United States
Environmental Protection
Agency
Environmental Research
Laboratory
Athens GA 30613
EPA/600/9-86/023
September 1966
Research and Development
Proceedings of
Stormwater and Water
Quality Model Users
Group Meeting
March 24-25, 1986
Orlando, FL
-------
EPA/600/9-36/023
September 1986
PROCEEDINGS
OF
STORMWATER AND WATER QUALITY MODEL
USERS GROUP MEETING
March 24-25, 1986
Orlando, FL
Edited by
Thomas 0. Barnwell, Jr.
Center for Water Quality Modeling
Environmental Research Laboratory
Athens, GA 30613
Wayne C. Huber
Department of Environmental Engineering Sciences
University of Florida
Gainesville, FL 32611
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE. OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATHENS, GA 30613
-------
DISCLAIMER
The Information in this document has been funded in part by the United
States Environmental Protection Agency. Papers describing EPA-sponsored re-
search have been subject to the Agency's peer and administrative review, and
the proceedings have been approved for publication as an EPA document. Mention
of trade names or commercial products does not constitute endorsement or
recommendation for use by the U.S. Environmental Protection Agency.
-------
FOREWORD
A major function of research and development programs is to effectively
and expeditiously transfer technology developed by those programs to the
user community. A corollary function is to provide for the continuing ex-
change of information and ideas between researchers and users, and among the
users themselves. The Stormwater and Water Quality Model Users Group,
sponsored jointly by the U.S. Environmental Protection Agency and Environment
Canada/Ontario Ministry of the environment, was established to provide such
a forum. The group has recently widened its interests to include models
other than t!ie Stormwater Management Model and other aspects of modeling
water quality in urban and natural waters. This report, a compendium of
papers presented at the users group meeting held on March 24-25, 1986 in
Orlando, FL, is published in the interest of disseminating to a wide audience
the work of group members.
Rosemarie C. Russo, Ph.D.
Director
Environmental Research Laboratory
Athens, Georgia
-------
ABSTRACT
This proceedings includes 22 papers on topics related to the develop-
ment and application of computer-based mathematical models for water quantity
and quality management. The papers were presented at the semi-annual meeting
of the Joint U.S.-Canadian Stormwater and Water Quality Model Users Group
held on March 24-25, 1986, in Orlando, Florida.
Several papers deal with using stormwater and water quality models on
microcomputers and interfacing microcomputer software such as spread sheets
and data base managers with these models. Specific programs discussed
include the Storm Water Management Model, DABRO, HSPF, a simplified water
quality program, HAZPRED, QUAL-TX, and OTTSWMM.
Other papers discuss statistical properties of point and nonpoint
pollutant sources, particularly hinhw*« runoff and the effectiveness of
detention/retention basins for mitigating pollution. Two papers discuss
trophic state models in lakes and
IV
-------
CONTENTS
Page
FOREWORD .............................. iii
ABSTRACT .............................. iv
ACKNOWLEDGMENT ........................... vi i
"PORTING MAINFRAME-BASED NUMERICAL MODELS TO MICROCOMPUTERS: A
CASE STUDY USING THE EPA STORM WATER MANAGEMENT MODEL" ...... 1
R.M. Baker and K.J. Brazauskas
PORTABILITY, MAINTENANCE AND FORTRAN PROGRAMING STYLE ....... 20
T.O. Barnwell, Jr. and D. Disney
CONSIDERATIONS ON USE OF MICROCOMPUTER MODELS FOR STORMWATER
MANAGEMENT ............................ 28
P. Wisner, D. Consuegra, H. Frazer, and A. Lam
PC SOFTWARE FOR COMPUTATIONAL HYDROLOGY .............. 45
W. James, M. Robinson, and M. Stirrup
DABRO: A BASIC LANGUAGE PROGRAM FOR HYDROGRAPH COMPUTATION .... 58
B.L. Golding
APPLICATION OF A LOTUS SPREADSHEET FOR A SWMM PREPROCESSOR ..... 80
S.W. Miles and J.P. Heaney
IMPACT OF EXTENSIVE IRRIGATION PUMPAGE ON STREAMFLOW BY HSPF .... 93
A.K. Nath
THE USE OF SWMM TO PREDICT RUNOFF FROM NATURAL WATERSHEDS IN
FLORIDA ............................. 109
W.C. Downs, J.P. Dobson, and R.E. Wiles
A SIMPLIFIED WATER QUALITY COMPUTER PROGRAM FOR REGIONAL
STORMWATER MANAGEMENT SITE EVALUATION .............. 121
J.M. Crouse and M.H. Helfrich
DEVELOPMENT OF THE HAZPRED MODEL .................. 128
G. Zukovs, J. Kollar, and M. Shanahan
ALTERNATIVE CALIBRATION OF THE QUAL-TX MODEL FOR THE UPPER
TRINITY RIVER .......................... 147
R. McCarthy
-------
CONTENTS (cont'd)
LOGNORMALITY OF POINT AND NON-POINT SOURCE POLLUTANT CONCENTTATIONS. . 157
E.D. Driscoll
POLLUTION FROM HIGHWAY RUNOFFPRELIMINARY RESULTS 177
P.E. Shelley and D.R. Gaboury
EFFECTIVENESS OF DETENTION/RETENTION BASINS FOR REMOVAL OF HEAVY
METALS IN HIGHWAY RUNOFF 193
H.H. Harper, Y.A. Yousef, and M.P. Wanielista
SIMPLE TROPHIC STATE MODELS AND THEIR USE IN WASTELOAD ALLOCATIONS
IN FLORIDA 219
R.W. Ogbunr, P.L. Brezonik, and B.W. Breedlove
MODEL COMPLEXITY FOR TROPHIC STATE SIMULATION IN RESERVOIRS 235
R.A. Ferrara and T.T. Griffin
F.D.O.T. DRAINAGE MANUAL: WHY "DRAINAGE" IN AN AGE OF "STORMWATER
MANAGEMENT" 255
E.G. Ringe
APPLICATION OF THE OTTSWMM MODEL FOR RELIEF SEWER STUDY IN LAVEL,
QUEBEC 263
R. Roussel, J.C. Pigeon, and J.R. Noiseux
APPLICATION OF INLET CONTROL DEVICES AND DUAL DRAINAGE MODELLING
FOR NEW SUBDIVISIONS 275
P. Wisner, H. Fraser, C. Kochar, and C. Rampersad
USE OF CONTINUOUS SWMM FOR SELECTION OF HISTORICAL RAINFALL DESIGN
EVENTS IN TALLAHASSEE 295
W.C. Huber, B.A. Cunningham, and K.A. Cavender
AN EXPERT SYSTEM PROTOTYPE FOR RECEIV-II USING TURBO PASCAL 322
R.E. Dickinsonson, I.B. Chou, and F.V. Ramsey
MODELING FLOOD HYDROLOGY USING HYMO 326
J.E. Scholl
LIST OF ATTENDEES 332
-------
ACKNOWLEDGMENT
The Stormwater and Water Quality Model Users Group relies on local
hosts to make arrangements for its meetings. The hosts for the meeting
reported in this proceedings were Dr. Larry A. Roesner of Camp, Dresser,
McKee, Inc. and Dr. Wayne C. Huber of the University of Florida. Dr. Huber
reviewed abstracts of papers and arranged the meeting agenda. Dr. Roesner
made local arrangements for meeting rooms and hotel accommodations.
-------
"Porting Mainframe-Based Numerical Models to Microcomputers:
A Case Study Using the EPA Storm Water Management Model"
Richard M. Baker and Karl J. Brazauskas
Metcalf & Eddy, Inc., Wakefield, Massachusetts
ABSTRACT
This paper describes porting of the United
States Environmental Protection Agency Storm
Water Management Model (SWMM Version 3) from an
IBM mainframe CMS environment to a DOS
environment on an IBM PC AT microcomputer.
Subsequent application of the micro SWMM model
during a study of a combined sewer system in a
major New England city is then briefly reviewed.
INTRODUCTION
As a result of the rapid evolution of micro hardware and
recent advances in the development of micro-resident Fortran
compilers, porting and application of traditionally mainframe-
based numerical models on micros has become a practical and
extremely cost effective method of computing. The advantages of
1
-------
computing on micros are numerous and not restricted to cost
considerations alone. They include: easier, more user friendly
access, possible reduced turn-around times and integration of
pre-processing, numerical modeling, post-processing, word
processing, graphics and software development capabilities
within one, relatively compact machine.
Prior to porting of the SWMM model to micros, engineers
at Metcalf & Eddy executed it on a remote IBM 3033 commercial
time-share computer. Input data sets were developed on an in-
house Digital Equipment PDF 11/70 and transmitted to the remote
IBM mainframe via telephone line. Following execution of SWMM
on the mainframe, model results were stored on disk and
subsequently transmitted back to the in-house POP 11/70, where
printing was accomplished using a line printer.
Use of this remote job submittal system was satisfactory
as long as only one or two users were active at once. However,
during periods of heavy use, turnaround times increased
drastically from the usual 30 minutes to as much as three hours.
In addition, if either the in-house POP 11/70, remote mainframe
or telephone communications were down, computing would come to a
halt.
It was recognized that several options were available for
decreasing turn-around times on the above remote job submittal
system. However, these options would not address cost and
dependability issues. Due to several previous successes in
porting relatively small numerical models to micros at Metcalf &
-------
Eddy, porting of the SWMM model to in-house micros was initiated
in an attempt to address the issues of cost, dependability and
turn-around.
THE PORTING PROCESS
The process of porting numerical models from mainframe
(or mini) computers to micros can be divided into several major
tasks, including:
Obtaining the mainframe source code,
Transferring source code to the target micro,
Selecting a micro-resident compiler,
Modifying and compiling the source code,
Selecting a micro-resident linkage editor,
Developing an overlay structure and linking object modules,
Testing and documenting the ported numerical model, and
Applying the model on projects.
The above tasks are more or less the same regardless of
the relative size or complexity of the mainframe model. In
addition, the porting process is independent of the high level
language in which the model source code is written, e.g.,
Fortran, Pascal or c. However, selection of a micro-resident
compiler and linkage editor and development of an overlay
structure are best described using a large, relatively complex
numerical model like SWMM. As a result, the following detailed
description of the porting process for SWMM should serve as a
useful guide to others interested in utilizing the rapidly
expanding capabilities of micros.
-------
OBTAINING MAINFRAME SOURCE CODE
Mainframe source code for the latest version of the SWMM
model was obtained on 9-track magnetic tape from the USEPA
Environmental Research Laboratory in Athens, Georgia. This tape
included the main SWMM program and all subroutines and
functions. It also contained test input data sets and results
for use in verifying model operation. A listing of all files on
the SWMM tape was produced on an in-house line printer. The
SWMM source code was examined and found to be complete.
Each subroutine contains its own COMMON BLOCK statements.
However, it is important to note that many models incorporate
COMMON BLOCK specifications and other source code segments,
e.g., PARAMETER statements, into the main body of code at
compile time through the use of INCLUDE statements. Thus, it is
necessary to obtain source code for all of these included files
too.
Model documentation and users manuals for SWMM were
obtained from Dr. Wayne C. Huber at the University of Florida
in Gainesville. From an examination of these manuals, it was
found that the mainframe SWMM model utilizes overlays at run
time to reduce the size of the region in which the model
executes to approximately 400 kilobytes (kb) of random access
memory (RAM). As will be shown later, use of the above overlay
structure for the micro SWMM model eliminated the need to
develop one from scratch.
A description of the job control language (JCL) used with
-------
SWMM on the IBM mainframe OS/VMS (Release 3.8) operating system
was also included in the documentation. This information was
useful in identifying input and output (I/O), e.g., scratch
files, used during SWMM model execution.
Documentation for the Fortran compiler used with the
mainframe SWMM model was also obtained. This information was
used later in the porting process during selection of an
appropriate micro-resident compiler. In general, if the
mainframe Fortran compiler used with the model to be ported uses
features not included in the Standard Fortran-77 Language (ANSI
X3.9-1978), then the mainframe compiler documentation should be
obtained as an aid in porting to micros.
TRANSFERRING SOURCE CODE TO THE TARGET MICRO
The 9-track tape containing the SWMM source code and test
data was mounted on an in-house PDF 11/70 tape drive and copied
to a user area on a peripheral disk. The numerous SWMM files
were then merged using the PDP 11/70 operating system APPEND
utility.
Transfer of the master SWMM file from the PDP 11/70 to
the target in-house IBM PC AT hard disk was accomplished using
Hayes Smartcom communications software and a micro-based
internal smart modeia. Due to the use of relatively slow (1200
baud-rate) modems at both the PDP 11/70 and target micro,
transfer of the master SWMM file required almost four hours.
Transfer of mainframe source code to target micros may be
accomplished using several other methods, including: direct
-------
modem-based communication between mainframes, minis and micros,
magnetic tape and a nJcro-based tape subsystem or micros hard-
wired to minis or mainframes.
Service bureaus may also be used to copy data from
magnetic tape to target micro-compatable floppy diskettes. As a
last resort, hard copy of the mainframe source code can be
entered manually into the target micro using a micro- resident
text editor. When micro-based tape subsystems become cheaper
and more versatile, their use for transferring mainframe data to
micros and for backing up micro-based hard disks and floppy
diskettes will likely become commonplace. It is important to
realize that whenever data is transferred between mainframes,
minis and micros using magnetic tape, the compatability of tape
read/write utilities on these systems must be addressed.
SELECTION OF A MICRO-RESIDENT COMPILER
Documentation for the mainframe Fortran compiler used
with SWMM was examined. As expected, it was found to be a
superset of the ANSI Fortran-77 standard. Information on the
various language extensions available with this compiler was
used later during evaluation of alternative micro-resident
compilers.
The Microsoft Fortran compiler (Version 3.2) had been
previosly applied successfully at Metcalf & Eddy during porting
of several small numerical models from Digital Equipment
minicomputers to Intel 8086 processor based 16-bit micros. As a
result, compilation of SWMM was initially attempted using this
-------
compiler. Unfortunately, Microsoft Fortran was unable to compile
several of the larger SWMM subroutines, e.g, RHYDRO and TRANS,
due to its 64 kb limit on program size. In addition, this
compiler's memory model forced all local data, i.e, variables
used locally within subroutines, in the entire SWMM model to
reside within one 64 kb local group (DGROUP) in micro RAM. The
64 kb limitation on program size was overcome by splitting up
the several very large SWMM subroutines. However, the above
DGROUP restriction resulted in DOS "stack overflow" and " heap"
errors at model run time.
Due to the above limitations of the Microsoft Fortran
compiler, a search was initiated for another compiler. This
search resulted in selection of the Ryan-McFarland Fortran
(RM/FORTRAN 2.0) compiler for use in porting SWMM. RM/FORTRAN,
which is also marketed by IBM as Professional Fortran, is
available for either DOS or XENIX operating systems. It uses a
memory model which overcomes the 64 kb DGROUP restrictions of
Microsoft Fortran and allows program sizes up to sixteen
megabytes. RM/FORTRAN is certified by the General Services
Administration (GSA) as the only micro-resident Fortran compiler
that meets the full ANSI X3.9-1978 Fortran Standard without
discrepancies. RM/FORTRAN includes many frequently used
mainframe Fortran language extensions, including those used in
Digital Equipment VAX/Fortran, IBM mainframe VS/Fortran and
Fortran-66. RM/FORTRAN applications run 30% to 40% faster than
those compiled with Microsoft Fortran Version 3.2. A numerical
-------
coprocessor chip (Intel 8087 for IBM PC or XT and Intel 80287
for IBM PC AT) is required when using RM/FORTRAN or when running
RM/FORTRAN compiled numerical models.
MODIFYING AND COMPILING SOURCE CODE
The master SWMM file on micro hard disk was separated
into its component files, i.e., data files, main program and
subroutines, using a BASIC language program developed on the
target micro. This program was also used to convert several
mainframe language extensions used in SWMM, which are not
available in RM/FORTRAN, and to separate out all COMMON BLOCK
statements and related PARAMETER statements into INCLUDE files.
Use of include files often results in considerable time
savings during model testing, debugging and modification because
COMMON BLOCK array dimensions and variable lists modified in one
INCLUDE file result in global changes to the model. However,
when splitting out COMMON BLOCK statements into INCLUDE files
one must be cognizant of the fact that some programmers rename
variables within subroutine COMMON BLOCK statements instead of
formally eguivalencing these variables. Several instances of
this practice were found in SWMM. These locally renamed COMMON
BLOCK variables were respecified using appropriate EQUIVALENCE
statements.
An examination of documentation for the mainframe SWMM
model indicated that control of external data files, e.g.,
interface and input data files, is maintained through the use of
IBM Job Control Language (JCL). In contrast, file control using
8
-------
the DOS implementation of RM/FORTRAN is accomplished by Standard
Fortran OPEN and CLOSE statements within the model source code.
Accordingly, the SWMM source code was modified to include
statements necessary for creation and control of all external
data files. Specifications for these files, including: logical
unit numbers, access modes (sequential versus direct), format
(formatted versus unformatted) and status (named versus
scratch), were determined from a detailed examination of their
corresponding source code READ and WRITE statements. The
example JCL specifications included in the mainframe SWMM
documentation and user manuals were also useful for establishing
the characteristics of these files.
All SWMM subroutines were then successfully compiled
using RM/Fortran. This compilation, which was automated through
the use of a DOS batch file, required approximately two hours to
complete. The total size of the resultant object modules was
over a megabyte.
SELECTING A MICRO-RESIDENT LINKAGE EDITOR
Selection of an 80286 processor-based linkage editor to
be used on the target IBM PC AT involved several considerations.
The requirement of compiler-linker compatability did not present
a problem, since RM/FORTRAN may be used with numerous linkage
editors. However, during the initial attempt at linking the
Microsoft compiled SWMM model, it was found that the Microsoft
Linker supplied with DOS had two serious deficiencies.
First, it was found that the DOS linker can only link up
-------
to 900 kb of object modules into an executable program. Thus,
if the DOS linker were to be used, SWMM would have to be
separated into several blocks and each block would have to be
executed as a separate model. This would likely prove to be
very inconvenient during model application.
Second, it was found that the DOS linker can not be used
to build multiply-nested overlay structures. As stated
previously, four overlay nesting levels were required for
execution of the mainframe SWMM model within a 400 kb region of
RAM. Since DOS can only address 640 kb of micro RAM, the need
for multiply-nested overlays with the micro SWMM model was seen
as a distinct possibility. Accordingly, a search was initiated
for a linkage editor which would overcome the overlay nesting
and object module size restrictions of the DOS linker.
Based on the linkage editor needs identified above, the
Phoenix Software Associates PLINK86 (Version 1.48) linker was
selected for porting of the mainframe SWMM model. PLINK86 is
compatable with RM/FORTRAN, supports up to 32 overlay nesting
levels and does not have any practical restrictions on the total
size of objects modules it can link.
The mainframe SWMM overlay structure was utilized as-is,
following its conversion to PLINK86 format, to create the
numerous overlays required for execution of SWMM within the 640
kb DOS addressable RAM partition. A memory diagram of a portion
of the SWMM overlay structure is shown in Figure 1. The
vertical scale represents the relative RAM memory addresses
10
-------
RUNOFF
TRANSPORT
ERROR
QHYDRO
CTRAIN
RHYDRO
BUILD
High
Memory
Address
Time
During Execution
HCURVE
WSHED
HYDRO
Low
FILTH INITIAL
INFIL DWLOAD
TSTRDT
SLOP
PRINTR
RUNOFF. RUBLOCK
LOP r
PRINTQ
PRINTF
TINTRP
TRANS
SCALE
GRAPH
MAIN
FIGURE 1. MEMORY DIAGRAM SWMM V3 III
occupied by SWMM subroutine object code and data during model
executioii. In contrast, the horizontal scale indicates memory
addresses (and subroutines) which are overlayed at various times
during a model run. It is seen that at run time many of the
memory addresses occupied by code and data of the SWMM Runoff
Block are replaced, i.e., overlayed, by code and data of the
SWMM Transport Block. This overlaying occurs at the instant
11
-------
when the SWMM Executive Block (subroutine MAIN) encounters a
RETURN from the Runoff Block (subroutine RUNOFF) and then
initiates a CALL to the Transport Block (subroutine TRANS).
Also, while subroutine QHYDRO of the Runoff Block is executing
in RAM, code and data for all of its ancestors, including:
RHYDRO, HYDRO, RUNOFF, RUBLOCK, MAIN and all the RM/FORTRAN
library routines used by SWMM must also be in RAM.
Linking of the SWMM object modules into a relocatable
execute module required approximately five minutes. The SWMM
execute module required 1.4 megabytes of hard disk storage.
However, use of overlays resulted in a maximum 507 kb RAM
requirement at run time.
TESTING OF THE MICRO SWMM MODEL
The micro SWMM model was then tested using the EPA
supplied mainframe test cases. Results were found to be
identical to those produced on the mainframe. Test case run
times varied between approximately two minutes for execution of
the Runoff Block to almost forty minutes for execution of the
Extran Block. Results were printed at the rate of eight pages
per minute on a Hewlett-Packard Laserjet+ Printer attached to
the target IBM PC AT.
Mainframe models which utilize highly iterative solution
techniques may exhibit significant computational error or even
numerical instability when ported to micros. The reason for
this is that round-off errors can accumulate to a higher degree
when using 16-bit (typical micros) processors than when using
12
-------
full 32-bit processors (mainframes). This problem should be
assessed during model testing. If deemed necessary, double
precision may be specified for certain key variables used in
iterative procedures of the micro numerical model. Test results
for the micro SWMM model suggest that increased precision is not
necessary.
When an overlay structure is developed from scratch for
use with a ported micro model, it may be desirable to optimize
that structure in an effort to speed up model execution. A
general guideline to be followed during development and
optimization of overlay structures is to set up overlays which
contain isolated functional groups of code that will execute to
completion and then will not be returned to for a long time, if
ever. Also, the simplest overlay structure required to run
within addressable RAM is the best. Stratcom Systems, Inc.
presently markets a program which creates optimized overlay
structures based on subroutine sizes and CALL sequence
information produced by the RM/FORTRAN compiler. Use of this
porting tool is currently being investigated at Metcalf & Eddy.
Following the successful testing of the micro SWMM model,
the SWMM documentation was modified to provide users with
instructions on its hardware/software requirements and its
execution on an IBM PC XT or IBM PC AT. Since the micro SWMM
model's functionality is identical to the mainframe
implementation, no additional modifications to the EPA
documentation and user manuals were required.
13
-------
APPLICATION OF THE MICRO SWMM MODEL
The micro SWMM model was then used to simulate storm-
induced flooding within a combined sewer system in a major New
England city. Figure 2 shows the general configuration of the
modeled sewer system, which was divided into 12 sub-catchment
areas feeding into a sewer pipe system discretized into 12 sub-
systems with a total of approximately 1500 computational
elements. A schematic of the main sewer system SWMM model given
in Figure 3 shows the hydraulic connections between the various
sewer sub-systems .
Single event, 24-hour simulations were made using the
SWMM Runoff and Transport Blocks and a modified version of the
SWMM Combine Block capable of accepting up to 16 input
hydrograph files. A ten minute time step was used in all
^r:~?\
x-rqT- v:r/' M-7 r-N^ _
/,----.!-. M-4-,-! v-----iJ--'-V;v-, 7" _..... >v
/V^'c4.v*^^T-K.->.r:.\
/^g^-y.^^^-^
^;] M'12
MAIN WWTP "S
FIGURE 2. MAIN SEWER SYSTEM
14
-------
M-2
1
I
*^^
-1
I
L
M-
1
M-:
1
J
^^^
4
3
^
f-
l e
1-O
M-
1
M
^^^m
' -i
!
- !
1
1
1
1
I
1
f
H
n^
|w\
r
I_Q
O
MN
NJP
1
M-9
1 1
1 1
1 1
1
1 1
t t
M-11
{
M_4 fl
1U
I
^^^1*
» «
M-12
! In Syst
^ Normal
FIGURE 3. SWMM SUBAREA FLOW SCHEMATIC FOR THE MAIN SEWER SYSTEM
simulations and only flow quantity was modeled. Each of the
twelve sub-systems of the total sewer system was simulated
during individual SWMM runs. Batch files were constructed to
control sub-system model execution and the cascading of outflow
and in-system overflow hydrographs throughout the sewer network.
The SWMM sewer system model was then successfully
calibrated and verified using two independent rainfall and in-
system flow data sets and synoptic field observations of sewer
system depths of flow and surcharging. The verified model was
subsequently used to simulate in-system flooding, i.e.,
15
-------
surcharging and street flooding, during 10-year and 50-year
design storm events. Based on the locations and severity of
surcharging and flooding during the 10-year storm event, several
sewer relief projects were developed. These projects were then
designed and tested using the micro SWMM model.
Both the SWMM model calibration and verification input
data sets were also run on the remote IBM mainframe commercial
time share system as a cross check of the micro SWMM model and
to provide comparative run cost and speed benchmarks.
Statistics for the verification runs on both the micro and
mainframe are presented in Table 1. Model run time on the micro
was much longer than that experienced on the mainframe computer.
However, run cost on the micro was only 12% of that incurred on
the mainframe.
TABLE 1: Benchmark Run Results
IBM PC AT IBM 3033
RUN TIME RUN COST RUN TIME RUN COST
2.9 Mrs. $20. 1.1 Min. $175.
The longer run times experienced when using the micro
SWMM model were not found to represent the critical path in the
engineering analysis of the above sewer system. In addition,
use of the in-house micro for SWMM modeling resulted in almost a
90% decrease in computing costs incurred during that project.
16
-------
FUTURE DEVELOPMENTS
The porting and application of traditionally mainframe
based numerical models on micros has some limitations. Very
often model array sizes must be decreased in an effort to "fit"
the model into addressable micro RAM. For example, the maximum
number of finite element grid nodes allowed in a two or three
dimensional hydrodynamic or water quality model may have to be
decreased from 900 to 400. This may result in a micro model with
rather limited applicability to the simulation of complex
systems.
Another limitation of micro models is their relatively
slow speed of execution. In most applications, e.g., simulations
using any SWMM block except Extran, the slower execution times
on micros do not represent the critical project path. However,
in certain applications, use of micros may be prohibitively
slow. Long-term, continuous-mode simulation of a large combined
sewer system using the SWMM Extran Block is a good example.
Fortunately, there are several methods for increasing the
potential model sizes and execution speeds on micros. The
following discussion briefly reviews two of these methods as
they relate to the micro SWMM model.
SWMM makes extensive use of external scratch data files,
both within individual computational blocks and for
communication between blocks. These scratch files, which may
require over a megabyte of storage, are written to and read from
high access speed drum disks on mainframe computers. These
peripheral disks are also usually controlled by fast, dedicated
17
-------
processors. In contrast, these I/O activities are controlled by
non-dedicated processors using relatively slow access floppy and
Winchester hard disks on micros. As a result, use of hard disk
scratch files by the micro SWMM model is very inefficient and no
doubt contributed to the relatively long run times experienced
on the micro.
Data transfer within micro RAM is much faster than data
I/O using hard disks or floppies for storage. To take advantage
of this fact, several virtual (RAM) disks could be configured on
the target IBM PC AT using the DOS device driver VDISK.SYS.
Resident RAM on the micro could then be increased from the
current 640 kb to over three megabytes through the use of plug-
in RAM chips mounted on its AST advantage board and an
insertable piggy-back RAM expansion board.
Following the above reconfiguration of DOS and addition
of RAM, file OPEN and CLOSE statements within the SWMM source
code would have to be modified to include the new DOS configured
virtual disk drive specifications for all scratch files. The
SWMM executable module would then have to be rebuilt.
The above method could be used to speed up SWMM runs by
eliminating much uneccesary, relatively slow I/O activity. It
could potentially be used to increase the maximum size of sewer
systems modelable on micros using SWMM by allowing the temporary
storage of certain groups of variable arrays in virtual disk
resident external data files located outside of the 640 kb DOS
addressable RAM partition.
18
-------
The same result could be more efficiently achieved
through the use of a Unix or Unix-like operating system instead
of DOS. Xenix operating systems are currently available for IBM
PC's and many other 16-bit (and 32-bit) micros. Unix and Xenix
systems do not have the RAM addressability restrictions of DOS.
They are capable of addressing the entire amount of micro-
resident RAM. Thus, all the SWMM scratch data files could be
stored in RAM and the size of variable arrays, and hence model
sizes, could be increased. It is important to note that the
micro-resident Fortran compiler used during model compilation
would have to be compatable with the target micro-resident Unix
or Xenix operating system. Fortunately, a Xenix version of
RM/FORTRAN has recently been released. Current efforts at
Metcalf & Eddy involve the migration of the micro SWMM model and
numerous other ported mainframe models to 16-bit and 32-bit
micro-based Unix and Xenix operating systems.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
19
-------
Portability* Maintenance and FORTRAN Programing Style
Thomas 0. Barnwell, ir.
David W. Disney2
Introduction
At the Center for Water Quality Modeling, we have evolved a set of pro-
graming conventions that make our FORTRAN programs easily maintainable and
transportable. These conventions have helped us 1n transporting many of our
supported programs from minicomputers and mainframes to the microcomputer
environment. This paper describes these programing conventions and develop-
ment tools and discusses how they have aided 1n moving our programs to micro-
computers while still providing a workable environment for the maintenance of
code on different hardware and software systems.
Program Design Considerations
It 1s appropriate to discuss two aspects of programing style design
considerations and coding conventions. Design relates to the overall struc-
ture of the program while coding conventions refer to the the actual FORTRAN
Implementation. Much has been written about structured programing and soft-
ware design and we have found this design concept to be useful. The follow-
ing paragraphs outline our approach to structured programing 1n FORTRAN.
Possibly the most Important design consideration 1s modularity of struc-
ture. Modularity 1s fostered by writing small subroutines. To borrow a rule
from the writing profession, be brief and concise. Treat a subroutine as you
would a paragraph and confine it to a single thought a subroutine should
only do one thing. A good rule Is to restrict a subroutine to two pages or
less; some prefer to limit a subroutine to a single page. But a single page
may be too restrictive 1f your style uses a lot of open space to Increase
understanding. Always keep 1n mind that both you and others will have to
read (and maintain) what you write.
1. Center for Water Ooiallty Modeling
2. Computer Sciences Corp.
Environmental Research Laboratory
U S EPA, Athens, Ga 30613
20
-------
COMMON Blocks are a major part of our design. Although FORTRAN does not
have code structures like many other languages* code structures such as a
common work or data area can be emulated with COMMON Blocks. We also prefer
that data be passed between subroutines using COMMON Blocks rather than Argu-
ments* particularly at higher levels of the program. Although there are some
arguable advantages to using arguments for this purpose (after all* that 1s
their sole ralson d'etre), they can be troublesome. As discussed later, we
have found them to be a frequent source of problems.
Of course, modular structure 1s much more than writing small subroutines
and using COMMON Blocks. A clear specification of objectives and plan for
achieving them 1s at the heart of structured programing. But these princi-
ples are a good beginning.
Although there may be good reasons for not considering 1t, portability 1s
another Important design consideration. Our programs are written to be used
by others, mostly on their own computers and the programs have a surprisingly
long lifetime. Although we cannot claim credit for the original design,
QUAL2E (Brown and Barnwell, 1985) 1s based on a structure set out over 15
years ago. We use only 1977 ANSI Standard FORTRAN and avoid extensions to
the standard.
If writing single-purpose code for a specific application, the program-
mer Is justified 1n using the FORTRAN extensions available on his computer.
In general, these extensions are provided to enhance the efficiency of both
the programmer and the program. But 1f there 1s the slightest Inkling that
the program will be used on different hardware, the portability Issue must be
considered.
To Insure portability, 1t 1s useful to Identify target compilers that
represent a reasonable spectrum of computers. For example, U S EPA staff can
count on the availability of at least three kinds of computer IBM PC
compatible, DEC VAX technology, and IBM 30xx-ser1es mainframes. We run our
code through compilers on each of these systems.
Where hardware-specific code must be used, the programmer must Isolate
and clearly Identify this dependent code. For example, DATE and TIME subrou-
tine calls are a common extension that are not part of the ANSI-77 standard
but are available on many systems. Although not hardware-specific, file
opening and closing conventions may vary among Installations, even for the
same type of hardware, and it 1s good practice to Isolate these statements 1n
a separate, clearly Identified subroutine.
User Interface
The user Interface 1s becoming more and more Important to the computer
programmer. The "friendliness" of programs 1n the microcomputer environment
such as LOTUS 123* have raised user expectations so high that the traditional
batch-oriented card file Input 1s fast becoming unacceptable. In fact, the
Center for Water Quality Modeling has begun a major effort to develop a
consistent Interactive user Interface for all our programs.
21
-------
In designing a user Interface, the programmer should strive to use com-
plete English sentences for such things as error messages, status
diagnostics, or run-time prompts. Although Inside jargon and computerease
may be expedient 1n the short run, the novice user will often be confused by
such cryptic statements as "PROGRAM STOP - INPUT ERROR." Even an experienced
programmer may wonder 1f this message was generated by the operating system
or the program Itself.
An Important function of the user Interface 1s verification of all
Inputs. Although the user Interface may not be able to catch subtle errors
1n program Input, several standard checks should be made. Perhaps the most
obvious 1s checking of character versus numeric Input. How many of us have
not spent considerable time tracking down an error to find that Its cause 1s
a letter "0" where the number "0" should be» or vice-versa?
Another good practice 1s to explicitly check for real and Integer num-
bers and to check that numbers are 1n a valid range. Also, default values
should be stated or displayed explicitly. Cases exist where widely used
computer programs with perhaps Inappropriate default parameter values have
Inadvertently established "standard practice."
It 1s useful to place Input verification criteria and defaults 1n tables
or external files. This makes the criteria easy to locate and change. The
use of an external file 1s particularly convenient as changes can be made
without recompiling the program.
An echo of Inputs should be provided as an option so that the user can
document Input when desired but switch 1t off when It 1s not needed. As most
programs have the option of running either Interactively or batch, the pro-
grammer should provide program traps to stop execution when Input, errors are
uncovered rather than let execution continue.
And when writing tabular output from a program, concise but descriptive
headings should be provided and output formatted for easy reading. Nothing
1s more frustrating than 10 pages of densely packed numbers with no Identifi-
cation. Include your documentation on the output, not In the users manual.
Documentation
Documentation 1s critical to a successful computer program. We think of
documentation targeted both for the user and for the programmer. User docu-
mentation should Include a description of the application or use of the
computer program. The exact equations and solution techniques used as well
as guidance for the user on ranges and typical values of Input variables
should be provided.
The actual Installation of the program on the users' computer system,
however, 1s often done by a programmer who 1s Interested 1n quite different
things. The programmer 1s Interested In what we call "Implementation docu-
mentation" that contains detailed Instructions on how to Install and maintain
the program on a user's system.
22
-------
Both the programmer and the user are Interested 1n sample Input and
output to verify that the Installation 1s correct and the program 1s operat-
ing correctly on their system. The system or Implementation documentation
also should Include a functional organization of source code files by name as
well as descriptions of program variables (particularly global variables) In
case the user and his programmer have to troubleshoot the software. Often*
the user may want to read the program 1n order to better understand what 1t
1s doing. This should be made as easy as possible by making available copies
of the source code and Identifying development and maintenance software.
Coding Conventions
Comments are an essential part of the source code and are often the only
source code documentation. Although pseudocode 1s useful when designing a
large program* we have found 1t difficult and costly to keep current. Each
subroutine or program segment should have Introductory comments that clearly
state Its purpose; describe Inputs, outputs and modified variables; and
Identify error conditions. Comments are useful to physically separate
sections of code that do not justify isolation in a subroutine and to explain
loops and conditional statements. These are often the hardest parts of a
FORTRAN program to decipher; particularly 1f the dreaded GO TO is used with
indiscretion. (We all know the rule on 60 TO ~ DON'TIII)
Complete English statements should be used; many programmers use a
shorthand that can be completely indecipherable. Others will have to read
what you write and base the maintenance and understanding of the source code
on your commentary. When trying to decipher a program* remember that
comments are not Infallible they can lie.
Just as indentation and physical separation are used to help delineate
parts of a composition, these devices should be used to graphically explain
program logical flow and subordination. Loops should be Indented and should
end in the same column they begin. IF statements and IF-THEN-ELSE structures
also should follow a consistent Indentation format. Comments may be helpful
in separating parts of loops and conditional statements.
Indentation 1s useful when it is necessary to continue COMMON, DATA and
other type and declaration statements beyond a single line. For example,
beginning the continuation lines at, say, column 10 helps to delineate blocks
of code. And if there 1s no alternative to using a GO TO construct (remember
the old maxim rules are for privates and lieutenants; experienced program-
mers know when to break the rules), indentation is useful to help Identify
the continuations.
Modularity and structure in code can also be enhanced by placing COMMON
Blocks in separate source code files that may be brought Into other source
code files via FORTRAN'S INCLUDE statement. In addition to producing smaller
subroutines, use of INCLUDE files for COMMON blocks will Insure that COMMONS
are consistent across a program. Don't dimension or name variables differ-
ently in different subroutines. If it 1s necessary to rename variables, use
the EQUIVALENCE statement.
23
-------
It also 1s useful to Include data type specifications 1n INCLUDE files.
If possible, group different type variables (eg, CHARACTER, INTEGER and REAL)
Into separate COMMON blocks. This practice will help avoid boundary align-
ment problems and enhances the maintainability of a program. Explicit decla-
rations of INTEGER*2 and INTEGERS data type are also helpful. Most com-
puters default to the INTEGERM data type but, on microprocessors 1n parti-
cular, the INTEGER*2 calculations are considerably faster than INTEGER*4
arithmetic (Microsoft, 1985) 1n addition to saving some memory.
We have found 1t desirable to Isolate file OPEN and CLOSE statements 1n
a central, clearly Identified routine because, as mentioned above, these
statements may be Installation-dependent. It also Is good practice to use
variables for unit numbers 1n Input-output statements and assign these vari-
ables 1n an "environment" INCLUDE file, or within 1n the file-open subrou-
tine. Al ways open terminal Input-output units expl 1c1tly. Although some
Installations assume units 5 and 6 for terminal Input and output, respective-
ly, others will use unit 5 for both and still others may use units 1 and 2.
Where possible, the user should be allowed to specify file names either
at execution or at run time rather than specify the names 1n the compiled
code. This will allow easier Installation of code. In addition, the OPEN
statement error return codes should be used to recover open errors grace-
fully, giving explicit error messages using complete English statements.
