sea level at
the southern end of the city. Thus, tidal influences are felt within the
city. The USAGE has recently finished construction of a salt water barrier
on Taylors Bayou just west of the city of Port Arthur. It is closed to
prevent saltwater from moving upstream into the rice fields. It is opened
during runoff events. However, since its top elevation is plus 2.5 feet
above msl, it is overtopped during major runoff events.
Drainage paths within the city consist of open channels, streets, storm
sewers, and open ditches adjacent to the streets. Because of the flat
slopes, tidal effects, and present channel capacities, water backs up
through the system resulting in street and home flooding, in some areas of
the city, the first floor elevations of the homes are at or below street
level. 226
-------
RESPONSIBILITY FOR DRAINAGE
Responsibility for drainage in Beaumont is divided between the city's
public works department and Drainage District No. 6 (DD6). Drainage
districts were authorized by state legislation in the early 1900s. They
have the power of eminent domain, can collect taxes and can issue bonds.
DD6 maintains their channels with their own crews, but they contract for
all engineering and construction services.
DD6 owns the rights-of-way for their facilities within the city.
Traditionally, DD6 has constructed and maintained all the major open
channels within the city. The city has constructed and maintained the
storm sewer systems which flow into the DD6 open channels.
This dual responsibility for drainage within the city of Beaumont
requires that these two political entities cooperate to best serve the
drainage needs of the city's citizens.
HYDROLOGIC MODELING
Most of the area within Beaumont drains south into Hillebrandt Bayou, a
tributary of Taylors Bayou. The northern portion of the city drains north
into Pine Island Bayou. A small portion of the city drains east directly
into the Neches River. Hydrologic modeling of this drainage system was
accomplished using the USAGE HEC-1 computer program (2).
A HEC-1 model developed by another consulting firm for the DD6 master
drainage plan (3) was utilized as the starting point for Wright Water
Engineers (WWE) work. The-127-square mile Hillebrandt Bayou watershed had
been divided into 98 subareas. The inputs for drainage area, loss rates,
time of concentration, percent imperviousness and discharge/storage data
for channel routing were reviewed and revised as necessary.
Because of the flat slopes, tidal effects, and wide floodplains which
result in slow flow velocities, both times of concentration for the various
subareas and the discharge/storage relationships in the various channel
reaches were modified to reflect these conditions.
Another paper to be presented at this conference by Dr. Mohammed Samad
(4) presents in detail how the USAGE HEC-1 and HEC-2 computer programs (2,
5) were used in an iterative process to determine the proper balance
between discharge rates, water surface elevations, and channel storage
volumes in this flat coastal region of Texas.
Additional HEC-1 computer models were also developed to portray the
storage, channelization, and interceptor channel alternatives investigated
during Wright Water Engineers' review of the adequacy of the Beaumont
drainage system.
HYDRAULIC MODELING
The USAGE HEC-2 computer program (5) was used to depict the water
surface profiles for the various alternatives and recurrence intervals
(10-, 25-, and 100-year) investigated during Wright Water Engineers' review
of the adequacy of the Beaumont drainage system. Again, the basic models
227
-------
developed for the DD6 master plan were used as the basis for WWE's
investigations.
As mentioned in the introduction, the existing system of channels
produces water surface profiles and elevations which are too high. Even
using large box culverts as storm sewers does not meet the city's
requirements of a water surface elevation below the top of curb during the
10-year event.
The objective of each alternative was to produce a system of detention
facilities and/or channels which contain the 100-year discharge within the
channel banks and have a water surface elevation for the 10-year event
which would allow the design of economical storm sewers flowing into the
open channels.
The two most important items involved in the determination of an
acceptable water surface profile in every channel were the downstream
backwater elevation and the head losses at the bridges and culverts.
The beginning water surface elevations at the downstream end of
Hillebrandt Bayou at its confluence with Taylors Bayou were determined from
the USAGE report on its Taylors Bayou project (6). These elevations were
4.2 and 4.5 feet above msl for the 10- and 25-year events, respectively.