We favor the use of the PARAMETER statement or a BLOCK DATA subroutine
to declare array dimensions for system-wide variables, DO loops, I/O units
and other appropriate system wide parameters. Proper use of the PARAMETER
statement will allow easy modification of program limits by modifying a
single subroutine or source code file brought Into the program with an
INCLUDE file. This feature is useful If one wishes to modify a program for
different machine memory configurations. For example, the IBM/PC version of
the QUAL2E program 1s distributed with array dimensions set for a 256K
machine. These array dimensions can be changed by modifying one line In a
single subroutine. One should be careful, however, not to use PARAMETER to
set values for variables. Remember that the program source code must be
edited, recompiled, and linked, to change any PARAMETER or a variable Initia-
lized 1n a BLOCK DATA subroutine.
In defining variables, explicit declarations of variable type should be
used rather than the default type. A useful extension on some compilers 1s
the IMPLICIT NONE statement. This statement will force explicit declaration
of all variables when writing code and can be easily removed for the distri-
bution version. Dimensioning variables 1n INTEGER and REAL Statements was
considered good practice but we have found this to be a problem with micro-
computer compilers that only allow dimensioning 1n the DIMENSION and COMMON
statements and require type specification 1n separate statements. As
mentioned above, the same name should be used everywhere. This helps avoid
later confusion when debugging the code. Prefix/Suffix-Type variable names
are also useful in helping readability. And a clear specification of Initial
value avoids problems with systems where a zero default Initial value 1s not
the convention.
24
-------
Downloading Problems
In the process of downloading our software to the microcomputer* we have
encountered several problems that relate to programing style and coding con-
ventions discussed above. In general* microcomputer FORTRAN compilers are
quite strict 1n following the ANSI standard for FORTRAN 77. Few 1f any FOR-
TRAN language extensions are available on microcomputers and practices that
are acceptable on larger computers are not acceptable with these compilers.
For example* the compilers are quite sensitive to statement order and
will not allow mixing of specification* declaration and executable state-
ments. Variables that are assigned character string values should be defined
as Character 1n a type specification. Programmers should be careful not to
use variable names longer than six characters as the more restrictive com-
pilers will not recognize differences beyond the sixth character.
A good practice 1s to separate character and numeric data Into Indivi-
dual COMMON blocks to avoid problems with boundary alignment. The same 1s
true of Integer* real and double precision variables. And one should avoid
mixing these different type variables 1n executable statements* always using
explicit type conversions 1n arithmetic statements.
Another good practice Is to Initialize variables 1n a BLOCK DATA subrou-
tine and avoid reinitializing with DATA statements. On micros* argument
types must be compatible 1n CALL and SUBROUTINE statements. Although com-
pilers on larger computers may convert between types* the microcomputer
compilers do not.
Some compilers will allow FORMAT statement field separators to be omit-
ted on micros* this will cause errors. ENCODE and DECODE statements
should not be used; they have been replaced by Internal READ and WRITE
statements 1n FORTRAN 77.
Development Tools
There are a number of FORTRAN compilers available for personal com-
puters. Two recent reviews 1n PC Tech Journal (Howard* 1985) and Computer
Language (Bensor et al.* 1986) discussed nine different FORTRAN compilers
(Table 1) for the MS-DOS operating system and the Macintosh.
Several letters responding to the Howard's review are contained In the
March 1986 Issue and provide additional Insight to the compilers. Of parti-
cular Interest 1s an editorial 1n the March 1986 PC Tech Journal (Fastle*
1986) that observes "the volume of comment on our coverage Indicates consid-
erably larger Interest 1n FORTRAN than expected ... confirming that FORTRAN
1s alive and well. ... One particularly surprising comment 1s about FORTRAN'S
portability. Because 1t 1s not a systems programmers language* programmers
are not as likely to exploit extensions supplied by a particular vendor*
opting Instead for a more standard approach. This portability extends beyond
the desktop to the minicomputers and mainframes crunching FORTRAN."
25
-------
TABLE 1. REVIEWS OF PC FORTRAN COMPILERS
Compiler PC Tech Computer
Journal Language
Oct. 85 Jan. 86
IBM Professional FORTRAN X X
(a.k.a. Ryan-McFarland)
Lahey X X
Microsoft X X
Digital Research X X
Intel FORTRAN-86 X
Prospero X
ProFORTRAN
Supersoft X
WATFOR-77 X
Microsoft X
MACFORTRAN
The Computer Language article 1s particularly Interesting because 1t
also discusses a number of add-on tools that provide graphics, string manipu-
lation, windows, scientific subroutines, and productivity enhancements.
Our experience 1s limited to the Microsoft and IBM Professional
(a.k.a. Ryan-McFarland") FORTRAN (ProFORT) compilers. Microsoft's compiler
1s flexible 1n that 1t can produce executable code that executes on PCs
without the 8087 Math Coprocessor chip, whereas ProFORT requires the co-
processor. However, Microsoft's FORTRAN syntax* particularly for INCLUDE
statements, and use of Metacommands for compiler options differ from conven-
tional usage.
We have found that ProFORT 1s compatible with our primary development
environment, DEC VAX FORTRAN. ProFORT 1s a full Implementation of FORTRAN 77
with few, 1f any, extensions and 1s often 100% compatible with other FORTRAN
development environments. This compatibility has been a key to our ability
to develop and maintain software targeted for a wide variety of computer
systems.
26
-------
The key to Installing large FORTRAN programs on the PC lies 1n the link-
age editor. The LINK command provided with both the Microsoft and ProFORT
compilers will allow only one level of overlay and will only allow overlay of
code segments. In trying to link a large program (25*000 lines of code) with
these linkage editors, we received a message Indicating that code segment
DGROUP exceeded 64K (Parker* 1986, discusses this problem). Phoenix Soft-
ware's PLINK86 linkage editor allowed us to bypass this problem by allowing
multiple overlay levels (up to 32 levels of overlay may be used). A complex
overlay structure may be required by large program size and proper software
design can make for a much more efficient overlay structure.
Summary
We have reviewed programing design and coding conventions that have
served us well 1n maintaining and transporting out FORTRAN programs among a
variety of computers. The development tools discussed above make 1t possible
to Install quite large FORTRAN programs on microcomputers and our experience
1s showing that the microcomputer can be an attractive environment for
running these programs. When execution times are excessive, the programs can
be easily moved to larger machines to take advantage of their faster speed.
References
Bensor, R. et a!., "Software Review: FORTRAN on the MICRO", Computer. Lan-
3iLfl£g> Computer Language Publishing Ltd., San Francisco, CA ,Jan 86, pp83-
110.
Brown L. C. and T. 0. Barnwell, Computer Program Documentation for
Enhanced Stream Water Quality Model QUAL2E. EPA/600/3-85/065, U S EPA,
Athens, GA, 30613, Aug 1985.
Fastle, W., "Language Surprises," PC Tech Journal, Z1ff-Dav1s, NY, Mar 1986,
p!2.
Howard, A., "FORTRAN Options," PC Tech Journal. 21ff-Dav1s, NY, Oct 1985,
pp!49-160.
Parker, P., 1n "Letters", PC Tech Journal. Z1ff-Dav1s, NY, Mar 1986, p22.
. _ , IBM Personal Computer Professipna.1 FORTRAN, by Ryan-McFarl and Corp.,
IBM Corp., NY 1985.
_ , Microsoft FORTRAN Reference Majiual, Microsoft Corp., Bellevue, WA.,
98009
The work described In this paper was funded by the U.S. Environmental Pro-
tection Agency and has been subject to the Agency's peer and administrative
review. It has been approved for publication as an EPA document.
27
-------
CONSIDERATIONS ON USE OF MICROCOMPUTER MODELS
FOR STORMWATER MANAGEMENT
"by: Wisner, P.
Professor, Dept. of Civil Engineering
University of Ottawa, Ottawa, Canada, KIN 9B4
Consuegra, D.
Research Assistant, Dept. of Civil Engineering
University of Ottawa, Ottawa, Canada, KIN 9B4
Frazer, H.
Gumming - Cockburn Assoc., 1735 Courtwood Crescent,
Ottawa, Ontario, K2C 2B4
Lam, A.
Andrew Brodie Assoc., 3?1 Gilmour, Ottawa, Canada,
K2P OR2
ABSTRACT
Widespread use of microcomputer hydrologic models raises the question
of their optimum design. Microcomputer models developed in North America
and Europe can "be classified into two categories:
1. Models with explicit incorporation of many parameters which can be
used as default values or modified for calibration purposes if data
is available.
2. Models reducing drastically the number of parameters and incorpora-
ting default values in the program without giving the user a
possibility for a change. In this category rainfall distributions
are also built in.
The paper gives examples of possible sources of errors associated with
the second approach. If relatively sophisticated techniques are used to
simulate rainfall runoff processes, these methods can be incorporated in a
microcomputer user friendly model and still provide enough flexibility to
allow non-modellers to modify default parameters in cooperation with a
specialist. This also provides a better understanding of hydrologic
principles. The paper also describes a Canadian package of microcomputer
models based on this principle and some implementation aspects. A practical
application is also described. The main runoff simulation techniques
presented in this paper give consistant results with detailed SWNM for urban
areas and with appropriate selection of parameters results can be compatible
with SCS type simulations for rural areas.
28
-------
1. INTRODUCTION
In the early 70's the profession started to consider urban hydrologic
models as a desirable replacement for empirical methods for runoff control,
such as the Extended Rational Method. The use of models was advocated
mainly because they can be tested against rainfall-distributions, real or
synthetic. This objective can be achieved using various principles of
rainfall-runoff transformation such as kinematic routing, parallel non-
linear reservoirs, single quasi linear reservoirs, etc. A discussion of the
advantages or disadvantages of each type of model and mainly of "routing
versus convolution", is beyond the objective of this paper. What all these
models have in common is that the user can modify several parameters for
calibration and verification of applicability for specific conditions. This
also applies for rainfall loss models such as Green-Ampt, Horton, Holtan,
etc.
Flexibility in selection of parameters or use of various storm distri-
butions is an obvious advantage for a hydrologist, but may be confusing for
a practising drainage engineer. Calibration is not always possible because
of lack of data. Under these conditions two approaches are used in practi-
cal applications:
A. Models In which a relatively large number of parameters can be used
either as default values or modified for calibration purposes.
B. Simplified models with default values incorporated in the program and
very few input parameters.
Examples of packages of microcomputer programs in the second category
are those based on the SCS Curve Number method with an empirical relation
for the initial losses (ie: la = 0.2S). Runoff is computed by means of the
SCS curvilinear unit hydrograph with empirical relations for the response
time incorporated in the program.
These models can be even simpler if the rainfall pattern is already
built in (Debo, 1985). Attempts of simplification are also made in a more
sophisticated model used in France (Chocat, 1984). Although it has interes-
ting features, runoff contribution from the pervious area in urban water-
sheds is neglected, which is not applicable for North American conditions.
Although the response of urban watersheds has a higher degree of non-
linearity as compared to rural areas, microcomputer simplified models based
on SCS techniques, Santa-Barbara method, etc., assume linearity in both
cases.
In most recent microcomputer developments a significant effort is made
only towards user friendly structures, attractive output displays and exten-
sive graphical capabilities. Although these are important features they
will not be discussed in this paper, which is mainly intended to advocate
the advantages of A (above). This approach is possible without reducing
the ease of use.
29
-------
2. SIMPLICITY versus ACCURACY
The pitfalls of oversimplifications can be illustrated by Canadian ex-
perience with the SCS methodology (SCS, 1971).
In the SCS procedures, relation (1) in figure 1 between total rainfall
and runoff, has only two parameters; soil storage and initial abstraction,
la. The initial abstraction is related to the soil storage by the empirical
relation la = 0.2S.
The SCS unit hydrograph is automatically defined by the response time
which is related to the time of concentration by an empirical relation.
By means of these simplifications, the practising engineer has to
specify the same number of parameters as in the Rational Method, namely;
1. The soil storage or the CN number which is equivalent to a runoff
coefficient.
2. The time of concentration.
Comparisons of simulations and measurements conducted by the IMPSWM
(Implementation of Storm Water Management) Program at the University of
Ottawa (Wisner et. al., 1984) have shown that the cascade of linear reser-
voirs unit hydrograph proposed by Nash (Nash, 1957),has an adequate perfor-
mance for small rural watersheds. This unit hydrograph is given by a gamma
function defined with two parameters: n, the number of reservoirs or shape
factor and t , the time to peak (relation(2), figure 1). fly proper selection
of n and t Several shapes of unit hydrographs can be reproduced. For ins-
tance the shape of the SCS unit hydrograph is a particular case of the gamma
function for n = 4.7J5.
Table 1 shows the values of n determined after calibration of the Nash
model for 6 rural areas (Wisner et. al., 1984). These results were obtained
using a modified SCS curve number method in which the initial abstraction is
an input (Ia<>0.2S). It can be seen that the values of the shape factor n
are less than 4.75, the value corresponding to the SCS unit hydrograph. For
similar conditions a default value of n = 3 seems to be more adequate.
The initial abstraction was also found to vary from 1 to 4 mm; this is
less, by orders of magnitude than the value given by the default value,
la = 0.2S. It was also verified that for some of these watersheds the use
of more sophisticated infiltration methods such as Green-Ampt or Holtan did
not significantly improve the results (Consuegra et. al., 1984).
The Nash model was also compared with a more sophisticated variable
unit hydrograph (VUH, Ding, 1974). Results also showed that after calibra-
tion the results with the Nash model and the non linear model were not very
different. Based on the satisfactory performance of the Nash model and of
the modified Curve Number procedure, it was decided to incorporate this
30
-------
Table 1. Comparison of calibrated values of n and tp for 6 rural watersheds with the Nash Model.
Comparison of calibrated tp and tp computed using Williams equation.
Watershed
Wixon Creek
Holiday Creek
North Morrow
Clement Menard
Collins Creek
Parimbot
Location
Ontario
Ontario
Ontario
Ontario
Ontario
Switzerland
Area ha
1016
3053
382
96
15500
380
Slope %
1.2
0.7
1.0
0.1
0.15
2.0
Tp Calibrated
hrs
3.7
5.7
6.0
2.0
69.0
1.7
Tp Williams
hrs
1.1*2
2.82
3.31
1.93
11.8
0.61
n
2.6
2.0
2.0
1.85
3-0
3.0
Williams formula: Tp = 6.5^ A°'39 S
Tp = hours
A = surface sq. m.
S = slope ft/mile
-------
technique in the microcomputer programs described in the next sections of
this paper. The resulting submodel is called NASHYD.
The SCS procedures initially developed for rural areas were later
recommended for urban areas (SCS TR55). IMPSWM studies for urban water-
sheds were conducted by first comparing, for a hypothetical watershed,
various linear and quasi linear models with EPA.SWMM (Huber et. al., 1982)
simulations under design conditions. It was found that the performance of
some single linear or quasi linear reservoirs could give results compatible
with EPA.SWMM for a limited range of conditions (high imperviousness, etc.).
Very good consistency with SWMM was obtained with a two parallel quasi
linear reservoir conceptual model in which the storage coefficient is a
function of the time of equilibrium (P'ng, 1982). The resulting URBHYD
model based on this principle uses relation (3) in figure 1, as proposed by
Pedersen et. al., (1980). URBHYD was also compared with measurements on 4
urban catchments with good results as shown in table 2. (Wisner and
Consuegra, 1986). Flows shown in table 2 were obtained with SWMM default
values. Later, testing studies on European watersheds showed the need for
slight modifications of default parameters specially for low rainfall
volumes.
The selection of meteorological input is also a controversial issue.
It is useful to compare results from design storms and series of historical
storms. Studies conducted by the IMPSWM program (Wisner and Fraser, 1982)
compared Chicago (Keyfer and Chu, 1957) and SCS (type 2, 24 hours) design
storms with a series of real intense storms in the Metro-Toronto area. Peak
flows were computed with URBHYD and NASHYD for urban and rural areas respec-
tively. For rural areas the Curve Number (CN) was obtained from a correla-
tion with API (Antecedent Precipitation Index).
For urban areas it was found that Chicago design storms generated peak
flows closer to observed values than the SCS storms (fig. 2).
For rural areas, flows are strongly dependent on moisture conditions
and infiltration losses. With an appropriate selection of CN, the SCS
storms performed better than the Chicago design storms. For urban areas and
under design conditions, peak flows obtained with SWMM and URBHYD were found
to be very sensitive to the maximum rainfall intensity. For the Metro
Toronto region, it was found that a time step of 10 minutes for the Chicago
design storm gives realistic peak flows.
The above examples show that microcomputer models used for comparison
of pre and post-development conditions can not be excessively simple. If
simplifications are introduced they should be tested with measurements for a
variety of conditions and their limitations clearly defined. As an example,
the default value la = 0.2S may be acceptable for high rainfall volumes. A
single quasi linear reservoir can give good results for high degrees of ira-
perviousness. Simplified traditional methods can be applied mainly for
small areas. Simplified microcomputer models should also give the user the
possibility to compare synthetic with real storms.
32
-------
HTOROLOCICAL
SUBMODEL
URBKYD
Urban Runoff
METHOD
Borton equation for
losses
Two parallel single
linear reservoirs
for pervious and
Impervious areas
EQUATIONS
f - Infiltration rate at tine t
f Initial Infiltration rate
o
{ - final Infiltration rate
k - dekay factor
For both reservoirs, Storage coefficient K
K a.
T0.6 0.6
L U
0-* °-3
O)
DEFAULT VALUES
f - 76.2 urn hi
o
f 13.2 mm ht
k - 4.14
pervious - 0.2S
a conversion factor L - length of vatershed
n manning surface roughness
1 maximum rainfall Intensity
S slope of the watershed
impervious - 0.013
L (1.5 Area)0'5
00
KASHYD
Urban Runoff
Modified SCS
technique
Nash Cascade of
linear Identical
reservoirs
(1)
Q » runoff (mm) P > precipitation (nun)
5 soil water storage
la" initial abstraction (mm)
25400 ..
C 254+S U;
CN= modified curve number related to Antecedent
Precipitation Index (API)
Unit hydrograph defined by two parameters,
n shape parameter
T " time to peak of unit hydtograph
,- qp
Ia<0.2.E.Ia - 3 ma
n 3
Fig. 1 Description of NASHYD and URBHYD
qp- unit hydrograph peak
-------
Table 2. Comparison of observed and simulated peak flows
with
Storm date
05-06-63
10-06-63
14-06-63
20-06-63
01-08-63
14-08-63
02-08-66
21-01-78
16-12-83
30-12-83
01-01-84
22-09-73
23-09-73
31-05-74
21-06-74
13-02-79
22-05-79
18-02-80
lumped OTTHYMO
Observed
(cms)
2.29
2.24
0.87
0.84
2.49
0.98
0.86
2.97
4.62
2.54
4.43
0.92
0.71
0.90
0.58
1.20
1.07
1.59
and detailed SWMM for 4 urban
Gray Haven
peak Simulated peak
OTTHYMO
(cms)
2.05
2.05
0.89
0.90
1.86
1.19
0.93
Hillbrow
3.41
4.53
2.46
4.41
Malvern
0.74
0.46
0.85
0.56
Pine town
1.22
0.74
1.35
catchments
Simulated peak
SWMM
(cms)
2.21
2.21
0.89
0.97
1.79
1.09
0.98
2.78
4.62
2.18
5.38
0.98
0.63
1.11
0.73
1.11
0.72
1.34
34
-------
RURAL WATERSHED
URBAN WATERSHED
100
D L«|n*ra*l Distribution
r«m hlicocic*! «v«nt«)
SO
40
O SCS Da«l|B Se*m Mows
& Cble*|o D««lfa Scon Flew*
501
- JJX
13
J.
3 10 25 100
HTUWf PEHIOD (y«»rp)
10
VI
PUtting fotUUn {hliterlc »toi«
flew*)
O U|n*iMl DUtrlfciitUn
(Mit«ric t*r« (Iowa)
A ChlcM* D««l«n Ctera CM*»5J
O ICS D««l|n lt*ra CH'»55
10
15
100
1000
CTUUI
l»ri.)
Figure 2. Comparison of design storms and a. series of real storms for
Metro-Toronto conditions
-------
3. USER FRIENDLINESS versus FLEXIBILITY
An example of a simple procedure for the incorporation of flexibility
in a user friendly program is the FASTHYMO model. FASTHYMO was developed by
a consulting firm participating in the IMPSWM program (Brodie et. al.,1984)
FASTHYMO can be used for simplified lumped watershed simulation with URBHYD
and NASHYD, using very simple inputs (figs. 3 and 4). The user of FASTHYMO
works with a series of design storms, given by municipal criteria or with
real storms. Once these storms are entered, they can be permanantly stored
on disk. The user can first check list of storms available on disk. Prior
to simulations, users can review all default values stored on disk and
eventually modify and store a new set of parameters on disk. Figure 5
displays a typical parameter file. As an example for rural areas a default
value of n = 3 and a value of la = 1.5 mm were recommended for Metro Toronto
conditions. The user can also specify n = 4.75 and la = 0.2S to ensure
complete compatibility with the SCS procedures.
However, the user must specify the time to peak of the Nash unit
hydrograph (fig. 3). No attempt was made to include one of the available
formulas for t . It was felt that this may limit the choices of the user
and promote a given empirical formula without testing it. Table 1 shows a
comparison between calibrated values of tp and computed ones using the well
known Williams formula (Willians and Hann, 1973) for the same rural areas.
It is evident that differences may be significant. Similar discrepancies
may be found using other relations. An analysis of travel time (overland
and channel), preferably by hand computation, is recommended. Providing t
as an input also allows for sensitivity analysis and a better understanding
of the model. The rest of the input data are very easy to determine
(fig. 3).
Figure 5 also shows a set of default parameters used to compute runoff
from urban areas. The meaning of these parameters was already presented in
fig. 1. A default relation between length and area is also provided. For
maximum flexibility, the user can also specify particular values for the
storage coefficient and use the model as a linear one. If a municipality
intends to standardize parameters it can use values from calibration studies
or those recommended by a specialist.
For a complete watershed analysis including detention reservoirs
channel and conduit routing, users require the more advanced models descri-
bed further on. The same principles were maintained for parameters and
design storms. PLANHYMO (Brodie and et. al.,1984) is a simple Interactive
microcomputer watershed model for simplified watershed simulation. The
input for each watershed is similar to that of FASTHYMO. At each junction
the user specifies all the information related to the different sub-areas.
He also indicates the type of routing (if any) to the next downstream junc-
tion. PLANHYMO has the following capabilities:
1. Determination of runoff hydrographs from contributing urban and
rural areas at each junction with URBHYD and NASHYD, respectively.
36
-------
FASTHYMO.. NASHYD
NOTE : PRESS 'RETURN' TO SO BACK. TO MENU
PLEASE ENTER :
FASTHYMO.. URBHYD
NOTE : PRESS 'RETURN' TO 60 BACK TO MENU
PLEASE ENTER :
U>
1) UNITS,
i) TIME TO PEAK , HOURS
7> OUTPUT ON PRINTER , Y/N
3J PRINT FINAL HYDROGRAPH, Y/N
1) UNITS, OR , M/I
2) GET STORM FROM DRIVE ?
-) NUMBER OF 'STORM' TO USE ,
4) DRAINAGE AREA (HECTARES),
5) IMPERVIOUSNESS RATIO,
6) SLOPE IN X,
7> WANT OUTPUT ON PRINTER , Y/N
S) WANT FINAL HYD. PRINTED , V.-M
Figure 3. Window display for NASHYD in FASTHYMO
Figure A. Window display for URBHYD in FASTHYMO
-------
< FASTHYMOi PARAMETERS >
Fo,
Fc ,
DECAY value
ACCUM. inf.
DEP.STO. IMP.
DEP.STO.PERV
STO.COEF. IMP.
STO.CQEF.PERV
MANN. 'n' IMP.
MANN. 'n' PERV
in, mm =
in, mm =
=
in, mm «
in, mm =
i n , mm «
in, fliffi **
=
=
0
3. OOO
0.520
4. 140
O.OOO
O.O62
0.104
0.000
O.OOO
O.013
O.250
76.200
13.208
4.14O
0.000
1.575
4.674
O.OOO
O.OOO
0.013
0.250
IN. ABS. in,mm = O.O59
# of NASH RES. - 3. OOO
1.499
3,000
> WANT TO CHANGE SOME VALUES, Y/N:
Figure 5. Parameter file in FASTHYMO
2. Addition of these hydrographs with other input hydrographs even-
tually determined from other computations.
3. Routing of hydrographs through reservoirs, channels and pipes.
4. Storage or printing of summary results.
Figure 6 shows a typical input display of PLANHYMO.
More recently, LOTHYMO (Consuegra et. al.» 1986) was developed as an
extension of PLANHYMO. The structure of LOTHYMO has been developed to meet
the following criteria: fully interactive input and output processes and
secondly, efficient executional performance. The first objective was
achieved by interfacing PLANHYMO with the powerful spread-sheet, LOTUS
1-2-3. Using ready made templates, input files can be easily created. The
systematic tabulation of all input parameters permits rapid entry, quick
review and .therefore, quick detection of any inconsistencies. In the out-
put stage, results can be quickly retrieved by LOTUS 1-2-3 for their review,
interpretation and management. Graphics can also be generated automatical-
ly. The second objective was satisfied by the development of a
driver program written in BASIC, which reads the input files created with
LOTUS and calls all the appropriate hydrological sub-routines for the entire
drainage network. At this stage absolutely no user interaction is required.
-------
PLANHYMO
- Press 'RETURN* after your INPUT
NOTE: Maximum No. of hydrographs @ one point is 4.
Please enter the following* Simply hit 'RETURN' to go back to MENU
*» ENTER the JUNCTION POINT NO. , =
Any ROUTED HYDROGRAPHS at this junction, (Y/N) =
1> How many UPSTREAM HYDR06RAPH to be added, =
2) How many RURAL WATERSHEDS <0- >, de-f«=o =
3) How many URBAN WATERSHEDS <0- ), def«O =
4) How many INPUT HYDROGRAPH <0- >, def=O =
5) Want the FINAL HYDROGRAPH to be ROUTED, (Y/N) =
Figure 6. Window display for PLANHYMO
Figure 7 shows the Highgate Creek (Ontario) watershed where LOTHYMO was
applied. The total area is 540 ha. The downstream area is urbanized and
drained by an open channel with a capacity of 16.4 cms. In the near future,
the entire watershed will be urbanized. If no runoff control measure is
undertaken, the 100 year post-development flow will raise up to 42.6 cms.
Three detention reservoirs are proposed to reduce the post-development flow.
Schematization for the post-development conditions is also shown in figure
7. Figures 8 and 9 show the output obtained with LOTHYMO and the
corresponding plots generated with LOTUS.
4. EXPERIENCE WITH THE MODELS
FASTHYMO and PLANHYMO are presently used on IBM PC and Apple He for
quick project verification by organizations such as the Department of
Engineering of the Town of Markhara, the Department of Public Works of the
City of Scarborough, the Project Review and Approval Group of the Ontario
Ministry of Environment, the Water Resources Service of the Metro Toronto
Conservation Authority, etc.
The following are examples of projects using the flexible microcomputer
models presented in this paper, illustrating the range of applications by
consultants:
i) Preliminary Master Drainage Plans. A consulting firm in Montreal
applied PLANHYMO for an area including two dry ponds and three wet ponds.
The application was done during a two day training session on the job with
an engineer without previous computer experience.
Safety study for a SWM reservoir. A consulting firm in Illinois used
PLANHYMO for the analysis of a reservoir during the maximum possible storm.
39
-------
urban area
R rural area
junction number
junction
Dummy junction
no effect
Figure 7. Highgate Creek watershed and post-development schematization
ill) Multiple on-site detention reservoirs for an Industrial pagk. A con-
sulting firm in Ottawa designed a drainage system with three storages for
large parking lots and a separate storage for the minor system with
PLANHYMO. The objective of this study was to drastically reduce peak flows
due to a lack of downstream capacity.
Drainage engineers involved in the review of projects presented above,
did not have previous experience with hydrologic modelling and initially
needed indications regarding the choice of default values. Two years of ex-
perience with these microcomputer models show that practising engineers can
easily modify the described microcomputer models to suit their own needs.
Some of them perform sensitivity analyses on their own initiative. Storms
have also been compared, and this led to useful discussions.
5. FINAL COMMENTS
As indicated in the Introduction, there are several models for rainfall-
runoff transformation, infiltration losses and design storms. While their
advantages and disadvantages are still discussed by hydrologists, there is
some agreement, that with proper parameters reasonable results can be
40
-------
UATERSHED
Rural
Urban
Urban
TOTAL"
ROUTE TO 2
WATEWSKB
U/S JN Ho 1
TOTAL"
to 3
UAT£R«ME1>
Urban
U/S JM No 2
TOTAL-
RffOUTC TO 4
UATERSrCD
U/S JN No 3
TOTAL"
to 3
DV P WIHY uvir
BullAH'T OUTPUT FOR
AREA IHP CN*
na. X
47.21 «" 41
40.93 27 *
37.30 70 a*
143.14 »»
149.24 ~
BUnTtARV OUTPUT fly*
AJtCA IHP CM*
ha. X
149.24 *
149.24 *4
149.24 4»
SUrWARV OUTFVT FOR
MtCA f^^^ CMv
ha. x
213.40 30
149.24 *»
371.44 a. a*
370. 44 «
ur*iAftv OUTPUT FOR
ARCA IHP CN»
na. X
370. *4 »
370.44 ^»
370.44 *
^f i rut r^E
JUNCTION
8
X
1.2
1.3
1.4
«
~
JUNCTION
S
X
*«
«
««
JUNCTION
9
t
I.I
*
«
JUNCTION
B
X
«
a
. l««wl«t
POINT NO.
PCAK
cm
U3»
».24
14.27
23.09
4.43
POINT NO.
PCAK
eaa
4.49
4.43
4.44
POINT NO.
ffftf
CM
27.42
4.44
20.»2
10.30
POINT NO.
cam
10.30
10.30
10.30
1
TPEAK
13. OO
11. BO
11.00
11.00
12.00
2
TPCAK
12.00
12.00
13.16
3
rr
11.00
13.14
11. BO
12.00
4
TPEAH
nrm
12.00
12.00
13.24
RUNOFF VOL
42.74
31.43
04.49
40.53
40.92
RUNOFF VOL
40.92
40.92
40.90
UNQpa VOL
93.93
40.90
94.00
94.00
RUNOFF VOL
Ml
94.flO
94.00
34.00
WATERSHED
Urban
Urban
U/B JN No 4
TOTAL"
HATER9HCO
U/8 JN No 9
TOTAL-
UATER6HEO
Urban
Urban
U/S JN HO 4
TOTAL-
RftDUTE TO 0
-ATEITBMCB-
U/6 JN No 7
TOTAL"
SUHIIAI1Y OUTPUT FOR
AREA IMP CN.
na. X
26. 7O 3O *
44.00 30
378.64 * 4«
44*. 34 a*
SUmARV OUTPUT FDR
AREA IHP C»«
444.34 »a »
44*. 34 »
SunnARV OUTPUT FOR
AREA ln» CN*
na. X
34. BO 30
3». 70 30 »
44*.34 w
923. 04 »
323.04 » .
BUrHARV OUTPUT FOR
AKEA TMP en*
ha. t
923.84 »« »
323.04
JUNCTION
fl
X
o.a
0.0
a«
-
JUNCTION
s
X
«
JUNCTION
X
0.0
0.0
«
JUNCTION
s
X
»»
POINT NO.
PEAK
4.44
4.04
10.30
13.08
POINT NO.
PEAK
cm
19.08
19. 0B
POINT NO.
PEAK
cam
3.98
4.2*
13.08
S7.72
19.07
POINT NO.
PEAK1
cm
13.07
I9.O7
3
TPCAK
nrm
11.00
11.80
13.24
11.00
4
TPCAK
11.00
11.00
7
TPCAK
hra
11.00
11.00
11.00
11.00
12. 4O
B
T^CAK
nrm
12. 4O
12. 4O
RUNOFF VOL
93.94
93. V4
34.00
94.34
RUNOFF VOL
94.34
94.34
RUNOFF VOL
a*
93.*4
34.34
34.00
94.00
-OlBMCyE- Hf"f
tnttn^r VDL
M
9*. 00
96. OO
Figure 8. Output generated with LOTHYMO for the Highgate Creek watershed
-------
a)
Ul
0
S.S
O
9
11
19
Time (hrs)
Figure 9. Plots of inflow and routed hydrograph generated with LOTUS
at junction 3 (a) and at junction 7 (b)
9
Time (hrs;
42
-------
obtained with several types of models. A key issue in Storm Water Manage-
mentis_ whether the rapidly developing microcomputer software for prac-
tising drainage engineers should or should not promote a better understan-
ding of basic hydrologic principles.
It is not yet known which of these options will prevail. The two years of
experience with some of the models discussed above, proves, that at least
the first approach is possible.
Many agencies require consultants to use standardized parameters, a
given unit hydrograph and a specific design storm. Experience with the
models proves that even in these situations there is some advantage in com-
paring results obtained with standardized parameters and other input
parameters*
Information on all models based on the IMPSWM research is available from
The IMPSWM group, Department of Civil Engineering, University of Ottawa,
Ottawa, Canada, telephone (613) 564-3911.
REFERENCES
1. Brodie, A., Lam, A.t Rampersand, C. Implementation of Microcomputers
for SWM Studies in Canadian Municipalities. Proceedings of the
Conference Stormwater and Water. Quality Management Modelling.
Burlington, 198^.
2. Chocat, B. Conception, evaluation et dessin des reseaux d'egouts
(CEDRE). Proceedings of the Third International Conference on Urban
Storm Drainage. Vol. 2. Gotebttrg, Sweden. 1984.
3. Consuegra, D,, Jaton, J.F., Wisner, P. Etude comparative de diffe-
rentes fonctions d*infiltration. Rapport Ecole Polytechnique Federal e
de Lausanne. Institut de Genie rural (IGR). Lausanne, Suisse. 1985.
4, Debo, T. Personal Computers and Stormwater Management Programs.
Proceedings of the Conference on Computer Application in Water
Resources. Buffalo. 1985.
5. Ding, J. Variable Unit Hydrograph. Journal of Hydrology. Vol. 22.
1974. pp. 53-69.
6. Huber, W.C., Heaney, J.P., Nix, J.J., et al. Stormwater Management
Model Users Manual. Version III. University of Florida, Gainsville,
Florida. 1982.
7. Keifer, C.J., chu, H.H. Synthetic Storm Pattern for Drainage Design.
Journal of the Hydraulics Division. Proceedings of ASCE. Vol. 83.
No. HY4. August 1957.
8. Nash, J.E. The Form of the Instantaneous Unit Hydrograph. Inter-
national Association of Scientific Hydrology. Publication 14-5. 1957.
43
-------
9. Pedersen, J.T., Peters, J.C., Helweg, O.J. Hydrographs by Single
Linear Reservoir Model. Journal of the Hydraulics Division ASCE.
Vol. 106. No. H5. Proc Paper 15^30. 1980. pp. 837-852.
10. P'ng, C. Conceptual Hydrologic Modelling for Master Plans in Urban
Drainage. MaSc Thesis. Dept. Civil Engineering, University of Ottawa,
Ottawa, Ontario, Canada. 1982.
11. Soil Conservation Service. National Engineering Handbook, Section 4,
Hydrology. United States Dept. of Agriculture. U.S. Government
Printing Office, Washington, D.C. 1971.
12. Williams, J.R., Haan, R.W. HYMO A Problem-Oriented Computer Language
for Hydrologic Modelling. ARS-5-9, U.S. Dept. of Agriculture. 1973-
13. Wisner, P., Frazer, H. Design Storms for Stormwater Management Studies.
Part V. IMPSWM Urban Drainage Modelling Procedures. 2nd Edition.
Dept. of Civil Engineering, U. of Ottawa, Ottawa, Ontario, 1983-
14. Wisner, P., Frazer, H., P'ng, C.E., Consuegra, D. An Investigation of
the Runoff Components in the HYMO, OTTHYMO and VUH Models. Heport to
the Ministry of Natural Resources, Ontario, Canada. 1983-
15. Wisner, P., Consuegra, D. Testing of OTTHYMO on Twenty Watersheds.
Proceedings of the International Conference on Storm Water Modelling.
Belgrade, Yugoslavia. 1986.
16. Wisner, P., Consuegra, D., Morse,B. Utilisation de Lotus 1-2-3 pour
1'amelioration de la gestion des donnees d'un modele hydrologique.
First Canadian Conference on Computer Applications in Civil Engineering/
Micro-Computer. McMaster University. 1986.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
44
-------
PC SOFTWARE FOR COMPUTATIONAL HYDROLOGY
by: W. James, M. Robinson and M. Stirrup
Computational Hydraulics Group
McMaster University
Hamilton, Ontario, L8S 4L7
Phone: 416/527-6944
ABSTRACT
The Computational Hydraulics Group at McMaster University has been
developing and using integrated software for personal computer systems for
the past 8 years. The software encompasses data collection, data transfer.
data base management, hydrologic modelling (storms, stormwater, water
quality, hydraulics, receiving waters) and real-time control. Most of the
programs are written in FORTRAN.
This paper describes the relationship between these activities and the
sequence of data processing necessary for various types of study. About 50
programs are listed and briefly described - most are geared to the acquisi-
tion of continuous high-resolution rainfall data, subsequent data base
management, continuous stormwater modelling and real-time control.
Performance of the programs in an IBM-PC compatible hardware environ-
ment is indicated in broad terms; integration of this range of hydrologic
software is now being completed.
INTRODUCTION
Long-term flow and pollutant records may be synthetically generated
using an elaborate, deterministic computer model such as the U.S. Environ-
mental Protection Agency's Stormwater Management Model (SWMM){1). Version
3 of SWMM has been adapted by our group to run on IBM-PC compatible
machines, so that a complete, very large and intricate model can now be run
on a computer costing less than $1500 US (2). There are no serious draw-
backs using PCSVMMJ in this environment. Ve have supplied this package to
a significant number of users, and the existing large SWMM user group
infrastructure assures wide-ranging, continuing support and interest in
this product. The package is being used world-wide.
PCSWMM3 has many advanced capabilities not available in other models
in use today, e.g. menu-driven design dialogue, input error checking,
diagnostic messages, transparent file handling, verification/validation
45
-------
External
data sources
Data decoding
data transfer
Time-series
management
Local
keyboard data
File
management
Input data
files
4-
Program
library
V
Storm models
Stormwater models
Flood plain models
Receiving waters
Figure 1. Hydrologic computing environment.
modules, continuous modelling, water quality, diversion structures, sedi-
mentation and scour, costing, screen-oriented graphics, statistical sum-
maries, to name a few. Nevertheless the package is easy to use, especially
because of the user-friendly documentation, simple menu-driven input and
graphical output. Probably the most important attribute is continuous
modelling, since this removes all the intrinsic shortcomings of the design
storm approach - all flows are ranked directly, giving correct probabili-
ties and risks (5). Difficulties due to the management of large amounts of
input and output data are eliminated by our special data base management
software (CHGTSM), also specially written for IBM-PC compatible environ-
ments (4). The general computing environment is shown in Figure 1.
One question that is seldom meticulously addressed is model valida-
tion. We have devised a methodology incorporating systematic verification,
sensitivity analyses, calibration and validation, that appears to work
smoothly and faultlessly (5). The procedure is set out in the PCSWMM3
Workbook Module. Of course, adequate field data is essential, and our
inexpensive raingauges and data loggers (6) render it easy to collect
sufficient summer thunderstorm activity in one or two months (in the north
during the period May-October) to satisfy model calibration for most criti-
cal modelling situations. The same Instrumentation is used in our real-
time controllers (7). This points to the advantages of considering an
integrated microcomputer-compatible system from the outset. A typical
sequence of activities described here is shown schematically in Figure 2.
There are other important benefits that accrue from a dense network of
field stations. Perhaps the best example occurs in storm modelling. It
should be obvious that urban flooding often results from the short, sharp,
local convective storms that are so common in warm weather. These storms
are much smaller in area than the catchments and have surprisingly short
lives, typically 50 minutes or so. Our software PCRAINPAC analyses data
46
-------
from a network of synchronized raingauges to compute storm speeds, direc-
tions, growth and decay (8). Output from PCRAINPAC is fed directly into
PCSWMM3 (9) to compute flood flows and pollution loads that are demon-
strably more accurate than any other hydrologic package known to us. This
is to be expected: the better the Input, the better the results.
Figure 2 presents the typical sequence of activities involved in
carrying out a hydrologic modelling study on a personal computer
environment. Any such study requires time series' of rainfall as input
to the model. A record of discharge is required in order to calibrate and
validate the model. These data are typically supplied by government
agencies such as the Atmospheric Environment Service (AES) and Water Survey
of Canada (₯SC) in Canada and the National Weather Service (NWS) in the
United States.
Large amounts of data are often required, especially for studies
involving continuous simulation. This data can cover periods in excess of
40 years. It would not be feasible to distribute or acquire such data on
media such as floppy diskette or over telephone lines. Consequently, the
most efficient method for supplying/acquiring this data is on 9-track
magnetic tape. Several commercially available subsystems exist which per-
mit interfacing of standard, 9-track, mainframe-compatible tape drives with
IBM-PC compatible microcomputers. These subsystems interface through a
Obtain data
from archive
Tape interchange
package
Compress data
Data base
management
Hydrologic
models (storms,
floods, levels,
receiving waters)
Final Report
9-track
tape
drive
30 Mb
Hard
disk
area
Figure 2. Typical sequence of activities.
47
-------
board installed in one of the expansion slots on the PC bus. The tape
interchange utility supplied with the subsystem downloads the data to the
user's system.
Depending on the amount of data required for the study, the files
created from the data supplied on tape can range from 1 to 60 Mbyte in
size. This necessitates having sufficient storage available on the users
system. It has been our experience that a hard disk drive with 4-0 to 50
Mbyte capacity can be acquired fairly inexpensively and will satisfy the
needs of most studies requiring large data storage.
In order to properly manage, retrieve and update this database of
hydrometeorologlc data special software is required. The software which
has been developed has the capability to take the data in its raw form and
store it in a "compressed" form in which only non-zero data is retained
thereby reducing the data file size. In an application performed by our
group, a rainfall database covering the period 1975-1984- with an uncompres-
sed size of 52 Mbyte was reduced to 18 Mbyte using the data compression
technique. The database, once constructed, is coherent and can be made
available to all members of a modelling group. Updating or correction of
the database is the responsibility of one person who needs only manipulate
a single file. This ensures that all modellers are using an identical
database. Increased study reliability results. The data is accessed by
users through menu-driven retrieval utilities and can be automatically
interfaced to simulation package such as PCSWMM.
Such a hydrologic database management system when being used by a
group of modellers has been found to function most efficiently in a network
environment. The database and application programs reside on a central
network computer, or server. The modelling group uses a number of PC's
which are also nodes on the network. Applications programs run locally and
independently on each node access data from the simulation. Results can be
returned to the database for use by other applications. The applications
programs are described subsequently.
Flood and pollution control may be effected by:
a) diverting overflows to storage, directly to receiving waters or
to some other sewer network,
*
b) pumping or draining overflows from detention storage,
c) warning responsible authorities of impending flooding and/or
pollution, for example for timely closure of swimming beaches or
underpasses, and
d) computing the extent of the swimming or recreational areas which
should be evacuated.
Real-time control of urban stormwater may be achieved by onsite micro-
computers rather than by a single, central computer. The microcomputers
need not be expensive; in fact, $30 hand-held Z-80 based microcomputers
48
-------
having BASIC in ROM are probably sufficient at each control site (7). The
site will generally include:
1) an instrument to measure flow, water level, rainfall and/or
pollutant concentration,
2) circuitry to transform the signal into information meaningful to
the microcomputer,
3) a small, replaceable microcomputer running a model derived from
statistical or similar analysis of data relevant to the control
site,
O circuitry to drive the electrical control mechanism, and
5) a data logger recording all incoming information, the control
action, and the precise date and time.
The microcomputers and associated circuitry, for onsite control, and
inexpensive raingauges measuring rainfall intensity, require large amounts
of software, also developed by the Computational Hydraulics Group at McMas-
ter University. The real-time control software (TSDASUTIL) has been based
on transfer-function models (TFM) of synthetic long-term flow and pollution
at each control site. The TFM is easily coded into the data logging
program, written in BASIC. TSDASUTIL is menu-driven and has, among other
attributes, data communication software, so that the data may be automati-
cally transferred to a central database (10) after manually retrieving the
tape. Other software that may typically be involved in such a study is
depicted in Figure 3.
The TFM software is derived directly from the synthetic long-term time
series generated by continuous PCSWMM3 at each site. It predicts expected
flow and pollution in the vicinity of the control site a few minutes ahead
(enough to complete a control action). There is a risk, in some drainage
systems, that bad timing of diversions could cause flows to coincide down-
stream such that flow conditions become worse than they need have been. In
these cases it is necessary to run the continuous simulation for the
overall drainage system with all local control software at each site simu-
lated. This is why an elaborate overall deterministic model is necessary.
PCSWMM3 also allows the full range of other stormwater management strate-
gies (separation of roof leaders, tile drains, etc.) to be evaluated (10).
The minimum functions that must be supported in a real-time flood
control system are:
a) data acquisition,
b) data storage/retrieval,
c) streamflow forecasting,
d) project operation.
The software which might be utilized a typical RTC application is
illustrated in Figure 3. These programs perform the four tasks listed
above.
49
-------
TSDASUTIL
4-
DASDECOD or
TSDASDECOD
INTERFACE
CHGTSH
DASTRAN
PCGRAPH3
RAINPAC
PCSWMM3
RTCSIH and
TS analysis
TOTSED
Figure 3. Typical Application (evaluation of RTC),
DAS3 and TSDASUTIL collect data from nearly raingauges and store it on
cassette tapes in the field. TSDASUTIL might also be controlling a diver-
sion structure, locally, based on the incoming real-time rainfall data.
After collecting the cassette tapes and transporting them to a central
site, the data is decoded and communicated to our local area network (LAN)
of IBM-PC compatibles. The LAN consists of a 52 Mbyte hard disk, a 9-track
magnetic tape drive, a Idne printer, and three microcomputers, each with
512 kbytes of RAM, 8087 co-processors, and two floppy disk drives. Program
DD1 decodes the data collected by DAS3, using a data decoder. TSDASUTIL
transmits the data via a RS232 interface. Data is received by DASDECOD or
TSDASDECOD through the serial port of one of the LAN stations, and stored
in a disk file.
The data can be processed immediately by DASTRAN or permanently ar-
chived in the CHGTSM for a variety of applications. DASTRAN translates the
raw data file created by DASDECOD or TSDASDECOD into event summaries. If
desired the program also produces two additional files: 1) a PCSWMM3
compatible interface file for input to the PCSTATISTICS3 module, useful for
50
-------
ranking rainfall events, based on peak intensity, total volume, duration,
etc., and 2) a PCGRAPH compatible data file for plotting hyetographs.
INTERFACE reformats the raw data for entry to the CHGTSM, our rainfall-
runoff data base, which resides on the LAN. This makes the data available
to our group for a number of applications, including data processing,
graphics, report generation, and hydrologic modelling studies.
Figure 3, presents a typical hydrologic modelling application, aimed
at evaluating proposed RTC schemes. First RAINPAC is used to prepare
rainfall input for the PCSVMM3 continuous model. STOVEL is used to analyse
a number of point rainfall records for direction, speed, and peak growth/
decay mechanisms. THOR4-DPT uses the results of the STOVEL analysis to
generate hyetographs for each subcatchment in the PCRUNOFF3 model. The
continuous water quantity and quality record generated by PCRUNOFF3 is then
input to the PCTRANSPORT3 model of the diversion structure(s). The results
of the continuous PCRUNOFF3 and PCTRANSPORT3 simulations can be used to
develop a simple rainfall-runoff forecast model which can be evaluated in
RTCSIM, and implemented in TSDASUTIL. TOTSED models the deposition and
scour of sediment at a combined sewer outfall.
Other programs which might be included in such an analysis are
PCSTORAGE3, CHGQUAL, OVRFL03, and TOTSED. At any point in the hydrologic
modelling, the rainfall, runoff., and pollutant time series output by these
programs can be stored in the CHGTSM for subsequent analysis.
The whole system may sound unduly complex, but runs very effectively
on IBM-PC compatibles. All of the programs are operational; the software
is listed below.
1. STORMWATER MANAGEMENT
The PCSWMM3 package for urban runoff quantity and quality in storm and
combined sewer systems comprises 11 modules:
1. PCRUNOFF3 module - simulates overland flow hydrographs and pollu-
tographs from urban and non-urban watersheds.
2. PCTRANSPORT3 module - generates infiltration, dry weather flow
and provides hydraulic routing model for flow and pollutants in
non-surcharged situations. Includes pipe design options and
various pre-programmed pipe shapes and regulators.
3. PCEXTRAN3 module - hydraulic routing model for flow in surcharged
situations. Includes regulators, pump stations, tides and some
pre-programmed pipe shapes.
4. PCSTORAGE3 module - simulates the flow routing and removal pro-
cesses of detention ponds and sewage treatment plants. Includes
cost estimation procedures.
5. PCSTATISTICS3 module - provides simple statistical analysis and
frequency analysis of continuous simulation results generated by
other modules.
51
-------
6. PCGRAPHICS3 module - provides a general plotting utility for
PCSWMM3 results and other data.
7. PCEXECUTIVE3 module - assists the user in data preparation for
PCSWMM, controls file handling.
8. PCCOMBINE3 module - provides a utility for merging files of
results from individual PCSWMM3 simulations for input to subse-
quent modules thereby allowing large areas to be simulated in
segments.
9- PCWORKBOOK3 module - leads novice user systematically through a
series of verification tests on several modules of PCSWMM3.
10. PCINSTAL3 module - provides an "official" check on the perfor-
mance of the users hardware and software system.
1 1 . PCSWMM3 - RUNOFFG - a modified version of the RUNOFF module which
includes interflow simulation. This provides a means for con-
tinuous computing of startup values for wet weather events.
12. CHGQUAL - modified version of the SVMM3 RUNOFF block water quali-
ty modelling algorithms, for pollutant buildup and washoff.
13. OVRFL03 - given the geometry of the main channel and side weir,
computes the water surface profile along a side weir, and the
over-spill, given incoming flow.
14. TOTSED - analyses quasi-transient sediment motion, deposition and
scour at a combined sewer outfall.
Expert System:
15. PCSWMM A-S (Auto Sensitivity) represents a group of program
modules forming a shell around the PCSWMM program. The program
modules work on an interactive basis with the user and the RUNOFF
module of PCSWMM. They provide sensitivity and uncertainty in-
formation on the effects of eleven physical input parameters
required by the RUNOFF module. The output generated through
PCSWMM A-S includes a ranking of sensitivity magnitudes for each
of the tested parameters and the uncertainty, expressed as a
percentage, that is associated with the output from the RUNOFF
module.
The program requires the standard minimum hardware configuration
specified for the PCSWMM3 package. This includes an IBM-PC or
compatible with at least 512k RAM, two 360k drives and an 8087
HDP. To operate the program the user responds to a series of
prompts contained in the batch "shell" and indicates an existing
file which is to be analysed by the A-S. Input required includes
confidence levels in the input parameters and certain utility
data. Analysed output data includes information on total volume
of runoff, peak flow, mean flow, time to peak flow, flows ex-
52
-------
ceeding prespecifled limits and the volume of any exceedance.
Output is directed to the console as well as to files on the disk
drive. The time required for a complete run of the package is a
minimum of 60 to 80 minutes for a single subcatchment with 12
data points for the input hydrograph. Internal processing of the
data requires approximately 80* of the total run time.
2. STORM MODELLING
RAINPAC is a package of 4 programs which analyses and models the
kinematic properties of urban rainfall events:
16. STOVEL - analyses a number of point rainfall records for direc-
tion, speed and peak growth/decay mechanisms
17. THOR4DPT - simulates the kinematic properties of a storm using
STOVEL results to generate hyetographs at point locations.
18. THOR4D - simulates kinematic and areal averages for storm events.
19. THOR3D - simulates the areal average of a line storm event.
3. FLOOD PLAIN ANALYSIS
20. PCHYMO - hydrologic modelling language for simulating non-urban
watersheds on a simple event basis using the SCS curve number
technique. Methodology not recommended for serious hydrologists.
21. HEC-2 - a steady state backwater curve calculation model which
simulates water surface elevations along open channels.
22. FASTHEC-2 - a preprocessing program for HEC-2 which assists in
data preparation.
4. TIME SERIES MANAGEMENT SYSTEM (TSM)
The following 5 programs are used to build and/or update the data
base:
23. OPENFL - opens a TSS file and optimizes record size and number of
records in a file.
24. OPELBL - opens a data set in the file opened by OPENFL.
25- PATCH - reads the contents of a data set opened by OPELBL
(created by user previously) and inserts it into the data base.
26. INSERT - similar to PATCH (replaces current data with new data).
27. UPDATE - similar to PATCH.
53
-------
The following 5 programs are used to retrieve data. QUERY and CONTINS are
similar to RETRIV, and could be classed as output handlers for specific
purposes,
28. RETYDIR - displays information regarding the types of data which
reside in the data base, and selects the appropriate TSS file,
containing the data requested by the user.
29. RETSSDIR - displays information regarding the locations for which
data reside in the directory,
3C. RETRIV - retrieves data for user specified period (1-365 days) in
constant or variable timestep format.
31. QUERY - retrieves and displays data for user specified site and
period. A disk file containing this data, at either constant or
variable timesteps may be created.
32. CONTINS - similar to QUERY, but the user may specify the format
of the output file, and thus can be used to prepare input data
files for other programs. Default format is for PCSVMM3 RUNOFF
data file (Rainfall data groups).
33. INTERFACE - utility for processing data (flow, rainfall) from
government agencies such as National Veather Service (US), Atmos-
pheric Environment Service (Canada), Vater Survey of Canada, or
private groups such as our group (CHG), into forms suitable for
input to PCSVMM3 or the CHG-TSM.
A. DATA MANAGEMENT
34. INTERP - processes rainfall from decoders or flow from chart
records into event summaries. Also provides input to FASTPLOT.
Similar version for Apple lie, written in Applesoft BASIC. Also
FORTRAN version for APIOS DCPS-D/VPS units.
35- DASTRAN - processes raw rainfall data as created by DASDECOD or
TSDASDECOD into event summaries (similar to INTERP). Also
creates PCSVMM3 compatible interface file of rainfall, and
PCGRAPH input file for plotting data.
36. TSDASUTIL - program package performs various tasks related to
data acquisition and RTC, including collecting rainfall data and
controlling a simple diversion structure in real-time. The
package also processes the data, plots hyetographs, and transmits
the data to IBM-PC compatibles for further processing; written in
Sinclair BASIC for the Timex/Sinelair 1CB0 microcomputer. Two
main modules handle these tasks: 1) RTCONTROL - data acquisition
and RTC, and 2) TSDECODER - data processing, plotting, decoding,
and communication.
54
-------
6. DAS PROGRAMS (DAS = Data Acquisition System)
37. LPDAS2.V2 - microprocessor controller program for timing and data
collection for DCPS and lov-pover data logger; written in macro-
assembler.
38. DAS3 - same as above for standard data logger.
39- TBDAS3 - same as above but for TBRG.
40. DASMOE1 - same as DAS3 but also records data from wet/dry pre-
cipitation sampler.
7. DAS Data Decoder Programs
41. DD1 - decodes data collected by DAS on cassette tape, and Trans-
mits it to computer (IBM-PC compatibles, TRS-80, PD? 11/23);
written in macro assembler.
42. DD1-APP - same as above but for Apple lie.
43- TSDECODER - retrieves data from storage, converts from Sinclair
to ASCII character set, and transmits data serially to IBM-PC
compatibles. Also performs some simple processing of the data in
the field, including simple graphics; part of TSDASUTIL package
(see 36).
44. DASDECOD - receives data sent by TSDASUTIL through RS232 device
and creates a disk file (same as DASDECOD) on an IBM-PC compati-
ble; written in GVBASIC.
45. TSDASDECOD - receives data sent by TSDECODER module of TSDASUTIL
through RS232 device and creates a disk file (same as DASDSCOD)
on an IBM-PC compatible; written in GVBASIC.
46. DASINTERFACE - receives data sent by decoder (DD1-APP), creating
a disk file on Apple He; written in Applesoft BASIC.
8. PLOTTING PROGRAMS
47. FASTPLOT - plots hyetographs or hydrographs of events, input from
INTERP program; for use with Houston Instruments plotter.
48. PCGRAPH - (same as GRAPHICS module of PCSWMM3), used to riot
hyetographs as created by DASTRAK; written for IBM-PC compati-
bles.
55
-------
8. REAL-TIME CONTROL
49. RTCONTROL - see 36.
50. BOXJEN (MITS originally) - program for the identification of a
discrete linear Transfer Function Model from input and output
time series; vritten by the Chemical Engineering Process Control
Group at McMaster University modified for use on IBM-PC compati-
bles.
51. RTCSIM - simulates the operation of a simple diversion structure,
continuously, in real-time. Uses the results of a TRANSPORT
simulation, and runoff forecasts to simulate the diversion of
CSO.
REFERENCES
1. Huber, Wayne C., Heaney, James P., Nix, Stephan J., Dickinson, Robert
E. and Polmann, Donald J. Stormwater Management Model User's Manual
Version III. U.S. Environmental Protection Agency, Cincinnati, Ohio,
1981 .
2. James, ₯. and Robinson, M.A. An affordable alternative to a mainframe
computer environment for continuous modelling. In: Proceedings of
the Stormwater and Water Quality Modelling Conference, USEPA, Gaines-
ville, Florida, 1985, pp. 13 - 30.
3. James, W. and Robinson, M. Continuous urban runoff modelling. Con-
ference on Urban Drainage Modelling, Dubrovnik, Yugoslavia, 1986.
4. Unal, A. and James, W. Distributed continuous hydrologic processing
using microcomputer networks. In: Proceedings of the Stormwater and
Water Quality Modelling Conference, USEPA and Ontario Ministry of
Environment, Toronto, Ontario, December 5-6, 1985, pp. 169 - 180.
5. James, W. and Robinson, M.A. Standard terms of reference to ensure
satisfactory computer-based urban drainage design studies. Canadian
Journal of Civil Engineering, Vol. 8, No. 3, 1981, pp. 294 - 303-
6. Haro, H., Kitai, R. and James, V. Precipitation instrumentation
package for sampling of rainfall. Institute of Electrical and Elec-
tronics Engineers (Transactions on Instrumentation and Measurement),
Vol. IM32, No. 3, 1983, pp. 423 - 429.
7. James, W. and Stirrup, M. Microcomputer-based precipitation instru-
mentation. International Symposium on Comparison of Urban Drainage
Models with Real Catchment Data, IAHR, IAWPRC, Dubrovnik, Yugoslavia,
1986.
56
-------
8. James, V. and Scheckenberger, R. Storm dynamics model for urban
runoff. International Symposium on Urban Hydrology, Hydraulics and
Sediment Control. University of Kentucky, Lexington, Kentucky, 1983,
pp. 11 - 18.
9. James, ₯. and Robinson, M. Conversion of the USEPA SWMM3 package to
microcomputers. ASCE Fourth Conference on Computing in Civil Engi-
neering, Boston, Massachusetts, 1986.
10. James, V. and Unal A. Evolving data processing environment for compu-
tational hydraulics systems. Canadian Journal of Civil Engineering,
Vol. 11, Mo. 2, 1984, pp. 187 - 195.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
57
-------
DABRO
A BASIC LANGUAGE PROGRAM FOR HYDROGRAPH COMPUTATION
by: Bernard L. Golding, P.E.
Consulting Hydraulics Engineer
Orlando, Florida 32819
ABSTRACT
In this paper a hydrologic model designated DABRO for the computation
of hydrographs from complex drainage basins is presented and discussed. In
this model, written in the BASIC language, rainfal1-excess increments,
computed by the SCS runoff curve number procedure, are applied to a
mathematically computed unit hydrograph of each subbasin to obtain the
runoff design hydrograph from each subbasin. These subbasin hydrographs are
then summed and/or routed downstream by the model to obtain the total basin
hydrograph. The hydrographs computed by the model are similar to those
computed by TR-20 and HEC-1. However, because of easy data entry and
editing capabilities, the model is extremely easy to use. Also, the code
may be easily changed by the user to fit particular situations.
Simulations of actual runoff events from three urban watersheds and
one large rural watershed by use of this model are also presented and the
results discussed.
INTRODUCTION
DABRO, in its "batch model format", actually consists of three
separate programs or parts. The first program (Part 1 - Data Input Files)
58
-------
is used to input the various subbasin properties (subbasin areas, runoff
curve numbers, unit hydrograph and routing parameters, etc.) on to a disk
file; the second program (Part 2 - Read Data Files) is used for reading
and/or editing the data placed in disk file by the first program; and the
third program (Part 3 - Hydrograph Computation) computes the rainfall-
excess, unit hydrograph and total hydrograph of each subbasin, and sums
and/or routes them downstream to the basin outlet. In this version of the
model, the total hydrograph of each branch is computed in a sequential
manner and stored on disk file. Later, during subsequent program operation
as the main channel hydrographs are being computed, these hydrographs are
read from disk file by the program in the same order in which they were
input. The program then adds the branch hydrographs to the main channel
hydrographs to produce the total hydrographs below the channel
intersections. A definitive numbering sequence is necessary to insure
proper program operation.
A separate program (RAIN.FIL) is used to input rainfall increments
comprising the design storm on disk file.
These programs, plus programs to perform the various operations
separately (compute hydrographs, sum hydrographs, flood route hydrographs),
constitute the book "Design Hydrographs by the Drainage Basin Runoff Model
(DABRO), published by Hilbern Engineering Software.
In the model, rainfall-excess increments are computed from successive
rainfall increments by a modification of the standard Soil Conservation
Service (SCS) rainfall-runoff equation which modification accounts for
runoff from three sources - from the urban directly connected impervious
area; from the urban pervious (grassed) and nondirectly connected
impervious area; and from the rural portion of each subbasin or from each
source only. The mathematically computed synthetic unit hydrograph used in
the computation of the design hydrograph of each subbasin, although similar
in shape to the unit hydrograph of F. F. Snyder, permits the user to vary
both its peak flow value (discharge) and the time to peak (lag time) of the
peak flow. Reservoir routing is accomplished by the storage-indication
working curve method and channel routing by the Muskingum Method.
59
-------
In addition to successive rainfall increments comprising the design
storm, inputs to the various programs are the number (designation) of each
subbasin; the properties of each subbasin including:
1. the previously computed peak flow (discharge) and time to peak of
the unit hydrograph of the subbasin,
2. the area of the subbasin,
3. the percentages of the directly connected impervious area and
rural areas of the subbasin,
4. the runoff curve number (CN) for the urban pervious and non-
directly connected impervious area of the subbasin,
5. the runoff curve number (CN) for the rural portion of the
subbasin,
6. the initial abstraction (IA) for the urban pervious and non-
directly connected and rural portions of the subbasin;
the properties of the various channel reaches through which flood routing
is being accomplished, including:
1. the length of the river reach (reach length) through which the
subbasin hydrograph or combined subbasin hydrographs are to be
routed,
2. the reach velocity,
3. the reach routing coefficient (X);
the properties of the storage reservoir through which flood routing is to
be accomplished, including:
1. discharge rates from the reservoir at various elevations above
the spillway,
2. the storage volume in the reservoir at these same elevations;
and the computational time interval.
Outputs from the programs are a table of the original successive
rainfall increments input to the programs and the resultant computed
rainfall-excess increments; a table of the computed unit hydrograph
ordinates of each subbasin; tables of the total combined (summed and/or
routed) hydrographs of the various subbasins; and a plot of the final total
basin hydrograph. Included with the tables of hydrograph ordinates (unit,
design, summed and routed) are the number of inches of rainfall-excess from
each subbasin (=1.0+" for the unit hydrograph).
60
-------
RAINFALL-EXCESS
In the DABRO Model, rainfall-excess increments are computed by a
modification of the standard rainfal1-runoff equation of the Soil
Conservation Service.(1) In this modification, as previously stated,
direct runoff from a drainage basin is considered to come from three or
less sources - from the urban directly connect impervious area; from the
urban pervious (grassed) and non-directly connected impervious (urban)
areas and from the rural area of the basin, each of which is considered
separately.
In the original SCS procedure, a runoff curve number CN for a
particular entire subbasin was first computed which enabled the computation
of the direct runoff Q up to a certain time from the basin knowing the
rainfall P which fell on the basin up to that time. In the originally
derived procedure, the following equations were used to compute direct
runoff Q from rainfall:
Q = (P-IA)2
P-IA+S (1)
CN = (1000)
10+S (2)
where
Q = Direct runoff (inches)
P = Total storm rainfall (inches)
IA = Initial abstraction (inches)
S = Watershed storage factor (inches)
CN = Runoff Curve Number (dimensionless parameter)
The runoff AQ in inches which occurred in a certain time interval At was
then equal to (Qt~Qt-l)»
The Runoff Curve Numbers for various types of agricultural, suburban
and urban land uses as originally developed by the SCS for computing direct
runoff from rainfall utilizing these equations are given in various SCS
manuals. As most engineers know, these Runoff Curve Numbers depend on the
61
-------
Hydrologic Soil Group, the land use or cover and the Antecedent Moisture
Condition for the particular drainage basin.
For purposes of computing direct runoff by the method presented in
this program, the following equation is used:
AQt = AP(%DCIA) +AQ,, (100-%DCIA-%RURAL) +AQr (%RURAL) (3)
10~0
where AQt is the total direct runoff from the basin in a particular time
interval £t (=Qt-Qt-l); _ A P(%DCIA) is the direct runoff from the
IfiO
directly connected impervious area (DCIA); £QU (100-%DCIA-XRURAL) is the
100
direct runoff from the pervious areas and non-directly connected impervious
(urban) areas in this time interval as computed by Equation 1,
AQr (%RURAL)
100 is the direct runoff from the undeveloped (rural) areas in this
time interval as also computed by Equation 1, and AP is the rainfall that
falls on the basin during this time interval. The ability to compute runoff
from three separate type areas within a subbasin enables the evaluation of
land use changes occurring on the basin.
In the computation of direct runoff from the directly connected
impervious area (DCIA) by this program, the first 0.1 inch falling on such
areas is assumed to fill up depression storage and, therefore, is not
assumed to contribute any runoff.
In the original procedure as developed by the SCS for use in rural
areas, an Initial Abstraction of 0.20S was used in Equation 1 to complete
direct runoff. However, in the application of the procedure as modified
herein the following values of IA can be used as being more representative
of the initial abstractions imposed on runoff by urban areas (2):
Hydro! ogic Soil Group IA
A 0.075S
B 0.10S
C 0.15S
D 0.20S (unchanged)
62
-------
Where runoff from undeveloped (rural) land is computed, the value of 0.20S
(unchanged) should be used.
In the DABRO Model, runoff can be computed from completely urbanized
basins in which case zero values are input to the program for the Runoff
Curve No.-Rural; the Initial Abs. Coef.-Rural and for the Percent-Rural. If
the basin is completely rural (no urbanization), zero values are input to
the program for the Runoff Curve No.-Urban Pervious & DCIA; the Initial
Abs. Coef.-Urban Pervious & DCIA; and the Percent-DCIA.
In the computation of runoff by the SCS procedure and as modified
herein, three Antecedent Moisture Conditions (AMC) are defined depending
upon the amount of rainfall (antecedent rainfall) that has fallen on the
basin prior to the storm from which direct runoff is being computed, which
AMC's are defined in Section 4, Hydrology of the SCS's National Engineering
Handbook.
In the computation of a Runoff Curve Number (CN) for a particular
basin or portion thereof, by the SCS procedure, a weighting process,
depending on the percentage of the particular land use cover, is used. As
previously mentioned, separate Runoff Curve Numbers for both the urban
portion of the Drainage Basin not directly connected - the pervious
(grassed) area and the non-directly connected impervious area of the urban
portion and the rural portion of the subbasins are both program inputs.
Since the DABRO Model uses the SCS Runoff Curve Number to compute
rainfal1-excess, it is, of course, a single event model. However, as
subsequently discussed, it has been and can be easily modified for
continuous analysis using an antecedent precipitation index (API) which
relates rainfall to S, the watershed storage factor, or CN, the Runoff
Curve Number.
UNIT HYDROGRAPH
In the DABRO Model, the ordinates of a synthetic unit hydrograph for
each basin are computed by the following equation developed by James A.
63
-------
Constant, formerly Chief Reservoir Regulation Unit, Albuquerque District,
Corps of Engineers{3).
(Log t)2
2
2cr
Q = 111.8094 A e
crt (4)
where
Q = Flow Rate (Discharge) in cfs
A = Drainage basin area in square miles
t = Time in hours from beginning of runoff
T = Time in hours at which 50% of flow has passed
= 10 (In lOcr2 + Log TmaxQ)
o" = Standard deviation of the log 10 of the values (see note below)
This equation produces hydrographs that generally conform to the shape
of the synthetic unit hydrograph developed by F. F. Snyder as presented in
Corps of Engineers Manual EM-1110-2-1405 - the rising limbs of the unit
hydrograph are parabolic and the recession curves are approximately
exponential and are similar to the unit hydrograph of the Soil Conservation
Service.
In the evaluation of Equation 4, Q is differentiated with respect to
time and set equal to zero - standard procedure for determining the peak
value of a function - which results in the following equation (after
transposing and rearranging):
-'(In 10)2 o- 2
2
e - TmaxQ Qmax (See note below) (5)
ff- 111.8094A
Note: cr = Greek letter SIGMA; letter S used for SIGMA in program listing.
(In 10)2 # 2 = 2.650949056 CT2
2
64
-------
The values on the right side of Equation 5 are generally easily computed
from standard equations as will be subsequently described. TmaxQ is the
time (in hours) from the beginning of rainfall-excess to the peak of the
unit hydrograph (TOTAL TIME TO PEAK-HOURS) which is equal to the lag time
(normally defined as the time from center of mass of rainfall-excess to the
time of peak) plus one-half of the unit hydrograph duration or tp + D/2.
In this program for computing unit hydrographs,
-------
ordinate of the design hydrograph. The program automatically performs the
required time increment lagging required for hydrograph computation by this
procedure. Flood routing through channel storage is accomplished by the
Muskingum Method and flood routing through reservoir storage by the
storage-indication working curve method or Modified Puls Method.
In the model, the total basin (outflow) hydrograph is computed by
summing up the various subbasin design (and/or routed) hydrographs. The
model automatically performs the time lagging required for basin hydrograph
computation. As previously stated, the DABRO Model will also plot up the
final outflow hydrograph from the total basin under consideration (time in
hours vs discharge in cfs) if desired. The final basin outflow hydrograph
can also be saved on disk for possible further manipulation by the other
programs in the book.
MODEL VARIATIONS
Because the DABRO Model program is written in Microsoft Basic and the
program can be listed (not copy protected), changes can easily be made. For
instance, data can be easily imbedded into the program so that manual entry
of subbasin variables does not have to be input to the model every time it
is used. Also, modifications to allow manual entry of a series of rainfall
increments so that a different set of rainfall increments can be applied to
each subbasin can easily be done. As subsequently described, this enables
the model to be used in real-time reservoir regulation.
MODEL APPLICATION
General: The DABRO Model has, to date, been used to simulate runoff
from four different basins. Three of the basins were small, fairly steep,
almost completely urban basins located in the Tallahassee area (hereinafter
referred to as Basin Nos. 1, 2 and 3). The fourth basin (Basin No. 4) was a
large, fairly flat, sandy, almost completely rural basin, located in
Manatee County, Florida (the Manatee River basin). All were well documented
as to their area, physical properties, land use, soil type, etc., and all
66
-------
had accurate gaged rainfall-runoff data including information on the amount
of antecedent rainfall prior to the event simulated. In all cases, the
Theissen polygon method was used to weigh the rainfall increments of the
various storms simulated.
Urban Basins: Basin Nos. 1, 2-and 3 are, as previously stated, all
highly urbanized basins, were gaged as part of the USGS's urban basin
gaging program, conducted in the Leon County, Florida area. Basin No. 1,
the Frenchtown Creek basin, 3.29 square miles in size, is essentially a
completely sewered basin consisting of the Frenchtown Creek area and the
Florida State University (FSU) campus. The lower portion of the Frenchtown
Creek basin consists of a 1200 feet long, 8 feet diameter circular conduit
which carries Frenchtown Creek under the FSU campus. Basin No. 2, the
Franklin Avenue drain basin, 2.06 square miles in size, is also essentially
a highly urbanized basin. It consists of two definitive portions or parts;
the upper one-half, which consists of a sewered residential area, and the
lower one-half, which consists of a portion of the downtown area of
Tallahassee and the State Capitol complex. The open channel draining this
lower portion of the basin is essentially retangular or trapezoidal in
shape, much of which channel is lined. Basin No. 3, 0.21 square miles in
size, is an essentially mixed residential and commercial area drained by a
fairly steep open channel. Soils in all three basins were classified by the
USGS as being in Hydrologic Soil Group B.
In all, sixteen rainfall events were simulated by the DABRO Model;
five on Basin No. 1, five on Basin No. 2 and six on Basin No. 3. The
results of these simulations are shown on Figures 1 through 16. In all
cases, the total impervious area and directly connected impervious area
were determined from a study of aerial photographs supplemented by field
inspection. Runoff Curve Numbers for the pervious and non-directly
connected impervious areas were determined in the usual manner by weighing
the grass and impervious areas. In all cases, the time to peak in hours and
peak flow rate of the ten-minute unit hydrograph of each basin was
determined by the equations developed by Espey, Altman and Graves(4). In
accordance with program requirements that the time step At be equal to the
67
-------
CT>
CO
OAOKO >v\OOEl_
-------
3O, 198
u.
\J
VJ
UJ
o
oi
u
in
a
it i£ i
(UOURS -V IO AAIM.
SAS//^ * 2 STOCM Of JULY II.
tL f
69
-------
81
4
I
0:
SIMULATION
t
DA&RO MODEL.
id£ASUR£Q DI5CHARCC
70
-------
^4,IB8O
PA0RO MOD£L_
10
TIME.
-r IO
71
-------
DRAINAGE -BASIN *3 ST&RM or JULV 24,
QA/SKO
Zl
23 2^ I
TIMS. ( HOURS v /O M
-------
DRAJWAGE. BASIXJ * 3 STORM OP SEPT 6.
DA6RO MODEL
MEA:>URED on CHARGE
fe
ORA/WAdE BASfM «» 5 S-TOR^N O^ S«^*T El.
CAflgO MDDEL.
73
-------
SAS1NJ *J 3
OF
, /9flO
IZO
DAB.^O MODEL
9 10 u
HOURS 4 IO -V»/MOr£
10
ORAPWA6E 6A51M a
STORM OP MARCH 3O, 1981
T(A1£ f UOURS -KO
74
-------
unit hydrograph duration D, ten-minute rainfall increments of the various
events simulated were Inputs to the program. No attempt was made to
calibrate the model to improve the simulation results.
As can be observed from these Figures, the results of the simulations
were quite good, except for the rainfall of March 31, 1981 on Basin No. 1,
shown on Figure 5, in which case it appeared that the 1200 feet long, 8
feet diameter conduit under the FSU campus was inadequate to convey the
peak runoff rate resulting from that particular rainfall.
Rural Basin: Simulation of various runoff events from the 130 square
mile Manatee River basin (Basin No. 4) using the DABRO Model was done as
part of a real-time reservoir regulation program implemented by Camp
Dresser and McKee for the Lake Manatee Dam(5).
For the prior design of a new emergency spillway for the dam and for
the purposes of a possible future real-time regulation program, the basin
had been previously subdivided into five subbasins and runoff curve numbers
based on land use in each of the subbasins for the AMC II condition
computed. Also, as part of this prior work, two or more rainfall gages with
telemetering equipment had been installed in each of the subbasins.
As part of the simulation work, the lag time of each subbasin was
computed by Snyder's Method using a Ct=2.2 and the time to peak Tp (=0/2
+ tp) for the two-hour unit hydrographs for each subbasin computed. The
peak flow rate of each of the two-hour unit hydrographs were then computed
by the equation Qp=CA/Tp. A "C" value ranging from 300 to 425 for each sub-
basin was selected based on the characteristics of the individual subbasin.
The results of- simulation of the June 17-18, 1982 rainfall, obtained
by summing and/or routing the various subbasin hydrographs, are shown on
Figure No. 17. This was a fairly major rainfall event in that an average of
approximately 6.6 inches of rainfall fell over the basin in a 38-hour
period, most of which fell in a 24-hour period. This resulted in 2.96
inches of runoff at the dam. The prior dry period to this rainfall was
75
-------
o
ir
IOVAAJ
14000
I200O
10000
8000
6030
4000
2000
0
1
L
J~-
?'
I/
/
/
/|
i
1
_r
f
'
'
*x
/
0 10 20
_
J---
^
\
(
N
\
1
\
\
"t
RESERVOIR GAT
OPERATION
i
T
^SERVED INFL<
IT RESERVOIR
SIMULATED INFL
AT RESERVOIR
V,
^^
F
S^
\
^
^
so
. 0
u
T
X1
VI
OVIL
JTFl
ANA
^
Fi
.OW
TEE
--,
gu
AT
DA
^
_...
re
M
^
1-
L
t
\7
~,
i
"" *
40 5<
TIME
reflected by the average computed basinwide CN of 67. In the computation of
the simulated inflow, shown on Figure 17, each of the previously computed
subbasin CN's for AMC II were adjusted down three units to account for the
difference between the originally computed average basinwide CN of 70 for
the AMC II condition and the actual average basi-nwide CN of 67 for this
rainfall event.