An elevation of 5.0 feet above msl was selected for the 100-year event.
After the water surface profile was obtained for Hillebrandt Bayou, the
water surface elevation was obtained at its confluence with each of its
tributaries: Bayou Din, Willow Marsh Bayou, Moore Street Drain, Usan Ditch,
llth Street Drain, Janes Gully, Hillebrandt Bayou Oxbow, and Keith Ditch.
After computing the water surface profiles for these tributaries, the water
surface elevations at each of their confluences with their tributaries were
obtained. These in turn became the starting downstream water surface
elevations for each of the minor tributaries.
The second important item is the head losses at bridges and culverts.
Because of the flatness of the channel slopes, the rise in water surface
caused by the head loss at a crossing is reflected for some distance
upstream. The land surface rises only about 1 foot per mile in some
portions of Beaumont.
There are a total of about 250 crossings on Hillebrandt Bayou and its
tributaries, 160 culverts and 90 bridges. The head loss at almost all of
the bridges for future developed conditions ranges from 0.1 to 0.3 feet per
bridge. The head loss at the culverts for future developed conditions
ranges from 0.2 to almost 2.0 feet per culvert. This is due to the use of
projecting conditions at the entrance and exits to the culverts with little
or no transition between the culvert and the trapezoidal channel.
Since there are sometimes 4 and 5 culvert crossings per mile, the
resulting head losses contribute to the high water surface elevations
experienced throughout the city.
The HEC-2 models were revised to include hydraulically efficient
transitions between the culverts and channels. Also, where necessary,
existing culverts were recommended to be enlarged to reflect both the
228
-------
increased discharges from future urbanization and the need to reduce the
head loss at most of the culverts.
MID-CITY INTERCEPTOR CHANNEL
Due to the flatness of the terrain, the backwater effects from rural
portions of Taylors and Hillebrandt bayous 20 miles downstream are felt in
the city of Beaumont as higher water surface elevations. For this reason,
a mid-city interceptor channel was envisaged to cut off the portion of
Beaumont north of the Union Pacific Railroad from these downstream
backwater effects.
This channel would intercept flows up to the 100-year event from upper
Janes Gully, Amelia Cutoff (which includes a portion of Keith Ditch),
Hillebrandt Bayou, and Hillebrandt Bayou Oxbow and convey them directly
into the Neches River.
The starting water surface elevation in the Neches River would range
from plus 2 feet above msl to plus 7 feet above msl depending on the
recurrence interval, as opposed to the backwater elevations ranging from
plus 14 to 16 feet above msl when tidal and other downstream effects are
taken into account. The interceptor channel would reduce the water surface
elevation in the above channels from 2 to 6 feet during the 100-year event.
Improved crossing hydraulics would reduce these elevations even more.
An additional benefit of this alternative is that the runoff from about
20 square miles of urbanized watershed is diverted into the Neches River.
This reduces the flow in Hillebrandt Bayou by a similar amount. This
reduced flow results in somewhat lower water surface elevations in the
rural portions of Hillebrandt Bayou and reduces the need for improvements
to some of the downstream channels.
SOUTHERN INTERCEPTOR CHANNEL
Another interceptor channel in southeast Beaumont would have similar
effects on the water surface elevations in some of the older neighborhoods
in Beaumont. This interceptor channel would begin at the Beaumont
wastewater treatment plant and intercept flow from four tributary
watersheds which total about 10 square-miles: llth Street Ditch, Usan
Drain, Ector Street Drain, and the Moore Street Drain. The intercepted
flow would be conveyed to the Neches River.
These older neighborhoods are vulnerable to flooding for two reasons:
(1) many of the homes have first floor elevations near or below street
level and (2) an existing railroad is located on a berm about 4 feet above
the surrounding land and only one small trestle exists to allow runoff to
slowly drain from the neighborhoods.
This interceptor channel would lower the water surface elevations in
these channels for the same reasons as listed above and would also reduce
the flows in the downstream rural channels.