As can be observed from Figure 17, reasonably good simulation of the
June 17-18, 1982 rainfall was achieved.
Verification of the various parameters input to the DABRO Model and
the model itself were limited by the lack of a second major rainfall event.
However, reasonable verification was achieved using the Storm of August 28,
1981 during which an average of 2.3 inches of rainfall fell over the basin
resulting in 1.1 inches of runoff. This large runoff volume (CN=85)
reflected the relatively large amount (3 inches) of rainfall which fell in
the three days prior to the August 28, 1981 event. Again, the originally
computed subbasin CN's for AMC II were adjusted (this time upward) to
account for the difference between the computed average basinwide CN of 70
for the AMC II condition, and the actual average basinwide CN of 85. Note
76
-------
that this value of CN=85 is very close to the CN value for AMC III as given
in Section 4, Hydrology of the SCS's National Engineering Handbook.
However, because of the obvious extreme importance of the antecedent
moisture condition in this sandy basin, the following equation, based on
the standard API, was derived to enable adjustment of the CN values in the
Manatee River basin.
S = 0.48 + 6.72/API
where
S = average basin storage coefficient
In the application of the adjustment factor, the average basin storage
coefficient, based on API, is first computed by Equation 6 and then the
basinwide runoff curve number CN, based on the API, computed by the
standard SCS equation. A basinwide correction factor, which is then applied
to the runoff curve number of each subbasin, is then computed by the
following equation:
CN = CNftpi-70
Simulation of the August 28, 1981 event using these equations is shown on
Figure 18.
Both of the above equations have been integrated into the DABRO Model
and the Model revamped {DABRO 1) such that inputs to the model are now only
rainfall increments and API.
ADVANTAGES OF THE DABRO MODEL
1. The DABRO Model is a simple, easy to use interactive model which
instructs the user as to parameter input.
77
-------
2. The model has been verified by application to both urban and
rural basins. However, like all models, careful determination of
subbasin parameters is necessary.
3. The model takes both the total impervious area and directly
connected impervious area into consideration.
4. The unit hydrograph parameters (time to peak and peak flow rate
of the various subbasins) can be changed by the user to fit the
local situation.
5. The program, written in Microsoft Basic and not copy protected,
is very easy to modify since almost any engineer can understand
what the various lines in the program do. For instance, A=area,
V=velocity, Q=flow rate, etc.
6. The program will operate on any microcomputer with 64K's of RAM.
However, with a 64K computer, the program is limited to six
subbasins.
Ul
o
K
10000
8000
6000
400O
2000
=t_
r
v
_r-
[/
?
,
ULA
RES
TED
ERV
r^
INFl
OIR
=^:
LOW
m^T"
~ ,
^^»
^^.
-L
*s
^w
^
.
^^^
N ^
^-
"" V
^ «
Fi
g^
re
18
OBSERVED INFLOW
JVT RESERVOIR
-T_
->_
"0 10 30 30 40 K
TIME (hri)
78
-------
REFERENCES
1. "SCS National Engineering Handbook, Section 4, Hydrology", Soil
Conservation Service, U. S. Department of Agriculture, Washington
D.C., 1972.
2. Golding, Bernard L., Discussion of Paper Titled "Runoff Curve
Numbers with Varying Site Moisture", Journal I and D Div.,
Proceedings ASCE, Vol. 105, IR4, December 1979.
3. Constant, James A., "A Mathematical Determination of the
Ordinates of a Unit Hydrograph", Proceedings of the Seminar on
Urban Hydrology, Hydrologic Engineering Center, Davis,
California, September 1970.
4. Espey, W. H., Jr., Altman, D. G. and Graves, C. G., Jr., "Nomo-
graphs for Ten-Minute Unit Hydrographs for Small Urban
Watersheds", ASCE Urban Water Resources Research Program,
Technical Memorandum No. 32, ASCE New York, N.Y., December 1977.
5. Powell, Robert A., "Lake Manatee Reservoir Regulation Manual",
Camp, Dresser and McKee, Inc., November 1983, 120 pp.
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.
79
-------
APPLICATION OF A LOTUS SPREADSHEET FOR A SWMM PREPROCESSOR
by: S. Wayne Miles and James P. Heaney
Department of Environmental Engineering Sciences
University of Florida
Gainesville, Florida 32611
ABSTRACT
This report describes a preprocessor for the EPA Stormwater Management
Model (SWMM) which is based on a Lotus 1-2-3 spreadsheet. Currently used
methods to input data into SWMM are described and are compared to the Lotus-
based preprocessor. The advantages and disadvantages of Lotus as a pre-
processor host are discussed. The internal structure of the preprocessor
commands as written in Lotus' macro language is described including the "Help"
command which allows access to a number of reference files. Future improve-
ments to the preprocessor are explained and conclusions as to its development
are made.
INTRODUCTION
Large mainframe computers have long been used to run hydrologic models*
One of the most popular models, the EPA Stormwater Management Model (SWMM),
has also been recently converted into a personal computer version available
from several sources. This conversion has prompted the use of personal com-
puters in many other areas of data preparation and handling in this field.
In" order to fully exploit the potential of the personal computer, soft-
ware packages may be used which create a structured format to store and handle
data. This paper will describe one of these software packages, Lotus 1-2-3,
and how it is being used to process input data for SWMM. The following
sections in this paper will describe current methods of parameter entry into
SWMM and will compare them with a Lotus 1-2-3 based preprocessor. The struc-
ture of each section of the preprocessor will also be described.
PARAMETER ENTRY IN ORIGINAL SWMM
The original method of parameter entry Into SWMM was very tedious. The
model user had to enter all parameters into the input file in the correct
80
-------
format. Each input card had to be entered Jn the correct order and each
parameter had to be entered in the correct space allotments within each card.
The user was totally dependent on the written documentation to determine the
correct format as is shown in the excerpt from the SWMM User's Manual (1984)
in Figure 1. It was easy to make errors in entering values and difficult to
determine where format errors had been made. Even though this method of
parameter entry is still being used, it is quickly becoming obsolete.
Card Card Variable Dfrfau.lt
Group Format Columns Description Mane- Val»«
HI BF5.0 16-20 Width of subcatchment, ft finj. This tern WCl)* Sana
actually refers to the physical width.
of overland flow ia the subcatchmeat and ^3
may be obtained as illustrated in the text.
21-25 Area of sub.catchment, acres [ha]. WAHEAs*«*(2)*- Hooe-
26-30 Percent imperviousneas of subcatchmeot, ZMH3f** None-
31-35 Ground slope, ft/ft [diaensionleas]. HSLQBE=WW{4>*- None-
36-40 Impervious area. 1 WW(S)*' Vane
I Roughness factor,
41-45 Pervious area. j (Manning's n) BW(6)* N6ne>
46-50 Impervious area.) WSTORE=^W(7)* Vontf
> Depression storage, is_
51-55 Pervious area. I [mm). USTORE=WW(8)* Sotic-
*** Horton equation parameters if INFILM = 0 (Card Bl) **
56-60 Maximum (initial) infiltration rate. WUIAX-WWC*)* None-
in./br [mm/hrj.
61-65 Minimum (asymptotic) infiltration rate-, WLMIJi=WW(10)* Notre-
in./hr [mm/hrj.
F10.5 66-75 Decay rate of infiltration in Hortoa's DECAT=WW(ll)* Hone-
equation, I/sec.
*** Greett-Ampt equation parameters if 1NTILM - 1 (Card Bl) ***
2F5.0 56-60 Capillary suction, inches [ma] of vater. SCCT-WWC9)*' Nane-
61-65 Hydraulic conductivity of soil, in./hr HT3':ON=UV(lf?)"-None-
[tnm/hr].
F10.5 66-75 Initial moisture deficit for soil, SttDMAX=W( 11)*None-
volume air/volume voids.
H2 Blank card (except for identifier) Co terminate-
subcatch/nent cards: one card.
Figure 1. Example of parameter format description in. SWMM
User's Manual (1984).
81
-------
IMPROVED PARAMETER ENTRY INTO SWMM
Improvements on the manual parameter entry method have been made. In the
personal computer version of SWMM currently being marketed by James and
Robinson (1985), the user is prompted to enter parameter cards in their cor-
rect order (see Figure 2.). Prompts for the correct number of each card type
are also included according to previously entered parameter values. The user
is still responsible, however, for positioning the parameter values within
each card. The user's manual must be consulted for each c'ard entry as to the
order in which the parameters are entered on each card and the default values
for the parameter. This program is helpful in that the user does not need to
enter correct spacing between each parameter on a card. The values need only
to be separated by commas. If a card is entered erroneously, however, this
program relies on outside editing procedures. Once the card entry process
begins, it must be completed. Then the file may be edited with a standard disk
operating system. This program is clearly an Improvement over the original
SWMM input procedure.
MENU-DRIVEN PARAMETER ENTRY
Another approach to this problem is used by the South Florida Water
Management District (SFWMD). They have been distributing a'menu driven pro-
gram which aids in performing calculations needed in their permitting pro-
cedures. The program is menu driven in that the user is offered choices
throughout the program. Initially, the user may choose from the Soil Con-
servation Method, the Santa Barbara Method, or his own method (Mass Route) to
perform runoff calculations (see Figure 3). This program is easy to use and
contains convenient editing procedures. The program does lack, however,
sufficient documentation for use without consulting the reference manual.
E3. ENTER SWMM BLOCK TO BE CALLED
RUNOFF
SWMM RUNOFF BLOCK REQUESTED
ENTRY MADE TO SWMM RUNOFF BLOCK INPUT
Rl. ENTER FIRST TITLE CARD (DATA GROUP Al)
A1SFWMD REPORT EXAMPLE
R2. ENTER SECOND TITLE CARD (DATA GROUP Al)
A1RUNOFF MODULE
R3. ENTER FIRST CONTROL CARD (DATA GROUP Bl)
610,0,0,1,0,0,0,12,0,1,7,85,0,0,0
R4. ENTER SECOND CONTROL CARD (DATA GROUP B2)
B260,60,30,0,0,0
Figure 2. Example of parameter entry mode in PCSWMM (James and Robinson,
1985).
82
-------
DESIRED STRUCTURE IN A SWMM PREPROCESSOR
In order to create a working SWMM input processing program, several ideas
must be addressed. Firstly, all of the problems previously mentioned with
current parameter entry methods must be solved. The program should include a
format with which the user may input all parameters directly into a space
allotted specifically for that parameter. This would necessitate a program
with clear documentation. The user should also have free editing capabilities
within the program. The user should be able to move back and forth between
cards while entering parameter values, and change any previously entered
values while continuing the current parameter entry mode. Also the choice of
the values themselves is important. The user must feel comfortable with the
choice of each value. Therefore, a complete program should include documen-
tation which aids the user in choosing parameter values. If calculations are
needed to choose any value, a format for these to be performed should be
included. Lastly, this program should be able to operate such that the begin-
ning user is comfortably led step by step while the experienced user is not
slowed by the entry process.
OTHER PREPROCESSORS
A good example of a model preprocessing program was one for the CREAMS
Model by Dennison and James (1985) using dBASE III. They present several ideas
which are important in the development of a working preprocessor for a model:
1) It should be flexible and easy for a beginning user while not hindering
expert users. 2) It should refer the user to the manual when additional
TYPE -
1 - TO EXECUTE "SCS" PROGRAM
2 - TO EXECUTE "SANTA BARBARA" PROGRAM
3 - TO EXECUTE "MASS ROUTE" PROGRAM
SELECT ONE -
1
***** ENTER LEGEND INFORMATION *****
ENTER PROJECT NAME ( )
SFWMD PROGRAM EXAMPLE
ENTER REVIEWERS NAME ( )
W. E. COYOTE
ENTER PROJECT AREA ( .0000000 ACRES)
999
ENTER GROUND STORAGE ( .0000000 INCHES)
ENTER TERMINATION DISCHARGE ( .0000000 CFS)
10
Figure 3. Sample from South Florida Water Management District
runoff and flow routing program with menu choices (SFWMD, 1983).
83
-------
information is needed. 3) It should provide a convenient edit mode. All of
these ideas are necessary features for a complete preprocessor.
A choice of calculation procedures for the user is an important aspect of
the South Florida Water Management District's program (1985). Their program
generates and/or routes hydrographs. The beginning of this program asks the
user to choose from the Soil Conservation Service method, the Santa Barbara
Unit Hydrograph method, or an alternative method for generating runoff hydro-
graphs. This feature allows the user to generate hydrographs by whatever
method he feels confident and comfortable. An interesting addition to a
program such as this might be provisions for a "knowledge base" from which the
chosen calculation procedure may extract information. In such a program,
however, the user should be able to check all intermediate calculations and
assumptions without having to review program lines whenever possible.
LOTUS 1-2-3 AS A PREPROCESSOR HOST
The idea of using a personal computer based system as a preprocessor for
the EPA Stormwater Management Model is examined in this paper. Whether this
system is used with a personal computer version of SWMM or interfaced with the
mainframe version, the advantages of the personal desktop computer for param-
eter entry into the model are many. Among the software packages which are
available for the personal computer, however, no clear favorite has been
generally accepted to host a SWMM preprocessor. The many software options
available all have their own advantages and disadvantages in terms of data
entry, calculation, and documentation capabilities. This report will describe
the mechanics and basic layout of a SWMM preprocessor which is housed on the
Lotus 1-2-3 spreadsheet software package (Lotus Development Corp., 1984).
The Lotus 1-2-3 package was chosen for its flexibility in all of the
areas needed for developing the preprocessor. As a spreadsheet Lotus is, in a
sense, a large electronic notebook which has approximately 3250 pages, all of
which may be randomly accessed at any time. This feature will allow the user
of the preprocessor free movement during the parameter entry mode and will
ease any editing of old parameter values which may be needed. This feature
will also allow the user to access reference files which may be needed during
the parameter entry mode.
Lotus 1-2-3 also has its own internal macro language which allows data
manipulation to be automated. Since the main feature of the preprocessor is
to format the model parameters and input data such that they are accepted by
SWMM, Lotus' simple but comprehensive macro language is a convenient tool.
This macro language allows for the creation of menus which facilitate the
execution of the macro programs. These menus create an atmosphere in which
the user may logically and systematically generate a SWMM input file while
randomly accessing reference files, and finally, formatting the parameters and
exporting the complete input file into a file to be read by SWMM.
Another point of flexibility with the Lotus package is its interaction
with other types of files in the personal computer environment. Lotus can
import other database and word processing files into its spreadsheet format as
well as exporting text and numbers out of the worksheet. This feature pro-
84
-------
motes Lotus as a preprocessor to be interfaced with other software packages
and allows the option of postprocessing SWMM output with the use of Lotus'
graphical and statistical capabilities. Depending on the size of the SWMM
output file, all or part of the file may be imported onto a Lotus worksheet to
be summarized, analyzed, and/or graphed.
Unfortunately, the Lotus spreadsheet also has a few drawbacks for use as
a SWMM preprocessor. One area which causes problems in the parameter format-
ting procedure is Lotus1 policy for placing numbers in a column. For the
visual purpose of keeping a space between columns of numbers, Lotrus 1-2-3 does
not allow an entire column width to be filled with the digits of a number. A
column with a width of five, therefore, can only hold a four digit number. A
five digit number entered into this column will cause a row of asterisks to be
displayed across the cell width. This does not cause problems while per-
forming calculations within Lotus since the cell will still contain the value
entered; the value simply cannot be displayed. This does cause problems when
exporting the SWMM file from the Lotus worksheet into an interface file as is
done with the preprocessor. Lotus will export the display on the screen and,
therefore, a row of asterisks instead of the parameter value. It is often
necessary when entering SWMM parameters to use the full number of digits
allotted to a parameter by the SWMM read procedure. This problem of space
allocation may be solved within Lotus by handling the parameters as text would
be handled. Lotus 1-2-3 allows any text entered into a cell to completely
fill the cell and even extend beyond. Treating the parameter as text, how-
ever, could increase the processing time up to a few minutes for a large
number of subcatchments and could also permit a number with too many digits to
be entered into a formatted input card for SWMM.
The spreadsheet format also seems, at this time, to impose certain size
limitations on a preprocessor. These limitations will not come in terms of
the number of cards (e.g. subcatchment) which may be entered into a module.
The number of rain gauges and number of data points on a rainfall hyetograph
may also be very extensive. The limitation is encountered because of the
manner in which the parameters must be formatted with specific space allot-
ments on a. card. These allotments are provided by stringing card formats end
to end along the top of the spreadsheet and adjusting column widths to receive
each parameter. Cards are separated with different range names. Since the
formatted cards span the entire width of the spreadsheet, this procedure
limits the spreadsheet to only process one module per file. This in turn
necessitates that each module be run singularly and that input, output, and
interface files be called and manipulated manually. Manual file control is
not difficult on small SWMM runs, but it can be cumbersome if larger runs
employing several modules are performed.
MECHANICS OF A LOTUS PREPROCESSOR
The example file that has been created is a preprocessing package for a
simplified version of the SWMM Runoff Module. In general, this preprocessor
creates a structured work area in which the user may, in a systematic and
orderly manner, enter the parameters necessary to run the Runoff Module. The
sections and capabilities of this preprocessor are summarized and linked with
a menu which was produced using the Lotus macro language. The menu allows the
85
-------
user to easiy move from the parameter entry mode to the help/reference mode
and access a number of help files. The menu also includes selections to
execute the Lotus macro programs which will format the parameters into a
Runoff input file and export this formatted file to an ASCII file. Once the
user has loaded the package onto a Lotus spreadsheet, the Introduction screen
as shown in Figure A will appear. By pressing the keys Alt M, a macro program
will begin which will display the menu choices ("Create" "Help" "Edit"
"Save" "Print"). Each of these choices will be explained individually in
the following sections.
Create
The "Create" mode allows the user to begin a new job file. The first page
of this selection will ask for input such as job initials and a job number
which will later be used to name the formatted file created by the pre-
processor. Items such as the number of subcatchments and number of gutters in
the simulation are also requested in order to allocate space within the work-
sheet. The user will next move to the parameter entry section of the pre-
processor. As can be seen in the sample of this section shown in Figure 5,
KEH-J
(CF:EflIE| HELP EDIT SAVE PRINT
Creates a n»n input file for a Runoff Module
Da DX DY DI Efl_
42
43
44
ED
46
47
48
49
50
5i
F1
JX.
c-^
JO
54
55
56
57
59
-RUNOFF MODULE PREPROCESSOR-
This Lotus file aids in the creation of an input file which ii
coapatible with the FCSUM Runoff Module.
flt any tise, press flit H to display the eenu choices.
When entering data, use the arroH keys to uove the cursor into
the appropriate cell. Enter additional cards in successive colusns to
the right of the previously entered cards.
To begin a new file, press Alt il and select the CREfiTE aoife.
60 j
tl '
CttD
ULC
CAPS
Figure 4. Introduction screen with menu choices displayed.
86
-------
the parameter value is entered just to the right of a brief description of the
parameter. The SWMM parameter name is also given at the far left of the
screen. In cases where more than one of a given card type may be needed, such
as with subcatchment cards, multiple card values are entered directly to the
right of the initial card values. If the brief description of the parameter
given in this section is not adequate to choose a. value for a parameter or if
calculations need to be performed to obtain a value, the user may access the
help/reference section through the main menu as described in the next section.
Help
At any time during the parameter entry mode, the user may display the main
menu by pressing Alt M, and then choose "Help" (see Figure 6). The user will
then be prompted to enter a variable name whereupon a help file for that
variable will be displayed, or to enter "index" whereupon the entire help file
index will be displayed. From the help index the user will have access to
help files for all parameters which will include further documentation, calcu-
lation aids, graphs, and any other material with which the user may make a
logical and defensible parameter estimate for a specific job location. A
-SUBCATCHHENT DATA-
REPEAT GROUP HI FOR EACH SUBCATCMENT
NAXIHUN OF 100 DIFFERENT SUBCATCHHENTB FDR SINGLE EVENT SWHH3,
ICRA1N=0, AND 30 FOR CONTINUOUS SHIM3, ICftfllH NOT = 0.
A BLANK LINE JS NEEDED TO TERMINATE SUBCATCKHENT DATA (H2>
GROUP ID HI
JK Hyetograph nunber 1
HAKEH Subcatchient nuaber <»x 100) 10
NGTO Gutter or inlet liianhole) nusber for drainage. 11
Kim) Width of subcatchnent, ft. 5280
WAREA Area of subcatchoent, acres. 640
HH(3) 1 ioperviousness of subcatchnent 30.0
HSLOPE Ground slope, ft/ft. 0.001
XW.I5) Iipervious area. Resistance factor, 0.05
nU(6) Pervious area. (Manning's n) 0.3
SSTORE Impervious area. 0.05
WSTORE Pervious area. Detention storage, in. 0.2
WH(9) Maxiiun infiltration or capillary suction 3
NNtlO) fliniaus infiltration or hydraulic conductivity 0.3
HHtil) Decay rate or initial aoisture deficit 0.0015
Figure 5. SWMM Runoff module preprocessor, parameter
ent.ry mode.
87
-------
documentation file may come directly from the manual as is the file shown in
Figure 7. Once the user is finished viewing any specific help file, pressing
return will display a menu offering the choices of returning to the parameter
entry mode or returning to the help index. In returning to the parameter
entry mode, the cursor is moved to the parameter for which the last help
screen was viewed.
Edit
The edit mode is simply a short macro language program that takes advantage
of Lotus' capability to move through the worksheet. This mode allows the user
to update an old file which has already been created on the preprocessor, or
to go back to a previous card on the present file and make a change. The user
is able to move directly to any card in the module by choosing edit and moving
the cursor to a card name as in Figure 8. A user more familiar with Lotus and
the arrangement of SWMM cards will most likely be able to use the Lotus "goto"
or "pageup/pagedown" keys to access a card more quickly than with this edit
mode. The edit mode does, however, provide a list of the card types in their
respective order for the less experienced user to choose from.
DT63: MENU
CREATE (HELP! EDIT SAVE PRINT
Provides assistance in choosing values to be entered.
DS DT DU
58 CONTROL PARAMETERS
59
60 USE ALT M, HELP FOR MORE INFORMATION ON EACH VARIABLE
61
62 FIRST CONTROL GROUP (Bl)
63
64 GROUP ID B1
65 ICRAIN Continuous SHMN parameter, 0=single event, l-4=contin. 0
66 rETRIC Enter 0 for U.S.units, 1 for Hetric units 0
67 1SKOH 0=no snon, l=single event SUCH, 2=continuous sno* 0
68 HR3AG * of hyetographs, iax=6, §ust be 1 for continuous SHMM 1
69 IMFILM Infiltration eq., 0 = Horton eq, 1 = Green-Aupt eq. 0
70 KWALTY Duality lor erosion) sinulated? 0 = no, 1 = yes 0
71 WAP Evaporation data, OsdefaultlO.l'/day), l=read group Fl 0
72 KSF: Hour to start stora (24 hr clock, iidnight = 0) 12
73 Nr!K ilinute of hour to start stor« °
74 NMY Day of *onth to start situation 1
75 f.ONTH Month to start simulation 7
76 IYRSTR Year to start simulation 85
77 IRPfiNT Print control paraneter, for ICRAIN = 1 or 4 only 0
CMD CALC
Figure 6. Referencing help file from parameter entry.
-------
Save
The "Save" command, along with the "Print" command, do not initiate user
convenience modes as did previously described commands, but are formatting
procedures which must be performed to record the latest version of a parameter
file. The "Save" command is the most time consuming macro program in its
execution since it calls a series of macro subroutines which perform the
formatting procedure. The basic idea of this procedure is to divide the
aforementioned parameter cards, which were strung end to end across the top of
the worksheet, and arrange them vertically in their correct order. As was
described previously, the parameter cards were strung end to end so that each
parameter could be entered in its own column whose width had been adjusted to
accept the correct parameter format. In order to stack these parameter cards
while preserving the correct spacing between parameter values, it was decided
to combine all values on each card into a single worksheet cell. The task was
performed by printing each range containing a card type separately into a
temporary ASCII file, and then Importing these files back into the worksheet
in their proper order while stacking them vertically. This procedure is
slightly tedious and takes about ninety seconds for a simple version of the
Runoff Module. The time to perform this procedure will increase in proportion
EH22;
[RETURN j KIBEX
Returns to variable sntry position.
EN EX EY EZ
23
24
25
26
2?
26
29
3<<
31
32
33
34
35
36
37
38
39
40
41
FA
FB
FC
CND HEi.Li
FD
G3i5'* ICRftIN continuous Sltfil paraaeter
continuous Stifltl paraaeter
=0 Single event SHHH, continuous StWN not used
tmmit Values greater than zero indicate continuous 'SfMIt
=1 Hourly precipitation values read as card iaso-ss
froB National Heather Service (NWS) tape. Input unit
is JINU) for NHS tape.
=2 Proceesed hourly precipitation values (and temperature
if ISNOH =2) are read froa unit NSCfiAT (2). These
values were generated and saved froi earlier run when
ICRflHM or 4.
=3 Read precipitation values froa cards, using o/raups
El and E2. Not useable Hith snoMielt, i.e., loNQK mist
equal zero.
=4 Saie as 1CRAIN=1, except that program stops after
processing precipitation (and temperature) data. The
onlyy RUNOFF Black input paraneters required are those
needed for this processing. Input ceases aft er Group
Dl.
CfiLC
Figure 7. Example help file from manual with return menu,
89
-------
to the number of card types used, but not to the number of each card type
used. For example, adding erosion cards and snow cards to a Runoff file would
increase the run time by a few seconds each while adding an additional fifty
subcatchment cards would add a minimal amount of run time.
Print
While the "Save" command is the workhorse of the preprocessor, the "Print"
command takes all of the credit. Once the formatted listing of cards has been
reviewed on the worksheet as seen in Figure 9, the user may invoke the "Print"
command which will print the formatted-cards from the worksheet into an ASCII
file which may be accepted by a personal computer version of the Runoff Module
as input. As mentioned earlier, the input file will be named using the job
initials and job number entered at the beginning of the "Create" mode. Here
again, however, a more experienced user may perform this print step manually
on Lotus and provide another choice of file name.
STEPS TOWARD A COMPLETE PREPROCESSOR
The preprocessor which has been completed does not yet contain all cards
which may be run with the Runoff Module. Though it can only format the param-
eters for a simple problem, it does contain all of the basic components
READ*
DX
DY
D2
EA
EB
EC
ED
2"?
23
24
25
26
27
25
29
30
31
32
33
T J
OT
35
36
37
38
39
40
4!
***f*i»****i*»»r»«»ft*t EDIT NODE
HOVE CURSOR TO li'SBE OF CARD
INTERFACE
SCRATCH
MODULE
CONTROL ONE
CONTROL TKD
GENERAL SNQM
MONTHLY HIND
IHPERV DEPLETION
PERV DEPLETION
AIR TEMPERATURES
CONTINUOUS DATA
RAINFALL
»tt**t*tt*f »**>* H*m Httmt
YOU WISH TO EDIT AND PRESS RETURN.
RAINFALL HYETO
EVAPORATION
GUTTER PIPE
SL'BCATCH DATA
SUSCATCH SiiOH
GENERAL SL'ALITY
LAND USE
CONSTITUENT
EROSION CROUPS
SUBCATCH SURF
BUTTER PRINT
Figure 8. List of card types in Edit mode.
90
-------
needed in a preprocessor of this type. Once all of the card types have been
introduced into the Runoff preprocessor, the only major step is to write a
macro language program to delete cards that are not necessary for any partic-
ular run. Lotus provides all of the logistical functions which will be needed
for this task and no problems are foreseen at this point in time.
The next step is to create a preprocessor for each SWMM module. Here
again returns the limitation that only one module will be able to be intro-
duced on a single worksheet preprocessor. Each preprocessor, however, will
have the same basic skeleton as'the Runoff Module preprocessor and should
require less effort in its creation. Other improvements in this type of
preprocessor may include providing a copy of the entire SWMM manual to be
referenced through the help index. This would in theory free the user from
leaving the personal computer to access information. The eventual interfacing
of this preprocessor with an "expert system" would further this idea by pro-
viding the user with calculation alternatives to SWMM which may feed from a
common "knowledge base" of Information. This "knowledge base" may be thought
of as a general source of data from which any model could extract necessary
ft99:
CREATE HELP EDIT SflVE IPRINT 1
Gives a printout of the loraatted Runoff input file cards
ft B C D E F G H I J K L H N
CUD H£«U
0 P Q
99
100
101
102
103
104
105
106
107
108
109
no
111
112
113
114
115
116
U7
118
0 0
1 2
RUNOFF
ftlSFHMD
ft2RUNOFF
Bl 0 0
D2
El
E2 0.
E2 0,
0.
61
HI 1
0
HI 1
H2 11
0
0
0
0
0
0
0
0 0
0
0
0
0
000000
EXAMPLE
MODULE
0
60
24
11 0.
54 4.
12 0.
0
10
0
1
0
1
60
60
11 0.
28 1.
12 0.
0
0
30
13
09
12
0
11 5280
0
0
0
0
0
0
0
0.13
0.48
0.12
640
0
0
0
0
0
0.14 0
0.3
0
0
30.00.
0
0
0
12
0
.16 0
0.3 0.
0
0
001 0.
0
0
0
0
.2
18
0
05
0
0
0
1
0,2
0.18
0
0
0.3
0
7
0.27
0.18
0
0.05
0
0
0
0
85 0 0 0
0.34
0.18
0
000
0.2 3 0.3 0.0
0 O 0
0000
0000
EHDPRQGR
Figure 9. Formatted version of parameter cards.
91
-------
information for its simulation. Random access to different methods of param-
eter calculation and years of data history through the help index would all
lead toward rendering the user independent of conventional referencing methods.
CONCLUSIONS
1. The Lotus 1-2-3 software package is a suitable host for an input
file/reference file preprocessing package that can be interfaced with the EPA
Stormwater Management Model (SWMM).
2. The Runoff Module preprocessor which has been created is a simplified
package, but it reflects well a basic skeleton which may be used to create
preprocessing packages for the remaining modules of SWMM. No technical dif-
ficulties In creating these packages on Lotus spreadsheets are foreseen.
3. The preprocessor idea may be the first step toward rendering the personal
computer user independent from conventional referencing techniques.
4. The "Create", "Help", and "Edit" modes of this preprocessor are suf-
ficiently documented for easy use by a beginner with a minimal knowledge of
Lotus 1-2-3 and SWMM, but do not hinder their use by an expert.
REFERENCES
1. Dennlson, K.D. and James, W., 1985, A Database Environment and Sensitivity
Framework For a Continuous Water Quality Models, In: Proceedings of Confer-
ence on Stormwater and Water Quality Management Modeling, McMaster University,
Hamilton, Ontario.
2. Huber, W.C., Heaney, J.P., Nix, S.J., Dickinson, R.E. and Polmann, D.J.,
1981, "Stormwater Management Model User's Manual, Version III," EPA-600/2-84-
109a (NTIS PB84-198423), Environmental Protection Agency, Cincinnati, OH.
3. James, W. and Robinson, M., 1985, PCSWMM3.2 User's Manual, Hamilton,
Ontario.
4. Lotus Development Corporation, 1984, Lotus 1-2-3 User's Manual, Release
1A, Cambridge, MA.
5. South Florida Water Management District, 1983, "Permitting Information
Manual - Volume IV, Management and Storage of Surface Waters," SFWMD, West
Palm Beach, FL.
6. South Florida Water Management District, 1985, Program for Generating
and/or Routing Runoff Hydrographs, SFWMD, West Palm Beach, FL.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
92
-------
IMPACT OF EXTENSIVE IRRIGATION PUMPAGE ON STREAMFLOW BY HSPF
By
Ananta K. Nath
Nebraska Natural Resources Commission
Lincoln, Nebraska 68509
ABSTRACT
The hydrology of a 2,700 square mile area of the Big Blue River Basin in
central Nebraska was simulated by the continuous process hydrologic simulation
program HSPF to investigate the impacts of extensive groundwater pumpage for
irrigation on streamflows. A long-term continuous simulation for a 23-year
period (1953-1975) was carried out by modeling water movement and storage
characteristics through the use of the PWATER section of PERLND module and HYDR
section of the RCHRES module of HSPF. The simulation indicated that surface
runoff and interflow from irrigated lands increase initially. Actual
evapotranspiration also increases. As pumping continues and the aquifer water
level drops, groundwater contribution to streamflow decreases. Finally, a
stable condition is achieved where the water table ceases to reach the stream
bed and measured streamflow is less than before pumping began. It is concluded
that HSPF can be used with a certain degree of success to simulate
stream-aquifer interaction and to assess possible effects of deep pumping on
surface water hydrology in a Nebraska basin. The accuracy of the results,
however, is limited by modeling complications because some algorithims of HSPF
were not specifically designed to simulate withdrawal of groundwater by pumping.
Coordination of some HSPF input parameters from groundwater modeling may produce
better results.
INTRODUCTION
Development of irrigation using groundwater started growing rapidly in the
1950's and expanded very rapidly in the 1970's in Nebraska. By 1980, more than
73 percent of water used for irrigation in the state was pumped from wells.
This development, however, has not been without consequences. The impact of
extensive groundwater pumpage for irrigation was noticeable in the flow regimen
of several streams, including the Big Blue River basin in central Nebraska. As
a part of the Nebraska Natural Resources Commission's (NRC) "Problem Analysis
and Area Planning" activity for the Big Blue River Basin, an attempt was made to
investigate the impacts of groundwater pumpage on the streamflows of the Big
Blue River and its tributaries using HSPF. This paper summarizes the
methodologies used in simulating by HSPF the effect of the time history of
93
-------
groundwater pumpage on the hydrology of the basin, arid the results obtained from
the simulation. It also evaluates the applicability of HSPF in simulating deep
groundwater pumpage and resulting groundwater-surface water interrelationships.
DESCRIPTION OF THE STUDY AREA
The Big Blue River Basin is located in south-central Nebraska and northeast
Kansas, and is a part of Kansas River Basin. A map of the basin with principal
tributaries is shown in Figure 1. The area of the Nebraska portion of the basin
totals about 4,600 square miles. The topography ranges from loess plains with
thick deposits of silt, sand, and gravel, to rolling hills and stream valleys.
The principal aquifer consists of unconsolidated sand and gravel layers of
Pleistocene age. These layers belong to three formations composed of sand and
gravel graded upward from coarser to finer material. Soils are of loess origin.
About sixty percent of the soils in the basin are of irrigation suitability
type A {i.e., with slight limitations only).
Land use is primarily agricultural. About 83 percent of agricultural lands
are croplands. Since the early 1950's the number of irrigated acres using
groundwater has increased every year. By 1980, approximately 1,061,000 acres in
the basin were irrigated. The greatest concentration of irrigation wells is in
four counties in the upper part of the basin. This intensive groundwater
development, coupled with occassional drought conditions, has caused widespread
water level declines. Water levels have declined as much as 30 feet in several
areas.
Upper reaches of the streams are characterized by small meandering channels
with intermittent flows. The streamflow is variable, being primarily derived
from precipitation runoff. Nearly all of the tributaries are intermittent.
Even the main stem of the Big Blue River does not become a perennial stream
until it drains about 450 square miles of the upper part of the basin. Base
flows in the streams are, however, relatively low. Return flows from increased
well irrigation have lengthened the period of time when flows are present in
many of the streams.
SELECTION OF SIMULATION MODEL
As a part of the problem analysis and area planning study of the basin
water supply, a two-dimensional finite difference groundwater model covering
most of the basin was developed. A large data base including irrigated lands
identified by remote-sensing, geologic data, and groundwater conditions was
generated for development of the groundwater model. However, since the
groundwater model could not accurately represent the effects of the continuous
changes on the surface water hydrology, it appeared that a long-term continuous
model was needed to simulate the hydrologic-hydraulic behavior of the stream
aquifer system prior to and during the period of extensive irrigation pumpage.
A search was made for an integrated and comprehensive surface water hydrology
model that would allow comparison of water movement, storage characteristics and
hydrographs for conditions before and after development of irrigation. Based
upon these requirements, the program package Hydrologic Simulation
94
-------
IO
en
FIGURE 1
BIG BLUE
RIVER BASIN
BASIN STREAM NETWORK
SCALE
-------
Program-Fortran (HSPF), originally developed by Hydrocomp, Inc. and revised by
Anderson-Nichols and Company for the EPA's Environmental Research Laboratory,
was chosen for this project. The project was started with Release 5.0, and
later updated with Release 7.0.
DATA BASE DEVELOPMENT
Data for time series storage management of HSPF consisted of meteorological
data, land data, channel-flood plain data, streamflow data, riverine structure
data and groundwater pumpage data. Meteorological data consisted of
precipitation, air temperature, dew point, solar radiation and wind movement.
Precipitation data were available at five hourly and 21 daily recording
stations. Daily maximum-minimum temperature data were available at three
stations. Daily evaporation, solar radiation, dew point and wind movement data
were available at one station each. Daily precipitation and temperature records
were disaggregated to hourly values. Daily groundwater pumpage data were
incorporated into the time series storage from the input data of the groundwater
model already developed for the basin. In the groundwater model, the seasonal
Crop Irrigation Requirements (CIR) for various crops and soil types in each node
were first computed by the Jensen-Haise methodology. The CIR was then adjusted
for irrigation efficiency to estimate the seasonal pumpage and distributed over
the irrigation months. The acre inches per acre of pumpage from the GW model
nodes were then disaggregated to daily values over the PERLND segments of the
HSPF model.
Daily streamflow data for calibration were available at the following USGS
gaging stations:
1) Big Blue River at Surprise (April 1964 - present)
2) Lincoln Creek near Seward (October 1953 - present)
3) Big Blue at Seward (October 1953 - present)
4) West Fork Big Blue near Dorchester (October 1958 - present)
5) Big Blue at Crete (October 1954 - Present)
Pervious hydrologic land segments were selected based on the unique combination
of meteorological characteristics identified by a Thiessen polygon network of
weather stations and land characteristics by soil type, topography and cover. A
total of 38 pervious hydrologic land segments were delineated for simulation
(Figure 2).
SIMULATION ALGORITHMS
The hydrologic simulation of the basin was conducted by the utilization of
the PWATER section of the PERLND" module. Water movement in the PERLND module
was modeled along three flow paths: overland flow, interflow and groundwater
flow. Storage processes that occur on the land surface and in the soil horizons
were simulated in six storage blocks: Interception Storage (CEPS), Surface
Storage (SURS), Interflow Storage (IFWS), Upper Zone Storage (UZS), Lower Zone
Storage (LZS), and Active Groundwater Storage (AGWS).
96
-------
Figure 2
BIG BLUZ RIVER BASIN
HYDAOLOOIC LAND SEGMENTS
IHCOftPORATED IM THE Bl6 «l_UE
WATCH SUPPLY UOOCi.
LE6ENO
WF3 ML I lOtHTVTIMI
A imtUIFKW MflE
(CM.IMATIOH
\
C » « itil< 1»»HMTO n M
-------
The processes of infiltration and overland flow are interdependent and
occur simultaneously. Water in surface detention will later infiltrate and
appear as interflow or it can be contained in the upper zone storage. Water
infiltrating through the surface and percolating from the upper zone storage to
the lower zone storage may flow to active groundwater storage, or may be lost by
deep percolation to inactive groundwater storage. Active groundwater eventually
reappears as baseflow, but the deep percolation to-inactive groundwater storage
is not accounted for in the simulation. Lateral external inflows to interflow
and active groundwater storage were also considered in simulation.
Evapotranspiration is simulated in all phases of the storages associated with
the process, i.e. from interception storage, upper and lower zone storages,
active groundwater storage, and directly from baseflow.
Hydraulic simulation for routing runoff from land surface and discharge
from groundwater though the stream system was performed by utilization of the
HYDR section of the RCHRES module. Simulation reaches were identified based on
homogeneity of cross-sectional shape, channel slope and channel-floodplain
roughness coefficients. Detailed surveyed data on channel-floodplain cross
sections and riverine structures on the main stem of Big Blue and West Fork Big
Blue were available from the Natural Resources Commission Flood Plain Studies.
Stage-discharge information computed by HEC-2 program in NRC Flood Plain studies
were also utilized in generating FTABLES.
In applying the model on this particular project, the basic fluxes without
and with the dynamic effect of groundwater pumpage on streamflow are illustrated
in Figure 3. Streamflow comes from surface runoff and interflow, controlled by
soil moisture storage, and from groundwater flow. For the conditions prior to
extensive irrigation development, the solid line linkages of Figure 3 represent
the hydrologic processes that existed in the basin. When irrigation pumpage was
introduced, another flow path was added, which is shown as the dashed line of
Figure 3. This flow path can be represented by various ways in HSPF, but the
dashed line in Figure 3 seemed to be a good approximation of actual conditions.
The layout of a typical pervious land segment incorporated into the Big Blue
model to represent the interrelationship of surface and groundwater flow
components under groundwater pumpage condition is illustrated in Figure 4.
METHOD OF INVESTIGATION
The primary strategy used in this study to evaluate the impact of deep
groundwater pumpage on strearaflow was to simulate the long-term
hydrologic-hydraulic behavior of the basin for the conditions prevailing prior
to the irrigation development conditions, and project that simulation through
the period undergoing irrigation development. The hydrologic-hydraulic behavior
of the basin was then simulated for the actual conditions prevailing in the
basin within the period of irrigation development by incorporating the
groundwater pumpage element into the model. Superimposing the series of
hydrographs representing pre-irrigation conditions on the series representing
irrigation development conditions would show the continual effect of streamflow
resulting from the extensive withdrawal of groundwater.
98
-------
CALIBRATION
Marty of the algorithms contained in the model are mathematical
approximations of complex natural phenomena. Before the model could be used to
reliably simulate streamflow behavior under various conditions, it was necessary
to calibrate the model by comparing simulation results with measured historical
data. If significant differences were found, parameters were adjusted to
calibrate the model to the specific natural and man-made features of the basin.
The model was calibrated for the conditions prevailing during two periods.
The first was prior to extensive groundwater pumpage, and the second was during
the period of the greatest irrigation development.
Though the development of irrigation in the basin began to take place in
1948, total development was minor compared to that in the late 1960's and the
1970's. Also, available continuous streamflow records for almost all of the
gages in the Big Blue basin do not start before water year 1954, preventing
calibration of the model prior to 1954. It was, therefore, decided to use
several periods between 1954 and 1960 for pre-irrigation development conditions,
and periods between 1960 and 1975 for irrigation development conditions to
calibrate the models of different stream networks, depending on the avilability
of measured streamflow records for that stream. The stream segment for the Big
Blue at Surprise, however, could not be calibrated for pre-irrigation
development conditions as recorded streamflow data at this gage are available
from only after October 1964. The months for which the model was calibrated are
as follows:
1. Big Blue at Surprise:
2. Big Blue at Seward:
October 1966-September 1969
October 1953-September 1955 and
October 1966-September 1968
(U
co
CO
Q)
4-1
I
o
Precipitation
Evaporation
Soil Moisture
-^-Surface Runoff
Interflow
Infiltration
Groundwater Storage
Groundwater
to Stream
Deep Percolation
Deep Groundwater
Figure 3 - Basic fluxes in HSPF simulation with groundwater pumpage.
99
-------
Figure
PBEOP
LAYOUT FOR A TYPICAL PERVIOUS LAND SEGMENT
INCORPORATED IN THE HSPF MODEL FOR THE BIG BLUE
3. Lincoln Creek at Seward:
4. Big Blue near Crete:
5. West Fork Big Blue:
October 1953-Septeraber 1955 and
October 1964-September 1967
October 1953-September 1955 and
October 1964-September 1967
October 1958-September 1959 and
October 1964-September 1967
The first step in the calibration procedure was to match the annual flow
volumes between simulated and recorded flows. Parameters related to the storage
capacities of the unsaturated soil zones, evapotranspiration and infiltration
are important determinants of runoff and volumes. Guidance for initial values
of such parameters as lower zone moisture storage (LZSN), upper zone nominal
storage (UZSN), infiltration (INFILT), lower zone evapotranspiration (LZET),
interflow (INTFW), and groundwater recession (AGWRC) were taken from the
suggested values in the Agricultural Runoff Model Users Manual (Donigian et al,
1978). Some guidance on parameter values was also taken from a concurrent HSPF
project on the Dee Creek Watershed in southeastern Nebraska conducted by the
Water Resources Research Center of the University of Nebraska (NWRC, 1982). The
UZSN, INFILT, and LZET parameters were varied monthly. In calibrating
parameters for annual flow volumes, adjustments were made in LZSN and INFILT.
100
-------
Since LZSN and INFILT indirectly affect the actual evapotranspiration rates,
decreasing LZSN and INFILT makes less water available for ET losses and
increases runoff. In simulating the daily and seasonal low flow volumes,
adjustments were primarily made in INFILT and AGWRC parameters. AGWRC controls
groundwater outflows, while INFILT controls inflow to the groundwater. However,
care needs to be taken in adjusting INFILT at this step. Since INFILT is
indirectly related to LZSN, significant change of INFILT affects the long term
flow volumes calibrated in the previous step. Consequently, most of the
adjustment needs to be in the AGWRC term. Once the annual flow volumes and
seasonal low flows are adequately calibrated, the shapes of the hydrographs were
adjusted by varying the interflow inflow parameter (INTFW). Some of the salient
calibration parameters for each of the stream segments simulated are summarized
in Table 1.
In calibrating the model for the two periods with and without irrigation
development, consideration was given to the change in evapotranspiration and
interception rates from irrigated and nonirrigated lands. ET is affected
considerably by the type of plant and plant density. Also, loss by interception
depends on the extent of vegetal cover and density of prevailing plants and
trees. It was, therefore, necessary to numerically represent the difference in
monthly INTERCPT and LZETPARM between irrigated and nonirrigated conditions.
The relationship between water loss and crop yield can range from linear to
curvilinear response functions (ASAE, 1980). A 1:1 correlation of ET loss to
crop yield was assumed based on a linear relationship of dry matter yield to
relative seasonal evapotranspiration. Since corn is the major irrigated crop,
the value of MON-LZETPARM and MON-INTERCEP for the months of June through
September under irrigated conditions were generated by prorating the ratio of
irrigated crop yield to nonirrigated crop yield. The results of a typical
calibration run are illustrated in Figure 5.
VERIFICATION
Verification of the hydrologic calibration was carried out by running the
models of each stream for the period October 1953 through September, 1960 for
the pre-irrigation development conditions, and for the period October 1960
through September 1975 for the irrigation development conditions. Figure 6
illustrates graphically the results of typical verification runs for a simulated
stream.
RESULTS OF THE CALIBRATION AND VERIFICATION PROCESS
The calibration process was particularly valuable in assigning values to
prominent land parameters which were seen to be dependent upon soil type, land
cover and on regional meteorological characteristics. The parameter sensitivity
runs indicate that simulated flows were most sensitive to the values of LZSN,
UZSN, and INFILT. In certain cases, the impacts of the monthly values of UZSN,
MON-LZETPARM, and MON-INTERCEP on monthly distribution of runoff were more
pronounced. In all the streams simulated, the first attempt to calibrate
included periods of high runoff extending over a period of about three months,
it was then extended to include longer periods of up to three years. The
101
-------
simulated monthly runoff volumes varied from the recorded*volumes by 10 to 20
percent for the wet months and by 10 to 50 percent for the drier months. For
the extended three-year calibration periods, the simulated annual runoff volumes
varied from 7 to 15 percent of the recorded volumes. The correlations between
measured and simulated flows were evaluated by plotting double mass curves of
cumulative runoff volumes and comparing them with a line of one-to-one
correspondence. A typical double mass curve of measured and simulated flows is
illustrated in Figure 7. The noticeable difference between recorded and
simulated annual flows resulted from relatively poor simulation of a few months
in which measured flows in the drier months were lower than simulated flows.
This is particularly evident during the initial months of simulation in which
the parameters representing the initial conditions may be responsible for poor
TABLE 1
SUMMARY OF CALIBRATED PWATER PARAMETERS
PARAMETER
LZSN
LZS
INFILT
IRC
INFILD
AGWRC
DEEPFR
AGWETP
IFWS
SURS
UZS
UZSN*
INTERCEP*
LZETP*
Big Blue
at Surprise
8.0
4.0
0.02
0.60
2.0
0.99
0.19
0.5
0.01
0.10
4.0
0.13-0.32
0.03-0.30
0.10-0.70
Lincoln Cr.
at Seward
6.0
3.5
0.01
0.60
2.0
0.999
0.19
0.5
0.01
0.10
4.0
0.13-0.29
0.03-0.30
0.05-0.42
Big Blue
at Seward
6.0
3.5
0.01
0.60
2.0
0.999
0.19
0.5
0.01
0.10
4.0
0.13-0.30
0.03-0.30
0.05-0.42
West Fork Big Blue
at Dorchester at Crete
6.0
3.5
0.015
0.40
2.0
0.999
0.15
0.35
0.01
0.10
4.0
0.05-0.15
0.03-0.20
0.03-0.30
4.0
6.0
0.01
0.20
2.0
0.99
0.10
0.30
0.01
0.10
4.0
0.19-0.29
0.01-0.30
0.03-0.70
* Varied monthly, only ranges shown in table.
LZSN - Lower zone nominal storage, inches
LZS = Initial lower zone storage, inches
INFILT = Mean infiltration rate index, in/hr
IRC = Interflow recession rate, per day
INFILD = Ratio of maximum to mininum infiltration rate
AGWRC = Active groundwater recession rate, per day
DEEPER = Percentage of groundwater to deep aquifer
AGWETP = Fraction of ET from active groundwater storage
IFWS - Initial interflow storage, inches
SURS = Initial surface detention storage, inches
UZS = Initial upper zone storage, inches
UZSN = Upper zone nominal storage, inches
INTERCEP = Interception storage capacity, inches
LZETP = Lower zone ET parameter
102
-------
simulation. In certain cases readjustments of interflow parameters and
roughness coefficients resulted in closer runoff volumes and improved timing of
peaks. The difference between observed and simulated flows in verification runs
was somewhat greater than in calibration runs because in calibration greater
emphasis was given to higher than normal runoff events, while the verification
period included both high and low periods of flows. Consequently, the
calibrated parameters were somewhat biased toward high and moderate flow
conditions.
The timing of snowmelt events in March and April was not well represented.
This is probably due to both the use of several precipitation data series with
data synthesized from other stations to substitute for missing records, and the
inability of the model to accurately represent frozen ground conditions. The
INFILT and UZSN parameters were reduced during the winter months, and the snow
input (SNOWCF-parameter) was increased to match the volume of measured surface
runoff. Additional research and model development work is needed to better
understand and represent frozen ground conditions and the timing and volume of
snowmelt runoff.
Some erractic behavior of the model in simulating runoff for some major
runoff producing storm events is primarily attributed to the characteristics of
those storm events. Review of precipitation data from several stations revealed
the frequent occurrence of scattered convective storms. These convective storms
Figure 5
RECORDED AND SIMULATED HYDROGRAPHS
FOR BIG BLUE RIVER AT SEWARD
PRE-IRRICATIOM DEVELOPMENT CONDITIONS CIOSO
259 B
22SO
2080
CO
LL.
O
use
CO
Q 1888
^ Tso
580
250
RECORDED
SIMULATED
JlllllllMlllllltl^J/lVl M
18 IS 28 25 30 5
JUNE
10 IS 2Q 25 SB
JULY
10 IS 20 25 30
AUGUST
103
-------
Figure 6
RECORDED AND SIMULATED ANNUAL
RUNOFF VOLUMES FOR
BIG BLUE RIVER AT CRETE
RECORDED
SIMULATEDO
S4 EE EB 67 68 60 68 61 62 63 64 65 OB 67 88 OB 70 71 72 73 74 75
* 4
UATER YEAR
result in very nonuniform distribution of rainfall over a relatively small area.
These types of storms, which produce localized flooding, are numerous in the Big
Blue basin. For such storms, adequate representation of the rainfall-runoff
relationship was not achieved because there were not enough weather stations in
the model area to accurately measure the rainfall. Correcting this deficiency
would require more precipitation gages in order to better represent and
understand the rainfall-runoff relationship.
In simulating deep aquifer pumpage, the available algorithms of HSPF can
not directly add or subtract specific volumes of water to or from a deep
aquifer. This is accounted for by the introduction of the parameter DEEPFR, the
fraction of groundwater lost to deep aquifer. The value of DEEPFR was
arbitrarily lowered for the pre-irrigation development condition without the
availability of any physically interpreted value. Secondly, the surface
water-groundwater interaction in HSPF is accounted for by the use of a time
series of continuous lateral outflow from active groundwater storage. A time
series of active groundwater inflow AGWI was used, but the impact of external
lateral groundwater inflow AGWLI, which could also have continuous infow, was
not accounted for. A percentage of the pumpage rate could have been used in the
time series to represent the possible increased leakage of water into the deep
aquifer. Consideration should have been given to linking these values of AGWLI
and DEEPFR with output from the groundwater model.
104
-------
Figure 7
DOUBLE MASS CURVE OF RECORDED AND
SIMULATED ANNUAL FLOW VOLUMES
L_
O
UJ
r-
H
CO
i eee
oee
sea
700
see
see
400
300
200
iee
LINCOLN CREEK AT SEWARD
100 268 300 400 6B8 000 70O BOO 800 1000
RECORDED FLOW CCFS>
OBSERVATIONS
Groundwater pumping has indeed had a significant effect on the baseflow
contributions to the streams in the basin. As pumping caused the water table to
decline, it decreased the former gradient towards the stream which in turn
decreased the discharge of the aquifer to the stream, or it reversed the water
table gradient between the aquifer and the stream, which induced streamflow to
seep to the underlying aquifer. These effects did not instantaneously reach the
streams, but rather lagged behind the operation of pumpage depending upon
aquifer properties and distance from the wellfield to the stream.
105
-------
The trend observed in the simulated hydrograph series as illustrated in
Figure 8 shows that the effect of irrigation pumpage was to increase surface
runoff and interflow from irrigated lands initially. Actual evapotranspiration
also increased. Initially, the effect of pumping is a net increase in the
quantity of streamflow due to surface runoff contribution from irrigated water.
This is apparent, as indicated by the early stages of the hydrographs, in almost
all of the streams. As pumping continues and the aquifer water level drops, the
deep percolation from the active groundwater storage that feeds streams
increases and no longer contributes to streamflow. Finally, a new steady state
balance is achieved where measured streamflow is less than before pumping began.
For instance, in Lincoln Creek, as illustrated in Figure 8, the wasted runoff
from irrigated lands increased the simulated average flows in the months
of August and September in 1970 by approximately 5 cfs. However, the reduction
in baseflow contribution due to aquifer level draw-down decreased the average
monthly flow by 14 cfs in October and by 2 cfs in November.
CONCLUSION
An attempt was made in this project to test the applicability of HSPF in
investigating the impacts on streamflow from deep groundwater pumpage and
provide some informative insights and conclusions regarding modeling of surface
and groundwater interactions. Certain phases of the calibration and
verification process in this project, however, could not produce the desired
results. Therefore, suggested improvements need to be discussed to provide
guidance for future use or development of the model. Some of the specific
conclusions derived from this study are:
(a) HSPF can be used with a certain degree of success to simulate
stream-aquifer interaction and possible effects of deep pumping on surface water
hydrology in a Nebraska river basin. The accuracy of the results, however, is
somewhat limited by the modeling complication because certain algorithms used in
HSPF are not specificaly designed to simulate withdrawal from groundwater by
pumping.
(b) Since HSPF can not directly add or subtract specific volumes of water
to or from the deep aquifer, accurate estimation of the parameters such as
DEEPFR, IGWI, AGWI, and AGWS, that play dominant roles in accounting for storage
and movement of groundwater, is very important in successful simulation of
groundwater pumpage impacts. If possible, the values of such parameters should
be estimated from, or determined by, a groundwater model.
(c) Meteorological time series data, particularly precipitation, is
critical to effective continuous hydrologic simulation. A denser network of
precipitation records is necessary for accurate representation of the
hydrometeorologic condition of a basin. Lacking this data, it is better to
compare duration and frequency information from the overall simulation rather
than data on an individual storm event.
(d) Equally important is the adjustment of infiltration parameters during
winter and early spring to represent frozen ground conditions for simulation of
106
-------
o
_l
LL.
7S
70 h
6S -
68 |-
55
50
15
40
35
30
25
20
15
to
5
0
CONTINUAL EFFECT OF GROUNDWATER PUMPAGE
ON STREAMFLOW 1970
LINCOLN CREEK NEAR SEUARO
WITH CU PUMPAGE
WITHOUT GU PUMPAGE
Illn ill in'iln ill in ill
5 10 IS 20 25 30
AUGUST
5 18 15 20 25 30
SEPTEMBER
5 te 15 20 25 30
OCTOBER 1
5 IB IS 20 25 30
NOVEMBER
-------
snowmelt runoff, and during summer to represent the degree of antecedent
saturation condition for simulation of runoff from intense storm events.
However, it was observed that while trying to adjust the parameters for
reconstitution of high runoff events, some parameters become biased towards high
and moderate flows, and do not respond to low flow events. Care should be taken
in adjusting such sensitive parameters.
(e) Effectiveness of continuous process hydrologic-hydraulic modeling is
affected by the size of the hydrologic land segment simulated. Due to the large
size of the Big Blue Basin, the hydrologic land segments had to be relatively
large in order to reduce computer costs. It was, therefore, felt that HSPF
would be more effectively used to model a medium-sized watershed.
REFERENCES
1. American Society of Agricultural Engineers. Design and Operation of Farm
Irrigation Systems, ASAE Monograph Number 3, 1980.
2. Donigian, A.S., Jr. and Davis, H.H., Jr. User's Manual for Agricultural
Runoff Management Model, EPA-600/3-78-080, 1978.
3. Donigian, A.S., Jr., Imhoff, J.C., Bricknell, B.R., and Kittle, J.L., Jr.
Application Guide for Hydrological Simulation Program-Fortan,
EPA-600/3-84-065 , 1984.
4. Johanson, R.J., Imhoff, J.C., Kittle, J.K., and Donigian, A.S., Jr.
Hydrologic Simulation Program-Fortan: Users Manual for Release 8-0.
EPA-600/3-84-066, 1984.
5. Nebraska Natural Reosurces Commission. Report on Big and Little Blue River
Basin Area Planning Study, Technical Appendix A-Development, Calibration,
Verification of Groundwater Models, 1983.
6. Nebraska Water Resources Center. Development of State Water Quality
Management Plan for State of Nebraska, Institute of Agriculture and Natural
Resources, University of Nebraska-Lincoln, 1982.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
108
-------
THE USE OF SWUM TO PREDICT RUNOFF FROM NATURAL WATERSHEDS IN FLORIDA
BY
Wayne C Downs
Jon Paul Dobson
Raymond E. Wiles
GeoScience Inc.
Gainesville, Florida
ABSTRACT
A large Florida watershed, the Deer Prairie Slough basin, was selected
to test the ability of the EPA Storm Water Management Model, Version III,
(SWMM III), to predict runoff from natural and undeveloped watersheds. The
Deer Prairie Slough watershed was chosen for calibration because it has a
USGS maintained stream gage, nearby National Weather Service raingage, and
nearby USGS monitored wells. Also, the site is near to a GeoScience
project in Sarasota County which is similar in size.
The model was calibrated to the measured runoff for the test area using
site specific hydrologic data, groundwater conditions at the time of the storm
event, and hourly rainfall data. A second recorded storm event was selected
to verify SWMM's ability to predict measured runoff once calibrated. The
model configuration for the calibration simulation was left unchanged in the
verification procedure with the exception of new hourly rainfall values and
new antecedent moisture conditions.
The modeling results show SWMM III to be quite accurate in predicting
both peak runoff rates and runoff volumes. The relatively simple modeling
strategy combined with SWMM's powerful runoff, storage, and routing features
Indicate the model to be a very effective tool in predicting runoff rates and
volumes in Florida.
109
-------
INTRODUCTION
The US Environmental Protection Agency (EPA) Storm Water Management Model
(SWMM) is a comprehensive, physically based mathematical model designed to
evaluate stormwater quantity and quality. The model has been widely used in
the United States, Canada, and Western Europe. It is supported by the USEPA,
and is updated and maintained at the University of Florida, Department of
Environmental Engineering Sciences.
The principal use of the model has been to evaluate stormwater runoff
from urban sewered systems. However, the rainfall, infiltration, evaporation,
overland flow, channel routing, and detention storage algorithms used in SWMM
are not peculiar to urban sewered catchments, but are basic to any stormwater
analysis. These features, in fact, make SWMM well suited for the evaluation
of before and after development runoff peak flows and volumes as required by
development permitting regulations of the State of Florida. Flow attenuation
by detention storage and channel routing through second and third order net-
works are two particularly strong features of the SWMM TRANSPORT module.
The purpose of this study was to determine SWMM's modeling accuracy on
pre-development sites in Florida. Of particular interest was the model's
ability to simulate large watersheds as is commonly required in large develop-
ment projects and regional or county-wide drainage studies.
SITE AND DATA DESCRIPTION
For this work, a large undeveloped watershed was selected which has a
history of recorded stream gage data monitored by the USGS. This basin is
Deer Prairie Slough located in the southern portion of Manatee County and the
northern portion of Sarasota County as shown in Figures 1 and 2. The water-
shed is composed of 27,016 acres of flat terrain interspersed with natural
depressions and wetland areas. Shallow groundwater conditions and sandy soil
types are characteristic of the area.
Rainfall data were taken from the Venice, Florida station, within ap-
proximately 10 miles of the site. The data are recorded in hourly increments
beginning in 1942 (NWS).
From this rain station a single storm event recorded June 17 and 18,
1982, was selected for calibration to the site. This 8.0 inch event over a
24-hour duration approximates the design storm required by Sarasota county of
a 9.5 inch, 24-hour event.
A storm event recorded September 12 through 26, 1982 was chosen to verify
the modeling results. This event occurred over 109 hours and contained a
depth of 6.6 inches (NWS). The major portion of the rainfall took place over
the first 26 hours.
Ground water conditions were simulated from data taken from the Big Slough
Shallow Well, ROMP 19ES, and ROMP 19WS shallow USGS monitoring wells near the
site (USGS, I982b). The well locations are shown in Figure 2.
110
-------
Data for the June 17 and 18, 1982, storm event used for the calibration
of the site, showed groundwater conditions at an average depth of 2.46 feet
below surface as recorded on June 15. For the verification trial, groundwater
data measured on September 20, 1982, indicated an average depth of 1.96 feet
below surface.
Figure 1. Location of the Deer Prairie Slough Watershed in Florida.
Ill
-------
METHODOLOGY
Proper simulation of runoff events in much of Florida requires an under-
standing of the effect of the ground water table on surface flow peaks and
volumes. Large storms, such as those approximating so called "design storms",
generate overland flow by filling the void space in the soil until the water
table rises to the surface. The hydraulic conductivities of Florida soils in
-
; i ^-' i[~.j>i ||ii.- i.
*; I [l&'^'-lftspiFW- \
H^rfSB^l&b '
V-r
Figure 2.
Location of Shallow Wells with respect to
Deer Prairie Slough.
112
-------
these high water table areas are greater than the peak intensities of even
design sized storms. Thus, runoff is generated by excess rainfall after the
water table has risen to the surface, not by high intensity rainfall on soils
of low hydraulic conductivity.
This condition may be simulated by modeling each sub-basin as a shallow
reservoir which must be filled before outflow, expressed as overland flow, can
occur. The simulation is done in the SWMM RUNOFF module by setting the Horton
or Green-Ampt infiltration parameters to low values, essentially turning them
off. Depression storage over the area is then employed as the reservoir which
must be filled before overland flow can occur.
Determination of the proper amount of depression storage is a function of
depth to water table and water storage capacity above the water table. Depth
to water table information may be obtained from the US Geological Survey Water
Resources Data Publications, the Water Management Districts or local data
sources. Soil moisture storage information is available in publications such
as Characterization Data for Selected Florida Soils available from the Insti-
tute of Food and Agricultural Sciences at the University of Florida (Carlisle,
et al., 1981). The South Florida Water Management District (SFWMD) recommends
the following typical values for storage above the water table in most South
Florida soils (SFWMD, 1984):
Depth of Water Table (ft) Storage (in)
1.0 0.6
2.0 2.5
3.0 6.6
4.0 10.9
These SFWMD values were plotted to aid in interpolation of soil moisture
storage values (Figure 3).
The watershed was divided into six sub-catchments along topographic
features and flow patterns. Figure 4 indicates the sub-basin delineations.
Sub-catchment widths were determined for model input by taking cross section
widths perpendicular to the direction of flow at the sub-catchment boundary.
Areas, widths and slopes were taken directly from topographic maps. Depth to
water table was determined by averaging values from wells near or in the
watershed, and water storage values were determined from the SFWMD curve.
These values were input as pervious depression storage. Wetland areas were
planimetered and input as containing 1.0 inch of depression storage within
the wet season hydroperlod elevations.
Flows were routed using the SWMM TRANSPORT module. Channels were
simulated as broad trapezoidal channels having side slopes typical of the
topography along the natural channels. A Manning's roughness value of
0.35 was used for overland flow to simulate undeveloped conditions
(Huber et al., 1981). Channel lengths and bottom slopes were taken from
the topographic maps. The watershed outflow hydrograph was plotted at
113
-------
one-half hour intervals. On the same hydrograph for comparison purposes
were plotted the measured data from the USGS gauge.
DISCUSSION
The 8.0 inch, 24 hour storm of 17 June 1982 was used in the calibration
of the outflow from Deer Prairie Slough. The only calibration parameter
used to adjust the predicted outflow was the sub-catchment width. The final
values used were similar to measured topographic values. The peak outflow
for the storm was the largest of the year for the gage and was recorded as
387 cfs, instantaneous (USGS, 1982a). The simulated peak was also
calibrated to 387 cfs. Table 1 lists pertinent data from this calibration
12.00
10.00
8.00
6.00
ct
I
4.00
2.00
0.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3,50 4.00 4.50
DEPTH TO WATO< TABLE (ft)
Figure 3. Soil Moisture Storage as a Function of Depth to
Water Table (as per SFWMD, 1984).
114
-------
simulation. The volume calibrations performed within the model indicated a
measured volume of 2.85 inches compared to a predicted volume of 2.51
inches. The measured and predicted hydrographs are shown on Figure 5.
Table 1: DEER PRAIRIE SLOUGH CALIBRATION DATA
WATERSHED SIZE:
AVERAGE DEPTH TO WATER TABLE:
TOTAL PRECIPITATION:
STORM DURATION:
PREDICTED PEAK:
MEASURED PEAK:
PREDICTED VOLUME:
MEASURED VOLUME:
27,060 acres
2.46 feet
8.0
24
387
387
2.51
2.85
inches
hours
cfs
cfs
Inches
inches
The second Deer Prairie Slough storm occurred 21-25 September 1982, when
6.6 inches of rain fell at the Venice raingage (NWS). No watershed parameters
were adjusted for this simulation. The only changes from the calibration
trial were the hourly rainfall values for the storm and the moisture storage
capacity above the water table. The measured and predicted outflow hydro-
graphs are shown in Figure 6. Features of the two hydrographs are similar to
those of the first storm. The predicted outflow of 264 cfs is 1.8% greater
than the measured 260 cfs which is a daily average (USGS, 1982a). The pre-
dicted volume is 9.4% less than the measured. Table 2 includes the data and
results.
Table 2: DEER PRAIRIE SLOUGH VERIFICATION DATA
WATERSHED SIZE:
AVERAGE DEPTH TO WATER TABLE;
TOTAL PRECIPITATION:
STORM DURATION:
PREDICTED PEAK:
MEASURED PEAK:
PREDICTED VOLUME:
MEASURED VOLUME:
27,060 acres
1.98
6.6
109
264
260
1.74
1.92
feet
inches
hours
cfs
cfs
inches
inches
There are several noteworthy features of the hydrographs shown on Figures
5 and 6. The shapes of both measured and predicted hydrographs correspond
very well to one another. The hydrographs in Figure 5 reflect runoff from
rainfall of varied intensity over 109 hours. SWMM allows a maximum of 200
rainfall values to be input for a single event simulation, so the measured
runoff "plateau" at about 260 hours which resulted from 2.90 inches of rain-
fall at that time is not reflected in the predicted hydrograph.
115
-------
The lag noticed in both predicted hydrographs when compared to the meas-
ured hydrographs is probably due to more immediate flow resulting from the
water table rising above the channel bed. The simulation generated flow when
the water table rose to the land surface.
uf :;: '\' ^7«- » i 11 :
- ."'"-f?. - (-^gis^v^j., ij U_;_j -i._
>..,6.' ^^df^l^i'^ ».|.;>|
>- "; J?'- ^i^Sil/U^^ ^ -
K-
Figure 4. Sub-basin Delineations within Deer Prairie Slough.
116
-------
IN
CF'a
*uu. uuuuu
330 OOOOO
240 OOOOO
160 OOOOO
SO OOOOO
O 0
0
++ *»*
+ »+» *
f » + *
* » + «
*»-»»
4- + «
* » + *
« * *
1- » fr
+ *
1- » » »
+ » +*
-» * I- *
+ » +**
+ * «»
+ * »+
+ * ** +
s **£»%
*. »» *+
4. »» +
4- *» +*
I. ** +*
1- ** ++
t »» ++
^ *« +++
1. *«* »+
v »»»» +.>+
«. »*»» ++++»
+. « *»»*» »+++
«. «. »»»**»» -f+-»-f++
K « *»»»*»»+++*
^ « »***
«. *
O 60 O 120. 0 ISO. 0 240. 0 300. 0 360. 0 420. O 480 0 54O 0 600.
0
TIME OF DAY. IN HOURS PREDICTED=*. MEASURED=+
OUTFLOW FROM DEER PRAIRIE SLOUCH LOCATION 65
Figure 5. Measured and Predicted Hydrographs for the Storm of 17 June 1982.
-------
00
3OO. OOOOO
Z4O. OOOOO
1BO COUOO
FLOU
IN
CFS
120. OOOOO
6O. OOOOO
0 0
0
I i .| 1 1 1 I 1 I
»»»«***
* # ++ «*
+ * * »»
f * -f »
+ » * »
* » >»»
+ * +++++++++++
* * » » +*»
» »- * *»*
»* * » »
* + * # + +
»» +»» *» +
» + *»
» +- **
»» *- *
* + **
» f *»
»* »£ **
++>. + + «**
»»+» *
+ ++ *
+ *
* *
+ *
+ *
+ »
+ *
* *
f »
* »
II- »*
I f »»»»
0 40.0 BO. O 1?0. O 160. O 200.0 240.0 280.0 320.0 360.0 400
TIME OF DAY, IN HOURS
OUTFLOW FROM DEFR PRAIRIE SLOUGH
PREDICTGD=*, MEASURED=+
LOCATION 65
Figure 6. Measured and Predicted Hydrographs for the Storm of 21-25 September, 1982.
-------
CONCLUSION
The Storm Water Management Model proved to produce good results when
compared with measured data for the site. The match of peak flows generated
by the storm events was in very good agreement with the measured peaks while
the volumes predicted by the model also compared favorably with the actual
data.
Uttle initial calibration was required to match predicted to measured
peak discharge values, with the width parameters being the only input value
that was adjusted. SWMM is able to predict runoff events for very large
watersheds, which gives it a distinct advantage over other modeling methods
such as the SCS and Rational Methods.
One of the more important factors for accurately modeling the events
appears to be the simulation of the water table depth prior to the storm.
Information about the water table depths is necessary in order to model the
antecedent moisture conditions which are critical to Florida runoff from large
storms.
It is also important for larger sites to be broken down into sub-
catchments. This allows the TRANSPORT module to account for flow routing from
one sub-catchment to another rather than producing one output hydrograph based
upon homogeneity within the watershed.
The use of SWMM allows physically measurable, site specific input to
be modeled rather than parameters derived from regression analysis of
regional data. The model also allows definition of worst case conditions
as well as wet season averages.
SWMM's capability to include site-specific data and antecedent moisture
conditions together with its channel network routing and detention storage
capabilities make it an excellent choice for accurately predicting runoff peak
discharges and volumes on high water table Florida watersheds.
REFERENCES
1. Carlisle, V.W., C.T., Hallmark, F. Sodek, R.E. Caldwell, L.C. Hammond,
and V.E. Berkheiser, 1981, Characterization Data for Selected Florida
Soils, Institute of Food and Agricultural Sciences, University of
Florida, Gainesville, Florida.
2. Huber, W.C., J.P. Heaney, S.J. Nix, R.E. Dickinson, and D. Polmann, 1981,
Storm Water Management User's Manual Version III, Environmental
Protection Agency, Project No. CR-805664.
3. National Weather Service, Venice Florida Station, Rainfall Tape, 1942-
1983.
119
-------
4. South Florida Water Management District, January 1984, Permit Information
Manual, Volume IV, Management and Storage of Surface Waters, pc-39, West
Palm Beach, Florida.
5. U.S. Geological Survey Water-Data Report, 1982(a), Water Resources Data -
Florida, Water Year 1982. Vol. 3A. Southwest Florida Surface Water, FL-
82-1A.
6. U.S. Geological Survey Water-Data Report, 1982(b), Water Resource Data -
Florida, Water Year 1982. Vol. 3B. Southwest Florida Ground Water, FL-82-
3B.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
120
-------
A SIMPLIFIED HATER QUALITY COMPUTER PROGRAM
FOR "REGIONAL STQRMWATER MANAGEMENT SITE EVALUATION
by: John M. Grouse, P.E.
Timothy C. McCormick
Greenhorne A O'Mara, Inc.
9001 Edmonston Road
Greenbelt, Maryland 20770
Michael H. He!frich
Montgomery County Stormwater Management Hi vision
101 Monroe Street
Rockville, Maryland 20850
ABSTRACT
Montgomery County, Maryland, has recognized the advantages of regional
Stormwater management facilities since the late 1970s. Until the last few
years, the design of SWM facilities was based on reducing the peak discharge
for the post-development conditions to the corresponding peaks for pre-
development conditions for the 2-, 10-, and often the 100-year storms. It
was known that SWM facilities also provide qualitative benefits. No
compatible method was available to quantify these benefits that would
require an effort similar to the work needed to compute the peak reduction.
The desire to produce a simple method to evaluate the water quality
benefits of SWM facilities and recent legislation concerning watershed
management policy in the State of Maryland have led Montgomery County and
their engineer, Greenhorne & O'Mara, Inc., to develop a simple computer
program to evaluate the water quality impacts of SWM ponds throughout a
watershed. The model was developed to utilize data prepared for USDA Soil
Conservation Service Technical Release Number 20 because TR-20 is frequently
applied throughout the State for quantitative design of SWM facilities.
A rainfall amount-frequency relationship was developed from data collected
by the County. Pollutant accumulation, pollutant washoff, and pond trap
efficiency information were obtained from previous works.
The computer model developed was applied during two watershed studies
conducted for the County by Greenhorne & O'Mara. The output from the model
aided in the evaluation of alternate SWM facility sites located throughout the
watershed. The model enabled a comparison of pollutant levels at various
points along the main stem and the larger tributaries similar to a comparison
of the peak discharges that had been accomplished for previous studies.
Since this water quality analysis required little additional data, the new
analysis was accomplished for a minimal cost.
121
-------
INTRODUCTION
Regional stormwater management facilities have been designed and
constructed in Montgomery County, Maryland, since the late 1970s. The
equations that govern the quantity of runoff that is stored and released
have long been known. Methodologies that utilize the storage and outflow
relationships have been developed into small dam and stormwater management
pond design procedures.
During the earlier years of the County stormwater management program,
the emphasis had been placed on pond release rates, downstream channel
velocities, and other hydraulic parameters. However, more recent concerns
about the levels of sediment and nutrients being carried to the Patuxent
River, the Chesapeake Bay, and other valuable bodies of water have modified
the rationale for the selection of regional stormwater management sites
towards sites that improve the quality of the runoff.
Water quality models have been in use for more than fifteen years. The
models, such as STORM, SWMM, NPS, WASP, HSP, SEDIMENT, and DEPOSITS, were
developed for various levels of planning and design applications. A model
was needed for which data could be easily developed and would be compatible
with the water quantity model most frequently used in the State of Maryland
for development impact studies, the USDA Soil Conservation Service TR-20
computer program (1).
RAINFALL AMOUNT-FREQUENCY
DATA ANALYSIS
Montgomery County has an on-going rainfall data collection program at
sites located throughout the County. The length of record for each site is
approximately the same. A site ne^ar the center of the County was selected
to be representative of the precipitation received County-wide.
Daily precipitation data were available from January 1980 through December
1983. Daily rainfalls greater than one-half inch were considered significant
for pollutant washoff. A tabulation of rainfall amounts exceeding this
value was compiled, and the rainfall amounts were plotted against frequency.