There are two additional benefits to be derived from this interceptor
channel. The city is under order by the EPA to divert the wastewater
treatment plant effluent from Hillebrandt Bayou to the Neches River. By
229
-------
constructing the interceptor channel with a bottom width of one-third to
one-half mile, the entire 100-year, 24-hour runoff can be stored for a
sufficient period of time to allow significant improvements in its quality.
Further treatment of the plant effluent would also be possible. The runoff
and effluent would be pumped into an existing channel at Port Arthur Road
and flow by gravity into the Neches River. Cost/benefit analyses will be
computed for all alternatives studied.
SUMMARY
This paper has discussed the investigations made by Wright Water
Engineers in the city of Beaumont, Texas located near the coast of the Gulf
of Mexico. This city of 120,000 people is beset by flooding problems
caused by a combination of heavy rainfall, clayey soils, flat slopes and
tidal effects.
The use of the USAGE HEC-1 and HEC-2 computer programs are described.
Both programs must be used together originally in an iterative process to
ensure a balance between discharge and valley storage. This must be done
for each alternative.
Existing master plan recommendations are not adequate because the
resulting water profile elevations are too high. Additional work needs to
be done in terms of storage, channelization, and interceptor channel
alternatives to lower these water surface elevations throughout the city.
Benefit/cost analyses will assist the city and DD6 to determine their best
approach to solving the city's current flooding problems.
The interceptor channel alternatives allow this reduction in water
surface elevations and the southern interceptor also has the potential for
water quality benefits under EPA's upcoming stormwater runoff quality
program rules and regulations.
The work described in this paper was not funded by
the U.S. Environmental Protection Agency, and
therefore, the contents do not necessarily reflect
the views of the Agency and no official endorsement
should be inferred.
230
-------
REFERENCES
1. Technical Paper No. 40, Rainfall frequency atlas of the United States.
U.S. Department of Commerce Weather Bureau, 1961.
2. HEC-1 Flood Hydrograph Package Users Manual. U.S. Army Corps of
Engineers Hydrologic Engineering Center, September 1981 (rev. January
1985).
3. Hillebrandt Bayou Watershed Report, Jefferson County Drainage District
Number Six. Bernard Johnson Incorporated, Bob Shaw Consulting
Engineers, October 1985.
4. Samad, M.A., Hydrologic Modeling of Watersheds Using HEC-1 and HEC-2.
paper presented at Stormwater and Water Quality Model Users Group
Conference, Denver, Colorado. October 3-4, 1988.
5. HEC-2 Water Surface Profiles Users Manual. U.S. Army Corps of
Engineers Hydrologic Engineering Center, September 1982.
6. Taylors Bayou, Texas, Drainage and Flood Control Project, U.S. Army
Corps of Engineers, Galveston, Texas, April 1969.
231
-------
STORMWATER AND WATER QUALITY
MODEL USERS GROUP CONFERENCE
Sponsored
by
Environmental Research Laboratory
Environmental Protection Agency
Athens, Georgia
Denver Urban Drainage
and Flood Control District
and
University of Colorado
at Denver
October 3 and 4, 1988
Denver, Colorado
List of Attendees
LASTHAHE FIRSTNAHE
ORGANIZATION
CITY
STATE
Albrecht
Bare
Barnwell
Belvin
Boyle
Brand
Brocknan
Brown
Chang
Chang
Cooke
Cunha
Cuninghan
Dahis
Diniz
Driver
Durrans
Eiffe
Fisher
John
Dan
Tonas D.
Laura K.
Jean
Cary
Clifford R.
Alice M.
George C.
Jim
Michael B.
Wolney C.
Brett A.
Doug
Elvidio
Nancy E.
S. Rocky
Michael A.
Debbie
HDR Engineering, Inc.
Donohue S Associcates,
Town of Castle Rock, CO
Environ. Research Lab.
Brown and Caldwell Consulting
Richard P. Arber 4 Associates
CH2HHILL
BCA Inc.,
Resources Consultants, Inc
City of Austin
Kiowa Engineering Corp.
Greenhorne and O'Hara, Inc.