Rainfall amounts for selected frequencies were read from the plot and
are presented in Table 1. Although a limited number of years of data was
available, the results obtained were judged to be reasonable because a large
number of the very frequent storms occurred during the period of record. In
addition, the data plotted well compared to rainfall amounts obtained from
Technical Paper 40 for 1 year and ?-year return periods (?.).
122
-------
TABLE 1. RAINFALL AMOUNTS FOR SELECTED RETURN PERIODS
IN MONTGOMERY COUNTY, MARYLAND
Frequency (months) Rainfall Amount (inches)
24
12
6
3
?.
1
3.2*
2.6*
1.8
1.2
0.95
0.6
*From TP-40 (2)
ANNUAL STORM DISTRIBUTION
In an average year, the following storms would be observed:
1. one storm greater than or equal to the twelve-month storm;
2. two storms greater than or equal to the six-month storm;
3. four storms greater than or equal to the three-month storm;
4. six storms greater than or equal to the two-month storm;
5. twelve storms greater than or equal to the one month storm; and
6. a larger number of storms less than the one month storm.
The rainfall amount associated with storms greater than or equal to the
one month storm was calculated to be 12.3 inches based on the above
information. Montgomery County receives an average annual rainfall of
about 39 inches. The difference between the rainfall values represents
a very large number of events that produce less than 0.6 inch of rain over a
twenty-four hour period. These storms were not considered further because
they produce insufficient runoff to wash significant amounts of pollutants
downstream.-
Rainfall in the County is spaced almost evenly throughout the year.
Large storms may and have occurred in every month. The storm distribution
developed assumed that the twelve storms greater than or equal to the one
month storm were equally spaced. Therefore, one storm occurred every 30.4
days. The twelve-month storm and the six-month storm were offset by six
months. The two remaining three-month storms were also placed six months
apart and were offset from the twelve-month storm by three months. In this
manner, the storms were distributed evenly over the twelve-month period.
123
-------
POLLUTANT BUILDUP AND WASHOFF
ACCUMULATION RATES
Daily, dry weather pollutant accumulation rates for pervious and
impervious fractions of urban and rural land uses were developed by the
Northern Virginia Planning District Commission (NVPDC) for the Metropolitan
Washington Council of Governments (COG) in 1979 (3). The NVPDC accumulation
rates were judged to be acceptable for Montgomery County because of similar
physiography and the proximity of the NVPDC study area and Montgomery County.
WASHOFF
The washoff function of Metcalf & Eddy, Inc., et al. was utilized as the
washoff mechanism (4). The limiting pollutant load value was found to occur
over a period of time shorter than the 30.4 days between each storm of
significance. Hence, the maximum accumulation of pollutants was assumed to
be obtained before the beginning of each runoff event.
Runoff in the washoff process was determined for a given soil type and
land use through the runoff curve number (RCN) methodology of the USDA Soil
Conservation Service (5). The RCN methodology was selected because the RCN
values were be available from the data prepared for the hydrologic portion
of the study.
POND TRAP EFFICIENCY
The principal reason for the development of the computer model was to
estimate the removal of pollutants by a regional SWM facility, which can be
related to the trap efficiency of the pond. In 1975, Chen developed a
series of curves of trap efficiency versus the ratio of basin area to
outflow rate (6). Chen's work was based on previous studies by Camp. The
curves reflect the size of the sediment and the settling velocity of the
sediment particles.
Settling velocity data for soils typically found in the metropolitan
Washington area were presented in the NVPDC study (3). The pond retention
time and the settling velocity are used to calculate the percent of each
particle size which settles in the pond. The computer model developed
employs a default soil gradation that reflects the general soil of the
County. A different gradation curve may be input by the user.
NVPDC (3) presents a table that assigns fractions of the suspended
portion of each pollutant included in the study to sediment particle sizes.
The pollutants were assumed to settle as their associated sediment particles
settled. The pollutants remaining in the discharge from the pond was
assumed to be the dissolved pollutant plus the portion of the suspended
pollutant associated with the sediment that failed to setfle in the pond.
124
-------
MODEL OPERATION
Execution within the model occurs in the following manner:
1. The total accumulation of sediment, BOD, total phosphorus, total
nitrogen, extractable lead, and extractable zinc on each subarea
prior to each storm is generated from impervious and pervious
fractions for each land use;
2. The amount of washoff from each subarea for each storm is calculated
based on Metcalf & Eddy's washoff function;
3. The pollutant washoff data are annualized by summing the data for
the number of occurances of each storm frequency;
4. The pollutant washoff from each subarea is added based on series and
parallel subarea relationships from upstream to downstream until a
pond is encountered;
5. The pond routine is called, and the amount of each pollutant deposited
in the pond is calculated. Basin area and outflow rates were input
based on the SCS TR-20 printout for the storm frequencies studied; and
fi. The subarea addition and pond routines continue until the study
outfall is reached.
MODEL APPLICATION
The simple computer model - titled "WATQUAL" due to the lack of a more
imaginative name - was applied in two watershed studies in the County. The
results for a portion of one of the studies, the Little Paint Branch study,
are included in this section.
A portion of the Little Paint Branch watershed is shown in Figure 1.
The anticipated urbanization of the watershed is reflected in the increase
in curve number from existing conditions to ultimate development shown on
the figure.
The data input reflects combinations of soil type and land use for
each subarea. These data, which had been digitized for the TR-20 model,
became the bulk of the input data required for WATOUAL.
Without the proposed pond in Subarea 7, the annual sediment load at
Subarea 5 would increase from 26 tons to 58 tons from existing to ultimate
conditions. With the proposed pond, the sediment loM would be reduced to
11 tons as shown on Table 2.
125
-------
Ex. 8O
a/t. do
7a/?6}/ewood
59
G3
76
"* /COO'
/. L/W
-------
TABLE 2. EXAMPLE STUDY SEDIMENT LOADS
Annual Sediment in Runoff (tons)
Land Use
Condition
Predeveloped
Existing
Ultimate
Subarea 9
0.39
6.97
13.73
Subarea 7
1.04
24.09
47.92
Subarea 8
0.20
0.92
1.06
Subarea 5
1.42
26.23
58.33
Without Site 1
Ultimate
With Site 1
13.73
0.34
1.06
10.76
MODEL LISTING
The length of the listing of the Fortran computer model prohibited the
publication of the listing with this paper. Those interested in obtaining a
listing may contact one of the authors.
REFERENCES
1. TR-20, project formulation-hydrology (1982 version) - technical release
number 20. USDA, Soil Conservation Service, Lanham, Maryland, 1982.
2. Hershfield, D.M. Rainfall frequency atlas of the United States, technical
paper number 40. U.S. Department of Commerce, Weather Bureau, Washington
D.C., 1961.
3. Guidebook for screening urban, nonpoint pollution management strategies.
Northern Virginia Planning District Commission, Annandale, Virginia
1979. 160 pp.
4. Metcalf & Eddy, Inc., et al . Stormwater management model. U.S.
Environmental Protection Agency, Washington, D.C., 1971.
5. National engineering handbook, section 4, hydrology. USDA, Soil
Conservation Service, Washington, D.C., 1972.
6. Chen, Charng-Ning. Design of sediment retention basins. Paper presented
at 1975 National Symposium on Urban Hydrology and Sediment Control
Lexington, Kentucky. July 23-31, 1975.
The work described in the 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.
127
-------
DEVELOPMENT OF THE HAZPRED MODEL
by: G. Zukovs1, J. Kollar1, M. Shanahan2
Canviro Consultants Ltd, Toronto, Ontario
2
Environment Canada, Toronto, Ontario
ABSTRACT
HAPZRED is an interactive, microcomputer based model capable of the
prediction of hazardous contaminant (HO concentrations and loadings in dry
weather sewage flows., and of runoff and CSO volumes and HC loadings. HAZPRED
has been designed to estimate both average quantities (expected values) and
the probability of specified events.
The paper presents an outline of the predictive techniques used in
HAZPRED. A case study of a combined sewer catchment located in the Toronto
area of Ontario is also presented. The case study illustrates model input
requirements and the nature of model outputs.
INTRODUCTION
HAZPRED is an interactive microcomputer based model designed to predict
the concentrations and loadings of selected hazardous contaminants (HCs) in
dry weather sewage flow (DWF), urban stormwater runoff, and combined sewer
overflow (CSO). The specific predictive capabilities included in HAZPRED are
summarized in Table 1.
The selected hazardous contaminants included in HAZPRED are based upon
the U.S. EPA list of 129 priority pollutants. The HCs which can be examined
using HAZPRED include compounds falling within five classifications: Volatile
Organics, Semi-Volatile Acid Extractables, Semi-Volatile Base/Neutral Extrac-
tables, Pesticides and PCBs, and Metals and Trace Elements.
128
-------
TABLE 1. SUMMARY OF HAZPRED PREDICTIVE CAPABILITIES
Waste Stream Predictions Performed
Dry Weather
Sewage Flow
Stormwater
Runoff
Combined
Sewer Overflow
HC concentrations in industrial wastewater
HC loadings in industrial wastewater
HC concentrations in dry weather sewage flow
HC loadings in dry weather sewage flow
Average runoff event volume
Exceedance probability of runoff volumes
Average runoff HC concentrations
Average runoff HC loadings
Exceedance probability of runoff HC loadings
Average annual runoff volume
Average annual runoff HC loading
Average CSO event volume
Exceedance probability of CSO volumes
Average CSO HC concentrations
Average CSO HC loadings
Exceedance probability of CSO HC loadings
Average annual CSO volume
Average annual CSO HC loading
MODEL STRUCTURE
Figure 1 presents the overall structure of HAPZRED. As is indicated,
HAZPRED supports two levels of sophistication in the dry weather flow (DWF)
model. The following discussion will focus on the Level II, or more sophisti-
cated approach. Reference may be made to the HAZPRED documentation and to the
User's Manual (Canviro, 1984, 1985, 1986) for additional details of Level I
procedures.
CATCHMENT - SEWERAGE CONFIGURATION
The conceptual models of the catchment-drainage system used in HAPZRED
for separate and combined sewerage are shown in Figures 2(a) and 2(b). A
simple catchment model comprised of three parallel sub-catchments is employed.
Open space areas are not treated per se but are assumed to be included in the
pervious portion of each sub-catchment.
The parallel catchment model assumes:
i) that catchments are identically affected by regional climatology, and
ii) that each of the three sub-catchments responds independently to the
same climatological input (no correlation of sub-catchment output
probability distributions)
129
-------
MAIN PROGRAM
HAZPRED
i
SUBROUTINE
OHF DWF
LEVEL I LEVEL 41
] '
SUBROUTINE
RUNOFF
i
!
*
SUBROUTINE
DATA FILE
EDITOR
t
SUBROUTINE
CSO
Figure 1. Schematic diagram showing general structure of HAZPRFD.
Given these assumptions, the probability distributions of catchment outputs
(flows and contaminant loads) are additive. Previous analysis of sewer flows
by Adams and Gemmell (1973 a, b) have shown that the above may be justified
since the impact of any covariance terms on probability estimates would be
quite small and hence could be ignored.
In dry weather (no precipitaiion), all land use areas contribute dry
weather sewage flows which are transported to treatment by either sanitary or
combined sewers. During periods of precipitation, rain falling on the catch-
ment is transformed to runoff after accounting for losses due to depression
storage and infiltration through pervious surfaces. Pollutants residing on
catchment surfaces are washed off and transported with the runoff to the
catchment outlet. In separate sewerage systems [Figure 2(a)], stormwater run-
off is directly transmitted to the receiving water body. In combined sewerage
systems [Figure 2(b], or nominally separate systems with a large wet weather
component, stormwater runoff enters the sewer system and mixes with the sani-
tary wastewater. Combined sewage flows in excess of sewer or treatment plant
capacity are overflowed at regulating structures; in the case of nominally
separate systems, flows are by-passed at pumping stations or at the treatment
plant.
ESTIMATION OF DRY WEATHER FLOW QUANTITY AND QUALITY
The quantity of dry weather sewage flow from a given area is dependent
upon the nature of land use activity within the area and upon the influx of
130
-------
PRECIPITATION
I I I 1 I I
SANITARY
WASTE WATER
TO TREATMENT
STORM RUNOFF
TO RECEIVER
A. SEPARATE SEWERAGE
PRECIPITATION
1 I 1 I I I
B. COMBINED SEWERAGE
COMBINED SEWAGE
REGULATOR
TO
TREATMENT
CSO
TO RECEIVER
Figure 2. Schematic of HAZPRED catchment sewerage configuration.
131
-------
groundwater through infiltration. For each land use, different techniques are
required to develop the actual sewage flow components. A distinction should
be made between flow estimates normally used for sanitary sewer or treatment
facility design, and those that have application in a predictive model.
Design values in general tend to be conservative, often producing higher than
observed flow.values and hence should not be employed.
Figure 3 shows the components making up sanitary sewage in dry weather
and indicates the factors that affect the magnitude of each component.
Residential Wastewater Flow
The residential per capita wastewater flow rate is developed from metered
water use data. Winter data is used since it reflects most accurately water
Summer time activities such as lawn watering,
use but not to sewer flows. Where household
use data can be developed through the instal-
data from a
loss in dis-
returned to the sanitary sewer.
car washing, etc., add to water
metering is not employed, water
lation of test meters in selected households. Alternately, flow
suitable pumping facility can be used. Some allowance for water
tribution must be applied to pumping station data. An allowance of 10-15% is
considered typical (AWWA, 1962) for "unaccounted" water (water lost in distri-
bution).
Population data for a given catchment is developed from the most recent
census figures. Since census and catchment boundaries may not match, the
catchment population is estimated from unit area population of the census
zone(s) and the catchment area.
RESIDENTIAL WASTEWATER
- Serviced
Population
Per Capita
Water Use
- Fraction water
use returned
INDUSTRIAL WASTEWATER
- No. of
Establishments
Water Loss
To Cooling,
Process, Product
DRY WEATHER INFILTRATION
- Sewer Age
- Sewer Network
Density
- Groundwater
CONMERCIAL WASTEWATER
- No. of
Establishments
- Fraction Water
Use Returned
DRY
WEATHER
SANITARY
SEWAGE
FLOW
Figure 3. Components of dry weather sanitary sewage
132
-------
Commercial Wastewater Flow
In the HAZPRED model, it is assumed that the quantity of wastewater from
commercial establishments is exactly equal to the commercial water usage.
Water usage data are, again, ideally derived from individual metered records.
Since the type of commercial establishments is not differentiated, commercial
water usage is simply summed for a given catchment area.
Industrial Wastewater Flow
The relationship of water consumed to wastewater generated is consider-
ably more complex for industrial establishments. Resident industries are
first identified either through listing services (ie. Dunn and Bradstreet) or
by field inspection. The latter method is more comprehensive, particularly
for small establishments, but can be quite manpower intensive. Once the
industrial inventory is complete, then water consumption records are accessed
and a total industrial water usage is computed. The industries listed are now
segregated into sublists based on industry type as defined by the Standard
Industrial Category code (SIC code). In the event that industries are identi-
fied by field inspection, the SIC category is determined by direct contact
with the industry. Major industrial water users (>1% of the total industrial
water use for the catchment) are identified in each SIC sub-list and are then
contacted in order to estimate the wastewater flow components of interest. In
the case of combined sewer catchments, the process and sanitary flows as well
as the cooling water flows are identified. For separated sewerage systems,
only the process and sanitary component is normally needed. Although where
municipalities allow cooling water discharge to sanitary sewers, estimation of
both components will again be necessary. Occasionally, the wastewater or
cooling water flow estimates will not be readily available from the industry.
Best estimates of these flow components, based upon previous experience or
knowledge of the industry, are then employed. Industrial wastewater flows
generated by minor water users are estimated for each SIC category using total
water usage data and a return factor of 0.85.
Dry Weather Infiltration
Dry weather infiltration is determined most accurately as the difference
between metered dry weather flow data obtained at the catchment outlet, and
the computed residential, commercial, and industrial wastewater flow compo-
nents. Since monitored data is, however, only rarely available, DWI rates are
usually estimated. Ontario experience (Cooper, 1984) has shown a reasonable
infiltration allowance to be between 91 and 227 Icpd. The exact magnitude is
determined subjectively from knowledge of the age of the sewer system and
population.
Estimation of Dry Weather Sewage Hazardous Contaminant Concentrations
Estimates of HC concentrations in the residential and commercial compo-
nents of wastewater are based on a 1979 U.S. EPA funded study carried out by
A.D. Little (1979). The objective of the Little study was to determine the
relative significance of source (residential, commercial or industrial) con-
tributions of the EPA designated priority pollutants to publicly owned treat-
ment works.
133
-------
For industrial flows, HAZPRED, wherever possible, uses contaminant con-
centrations specific to the categories of industry present in the catchments.
However, only a limited number of industrial effluents have to date received
characterization. The most comprehensive industrial effluent data base is the
EPA's Treatability Manual (U.S. EPA, 1980). The manual is the product of
extensive monitoring and gives the incidence and effluent concentration of EPA
priority pollutants for certain types of industry. Table 2 presents the
industrial categories for which HC data is presently included in the HAZPRED
model.
Contaminant concentrations in wastewater from all other industries are
based upon average data obtained during the A.D. Little study.
WET WEATHER FLOW QUANTITY AND QUALITY
Analytical Models
Derived probability (analytical) models are used in HAZPRED to describe
the wet weather behaviour of both separated and combined drainage systems. The
models employ as input single parameter exponential distributions to describe
the statistics of rainfall (ie. volume, intensity, duration, inter-event time)
for a given locale. These distributions are mathematically transformed into
the probability distributions of interest, such as runoff and CSO volumes, and
pollutant loadings. The derived distributions are then employed to estimate
distribution moments (ie. expected values) and the exceedance frequencies of
selected events.
The theory and development of these models is well documented. Reference
should be made to Adams and Bontje (1983) for details.
Runoff Quantity
A modified version of the coefficient method is used to describe the
transformation of rainfall to runoff. The relationship between runoff and
rainfall is expressed as:
QR = * (i - $) (1)
where QR = runoff rate, mm-ha/hr
<£ = modified runoff coefficient, ha
i = average rainfall intensity, mm/hr
$ = continuous abstraction rate, equivalent to depression storage,
mm/hr
The value of the modified runoff coefficient, $ , is determined from the
area weighted runoff coefficient for each sub-catchment. The overall expres-
sion for
-------
TABLE 2. INDUSTRIAL CATEGORIES INCLUDED IN HAZPRED
SIC DESIGNATION
2200 - 2299
2834
2841
2844
2851
2893
3011 - 3079*
3111
3312
3321 - 3325
3431
3471
3479
INDUSTRIAL DESCRIPTION
Textile Mill Products
- yarn, thread, webbing manufacture
- fabric manufacture and finishing
2821,. 2823, 2824 Plastics Manufacture
- plastic materials, synthetic resins
- cellulosic man-made fibres and synthetic organic fibres
Pharmaceuticals
- other than perfumes and cosmetics
Soaps and Detergents
-manufacture soap, synthetic*organic detergents, organic
alkaline detergent, and crude and refined glycerine
Cosmetic Preparations
- preparations which function as skin treatments,
excludes those used to enhance appearance
Paint Manufacture
- trade-sales paints, chemical coatings, varnishes,
lacquers, epoxy coatings, and paint removers
Printing Ink
- letterpress, lithographic, flexographic and gravure
inks, and varnish
Rubber Processing
- tire arfd tube manufacture and other rubber products
Leather Tanning and Finishing
- conversion of animal skins or hides to leather
Coke By-Products
- manufacture of light oils, tars, phenolates, etc. as
by-products of coking
Foundries
- melting, moulding or finishing of metals
Procelain Enamelling
- application of glass-like coating to steel, iron,
aluminum or copper
Electroplating
- application of metallic surface coating by
electrodeposition
Coil Coating
- painting of coiled sheet metals
* SIC-3079 included in this group since products are related
135
-------
3 (2)
* I = modified runoff coefficient
. = runoff coefficient for the jth sub-catchment
J
A. = area of the jth sub-catchment, ha
j
The values of 4>j for each sub-catchment are calculated from the runoff
coefficients for pervious and impervious surfaces within the catchment, and
the relative proportion of these surfaces within each sub-catchment.
The continuous abstraction rate, $ , is related to the depression stor-
age, average rainfall volume, and average rainfall intensity. The expression
used to calculate $ is :
= C_!d (3)
B
where * = continuous abstraction rate, ntm/hr
t, = reciprocal of average rainfall volume, I/mm
S
-------
The washoff constant, K, takes into account catchment hydrology and sur-
face characteristics, and is calculated for each sub-catchment as follows:
K = $ A.C./100 (5)
J J v'
where K = washoff constant, gm/mm
4>'. = runoff coefficient for the jth sub-catchment
A- = area of the jth sub-catchment, ha
J
C.j = average concentration of a single HC in runoff originating in
the jtfl sub-catchment, ug/L
Average hazardous contaminant concentrations in stormwater runoff (C-)
were obtained from an Ontario study conducted by Marsalek and Greek (19837 and
from studies undertaken as part of the U.S. EPA Nationwide Urban Runoff
Program (NURP) (U.S. EPA, 1982).
The study conducted by Marsalek and Greek (1983) examined the frequency,
concentration and loadings of 51 selected persistent toxic substances in urban
runoff in the Niagara River area. Samples of urban runoff were collected at
sites in Fort Erie, Niagara Falls and Well and. Each site possessed a charac-
teristic land use; of the eight permanent sampling stations, two were located
in residential areas and six were located in industrial areas (sampling sta-
tions were not located in any commercial areas).
The preliminary NURP priority pollutant data base was used to obtain
information on contaminant concentrations in runoff from commercial areas
(U.S. EPA, 1982). A review of all NURP sampling sites identified five which
listed 10056 of land use to be commercial. Pollutant concentrations reported
for each of these sites were abstracted from this data base.
CSO Quantity and Quality
Combined sewage quantity and quality are estimated in HAZPRED through
simple mass and volume balance relationships based upon the relative propor-
tions of runoff and dry weather sewage. Probability density functions derived
from the mass and volume balances are then used to describe the distributions
of CSO volumes and pollutant loads.
HAZPRED IMPLEMENTATION
The HAZPRED model is written in BASICA for the IBM PC microcomputer and
compatibles, and requires the following hardware:
o IBM PC, PC/XT, PC/AT or compatible microcomputer MS-DOS 2.1 or higher
o 128 kilobytes of random access memory (128K of RAM)
o one double sided disk drive
o a keyboard
o monochrome or colour display monitor
o graphics adapter (to display piecharts)
137
-------
In addition to the above requirements, HAZPRED will also support dot-
matrix printers (EPSON FX-series) which can be used to generate hard copies of
program text.
Three DOS system files are required to support the HAZPRED program; these
are BASICA.COM, COMMAND.COM and GRAPHICS.COM. In order to run HAZPRED, the
three system files must be resident in the same location as the HAZPRED pro-
gram, either on the HAZPfcED diskette or in the DOS sub-directory containing
HAZPRED.
CASE STUDY
The HAZPRED model was employed to study a combined sewer catchment loca-
ted in the City of York, Ontario (one of the lower tier municipalities making
up Metropolitan Toronto). The catchment is 834 ha in area and is wholly ser-
viced by combined sewers. The area is approximately 80% devoted to residen-
tial land-use with the remainder equally shared between industrial and commer-
cial land-use.
DRY WEATHER PREDICTIONS
Dry Weather Wastewater Quantity
Dry weather flow quantities for the catchment required as 'HAZPRED input
are presented in Table 3.
TABLE 3. DRY WEATHER FLOW WASTEWATER QUANTITIES
Total Wastewater Flow (I/day)
Residential Wastewater Flow (I/day)
Commercial Wastewater Flow (I/day)
Industrial Wastewater Flow (I/day)
Dry Weather Infiltration Flow (I/day)
Industrial Cooling Water Flow (I/day)
34,118,270
15,116,960
48,890
503,220
18,449,200
334,300
Details of the required industrial component flows are provided in Table
4. In this case, the catchment contains only seven significant industries of
which two were represented by specific HAZPRED contaminant concentration data.
(ie. 2851 - paint manufacture, 3011-3079 - rubber processing).
Dry Weather Wastewater Contaminant Concentrations and Loadings
Dry weather concentrations of priority pollutants predicted by HAZPRED
are presented in Table 5. The < sign indicates that the predicted concentra-
tions are below the Method Detection Limit, implying that the compounds would
be detectable but not quantifiable with routine analytical techniques.
138
-------
TABLE 4. YORK INDUSTRIAL FLOW COMPONENTS
Industrial Flow by SIC (I/day)
0
0
0
0
0
10,680
0
3011-3079 : 14,140
3111
3312
3321-3325
3431
3471
3479
0
0
0
0
0
0
Other SIC : 144,100
2200-2299
2821-2824
2834
2841
2844
2851
2893
Table 6 presents the predicted contaminant yearly dry weather loadings
indicating both component loadings (ie. residential, commercial, industrial)
and total loadings.
For the York catchment, depending upon the contaminant, either residen-
tial or industrial sources dominate the total loads. For example, in the case
of the popular cleaning solvent tetrachloroethene, residential discharges are
predicted as the major source of this substance. In contrast, acrylonitrile
is predicted as exclusively originating from industrial sources.
WET WEATHER PREDICTIONS
Stormwater Runoff Volumes and Loadings
Input data needed for runoff and CSO predictions pertaining to both
catchment and rainfall characteristics are presented in Table 7. A noteworthy
feature of the catchment sewerage is the relatively high ratio of interception
capacity to dry weather flow (6.2:1).
HAZPRED output for both storm and CSO predictions includes summary sta-
tistics as well as expected values and exceedance probabilities of loadings
and volumes. Table 8 presents the summary runoff statistics for the catch-
ment.
In total, 72 of the 89 rainfall events are of sufficient magnitude to
produce runoff. The residential area, as may be expected, contributes most to
runoff volumes. In addition to average runoff characteristics, HAZPRED
provides the exceedance probabilities of annual runoff volumes, which are pre-
sented in Table 9.
These data may be generated for any range of desired volumes and may be
used in a number of ways, including calculation of the return period of
various volumetric events, according to the following equation:
(6)
e PCV > vn]
139
-------
TABLE 5. YORK OVERALL CONTAMINANT CONCENTRATION PREDICTIONS
Contaminant Concentration (ug/1)
Carbon tetrachloride < 1.00
Chloroform 1.44
Toluene 2.20
Acrylonitrile 2.49
Benzene < 1.00
Chloroethane < 5.00
Ethyl benzene < 1.00
Methylene chloride < 1.00
Tetrachloroethene 3.19
1,1,2-trichloroethane < 1.00
1,2-dichloropropane < 1.00
Trichloroethene < 1.50
Phenol < 10.00
Butyl benzyl phthalate < 10.00
Diethyl-phthalate < 10.00
Nitrobenzene < 15.00
1,2-dichlorobenzene 1.25
Naphthalene < 10.00
Bis(2-ethylhexyl)phthalate < 10.00
Di-N-butyl phthalate < 10.00
Antimony < 2.00
Arsenic < 3.00
Cadmium < 2.00
Chromium < 33.50
Copper 32.76
Lead 44.96
Mercury < 1.50
Nickel < 15.50
Selenium < 3.00
Silver < 2.00
Zinc 103.52
where TR = average return period (years)
6 = average number of rainfall events/year
p[V > V ] = the probability of observing a runoff volume greater
than VQ
For example, for the York catchment, a runoff volume of 10,000 m^ has an
average return period of 0.03 years, or approximately 1.5 weeks.
HAZPREO also predicts the expected values and exceedance probabilities of
runoff pollutant loadings, both per average event and per average year.
140
-------
Table 10 gives the expected values of annual runoff loadings for various
contaminants, from the sub-areas of the catchment, and for the catchment as a
whole. As was the case with the dry weather sewage, the different land-use
areas contribute varying amounts of a given contaminant to the total annual
mass.
CSO Volumes and Loads
HAZPRED summary statistics for combined sewer overflow events are presen-
ted in Table 11. As is evident, the large interception capacity results in a
moderate number of annual overflow events.
TABLE 6. YORK PREDICTED DRY WEATHER FLOW YEARLY LOADINGS
Contaminant
Carbon tetrachloride
Chloroform
Toluene
Acrylonitrile
Benzene
Chloroethane
Ethylbezene
Methylene chloride
Tetrachloroethene
1,1,2-tr i ch1oroethane
1,2-dichloropropane
trichloroethene
Phenol
Butyl benzyl phthalate
Diethyl-phthalate
Nitrobezene
1,2-dichlorobezene
Naphthalene
Bis(2-ethylhexyl)phthalate
Di-N-butyl phthalate
Antimony
Arsenic
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Selenium
Silver
Zinc
Res
(kg/yr)
0.00
16.55
14.35
0.00
1.10
0.00
2.21
0.00
34.76
0.00
0.00
2.21
32.00
37.52
54.07
0.00
15.45
11.59
37.52
49.66
14.90
26.48
9.93
89.94
397.83
536.87
2.21
23.17
20.97
12.14
1,180.79
Com
(kg/yr)
0.00
0.10
0.16
0.00
0.04
0.00
0.04
0.00
0.31
0.00
0.00
0.19
0.07
0.16
0.08
0.00
0.11
0.04
o.ll
.17
,.00
0.04
0.01
0.83
0.80
0.73
0.01
0.18
0.05
0.04
2.03
Ind
(kg/yr)
1.72
0.84
9.01
21.63
3.42
8.91
6.97
2.22
3.19
0.01
0.03
1.07
5.96
6.18
0.00
0.30
0.00
2.01
2.99
3.17
0.13
0.26
1.36
27.76
6.43
15.48
3.07
5.10
0.07
5.54
73.98
Total
(kg/yr)
1.72
17.49
23.52
21.63
4.57
8.91
9.28
2.22
38.27
0.01
0.03
3.47
38.03
43.86
54.16
0.30
15.56
13.64
40.62
53.00
15.03
26.78
11.30
118.53
405.06
553.08
5.29
28.46
21.08
17.72
1,256.79
141
-------
TABLE 7. YORK CATCHMENT AND PRECIPITATION CHARACTERISTICS
Total Area (ha); 834
Combined Sewer Interception Capacity (1/d); 209,950,000
Coefficients
Pervious:
Impervious:
Runoff
0.25
0.90
Depression Storage
2.0 mm
0.25 mm
Percentage Impervious Area
Residential (%): 45.5
Commercial (%): 71.6
Industrial (X): 84.8
Precipitation Characteristics
Reciprocal Average Intensity
Reciprocal Average Volume
Reciprocal Average Duration
Average No. of Annual Rainfall Events -
(hr/mm): 0.716
(I/mm): 0.193
(1/hr): 0.288
: 89.4
Both expected values and exceedance probabilities of CSO contaminant
loadings and volumes are available as HAZPRED output options. Table 12 illus-
trates the average values of annual CSO loads including the runoff and dry
weather flow contributions.
HAZPRED has, as well, graphic output capability in the form of pie charts
which present for specific contaminants the relative fractions originating
from sewage and runoff. Figure 4 shows the pie chart for the York catchment
representing total PCBs.
TABLE 8. PREDICTED YORK RUNOFF STATISTICS
Average Annual Number of Rainfall Events :
Average Annual Number of Runoff Events :
Average Annual Hours of Runoff :
Average Runoff Volume per Event
Residential Runoff Volume (m;?)
Commercial Runoff Volume (no
Industrial Runoff Volume (nr)
89
72
251
Total Annual Runoff Volume (m3)
13,030
2,105
2.096
Total Runoff Volume from Catchment (m3): 17,233
1,540,661
142
-------
TABLE 9. PREDICTED YORK ANNUAL RUNOFF VOLUME EXCEEDANCE PROBABILITIES
Total Volume (nr*) Probability of Exceedance
1
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
10,000
0.81
0.70
0.62
0.57
0.53
0.50
0.47
0.44
0.41
0.39
0.37
TABLE 10. PREDICTED YORK ANNUAL RUNOFF POLLUTANT LOADINGS
Contaminant
1,2-Dichlorobenzene
1,2,4-Trichlorobenzene
1,3-Dichlorobenzene
1,4-Dichlorobenzene
Hexachlorobenzene
A-Endosulfan
B-Endosulfan
4,4-DDT
4,4-DDE
Aldrin
A-BHC
Chlordane
Dieldrin
Endrin
Heptachlor
Heptachlor epoxide
Total PCB
Arsenic
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Selenium
Zinc
Res
(gm/y)
30.06
2.56
3.61
5.94
0.93
1.51
1.40
1.28
0.35
0.35
19.22
1.40
0.82
1.05
0.47
1.05
13.86
2,310.47
999.89
1,106.72
10,096.39
5,000.03
6,582.05
6,989.78
1,766.90
31,745.26
Com
(gm/y)
0.00
0.00
9.00
0.00
0.00
4.95
0.00
6.65
0.00
0.00
12.56
6.59
6.46
0.00
0.00
0.00
0.00
1,004.11
0.00
0.00
8,378.14
37,390.96
18.83
596.26
125.58
135,179.80
Ind
(gm/y)
7.91
0.52
0.97
0.86
0.09
0.13
0.28
0.21
0.06
0.06
2.49
0.09
0.11
0.43
0.07
0.17
2.23
562.84
326.44
433.01
2,308.96
2,144.96
1,159.54
1,634.76
661.67
13,442.51
Total
(gm/y)
37.97
3.09
4.59
6.80
1.03
6.60
1.68
8.13
0.41
0.41
34.27
8.08
7.39
1.48
0.54
1.22
16.09
3,877.42
1,326.33
1,539.72
20,783.49
44,535.95
7,760.41
9,220.80
2,554.14
180,367.60
143
-------
TABLE 11. PREDICTED YORK CSO SUMMARY STATISTICS
Average Annual Number of Rainfall Events
Average Annual Number of Runoff Events
Average Annual Number of Overflow Events
Average Annual Hours of Runoff
Average Annual Hours of Overflow
Average Overflow Volume per Event:
Contribution from DWF (m3) -
Contribution from Runoff (nr)
Total Overflow Volume (m3)
Total Annual Overflow Volume (m3)'
5,102
456,186
TABLE 12. PREDICTED YORK ANNUAL CSO POLLUTANT LOADINGS
1,2-dichlorobenzene
1,2,4-trichlorobenzene
1,3-dichlorobenzene
1,4-dichlorobenzene
Hexachlorobenzene
A-endosulfan
B-endosulfan
4,4-DDT
4,4-DDE
Aldrin
A-BHC
Chlordane
Dieldrin
Endrin
Heptachlor
Heptachlor epoxide
Total PCB
Arsenic
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Selenium
Zinc
0.
0.
0.
1.
DWF Contr.
(gm/y)
10.16
0.16
0.24
0.35
0.05
,34
.09
0.42
0.02
.02
76
0.42
0.38
0.08
0.03
0.06
0.83
219.09
81.33
298.81
1,283.75
2,585.38
408.98
575.87
151.03
9,955.49
Runoff Contr,
(gm/y)
43.85
0.69
1.02
1.51
0.23
1.46
0.37
1.81
0.09
0.09
7.61
1.79
1.64
0.33
0.12
0.27
3.57
945.85
351.10
1,289.99
5,542.02
11,161.24
1,765.57
2,486,06
652.03
42,978.51
CSO Loading
(gm/y)
54.00
0.84
1.25
1.86
0.28
1.80
0.46
2.22
0.11
0.11
9.37
2.21
2.02
0.40
0.15
0.33
4.40
1,164.94
432.43
1,588.80
6,825.76
13,746.62
2,174.55
3,061.93
803.06
52,934.00
144
-------
HAZPRED STATUS
At present, the dry weather portion of the model has been verified in two
Ontario catchments. It is planned to carry out further verification (dry
weather only) in nine additional Ontario catchments. Wet weather data collec-
tion in three Toronto area catchments will be completed this year and will be
used in connection with simulation modelling to verify the wet weather model
alogrithms.
YORK
RUNOFF
DRY WEATHER
TOTAL PCB 4
Figure 4. Total PCB loading distribution for the York catchment.
REFERENCES
Adams, B.J., and Bontje, J.B. Microcomputer applications of analytical models
for urban drainage design. _[n: W. James (ed.), Emerging Computer Techniques
for Stormwater and Flood Management. ASCE, N.Y., 1983. pp. 138-162.
145
-------
Adams, B.J., and Gemmell, R.S. An analysis of individual and grouped perfor-
mance of regionally-related wastewater treatment plant. TP-73-06-01, Urban
Systems Engineering Centre, Northwestern University, Evanston, Illinois.
1973a.
Adams, B.J. and Gemmell, R.S. Performance of regionally-related wastewater
treatment plants. Journal of the Water Pollution Control Federation. 45:10,
2088, 1973b.
American Water Works Association. A training course in water distribution.
Manual MS, 1962.
Arthur D. Little, Inc. Sources of toxic pollutants found in influents to
sewage treatment plants: VI - Integrated interpretations. EPA 68-01-3857.
U.S. Environmental Protection Agency, Cincinnati, Ohio, 1979.
Canviro Consultants Ltd. Development of the HAZPRED model phase I. Prepared
for Ontario Ministry of the Environment, Toronto Area Watershed Management
Study, Toronto, Ontario, 1984.
Canviro Consultants Ltd. Development of the HAZPRED model phase II. Prepared
for Environment Canada. Toronto, Ontario, 1985.
Canviro Consultants Ltd. HAZPRED - user's manual. 1986.
Cooper, B.O. Ontario Ministry of the Environment, Toronto, Ontario, 1984.
Personal communication.
Marsalek, J., and Greek, B. Toxic substances in urban land runoff in the
Niagara River area. National Water Research Institute. 1983. Draft report.
U.S. EPA. Treatability manual. EPA-600/8-80-042b, U.S. Environmental Protec-
tion Agency, Office of Research and Development, Washington, D.C., 1980.
U.S. EPA. NURP priority pollutant "monitoring program, Volume I: Findings.
U.S. Environmental Protection Agency, Washington, D.C. 1982.
Zison, S.W. Sediment-pollutant relationships in runoff from selected agricul-
tural, suburban and urban watersheds. EPA-600/3-800-022, U.S. Environmental
Protection Agency, Environmental Research Laboratory, Athens, Georgia, 1980.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official, endorsement should be inferred.
146
-------
ALTERNATIVE CALIBRATION OF THE QUAL-TX MODEL FOR
THE UPPER TRINITY RIVER
by Robert McCarthy
Dallas Water Utilities
Dallas, Texas 75201
ABSTRACT
The objective of model calibration is to establish the values of
free parameters by forcing an agreement between model predictions and
field measurements. In water quality modeling, the kinetic rate
coefficients are usually the model parameters of concern. It is
frequently the case, however, that there are more model parameters
requiring definition than there are data sets available. In this study,
Che QUAL-TX model, a version of QUAL-II, was calibrated for a 210 mile
portion of the Upper Trinity River which flows through and south of the
major metropolitan areas of Dallas -and Fort Worth, Texas, and into which
five major wastewater treatment plants discharge a combined average daily
flow of over 400 MGD.
Only one set of field data was available that was satisfactory for
model calibration, namely an intensive survey performed by the Texas
Water Commission. To emphasize the role of subjective judgement in
choosing coefficients when there is a lack of direct measurement defining
real river processes, two distinctly different calibrations were
performed based upon the observed dissolved oxygen measurements. The
first calibration was based upon the postulate that nitrification is
largely responsible for dissolved oxygen depletion in the river, and
employed a carbonaceous decay coefficient of 0.1 per day, a coefficient
of nitrification ranging from 0.3 to 0.5 per day, and sediment oxygen
demand rates approximately 0.1 gram/m^/day throughout the river
length. The second assumed that carbonaceous decay and sediment oxygen
demands are more important than nitrification. For this calibration a
carbonaceous BOD decay coefficient of 0.3 per day, a coefficient of
nitrification of 0.1 per day, and sediment oxygen demand rates in the
order of 2 grams/m2/day applied principally in the urban areas, were
assumed with all other coefficients the same as in the first calibration.
No significant difference in goodness of fit between predicted and
observed DO values was obtained between the two calibrations. Both
postulates are plausible; however, the implications on control strategies
of employing one set of coefficients, over the other are quite different.
147
-------
The first set would imply the need for advanced treatment with strict
ammonia removal* while the second would require strict processes for
carbonaceous BOD and sediment control, and perhaps even nonpoint source
controls.
INTRODUCTION
The Upper Trinity River flows through and south of the major metropolitan
areas of Dallas and Fort Worth, Texas. There are four major forks of the
Upper Trinity River - all of which are highly regulated by reservoirs
utilized for water supply by municipalities in the Dallas/Fort Worth
area. The Clear Fork of the Trinity is regulated by Benbrook Lake, the
West Fork by Lake Worth and Eagle Mountain Lake, the Elm Fork by Lakes
Grapevine and Lewisville, and the East Fork by Lakes Ray Hubbard and
Lavon. Numerous tributaries, a few of which are also controlled by
reservoirs, complete the drainage basin system for the Upper Trinity.
Figure 1 shows the major features of the Upper Trinity River basin system.
A water quality modelling study was performed covering the 210 mile
portion of the river from Beach Street in Fort Worth to Highway 31 at
Trinidad, south of Dallas. Five major municipal treatment and
approximately 30 minor municipal and industrial treatment plants
discharge wastewater effluent into the modelled reach. In addition two
major municipal plants discharge effluent into the East Fork which flows
into the modelled reach. The major dischargers and their treatment
capacities are shown in figure 2. Figure 2 also shows the locations of
United States Geological Survey flow gaging stations and existing and
planned water quality monitors.
BACKGROUND
Flows in the Trinity basin are highly variable. Under low flow
conditions the river flow can consist completely of wastewater effluent.
A frequency analysis performed by the Texas Department of Water Resources
determined the 7 day 2 year low flow, less discharger return flows, to be
0.0 (zero) cfs at USGS gaging stations located in Grand Prairie between
Fort Worth and Dallas, at Commerce Street in Dallas, and at Highway 34,
78 miles south of Commerce Street (1). At other times, immediately after
rainstorms or later when flood water is released from upstream
reservoirs, wastewater flows represent only a small percent of the total
flow in the river.
Periods of low flow have always caused water supply and water quality
problems for residents of the area. Early residents in this region were
compelled to construct impoundments upstream of the cities to store water
to improve water supply dependability. However, these impoundments have
further aggravated the water quality problems by reducing the already
insufficient natural flow in the river.
148
-------
I LEWISVILLE
LAKE
-I
LAKE I
RAY HUBBARO
EAGLE
MOUNTAIN
LAKE !
, LftKE _,.
ARLINGTON
BENBROOK
> I-KIHT - T
CEDAR CREEK
LAKE
FIG. I. TRINITY RIVER / RESERVOIR MAP
Low flow summer conditions, when most of the river flow is wastewater
effluent, have caused oxygen depletion in the river near the cities for
many years. The return flows have effectively created an artificial
hydrological environment inviting fish to swim upstream in the
effluent-augmented streamflows. Fish kills have occasionally been
observed downstream after rise events associated with rainfall occurring
in the Dallas/Fort Worth areas. Much speculation over the cause of the
kills is ongoing even now and many studies are being planned to address
this issue. At the present time it appears that the kills are rainfall
related and occur within a certain range of rising stage in the river at
Highway 34 south of Dallas. The automated water quality monitors show
slugs of water having extremely low dissolved oxygen levels traveling
downstream; however, the source of the low DO, either from nonpoint urban
runoff, or resuspension of bottom sediments from wastewater effluent or
nonpoint runoff deposits, is unknown. It is known that fish kills
generally do not or have not been known to occur for rise events below
2300 cfs or above 10,000 cfs and generally fall in the 2300-3500 cfs
range.
149
-------
The water problems in the Trinity are difficult to solve. The small
river is inadequate to support the large population of the river basin.
Water supplies are being supplemented by transporting water from other
river basins. The resulting return flows further increase the wastewater
loads on the stream.
In recognition of the effluent domination and high summer water
temperatures, the stream standard for dissolved oxygen has been set at
1.0 ppm for flows below 80 cfs above Beach street and set at 3.0 ppm for
flows above 80 cfs above Beach Street. This standard changes to 5.0 ppm
dissolved oxygen for all flow conditions at the Trinidad boundary.
Recognizing the major contribution that wastewater effluent has in
maintaining the base flow in the Trinity River, the major plant operators
have long been interested in the water quality of the river. In October,
1975, the cities of Dallas and Fort Worth, the Trinity River Authority,
and the North Texas Municipal Water District formed the Upper Trinity
River Basin Water Quality Compact for the purposes of carrying out
cooperative programs related to water quality in the Upper Trinity
WISE. CO.
OEWTON CO.
LEWISVILLE LAX
EAGLE MOUNTAN GRAPEVINE
LAKE LAKE
BENBROOK
LAKE
LEGEND
^ Waetewater Treatment Plant
USGS Flow Gaging Station
A Automated Quality Monitor
A Planned Automated Quality Monitor
Waitewater Treatment Plants
1 Ft. Worth 100MGD 4 Dalits Southslde 30MQD
2 TflA Central 100MOO 6 TRA 10 Mil* 10MQD
3 Dallas Cantral 150MQD 6 Garland 30MQD
7 Mosquito 12MOD
FIG. 2 MAJOR WASTEWATER DISCHARGERS AND
USGS MONITORING SITES IN THE
UPPER TRINITY RIVER BASIN
150
-------
River. Under this agreement four continuous automated monitors (CAMS)
were installed to measure select water quality parameters. The Compact
members and the US Geological Survey have had joint funding agreements
for these monitors since 1977. A fifth monitor funded by the City of
Dallas and the USGS was installed in February 1984. Arrangements have
been made between the Compact and USGS for two additional monitors to be
installed in spring of 1986, and the City of Garland and USGS should
install an 8th monitor later in 1986.
These monitors 'continuously measure four water quality parameters,
recording data at hourly intervals. These parameters are: dissolved
oxygen, water temperature, specific conductance, and pH. Discharge
measurements are also recorded. In addition to the continuous monitors,
the Cities of Dallas and Fort Worth and the Trinity River Authority have
long had various individual sampling activities within the modelled reach
in which weekly or monthly grab samples were taken at different sites on
the river. While the above data are quite useful in showing that an
improvement in overall dissolved oxygen levels and other water quality
parameters has taken place in the Trinity in recent years as the major
wastewater plants have been upgraded, they are inadequate to develop an
understanding of river processes.
In 1983 the Compact realized it was not enough just to know what the
status of certain water quality parameters was in the Trinity. Rather,
an understanding of river processes was desired. The Qual-Tx model, a
version of Qual-II, was acquired from the Texas Water Commission (TWC,
then the Texas Department of Water Resources) and implemented on the City
of Dallas computer system. While the aforementioned sampling programs
provided a wealth of water quality information, due to the uncoordinated
grab-sample nature of the surveys and the fact that no BOD or N^-N
data were acquired by the CAM monitors, inadequate data existed for
calibrating the model. There were however, two intensive surveys
completed by the TWC at the time of our study (2). The intensive surveys
included sampling at 35 sites from Beach Street in Fort Worth to Highway
31 at Trinidad in addition to sampling the effluent at the major
treatment plants. A composite sample was made from the four to five grab
samples taken at each site spaced over a sixteen hour period from predawn
to after dark. In-situ readings at the time of each grab sample were
also recorded. The TWC generously made these data sets available to the
Compact for use in studying the river. Under the guidance of Dr. George
Ward of Espey Huston & Associates, the Compact attempted to utilize the
data sets in calibrating the Qual-Tx model.
DATA SET REVIEW
The first step in employing these data sets was a review of the
conditions under which they were obtained. The Qual-Tx model is designed
to be applicable only to steady-state conditions. This means a condition
whereby all components of the system, including streamflow, wasteloads,
water temperature, etc., are steady in time throughout the modelled
reach, and the dissolved oxygen profile has equilibrated to these steady
components. For the data sets to be applicable for calibration and
151
-------
verification purposes they must represent a reasonable steady state
condition.
The first and easiest check for steady state conditions for the two
available data sets was a review of antecedent streamflows in the river.
If flow was too highly variable immediately prior to or during the
intensive survey sampling activities, then the sampling would not have
been made under steady state conditions and the data sets would not be
applicable for calibration. In fact this was found to be the case for
one of the data sets.
Figure 3 shows the daily streamflow values at USGS gaging stations for
the 30 days prior to and during the sampling for the September, 1982
survey. As it turns out the survey was conducted just a few days after
flood releases from upstream reservoirs were discontinued. The declining
flows shown on figure 3 are characteristic of gradual decreases in flood
releases as reservoir levels approach the top of conservation storage.
Further review showed that in fact very high releases had been
continuously made for a full thirteen months immediately prior to the
survey.
8 i
7 -
to ~
13 15 17 19 21 23 26 27 20 31 2 468 10 12
DAY
D BEACH ST. GAGE + COMMERCE GAGE O LOOP 12 GAGE A TRINIDAD GAGE
FIG. 3 USGS STREAMFLOW
13 AUG TO 13 SEPT 1982
152
-------
While streamflow had steadied for two or three days before the survey,
this was insufficient to ensure steady state hydrological conditions.
For this to have occurred, steady flows should have been achieved
throughout the model reach for a period at least equal to the travel time
from the head to the end of the reach. This would have been on the order
of 10 days for the normal expected low flows. The September, 1982 data
set was therefore judged to be inapplicable for calibration purposes.
Analysis then turned to the October, 1982 data set. Figure 4 shows
streamflows for the 30 days during and prior to this survey. While some
minor variations in flow were noticeable, the data set was judged
applicable for calibration purposes.
CALIBRATIONS
Since there were more model parameters requiring definition than data
sets available to ensure the assumed rate coefficients mirrored real
world conditions, and to emphasize the importance of correct subjective
judgement in choosing from a range of "reasonable" values, two sets of
rate coefficients, both within ranges cited in the general literature
7 -
~ 6-
w^
fel
z -
0 11 13 IS 17 19 21 23 25
25 27 29 1
DAY
o BEACH ST. GAGE + COMMERCE GAGE O LOOP 12 GAGE
A TRINIDAD GAGE
FIG.4 USGS STREAMFLOW
25 SEPT TO 25 OCT 1982
153
-------
were utilized in developing, two distinctly different calibrations based
upon the observed dissolved oxygen measurements.
The first calibration assumed, at 20 C, a carbonaceous BOD decay rate
of 0.1 per day, a coefficient for nitrification ranging from 0.3 to 0.5
per day, and a benthal oxygen demand on the order of 0.1 gram/m2/day.
This last coefficient would assume that benthal oxygen demand in the
Trinity primarily originates with point source dischargers and the
advanced waste treatment permits in effect for area operators (10 rag/L
BOD, 15 mg/L TSS) keep these demands very low.
The second calibration, assumed a carbonaceous BOD decay rate of 0.3 per
day, a coefficient for nitrification of 0.1 per day and a benthal oxygen
demand rate within the urban corridors on the order of 2 grams/m^/day.
The 2g/m2/d is a moderate demand and its use would assume that there
were sources of oxygen demanding deposits other than point sources. All
other coefficients were unchanged between the two calibrations.
Figures 5, 6 and 7 display the measured dissolved oxygen values from the
October, 1982 intensive survey. The vertical bars represent the range of
values obtained throughout the survey and the small circles indicate the
arithmetic average of the readings at each sample station. The modelled
dissolved oxygen profiles from the two alternate calibrations are also
plotted. Both profiles could be considered as good calibrations.
The Root Mean Square value, the square root of the average of the sums of
the differences between the observed and the predicted values squared,
provides one measure of the "goodness of fit". These values were 0.538
and 0.356 respectively indicating the second calibration was somewhat
better than the first.
z
HI
o
X
Ujg
o
-------
DISSOLVED OXYGEN
ppm
tf-i
KILOMETERS
CALIBRATION 1
CALIBRATION 2 FIG. 6 ALTERNATE CALIBRATION RUNS
COMPARED TO DATA POINTS
CONCLUSIONS
With two essentially equally good calibrations, which is better? The
answer is that we cannot tell from the available information. Of course
it was known at the outset that this was the case. The purpose of the
exercise then was not to have a completely calibrated model which one
could confidently employ in answering management decision questions. On
the contrary, the purpose was to show that Trinity River processes are
far from understood and that decisions made from one set of calibration
assumptions could be very different from decisions made from the other
set of calibration assumptions.
For example, a control strategy decision made from the first calibration,
would call for advanced treatment with strict ammonia removal. In
contrast, a strategy based on the second calibration, which emphasizes
carbonaceous kinetics and sediment demands, would call for strict
carbonaceous BOD and sediment controls, and perhaps even nonpolnt source
controls. Either strategy requires substantial capital Investments. At
this time the major plant operators on the Upper Trinity are committed to
achieving 10 mg/L CBOD, 15 mg/L TSS, and 3 mg/L NH3-N warm weather, 5
mg/L NH3-N cool weather limitations as called for in the most recent
wasteload allocation (3).
A concerted data collection effort is now underway to collect additional
data sets to help establish "correct" coefficients for future modelling
efforts. Future calibrations will center not only on predicted versus
actual measurements of dissolved oxygen, but also on comparisons of BOD,
155
-------
DISSOLVED OXYGEN
ppm
tt-
10-
B-
7- ^_
*" -^^-- ..*-""*>^ v 5
«- 5
3- £
'"I I
H *
KILOMETERS <
g
FIG. 7 ALTERNATE CALIBRATION RUNS 1
SiuS2i2 I COMPARED TO DATA POINTS
PALI On A I IVJN Z
NH3~N, NC-3-N+N02~N, and perhaps chlorophyll-a. Three intensive
monitoring surveys have already been performed by Compact members under
fairly steady state flow conditions. More will be completed as
conditions permit. Additional special studies directed toward better
definition of the nitrification and CBOD decay processes, an
investigation of observed in-stream nitrogen losses, algal kinetics and
photosynthesis/respiration, as well as a reanalysis of hydrology and
hydraulics relationships are also planned.
REFERENCES
1. Texas Department of Water Resources. Waste Load Evaluation For The
Upper Trinity River System In The Trinity River Basin. Texas
Department of Water Resources, June 25, 1985. P.22.
2. Davis, J. W. Intensive Survey of the Trinity River Segment 0805.
IS-53, Texas Department of Water Resources, June, 1983. 62pp.
3. Texas Department of Water Resources. Waste Load Evaluation For The
Upper Trinity River System In The Trinity River Basin. Texas
Department of Water Resources, June 25, 1985. 299pp.
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.
156
-------
LOGNORMAL1TY OF POINT AND NON-POINT SOURCE
POLLUTANT CONCENTRATIONS
by : Eugene D. Driscoll
(Wood-ward Clyde Consultants)
Oakland, New Jersey 07436
ABSTRACT
This paper presents a series of probability plots of water quality data
from a variety of discharge sources. It Is intended to provide a visual display
of the appropriateness of characterizing the variable pollutant concentrations
by a log normal distribution.
Representative examples of observed data that have been analyzed and
plotted to test whether they can be treated as lognormally distributed random
variables, are presented for data sets from the following applications:
Highway stormwater runoff
Combined sewer overflows
Urban runoff
Point Source discharges from POTW's
Agricultural runoff
Such examination suggests that a lognormal distribution either actually
defines the underlying population of pollutant concentrations, or is at the least
a satisfactory approximation for most environmental analyses.
INTRODUCTION
It has always been recognized that natural processes are variable. The
value of quantifying this variability in some appropriate way has Increased
157
-------
substantially in recent years. This is because many of the issues, problems,
situations that the environmental engineering community is now dealing with
are more effectively addressed in a probaballstlc context.
A number of articles have appeared in the technical literature over the
past five years or so, in which the authors report that the particular data
they are dealing with has a log normal distribution. Almost invariably, such
statements carry a qualifier. Some examples
The log normal distribution was found to fit .... most consistently.
(Three distributions). . . were found to be adequate, -with log normal
preferred because of ease of application.
Visual examination . . . indicated that data were best described by a log
normal distribution.
Generally log normal distributions -were observed for all but the data
extremes.
Overall, the . . . data tend to fit a log normal distribution.
Space constraints will nearly always preclude the inclusion of any
significant number of distribution plots in a paper or report, so the reader has
no independent basis for deciding how -well, or how poorly, the general
conclusion on lognormality really applies.
The simple objective of this paper is to present a set of probability plots
for -water quality data from a number of different applications, to provide the
reader -with a visual picture of the extent to which the sampled observations
fit a log normal distribution.
During the inspection of the probability plots, it will be useful to bear in
mind the following considerations. 1 submit that the important issue is not
whether a specific data set can be reliably concluded to have a log normal
distribution. The real issue is -whether it is appropriate or reasonable to infer
that the underlying population of events represented by the sample of
observations is lognormally distributed.
Each data set is a small sample of the much larger population of values
represented by the sample. The particular sample taken will be representative
of the underlying population to an unknown extent, and may be fairly good or
rather poor. There is no way to resolve this satisfactorily when it is the only
sample available for Inspection. However, -when similar pollutant data are
available for inspection from a large number of comparable sites, or for a
variety of other pollutants at the same site, the information to guide the
desired inferences is extended.
158
-------
HIGHWAY STORMWATER RUNOFF
The probability distribution of event mean concentrations (EMC's) of four
pollutants are shown for each of four highway sites. Three of the sites,
Nashville, Milwaukee and Denver are urban highways. The Harrisburg site is
in a rural setting. Figure 1 is for TSS, Figure 2 for Total N, Figure 3 is for Lead
and Figure A is for Zinc. These data are from a study currently in progress
for the Federal Highway Administration.
COMBINED SEWER OVERFLOWS
The probability distribution of event mean concentrations (EMC's) of BOD
(Fig 5) and suspended solids (Fig 6) are shown for four CSO sites. Three of the
sites, are in Richmond VA, the fourth in Toronto.
Figure 7 shows the distribution of the site median concentrations of BOD
and TSS at the six CSO sites that were monitored in Richmond, and at 13 sites
from 6 cities. All data is from the University of Florida Data Base.
URBAN RUNOFF
The probability distribution of event mean concentrations (EMC's) of
Total P (Fig 8) and COD (Fig 9) are shown for six urban runoff sites. Figure 10
shows the distribution of the site median concentrations of Total P at 69 of the
urban runoff sites that were monitored under the Nationwide Urban Runoff
Program (NURP)
POINT SOURCE DISCHARGES - POTW's
There have been a number of studies over the past 6 or 8 years that
have looked at the variability of daily average effluent concentrations from
municipal sewage treatment plants. The authors have, as cited earlier,
concluded that distributions of daily average concentrations are at least
adequately approximated by a log normal distribution. I haven't had access to
such daily values and have no plots to present to illustrate this level of detail.
However the papers and reports do present summaries of the mean and
coefficient of variation (COV) of daily values for a sample that generally covers
one to three years of plant records.
Figure 11 shows the distribution on mean influent concentrations of
Cadmium for a large sample of POTW's. Figure 12 shows the distribution of
the overall mean effluent BOD concentration, and the COV of daily values for 66
Conventional Activated Sludge plants. The correlation plot shows that there is
no relation bet-ween the mean effluent that a particular plant produces and the
degree of variability of daily concentration values - for all plants in this
process category.
159
-------
I
HIGHWAY RUNOFF - SITE EMC's (TSS)
6U6PENDCO SOLIDS
-------
HIGHWAY RUNOFF - SITE EMC's (TOTAL N)
TOTAL N (mg/l)
10- tT NA6HVILLC 1-4
\0'0
-2
-I
0
Z SCORE
TOTAL N (fin/I)
10* IT MILWAUKEE HWV4
10*0
10'-
-2
0
Z SCORE
MEDIAN > O.D2
COV 3 0.68
N=2I
MEDIAN 3
COV >
N:23
1.38
0.64
TOTAL H (mg/t)
10- IT DENVER HIGHWAV SITE
10*0
I DM
-2
TOTAL N
10'I
10*0
10*->
-2
MEDIAN =
COV s
H=I6
0.93
0.66
-I 0
Z SCORE
HARRIS8UR6 1-81
MEDIAN 3
COV >
NsIB
0.70
O.S9
0
Z SCORE
-------
HIGHWAY RUNOFF - SITE EMC's (LEAD)
LEAD
-------
HIGHWAY RUNOFF - SITE EMC's (ZINC)
CA>
ZINC (mg/l)
10* Or
NASHVILLE 1-40
ZINC (mg/l)
10* IT MILWAUKEE HWV4S
i
-------
COMBINED SEWER OVERFLOWS - SITE EMC's (BOD)
eOO-5
1C' 3
10'2
-------
COMBINED SEWER OVERFLOWS - SITE EMCs (TSS)
en
T65 (mq/1)
10-3i
I0'2
10'1
-2
TSS (mo/I)
10-3
10*2
lO'l
-2
BLOOD v
-i
0
Z SCORE
0
Z SCORE
O. 80
N s 19
TSS «mg/l)
10'
10'2
10* I
-2
T66
10'3T
10-1
-2
-I
0
Z SCORE
BAST YofcK
N a 14
N a 22
-I
0
Z SCORE
-------
10*
-2
TSS (mg/l)
10'
10-2
-2
COMBINED SEWER OVERFLOWS - SITE MEDIANS
57
0.33
N a 6
0
Z SCORE
EULHMOOD tfA
0.34-
N 3 6
1 0
Z SCORE
BOD-S (mg/l)
10-2T
10*1
-2
TGB (mg/l)
1C'3r
KT2
-2
13 CSO
«a (, CITIES
MLDMAJ 5?
N a 13
0
Z SCORE
IS C5o SITES /u 6 CITIES
N a 13
I 0
Z SCORE
-------
URBAN RUNOFF
DRTR FROM NURP STUDY
EVENT MEAN CONCENTRATIONS OF CHEMICAL OXYGEN DEMAND
AT SIX STUDY SITES
COD Owj/ll
ID" 5 CO I NORTH AVE DRAIN
ID'?
10'I
WOIAM t 279 22
CDV > 0.74
-2
I 0
I SCORE
coo (ma/it
IQ-3 m I PARKING LOT
10-2
ro-i
MtDlAHt 7933
COT . 0.72
N.JJ
2
-I
0
t SCORE
COP (roj/0
10- 3 HI I WAVIRIY H11L6 1W.ET
ID1 2
10-
lO'O
MEDIAN t '4973
CUV O.M
Ht27
-2
-t 0
t scon
18-3* WAI UK H1UI
10-2
10-1
COO (mt/0
10- 3 TI I HA8I IANE
10-2
10-1
KDIANt SII8
COV i 08J
Mill
2
0
I SCORE
COD
10-J
10-2
to-1
DC I WtCUtIGM IM.ET
KDIAN » 45 02
COV 0,«
2
t ectvtf
167
-------
URBAN RUNOFF
DBTfl FROM NURP STUDY
EVENT MEAN CONCENTRATIONS OF TOTAL P
AT SIX STUDY SITES
TOTAL P (mo/I)
tO" 1 CO t NORTH AVt MAIN
10" 0
10'-
-I -I
0
I ECORt
TOTAL P (mj/l)
1C' 0» NH I PARttrHGLOT
10'-
10'-7
HEDIAN' 017
COV 1.21
-1 -1
I SCOOT
101 Al P (iM/1)
tO'P HI I WAVIRLY H1UC INLtT
W-l
IO'-3
MEDIAN. 017
COV 0.64
NiJS
-i
1 «
t KDM
TOTAL P |mf/D
10-1, WA I
10-0
lO'-J
LAKE HILtS
fUDIMI. 020
COV 0.01
H.I127
-2 -1
I 2
t KOBE
TOTAL P
-------
11
URBAN RUNOFF
DflTB FROM NUHP STUDY
SITE MEDIAN CONCENTRATIONS OF TOTAL P
FOR 69 URBAN RUNOFF STUDY SITES
LOO MEAN
LOO SIGMA
MEAN
SIGMA
MEDIAN
COEF VAR
5,787
400.143
284.545
324.092
0.711
****#***»**»*********»**»*»»»##**»»»*»*»
TOTAL P (ug / I)
10* 4T
NT 3
IOA2
ICT I
N = 69
-2
"H o~~
Z SCORE
169
-------
99.99
99.95
99.9
0.01
I I I IlllllTTT
o.ooor o.oo i 001
i
Log Concentration CO (mg/IJ
to*
r i i nun
1.Q
POTLJ- INFLUENT
MEDIHN
CHOMIUM CONCENTRRTIONS
DATA BASE FOR INFLUENT
HEAVY METAL CONCENTRATORS
IN POTW3
EPA 600/52-81-220
2 YEAR STUDY
DATA FROM 239 PLANTS
'Generally, log normal distributions
war* observed for all but ine data
txtrvmts.*
Overall, tho Individual plant m*an
and median metal concentration
data tend to nt a log normal distribution."
FIGURE SHOWS
LOG PROBABILITY PLOT
for reported
MEDIAN CADMIUM CONCENTRATIONS
-------
UJ
f^^
^%
rv =
1 1 1 1
; **(a
, 1
-3 -2 -19 1 2 3
Z SCORE
*n &. -
O
o. -
-3 -2
-10 1
Z SCORE
POTUI
EFFLUENT CHRRflCTERISTICS
BOD
.*-
o r
r - o. o i
PLT MERNS
-------
AGRICULTURAL RUNOFF
Figure 13 shows the distribution of the EMC's of three pollutants
(ammonia, nitrate and TKN) at two study sites in Watkinsville^Georgia.
The ammonia nitrogen EMC's for agricultural runoff from four study
sites in the Four Mile Creek watershed in Iowa are presented in Figure 14.
Total Phosphorus EMC's at two of these sites are shown by Figure 15. The
second of the plots for each site excludes 4 observations during one of the
winter periods when manure was applied on frozen ground and not worked
into the soil.
Figure 16 displays the distribution of annual average values for the
concentration of soluble phosphorus and nitrogen in both surface runoff and
subsurface storm-water discharges at the site of a long term study at a
different location in Iowa. The final plots describe the variability of the
concentration ratio between subsurface and surface discharges.
CONCLUSIONS
1. Whether or not it Is rigorously true that the distribution of the
underlying population of pollutant concentrations from the indicated sources ar
log normal (as suggested by the material presented), it appears to be an
acceptable approximation that will be adequate for many of the uses to which
it may be put.
2. Lognormality can be assumed for the distribution of EMC's at a site,
or for the distribution of mean or median concentrations from a number of
different sites that are in the same category.
The work described in this paper was not funded by the U. S. Environmental
Protection Agency and therefore the contents do not necessarly reflect the views of
the Agency and no official endorsement should be inferred.
172
-------
AGRICULTURAL NPS RUNOFF
UlflTKINSUILLE 6fl STUDY
EVENT MEAN CONCENTRATIONS OF NITROGEN FORMS
( NH4 - N , N03 - N , TKN )
STATION P2
NH4-N (mg/1)
to-1
10* 0
lO'-
Itr-
H03-H
10' 7
io- 1
to-o
10
,o-r
TWI
10'
lo-i
(0 0
ID'
_-
,^r
<
-2 -»
Z SCQRt
-»-
-» 0
Z SCORE
1.93
N 3»
Pz.
MEM/»»J
l.fcl
H 39
P2.
B . 37
-»-
I b
Z SCORE
STATION P4
NH4-N (mg/1)
10'0
10-
IT-3
N i M
-I
-H
«
Z SLORL
H03-N (mg/l)
'
P4-
to-
ID-
I- 50
H IT
-I 0
Z-SCORE
TKN (rog/1)
lO"
P4-
10'0
10'-
N If
I -I
-t-
0
Z SCOKE
173
-------
AGRICULTURAL NPS RUNOFF
FOUR MILE CREEK IOUIR STUDY
EVENT MEAN CONCENTRATIONS OF AMMONIA NITROGEN
AT FOUR MONITORING STATIONS
NH4-N
10- 1
10' 0
IC'-
W-}
10-
1
NH4-N
In 1
10' >
w-
w-
+-
0
Z SCORE
j
-+
o
Z SCOKE
NH4-N
lO'l
lo-o
lO'-
w-
2
NH4-N (mg/1)
10-0
10'-
In-.
STAft
» M
-1 0 1
Z SCORE
0.148
0.61
h t tt
I *
Z SCORE
I
174
-------
fl
AGRICULTURAL NPS RUNOFF
FOUB MILE CREEK lOUJfl STUDY
EVENT MEAN CONCENTRATIONS OF DISSOLVED PHOSPHORUS
AT TWO MONITORING STATIONS
STATION 1 STATION 2
OISSPWO f (mg/l>
10-1
10'0
10'-
w-i
-i
4-Mil£. CfUEJZK.
ITff-life
10'1
10'0
to-i
ID-:
t STOW
biss P UHI wmrtR ISBQ)
10'0
10'-
10'-
-t
4 Mile
0
1 ttORt
- r?flo
DISS f iKHf WIHItR 801
10'0
10'-
4MI e/Z£e£
-J
175
-------
u
SURFACE AND SUBSURFACE RUNOFF
CONVENTIONAL TILLAGE
ALBERTS md SPOONER. J SOIL AND WATER CONSERVATION. FED 1905
ANNUAL AVERAGE CONCENTRATIONS OF SOLUBLE POLLUTANTS
NITROGEN (N03)
PHOSPHORUS (P04)
IP .'
BO ( pt/l )
I0'7
rp.i '.out
10-I
tfO
lO'O
w-
B t 10
-I «
i
/W*ju*t.
CTOM
r.el
N « ID
-I 0
i ecm
mu cone-sum Act m
to-1
w-
2.44
-J '-I
t stoet
IW CVR CUPflfRf TO nif/l
'
10' I
.10-
5. IS
2 -I
I SCOfit
o.ipnrt/cu-r tot-c«*no
10-1
-j
I tton
176
H . 10
N 10
-------
POLLUTIOM FROM HIGHWAY RUNOFFPRELIMINARY RESULTS
by: Philip E. Shelley
EG&Q Washington Analytical Services Center, Inc.
Rockville, Maryland 20850
and
David R. Gaboury
Woodward-Clyde Consultants
Walnut Creek, California 94596
ABSTRACT
This paper presents the preliminary results of a project aimed at developing
models that can be used by planners and highway engineers for predicting pollutant
runoff from highways. It is a part of an ongoing research program being conducted
by the Federal Highway Administration of the U. S. Department of Transportation to
characterize stormwater runoff from highways, assess its potential impacts on re-
ceiving waters, and determine the effectiveness of various control measures.
A brief review of different approaches to predicting pollutant runoff loads
from highways is followed by a description of the data base being assembled as a
part of this project. The probabilistic data analysis methodology that is being used
to characterize highway storm-water runoff is then described in some detail. Pre-
liminary analytical results are presented in two main areasrainfall and runoff
data and water quality data.
For the sites examined so far, it is shown that taking the percent impervi-
ousness of an unmonltored site as Its runoff coefficient, with an upper bound of
around 0.9 to account for initial abstraction, offers a reasonable first estimate.
Thus, the runoff quantity from an ungaged highway site can be estimated from a
simple analysis of rainfall records from a gage near the site.
Based upon the their overall characteristics, the highway sites in the data
base have been classified as being either "urban" or "rural." Best estimates for
pollutant concentrations for each of six pollutants (SS, COD, TKN, TPO4, Pb, and Zn)
for highway sites are presented and compared with corresponding values for urban
runoff. Lacking site specific data, they represent reasonable first estimates of high-
way stormwater runoff quality. It is also shown that suspended solids can be a
reasonable indicator for other pollutant concentrations in highway runoff.
177
-------
INTRODUCTION
For over ten years the Federal Highway Administration (FHwA) of the United
States Department of Transportation has been engaged in a research program to
characterize stormwater runoff from highways, assess Its impacts on receiving
waters, and determine the effectiveness of various control measures for possible
Instances where their use to mitigate impairment of designated beneficial uses of
receiving waters might be required. As a part of this research program, a con-
tractor team headed by Woodward-Clyde Consultants, and including EG&G Washing-
ton Analytical Services Center, Inc. and the University of Florida, has been tasked
with developing models that could be used by planners and highway engineers for
predicting pollutant runoff loads from highways.
Approaches to predicting pollutant runoff loads from high-ways can be
generally grouped into one of two classes;
those that are based on regression equations, and
those that use some form of simulation model.
Approaches that use simulation models can be further characterized mechanistical-
ly into three types:
those employing rating curve methodologiesactually
an extension of regression analysis based on a power
law relationship between flow and sediment load or
concentration;
those where pollutant build-up and -wash-off
relationships are used as, for example, in SWMM,
STORM, HSPF, and the FHwA Urban Highway Storm
Drainage Model; and
probabilistic based approaches, wherein probability
density functions are assigned to runoff flows and
concentrations, and the resulting pollutant loads are
characterized statistically, as was done, for example, in
the Environmental Protection Agency's Nationwide
Urban Runoff Program.
Thus, prediction of highway stormwater runoff quality may be performed using a
number of procedures that range from the very simple to detailed deterministic
simulation models, and a large body of literature exists describing them. However,
all of the approaches have one common aspecttheir predictive capabilities tend to
be rather poor -without suitable site-specific data for calibration. For this reason,
one of the first tasks in the present effort was to assemble a data base of highway
runoff characteristics. In the following sections of this paper, that data base will
be described, the data will be summarized, and best preliminary estimates of high-
way stormwater runoff will be made.
178
-------
HIGHWAY RUNOFF DATA BASE CHARACTERISTICS
The data in the highway stormwater data base is taken from monitoring
projects supported by the FHwA and several states. At the present time, there is
coverage of twelve sites, with several more to be added as the data become avail-
able. There are from seven to 139 monitored events at each site. The data base
consists of three categories of data;
rainfall and runoff,
water quality, and
fixed site.
Each of these categories will be discussed in turn.
RAINFALL AND RUNOFF DATA
Rainfall data were taken with recording raingages. For each rainfall event
there is a start time and date, five-minute raingage readings that describe the
hyetograph, and a stop time and date. In a similar fashion, runoff data were
taken from flowmeters and consist of a start time and date, (typically) five-minute
instantaneous flow readings that describe the hydrograph, and a stop time and
date. The lag between the time rainfall starts and the time runoff starts, although
highly variable for a given site due to antecedent conditions, provides an indication
of basin response time or "flashiness."
Taken together, the rainfall and runoff data describe the water quantity
characteristics of the site. The rainfall data can be analyzed to provide informa-
tion on rainfall frequency (i.e., the time between storms), intensity, duration, and
total amount. For example, the total quantity of rainfall for the event is simply
the difference between the raingage readings at the beginning and end of the event.
The flowmeter data can be analyzed to determine peak flow, time to peak, total
quantity discharged, etc. The total quantity of runoff discharged during the event
is computed by integrating the flow rate record over the time period of the event.
Of particular interest in this preliminary analysis are the total rainfall and runoff
quantities.
WATER QUALITY DATA
The highway runoff quality data consist of the results of laboratory analyses
of sequential discrete samples that were taken at recorded times throughout the
runoff event, plus the results of laboratory analyses of flow-weighted composite
samples taken over the entire runoff event.
Although over 24 analytical determinations were made for the different sites
contained in the data base so far, there was consistent coverage of up to eighteen
pollutants, and these were selected for this preliminary analysis. By category,
they are:
Solids TS, TSS, VSS
Physical Parameters pH, Cl~
179
-------
Oxygen Demanding Substances BOD, COD, TOC
Nutrients TKN, N02+N03, TP04
Metals (Total) Cu, Pb, Zn, Fe, Cr, Cd
Hydrocarbons Oil and Grease
FIXED SITE DATA
The fixed site data, so called because they tend to be fixed rather than var-
iable on an event to event basis, fall into three general categories. They are:
Highway Site Data
configuration (e.g., elevated, ground level, depressed)
pavement composition, quantity, and condition
design, geometries, cross-sections
vegetation types on right-of-way
drainage features
Operational Characteristics
traffic characteristics (e.g., density, speed, braking)
vehicle characteristics (e.g., type, age, maintenance)
maintenance practices (e.g., sweeping, mowing, weed control)
institutional characteristics (e.g., litter laws, speed limit
enforcement, vehicle emission regulations)
Surrounding Land Use Characteristics
land use type (residential, commercial, industrial, agricultural,
forest)
geologic features (relief, soil types and horizons, groundwater
characteristics)
agricultural practices (e.g., tillage, irrigation, and cropping practices)
In this preliminary analysis, the chief fixed site data to be used are the size of the
drainage area, the average traffic density expressed as vehicles per day, the per-
centage of the basin area that is impervious, and three roadway featuresthe
surface type, number of lanes, and the presence of curbs.
180
-------
DATA ANALYSIS METHODOLOGY
Pollutant runoff from highway sites (as well as from other types of nonpoint
source sites) is quite variable, especially as compared to discharges from point
sources. Therefore, new methodological frameworks are required for analysis of
data from nonpoint sources in general and highway runoff in particular, especially
with respect to characterization of pollutant loads and their possible effects upon
receiving -waters. New approaches to water quality criteria and standards also
seem warranted when one is dealing with nonpoint sources, due to the high vari-
ability and intermittent nature of the latter.
When one has flow and polutant concentration data for runoff from a moni-
tored site, there is a need for a sensible way to reduce the data to summary form
in a way that will be useful to decision makers. For this task, it was decided to
follow the general data analysis approach developed for the Environmental Protec-
tion Agency's Nationwide Urban Runoff Program (NURP).