Canp Dresser $ HcKee
Rocky Mountain Concultants
Resources Technology Inc.
Federal Center
Herrick & Company
Enviromental Consulting Eng
York $ Associates
Omaha NE
Sheboygan Wi
Castle Rock
Athens
Seattle
Denver
Denver
Parker
Ft Collins
Austin
Denver
Aurora
Ft Collins
Annandale
Denver
Albuquerque
Lakewood
Denver
Alabama
CO
CA
WA
CO
CO
CO
CO
TX
CO
CO
CO
VI
CO
NH
CO
CO
Al
232
-------
Frye
Theresa
The EDGE Group
Gildersleeve Perm
Gingery
Harp
Harrell
Heiat
Huang
Huber
Hughes
Jennings
Johanson
Johnson
Kindred
Kohlenberg
Krawczyk
Lake
Liberti
Liu
Love
Hack
Halldry
HcGhee
HcLaughlin
Meyer
Mhauglin
Ohlrogge
Pahl
Paulson
Peaman
Perasso
Rau
Rios
Ronberg
Rossniller
Rumann
Ruzzo
Sabol
Samad
Stepanek
Strecker
Terestriep
Terrell
Thomas
Tucker
Wells
Wells
Whitt
Wu
Yasenchak
Kevin W.
Jimiy
Sidney L.
Mohamad R
Poshu
Wayne C.
William
Marshall E
Robert C.
Lynn
Karolette
Bryan
Steve
Gary
Angelo S.
Phillip
Nancy B.
Brian W.
David
Terence J.
Ronold
Tiffl
Hloyd G.
Dennis
Randy
Cindy L.
Vicky
Paul
David M.
Erailio
Richard
Ronald L.
Joe M.
William P.
George V.
Mohammed
Vera
Eric W.
Mike
Pat
Mike
L. Scott
Richard A.
Paula B.
Gary
Jy
Moira
Gingery Engineering Co.
University of Oklahona
State of North Carolina
. Dept of Highways
Najarian S Associate. Inc
University of Florida
U of Colorado at Denver
. Federal Center
U of the Pacific
U of Colorado at Denver
Denver Water Dept.
Centenial Engineering
Jeferson County
HDR Engineering Inc.,
R.I. Dept of Enviro Managnent
Dept of Environ. Protection
David Love & Associates
Camp Dresser i McKee
Centential Engineering
Tulane University
Mclaughlin Water Engineers, T
Black & Veatch
RBD Engineering Consultants
Centenial Engineering
Western Water Consultatnts, In
Brown and Caldwell
The EPA Office
Brown and Caldwell
Stewart Environiental Consult.
J.F. Sato & Associates
Lane Engineering Service Inc.
Wright Water Engineers, Inc.
Flood Control District
Wright Water Engineer, Inc
Wright Water Engineers, Inc
Rocky Mountain Consultants
Woodward-Clyde Consultatns
Illinois State Water Survey
Western Water Consultants Inc.
Jerferson County
UD4FCD
Rapid City
Wells Engineers, Inc
U of North Carolina Charlotte
Tulane University
Nashville
Evergreen
Denver
Norman
Raleigh
Denver
EatonTown
Gainsville
Denver
Denver
Stockton
Denver
Denver
Arvada
Golden
Denver
Providence
Trenton
Boulder
Annandale
Arvada
New Orleans
Denver
Kansas City
Ft Collins
Arvada
Laramie
Denver
San Francisco
Denver
Ft Collins
Littleton
Lakewood
Denver
Phoenix
Denver
Brighton
Denver
Denver
Oakland
Champaign
Laramie
Golden
Denver
Rapid City
Omaha,
Denver
Charlotte
New Orleans
TN
CO
CO
OK
NC
CO
NJ
Fl
CO
CO
CA
CO
CO
CO
CO
CO
RI
NJ
CO
Vi
CO
Lo
CO
MO
CO
CO
WY
CO
CA
CO
CO
CO
CO
CO
Az
CO
CO
CO
CO
CA
IL
WY
CO
CO
SD
NE
CO
NC
LA
233
------- |