In NURP, flow and concentration data for a site were first analyzed to deter-
mine event mean concentrations (EMCs). The EMC is defined as the concentration
that would result if the entire storm event discharge were collected in a container,
and its concentration determined; i.e., it is the total mass of pollutant discharged
during the event divided by the total quantity of water discharged during the
event. In practice the EMC is simply taken to be the concentration of a flow-
weighted composite sample collected over the duration of the runoff event. If
sequential discrete samples are taken over the period of the runoff event, an ac-
ceptable approximation to the EMC can be formed by manually compositing aliquots
that are sized In proportion to flow (manual flow-weighted composite sample) and
performing analytical determinations on the resulting composite sample. If separ-
ate analytical determinations have been performed on the individual sequential
discrete samples, an acceptable approximation to the EMC often can be computed
from the individual concentration values and the corresponding flow values.
The main point is that, In dealing with nonpoint sources, the basic unit of
water quality information is the average concentration of each pollutant of interest
in the total volume of runoff produced by each individual storm event. Thus,
within-storm fluctuations in pollutant concentration are completely ignored.
Hence, for a particular pollutant and a particular site, there will be one EMC value
for each storm event monitored, and the EMC is considered to be the random
variable.
Given a set of nonpoint source monitoring data for a site, the first step is to
determine the EMC values for the pollutants of interest. This set of EMC values is
then analyzed to compute the statistics of runoff quality for the site. However, in
order to properly interpret the site statistics, it is usually convenient to know
something about the nature of the underlying distribution function (or, equivalent-
ly, the underlying probability density function). From physical considerations, it
can be shown that a normal distribution cannot be exactly correct for EMCs (e.g.,
pollutant concentrations can never be negative, they are always skewed, etc.). It
can also be shown, by appeal to the central limit theorem, that when a series of
multiplicative processes are involved (as is the case with nonpoint sources), the
resulting distribution function 'will be asymptotically lognormal. The adequacy of
this lognormality assumption has been checked with data from urban runoff,
highway stormwater runoff, combined sewer overflows, and limited sets of agri-
181
-------
cultural runoff data. In no case has the lognormal distribution been clearly Inap-
propriate, and in most cases it seems to fit the data better than other possible
distributions.
Under the asumptlon that the lognormal distribution would be an adequate
representation for the highway runoff data sets in question, it remained to com-
pute the EMC values for each event at a site, transform these values into the log-
arithmic domain, compute the mean and variance in log-space, and then properly
transform these values into the appropriate statistics in arithmetic space (typically
the median, mean, and coefficient of variation). For surface runoff data, the two-
parameter lognormal distribution is adequate. It is completely specified by a cen-
tral tendency parameter (say, the median) and a dispersion parameter (say the
coefficient of variation).
Having determined the two EMC statistics for a pollutant at a site (say, the
median and coefficient of variation), they completely describe the variable runoff
characteristics for that site. They can then be compared with similar data from
other sites to evaluate similarities and differences in the context of physical site
characteristics. This amounts to comparing the underlying probability distribu-
tions for the sites in question. The main point is that the foregoing methodology
recognizes that the natural processes that lead to highway runoff are highly vari-
able and provides an appropriate way of quantifying this variability.
Each data set is a small sample of the much larger population of values rep-
resented by the sample. The particular sample taken will be representative of the
unknown population to an unknown extent, and may be fairly good or rather poor.
There is no way to resolve this satisfactorily when one has only one sample
available for inspection. However, when similar pollutant data sets are available
for inspection from a large number of comparable sites, or for a variety of pol-
lutants at the same site, the information to guide the desired inferences is extend-
ed. One way to shed further light on the representativeness of a nonpoint source
data set is to compare the rainfall statistics for the monitored events with similar
statistics computed from long-term rainfall records for the site. If the average
rainfall volume for monitored storms is, for example, 0.80" while the long-term
average for all storms is only 0.40,.", one can be reasonably certain that the mon-
itored storms are not representative of the total population of storms at that site.
For reasons just alluded to, as well as the fact that an integral part of the
assessment of the impact of storm loads on receiving water quality is the statis-
tical evaluation of rainfall records, a program to summarize the important rainfall
variables was set up on FHwA's computer as a part of this effort. The purpose of
the Synoptic Rainfall Data Analysis Program (SYNOP) is to provide the user with a
tool that can be used to summarize and statistically characterize a rainfall record
in terms of its important variables (volume, intensity, duration, and time between
storms). Since hourly rainfall records of many years duration are cumbersome
and difficult to analyze, SYNOP provides an easy to use tool to facilitate the deter-
mination of, for example, seasonal trends important to the assesment of impacts
and selection of control alternatives for storm-related loads.
SYNOP summarizes the hourly rainfall data by storm events, each -with an
associated volume (inches), duration (hours), average intensity (inches/hour), time
since the previous storm (hours) as measured from the end of the previous storm,
the antecedent rainfall (inches) for the last 24, 48, 72, and 168 hours, the hours of
182
-------
missing data, and the hours that the gage did not read. A storm definition, or
interevent time, must be established to determine when, in the hourly record, a
storm begins and ends. SYNOP delineates storm events as rainfall periods
separated by a minimum number of consecutive hours without rainfall, a user
supplied number. After the entire record has been read, SYNOP computes the
statistics of relevant storm parameters by month and year and for the entire
period of record.
PRELIMINARY RESULTS
RAINFALL AND RUNOFF DATA
Let us first consider the relationship of runoff to rainfall. Figure 1 Is a
scatter plot of the rainfall and runoff expressed in inches for a site on Highway
Number 45 in Milwaukee, Wisconsin. There it can be seen that, although there is
some scatter, there does tend to be a rather strong linear correlation, supporting
the notion that a runoff coefficient might adequately characterize the site and
allow runoff quantity to be predicted from rainfall data with a known degree of
uncertainty. To examine this notion further, a scatter plot of runoff coefficient
(Rv = runoff/rainfall) versus rainfall is presented in Figure 2. The lack of any
correlation further supports the notion of a single runoff coefficient.
Although the claim for a single runoff coefficient is Indicated in the foregoing,
it is less than totally compelling. One of the reasons for the scatter might lie in
the limited sample size for that site. We now turn to a site on Highway 94 in
Milwaukee for which there are 137 monitored events. The corresponding scatter
plots are presented as Figures 3 and 4. As can be seen, the argument becomes
much more compelling.
A summary of the highway site rainfall and runoff data is presented in
Table 1. The computed statistics (mean, median, coefficient of variationthe
standard deviation divided by the mean and abbreviated as COV, and the number
of samples N) are indicated for rainfall quantity, storm duration, runoff quantity,
and runoff coefficientthe runoff divided by rainfall and abbreviated as Rv. The
results of regression analyses of the form
Runoff = A x Rain + B and Rv = A x Rain + B
are presented in Table 2. In the first case, the value of A is the desired runoff
coefficient, and one wants the residual term (B) to be small. In the second case,
the value of B Is the runoff coefficient, and one wants the value of A to be small.
As can be seen from Table 2, the foregoing tends to be the case for all but the last
two sites. The mean, median, and coefficient of variation for the runoff coefficient
for each site are presented for comparison purposes in Table 2.
Since one would expect that the percent impervious area will affect the run-
off coefficient for a site, this is examined in Figures 5 and 6. Taken together, these
figures suggest that, lacking any other information, it is not unreasonable to assign
a runoff coefficient equal to the percent impervlousness for a site, with an upper
bound of around 0.9 or so to account for initial abstraction.
183
-------
RO
0.90-r
0.80
0.70-
0.60
0.50
0.404
0.30
0.204 . .
0.10-
0.00
. *
H 1-
_, ^
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60
Rainfall
Figure 1. Runoff (RO) versus rainfall for Milwaukee Highway 45
1.60
1.40
1.20
1.00
Rv 0.80-
0.60
0.40
0.20-
O OA-
*
.
* *
*
* *. * *
^ ^
t
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60
Rainfall
Figure 2. Runoff coefficient (Rv) versus rainfall for Milwaukee Highway 45
184
-------
RO
1.40
1.20
1.00
0.80
0.60
0.40
0.20
**
0.00 44MC^ «-
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40
Rainfall
Figure 3. Runoff (R0> versus rainfall for Milwaukee Highway 94
1.60-
1.40
1.20
1.00 «»
Rv
0.80
0.60
0.40
0.20
«*
*
«
.* * * .*
» ***** *
* «
«
0.00 0.20 0.40 0.60 0.80
Rainfall
1.00 1.20 1.40
Figure 4. Runoff coefficient (Rv) versus rainfall for Milwaukee Highway 94
185
-------
TABLE 1. SUMMARY HIGHWAY SITE RAINFALL/RUNOFF DATA
RAIN
On.)
DUUTMX
(hr)
OJNOFF
(In.]
nun
tadm
COV
N
ritm
1Mi«n
COV
N
rw,
tadltn
COV
N
Mtm
tedltn
COV
N
Dffivtr r
1-25
042
027
1,14
16
523
1.93
251
16
0.194
0.066
2.01
16
0.36
0.31
054
16
III*****
1-94
0.27
0.14
1.64
137
024
0.11
169
139
0.84
0.81
0.28
137
NHwlItt Si
MO
0.63
0.44
i.se
31
5.34
2.76
1.66
31
0.34
0.16
1.97
31
0.41
045
0.58
31
crimmto IH1wM*M n
Kwy.SO Hwy.45
052
0.32
Ut
34
0.45
026
1.43
34
042
0.91
0.18
34
0.54
0.37
1.0S
29
729
3.34
1.94
29
0.23
0.13
1.53
29
0.42
0.35
0.60
29
ltlwiuk«« KKTlibirg
I-79S 1-flHPrUI)
0.63
0.39
126
35
556
328
1.36
35
0.53
0.32
1.30
34
0.66
0.65
0.14
35
0.71
0.49
1.07
21
O.IB
0.02
6.63
21
0.11
0.04
2.32
21
Eflind Htrrbturg
UK l_Al /Ok t\
i u«j 1^0 1 \KH.U
0.86
0.66
0.64
36
0.64
0.33
2.33
39
0*7
0.47
1.03
36
0.76
0.61
0.74
23
6.14
4.35
1.58
23
0.40
0.16
2.03
23
0.44
0.31
1.72
23
RUNOFF QUALITY DATA
To compare the runoff data from the highway sites in the data base, data for
six of the eighteen water quality parameters are presented In Table 3. Since data
from NURP and other nonpoint source studies have indicated that data from indivi-
dual sites tends to be lognormally distributed, the summary statistics given in
Table 3 were computed in the same way as the summary statistics for an indivi-
dual site. In order to provide a basis for comparison, the median values computed
from the NURP data base are also presented in Table 3, along with the ratio of the
highway median values to them. Thus, it can be seen that the concentrations of
suspended solids and phosphorus in highway runoff are virtually the same as
those in urban runoff, concentrations of COD and zinc are slightly higher In high-
way runoff, and concentrations of lead and TKN are significantly higher in high-
way runoff. In reviewing these numbers, it is important to remember that we
TABLE 2. RUNOFF/RAINFALL REGRESSIONS
Milwaukee. 1-94 Runoff-
Milwaukee, 1-795 Runoff-
Socremen (o. Hwy 50 Runoff-
Milwaukee, Hwy 45 Runoff-
Nashville, 1-40 Runoff-
Oanver, 1-25 Runoff-
Hvrljfigro Ph II. 1-81 Runofr-
EHond, 1-85 Runoff-
Harrlsourg (>h 1.1-81 Runoff-
0.86 RAIN- 0.00 Rv- 0.04 RAIN » 0.83
0.84 RAIN O.Ot Rv- 0.06 RAIN * 0.80
0.88 RAIN- 0.02 Rv- 0.06 RAIN » 0.79
0.45 RAIN- 0.02 Rv- -0.02 RAIN » 0.43
0.33 RAIN- 0.04 Rv- -0.02 RAIN* 0.41
0.60 RAIN- 0.06 Rv- 0.37 RAIN » 0.21
0.29 RAIN- 0.08 Rv- 0.IS RAIN- 0.01
0.94 RAIN- 0.20 Rv- 0.20 RAIN » 0.41
0.84 RAIN- 0.2S Rv- 0.39 RAIN » 0.12
Maen
0.84
0.66
0.82
0.42
0.42
0.36
0.11
0.67
0.44
Median
0.61
0.85
0.81
0.35
0.35
0.31
0.04
0.47
0.31
COV
0.26
0.14
0.16
0.66
0.58
0.54
2.32
1.03
1.72
186
-------
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Runoff Coefficient
Figure 5. Percent imperviosness versus mean runoff coefficient
arc talking about concentrations and not loads. On a load basis, highways would
be much smaller contributers, because they account for only a small percentage of
the land in an urban area.
Ratio techniques are frequently used to facilitate comparison among similar
data sets, since they provide a simplistic form of normalization. In the present
case we have divided the median value for a pollutant at a sjte by the median
value for all of the sites to form the ratio. A value greater than unity indicates
that that particular site is "dirtier" than average, while a value less than unity
indicates that the site is "cleaner" than average. The results are presented In
0.00 O.!0 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90
Runoff Coefficient
1.00
Figure 6, Percent imperviousness versus median runoff coefficient
187
-------
TABLE 3. SITE MEDIAN DATA SUMMARY AND COMPARISON
SITE SS COD TKN TP04 Pb Zn
(mg/l)(mg7l)(mg/l)(mg/l)(mg/l)(mg/l)
DENVER 410 289 3.38 0.82 0.68 0.62
MILWAUKEE HWY 795 183 130 2.52 0.38 2.03 0.46
LOS ANGELES 172 196 4.22 0.45 0.99 0.55
MILWAUKEE HWY 45 343 134 2.76 0.45 0.88 0.44
NASHVILLE 190 113 1.90 1.69 0.41 0.26
MILWAUKEE HWY 94 161 122 3.20 0.30 0.90 0.52
WALNUT CREEK 224 120 224 0.41 0.75 0.30
HARRISBURG (Ph. II) 184 34 2.20 1.08 0.03 0.17
SACRAMENTO 90 51 1.90 0.12 0.28 0.27
HARRISBURG (Ph. I) 31 31 1.20 0.29 0.10 0.06
EFLANO 19 49 250 0.13 0.01 0.06
BROWARD COUNTY 9 38 0.68 0.06 0.23 0.07
Mean 221 113 2.46 0.55 1.03 0.35
Median 108 86 2.18 0.35 0.31 0.24
COY 1.78 0.85 0.52 1.21 3.21 1.09
NURP MEDIAN 100 65 0.68 0.33 0.14 0.16
HIGHWAY/NURP 1.08 1.32 3.21 1.07 2.18 1.48
Table 4. The last column in Table 4 (labeled mean) is a sort of site cleanliness
index, formed by simply averaging the ratio values for each of the six pollutants at
the site.
Based upon data in the fixed site data base, we have chosen to simplistically
group the highway sites into one of two categories, urban and rural, and repeat
the ratio analysis just described. The results are presented in Tables 5 and 6.
From these data, it appears that median pollutant concentrations in runoff from
urban highway sites are from two to four times greater than those found in ur-
ban runoff, while those from rural highway sites tend to be around one quarter to
a little over one half of that found in urban runoff. TKN is an exception to this
last observation, and the data will be examined more carefully to attempt to find
an explanation.
Many workers have tried to relate pollutant concentrations in runoff to sus-
pended solids values, attempting to use the latter as an indicator parameter. We
have done this on an event basis, and sample results for the Denver highway site
are presented in Table 7. The relatively low coefficients of variation suggest that,
for this site at least, such an approach has some validity. Summary results for
all of the urban highway sites are given in Table 8. Only the coefficient of varia-
tion for lead seems surprisingly high. The two suspect values are those for the
Milwaukee High-way 795 and the Harrisburg Phase II sites, and they will be exam-
ined critically in subsequent analyses. Similar results for the rural highway sites
are given in Table 9.
188
-------
TABLE 4. SITE MEDIAN RATIO SUMMARY AND COMPARISON
SITE SS COD TKN TP04 Pb Zn MEAN
DENVER 3.78 3.36 1.55 2.32 2.23 2.62 2.64
MILWAUKEE HWY 795 1.69 1.51 1.16 1.08 6.65 1.94 2.34
LOS ANGELES 1.59 2.28 1.94 1.27 3.24 2.32 2.11
MILWAUKEE HWY 45 3.17 1.55 1.27 1.26 2.87 1.86 2.00
NASHVILLE 1.75 1.31 0.87 4.77 1.34 1.10 1.86
MILWAUKEE HWY 94 1.49 1.42 1.47 0.85 2.95 2.20 1.73
WALNUT CREEK 2.07 1.40 1.03 1.16 2.46 1.27 1.56
HARRISBUR6CPh.il) 1.70 0.40 1.01 3.06 0.09 0.71 1.16
SACRAMENTO 0.83 0.59 0.87 0.34 0.92 1.14 0.78
HARRISBURG (Ph. I) 0.28 0.36 0.55 0.82 0.32 0.26 0.43
EFLAND 0.18 0.57 1.15 0.37 0.04 0.25 0.42
BROWARD COUNTY 0.08 0.44 0.31 0.17 0.75 0.30 0.34
TABLE 5. URBAN SITE MEDIAN DATA SUMMARY AND COMPARISON
SITE SS COD TKN TP04 Pb Zn
(mg/1) (mg/l) (mg/1) (mg/l) (mg/1) (mg/1)
DENVER 410 289 3.38 0.82 0.68 0.62
MILWAUKEE HWY 795 183 130 2.52 0.38 2.03 0.46
LOS ANGELES 172 196 4.22 0.45 0.99 0.55
MILWAUKEE HWY 45 343 134 2.76 0.45 0.88 0.44
NASHVILLE 190 113 1.90 1.69 0.41 0.26
MILWAUKEE HWY 94 161 122 3.20 0.30 0.90 0.52
WALNUT CREEK 224 120 2.24 0.41 0.75 0.30
HARRISBURG (Ph. II) 184 34 2.20 1.08 0.03 0.17
Mean 234 149 2.81 0.70 1.31 0.42
Median 220 124 2.72 0.59 0.55 0.38
COV 0.36 0.67 0.27 0.66 2.14 0.47
NURP MEDIAN 100 65 0.68 0.33 0.14 0.16
HIGHWAY/NURP 2.20 1.91 4.00 1.78 3.94 2.40
189
-------
TABLE 6. RURAL SITE MEDIAN SUMMARY AND COMPARISON
SITE SS COD TKN TP04 Pb Zn
(mg/l) (mg/l) (mg/1)(mg/l)(mg/l)(mg/l)
SACRAMENTO 90 51 1.90 0.-12 0.28 0.27
HARRISBURGCPh. I) 31 31 1.20 0.29 0.10 0.06
EFLAND 19 49 2.50 0.13 0.01 0.06
BROWARD COUNTY 9 38 0.68 0.06 0.23 0.07
Mean 42 43 1.65 0.16 0.26 0.12
Median 26 41 1.40 0.13 0.09 0.09
COV 1.24 0.23 0.62 0.71 2.64 0.85
NURP MEDIAN 100 65 0.68 0.33 0.14-0.16
HI6HWAY/NURP 0.26 0.64 2.06 0.39 0.67 0.56
TABLE 7. DENVER SITE SOLIDS RATIO RESULTS
EVENT SS/SS COD/SS 100TKN/SS 100TP04/SS lOOPb/SS lOOZn/SS
1
3
4
5
6
7
6
9
10
11
12
13
H
15
16
Mean
Median
COV
N
1.00
1.00
1.00
1. 00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1. 00
0.00
15
0.84
0.56
0.42
0.56
0.72
0.62
0.60
0.52
0.63
1.52
1.29
0.43
0.67
0.88
0.58
0.75
0.70
0.37
15
1.27
1.12
0.82
0.29
0.59
0.42
0.61
1.36
0.34
0.95
3.20
089
082
1.05
1.03
0.99
0.82
0.66
15
0.29
023
0.19
0.12
0.17
0.15
0.17
0.14
0.19
0.19
0.48
0.19
0.21
0.20
0.25
0.21
0.20
0.34
15
0.18
0.19
0.21
0.14
0.14
0.16
0.18
0.17
0.16
0.17
0.16
0.09
0.19
0.22
0.15
0.17
0.16
0.22
15
021
0.16
0.13
0.11
0.13
0.12
0.12
0.15
0.15
0.22
0.25
0.13
0.15
0.21
0.13
0.16
0.15
0.26
15
190
-------
TABLE 8. URBAN SITE SOLIDS RATIO RESULTS
SITE SS/SS COD/SS 100TKN/SS IOOTP04/SS lOOPb/SS lOOZn/SS
DENVER 1.00 0.70 0.82 0.20 0.17 0.15
MILWAUKEE HWY 795 1.00 0.71 1.38 0.2! 1.11 0.25
LOS ANGELES 1.00 1.14 2.45 0.26 0.58 0.32
MILWAUKEE HWY 45 1.00 0.39 0.80 0.13 0.26 0.13
NASHVILLE 1.00 0.59 1.00 0.89 0.22 0.14
MILWAUKEE HWY 94 1.00 0.76 1.99 0.19 0.56 0.32
WALNUT CREEK 1.00 0.54 1.00 0.18 0.33 0.13
HARRISBURG (Ph. II) 1.00 0.18 1.20 0.59 0.01 0.09
Mean 1.00 0.65 1.34 0.33 0.59 0.19
Median 1.00 0.56 1.23 0.27 0.25 0.17
COV 0.00 0.59 0.42 0.73 2.15 0.50
TABLE 9. RURAL SITE SOLIDS RATIO RESULTS
SITE SS/SS COD/SS 10OTKN/SS 10OTP04/SSI OOPb/SS 10OZn/SS
SACRAMENTO 1.00 0.57 2.11 0.13 0.31 0.30
HARRISBURG (Ph. I) 1.00 1.02 1.33 0.32 0.11 0.07
EFLAND 1.00 2.58 2.78 0.14 0.01 0.06
BROWARO COUNTY 1.00 4.22 0.76 0.07 0.26 0.08
Mean 1.00 2.38 1.83 0.18 0.29 0.13
Median 1.00 1.58 1.56 0.14 0.10 0.10
COV 0.00 1.12 0.62 0.71 2.64 0.85
TABLE 10. URBAN SITE COEFFICIENT OF VARIATION RESULTS
SITE SS COD TKN TP04 Pb Zn
DENVER 0.60 0.71 0.76 0.51 0.56 0.59
MILWAUKEE HWY 795 1.13 1.55 0.87 0.77 0.87 0.94
MILWAUKEE HWY 45 0.59 0.58 0.64 0.60 0.62 0.51
NASHVILLE 0.57 0.60 1.02 0.56 0.90 0.46
MILWAUKEE HWY 94 1.04 0.74 0.50 0.63 1.30 0.84
HARRISBURG (Ph. 11) 1.16 0.40 0.51 0.52 3.93 0.51
Mean
Median
COY
191
0.86
0.81
0.36
0.77
0.70
0.47
0.72
0.69
0.29
0.60
0.59
0.15
1.36
1.06
0.81
0.65
0.62
0.30
-------
Thus far we have only been dealing with pollutant concentration median
values. The same types of analyses could be performed on the site coefficients of
variation, and an example for the urban highway sites is given in Table 10. Here
again, the lead values for the Harrisburg Phase II site seem suspect and warrant
further examination. Otherwise, the results seem quite reasonable and fairly
consistent.
The "bottom line" of this preliminary data analysis is presented in Table 11.
If a planner has to estimate the pollutant load coming from an ungaged highway
site, the values for concentration and coefficient of variation for each pollutant
that are given there represent the best initial estimates for screening purposes.
As our work continues, we plan to refine the foregoing analyses and to add
the remaining twelve pollutants as well as other highway sites. We also plan a
rather extensive data quality assurance and quality control effort (unscreened raw
data have been used up to this point). Hopefully, this will reduce the number of
seeming anomalies pointed out earlier.
TABLE 11. HIGHWAY RUNOFF CHARACTERISTICS
URBAN RURAL ALL COEFFICIENT
SITES SITES SITES OF VARIATION
SS (mg/1) 220 26 108 0.8-1.0
COD (mg/1) 124 41 86 0.5-0.8
TKN (mg/1) 2.72 1.4 2.18 0.7-0.9
TP04 (mg/1) 0.59 0.13 0.35 0.6-0.9
Pb (mg/1) 0.55 0.09 0.31 0.7-1.4
Zn (mg/1) 0.38 0.09 0.24 0.6-0.7
The work, described in this paper was not funded by the U. S. Environmental
Protection Agency and therefore the contents do not necessarly reflect the views of
the Agency and no official endorsement should be inferred.
192
-------
EFFECTIVENESS OF DETENTION/RETENTION BASINS FOR
REMOVAL OF HEAVY METALS IN HIGHWAY RUNOFF
by: Harvey H. Harper, Yousef A. Yousef, and Martin
P. Wanielista
Department of Civil Engineering and Environmental
Sciences, University of Central Florida
Orlando, Florida 32816
ABSTRACT
The movement and fate of heavy metal inputs (Cd, Zn, Mn, Cu, Alr Fe, Pb,
Ni and Cr) from highway runoff were investigated in a three-year study on 5
1.3 hectare retention facility near the Maitland Interchange on Interstate 4,
north of Orlando, Florida. Stormwater characteristics were compared with
average retention pond water quality to determine removal efficiencies for
heavy metals within the pond. A total of 138 sediment core samples were
collected in the pond over a three-year period to investigate the horizontal
and vertical migrations of heavy metals within the pond. Core samples were
also carried through a series of sequential extraction procedures to examine
the type of chemical associations and stability of each metal in the
sediments. An apparatus was built which allowed sediments to be incubated
under various conditions of redox potential and pH to investigate the effects
of changes in sediment conditions on the stability of metal-sediment
associations.
INTRODUCTION
Within the past decade, a substantial amount of research has accumulated
relating to the water pollution caused by the operation of motor vehicles.
This concern is based largely on the potential aquatic toxicity of heavy
metals such as lead, zinc, and chromium. Heavy metals have been proposed by
several researchers as the major toxicant present in highway runoff samples
(Shaheen, 1975; Winters and Gidley, 1980). Many heavy metals are known to be
toxic in high concentrations to a wide variety of aquatic plants and animals
{Wilber and Hunter, 1977).
On a nationwide basis, the two most commonly used techniques for
management of highway runoff are roadside swales and detention/retention
facilities. As these facilities receive continual inputs of Stormwater
193
-------
containing heavy metals, processes such as precipitation/ coagulation,
settling, and biological uptake will result in a large percentage of the
input mass being deposited into the sediments. However, no previous
definitive studies have been conducted to determine the fate of toxic
species, especially heavy metals, in these stormwater management systems. In
particular, no studies have been conducted to investigate if physical and
chemical changes which may occur in these systems over time may mobilize
certain species from the sediment phase back into the water phase.
The purpose of this research was to investigate the fate of heavy metals
within stormwater management systems. The site selected for these
investigations was a series of stormwater management facilities located at
the Maitland Interchange on Interstate 4 north of the city of Orlando,
Florida. A retention pond (West Pond) with relatively defined inputs and
outputs was chosen as the primary study site.
SITE DESCRIPTION
The site selected for this investigation is the Maitland Interchange on
Interstate 4. This interchange, located north of the city of Orlando, was
constructed in 1976 (Figure 1). Three borrow pits dug to provide fill for
constructing the overpass remain to serve as stormwater detention/retention
facilities. The ponds are interconnected by large culverts and the
northwestern (Pond B) has the capability to discharge to the southwestern
(referred to hereafter as the West Pond) when design elevations are exceeded.
However, under normal conditions, the only input into the West Pond is by way
of a 45 cm concrete culvert that drains much of the Maitland Boulevard
overpass. Discharge from the West Pond travels to Lake Lucien through a
large culvert. A flashboard riser system regulates the water level in the
Wast Pond, and a discharge rarely occurs to Lake Lucien.
The West Pond has an approximate surface area of 1.3 ha and an average
depth of 1.5 m. The pond maintains a large standing crop of filamentous
algae, particularly Chara, virtually year round. Because of the shallow
water depth and large amount of algal production, the pond waters remain in a
well oxygenated state. The sediment material is predominately sand which is
covered by a 1 cm layer of organic matter.
Maitland Boulevard crosses over Interstate 4 by means of a bridge
overpass created during construction of the Interchange. The Maitland
Boulevard bridge consists of two sections, one carrying two lanes of
eastbound traffic and an exit lane, the other carrying two lanes of westbound
traffic and another exit lane. Traffic volume on Maitland Boulevard is
approximately 12,000 average daily traffic (ADT) eastbound and 11,000
westbound. Traffic volume on 1-4 through the Maitland Interchange is
approximately 42,000 ADT eastbound and westbound each.
194
-------
I Wait land Blvd
West Pond ^
It2
Exit Culvert to
Lake Lucien
Lake Lucien
SCALE: icm«50m
Figure I. Study Site at Malt!and Interchange.
195
-------
FIELD INVESTIGATIONS
Field investigations conducted during 1982 and 1983 at the West Pond
were divided into the following tasks: 1) determination of the quantity of
heavy metals entering the West Pond by way of stormwater runoff; (2)
determination of the average heavy metal concentrations in the retention
basin water; (3) assessment of the accumulation of heavy metals in the
sediments of the pond; and (4) monitoring of heavy metal concentrations in
ground waters beneath the retention basin. To determine the quantity of
heavy metals entering the West Pond by way of stormwater runoff, an Isco
automatic sampler was installed on the 45 on stormsewer line. Flow-weighted
composite samples were collected over a 1 year period for 16 separate storm
events representing a wide range of rainfall intensities and antecedent dry
periods. Samples were analyzed for heavy metals using argon plasma emission
spectroscopy.
Water samples were collected on a biweekly basis for 1 year in the West
Pond to document average retention pond water quality. Each of the five
samples was analyzed separately for the heavy metals listed, and an average
value was calculated for each metal on each sampling date.
The accumulation and vertical distribution of heavy metals in the
sediments was examined by collection of a series of 2.5 cm diameter core
samples to a depth of 15 cm. Forty-three separate core samples were
collected in the 1.3 ha West Pond, and metal concentrations in sediment
layers 0-1 cm, 1-3.5 cm, 3.5-6.0 on, 6.0-8.5 cm and 8.5-13.0 were measured
for each core sample. Metal concentrations in the 0-1 on layer were used to
investigate horizontal movement of heavy metals from the point of discharge
into the pond. Average metal concentrations in each of the sediment layers
were used to determine the extent of vertical migration.
Multiport monitoring wells were installed at the locations indicated in
Figure 1 to investigate the possibility of groundwater contamination by
leaching of heavy metals from the stormwater management system. Two of the
monitoring wells were installed at the edges of the West Pond with the
remaining three installed at various locations surrounding the stormwater
management system. The wells were designed so that all of the sample ports
were housed in a single casing to minimize soil disturbance and reduce
recovery time for obtaining representative groundwater samples compared to
other monitoring well designs such as cluster wells.
All wells were installed to a depth of 6 meters with sample ports at 0.1
m, 0.5 m, 1.0 m, 3.0 m, and 6.0 m below the average water table depth in the
area of the well. Groundwater samples were collected from each sample port
on a monthly basis using a peristaltic pump. Approximately 10 liters of
groundwater were pumped and discarded from each port before a sample was
collected. Collected groundwater samples were analyzed for heavy metals as
described previously.
196
-------
CHARACTERISTICS OF HIGHWAY RUNOFF AT THE MAITLAND SITE
A total of 16 storm events/ including a total of 150 separate runoff
samples were collected and analyzed over this period for dissolved and total
heavy metals. Total rainfall amounts for sampled storm events ranged fron
0.33 to 3.23 inches with antecedent dry periods of 0.24 to 25.4 days.
Average flow rates in the 45 cm storrasewer, an indirect measure of rainfall
intensity/ ranged from 0.085 to 59.4 liters/sec.
A summary of mean flow-weighted heavy metal concentrations measured at
the Maitland site is given in Table 1. Measured concentrations of heavy
metals in highway runoff collected at the Maitland Interchange during 1983-84
showed considerable variability between storm events as well as during storm
events. Average dissolved concentrations of all heavy metals, with the
exceptions of iron and aluminum, were less than 70 ug/1, with nickel,
chronium, manganese, and cadmium less than 3 ug/1. Measured mean
concentrations of total metal species were in excess of 100 ug/1 for lead,
aluminum and iron, while nickel, chromium, manganese, and cadmium were all
present in average total metal concentrations of 10 ug/1 or less. In
general, the variability of mean flow-weighted dissolved metal concentrations
appears to be much less than that observed for mean total concentrations,
with most dissolved species exhibiting a five-fold difference in range of
concentrations, while total concentrations exhibited over a ten-fold range in
most cases.
Of the heavy metals which were measured, the following orders were
observed for mean concentrations of dissolved and total metal species:
Dissolved: Al > Fe > Zn > Pb > Cu > Mn > Ni = Cr > Cd
Total: Al > Fe > Pb > Zn > Cu > Mn > Cr > Ni > Cd
However, the metal species aluminum, iron and manganese are common
constitutents of soils and may not be correlated with vehicle usage and
highway operation, as would be expected for lead, nickel, chromium, copper,
zinc, and cadmium. Therefore, the most common vehicle related heavy metals
found in highway runoff at the Maitland site were lead, zinc, and copper in
ratios of 4.70:1.91:1.0, respectively, for total concentrations, and ratios
of 0.85:1.04:1.0, respectively, for dissolved species. Together these three
metals accounted for approximately 91 percent of the dissolved heavy metals
present and 94 percent of the total metal concentrations, excluding aluminum,
iron and manganese.
With the exceptions of lead, iron, manganese, and aluminum, all heavy
metals in the highway runoff samples appeared to be present predominantly in
a dissolved form. Cadmium, nickel, and copper were all present in dissolved
fractions which were near 75 percent of the total metal measured. On the
other extreme, lead, iron, manganese, and aluminum were predominantly
particulate in nature with dissolved fractions of only approximately 20
percent. Zinc and chromium appeared to be approximately equal in dissolved
and particulate forms.
197
-------
TABLE 1. HEAVY METAL CONCENTRATIONS IN SEQUENTIAL HIGHWAY RUNOFF SAMPLES
COLLECTED AT THE MAITLAND WEST POND DURING 1983-84
00
HEAVY
METAL
LEAD
ZINC
COPPER
NICKEL
CHROMIUM
IRON
ALUMINUM
MANGANESE
CADMIUM
METAL
SPECIES
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
NUMBER
OF
SAMPLES
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
STANDARD
MEAN DEVIATION
(ug/1)
33.0
181
40.0
73.9
28.6
38.6
2.5
3.4
2.5
4.2
77.9
378
125
561
2.70
9.53
1.7
2.2
40.2
331
42.6
71.2
24.7
28.8
2.4
2.8
2.2
3.2
59.7
354
124
563
6.7
10.8
2.0
2.4
RANGE OF
VALUES
(ug/1)
7.0 -
11.0 -
1.0 -
5.0 -
6.0 -
6.0 -
0.5 -
0.5 -
0.5 -
0.5 -
11.0 -
44.0 -
19.0 -
53.0 -
a -
<1 -
<1 -
<1 -
413
3,596
324
372
175
176
15
18
16
18
466
2,172
832
3,499
59
62
12
12
PERCENT
DISSOLVED
(%)
18.2
54.1
74.1
73.5
59.5
20.6
22.3
28.3
77.3
pH
101
6.30
0.88
4.95 - 8.49
-------
Probability distributions of mean flow-weighted heavy metal
concentrations in the 16 measured storm events were examined by plotting
flow-weighted mean metal concentrations for each measured event on
probability paper. Total concentrations of heavy metals appeared to best
approximate a straight line relationship with a log-normal distribution of
concentrations for the events measured. Dissolved concentrations of heavy
metals also seem to approximate a log-normal distribution. However/ cadmium
appears to exhibit a convex curvilinear relationship.
A "first flush" effect was observed for total concentrations of lead,
zinc, iron, and aluminum. In general, approximately 50 percent of the total
mass of these metals was found to be transported during the first quarter of
a storm event, 25 percent during the second quarter, and the remaining 25
percent divided between the third and fourth quarters. This trend was not
observed for total concentrations of the other metal species or for dissolved
species of any measured metals.
FATE OF HEAVY METALS ENTERING THE MAITLAND POND
Although stormwater inputs into the pond were characterized by a
considerable degree of variability, heavy metal concentrations measured in
the pond were, in general, relatively consistent and low in value. Dissolved
concentrations of all metal species in the pond with the general exception of
aluminum, were never found to exceed 50 ug/1 with dissolved concentrations of
cadmium, zinc, manganese, nickel, and chromium rarely exceeding 10 ug/1.
Total metal concentrations followed a similar pattern with only manganese,
aluminum, iron, and on one occasion lead, exceeding 100 ug/1 on any given
sample day at any of the five sampling stations within the retention pond.
The mean pH value of the retention pond water was 7.46 with a range of
6.62-8.46. Measurements of dissolved oxygen indicated an aerobic water
column on all sample dates. The mean value for dissolved oxygen was 5.6 mg/1
with measurements of ORP generally in excess of 500 mv.
The Maitland pond was found to be very effective in removal of heavy
metal inputs from highway runoff. A comparison of summary statistics for
highway runoff and Maitland pond water is given in Table 2. Heavy metal
concentrations in this table have been given in terms of the concentrations
of dissolved and particulate species rather than dissolved and total species.
This was done so that the removal characteristics of soluble and particulate
species could be examined separately.
Upon entering the Maitland retention pond, chemical, physical, and
biological processes begin to occur which, for most metal species, results in
substantial reductions in concentrations. The most noticeable removals for
heavy metals occurs for the particulate species. Particulate species of lead
and zinc are reduced in excess of 95 percent, cadmium and iron near 85
percent, with copper and aluminum averaging near 75 percent. Reductions of
particulate nickel and chromium, however, were much less, with a removal of
only 25-35 percent. The order for reduction of particulate metal species
upon entering the West Pond is:
199
-------
TABLE 2. COMPARISON OP SUMMARY STATISTICS FOR HIGHWAY RUNOFF
AND MAITLAND RETENTION POND MATER
HEAVY
METAL
LEAD
ZINC
COPPER
NICKEL
CHROMIUM
IRON
ALUMINUM
MANGANESE
CADMIUM
MKTAL
SPECIES
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
STORMHATER
MEAN
(ug/1)
33.0
148
40.0
33.9
28.6
10.0
2.5
0.9
2.5
1.7
77.9
300
125
436
2.7
6.8
1.7
0.5
RUNOFF1
PERCENT
OF TOTAL
18.2
81.8
54.1
45.9
74.1
25.9
73.5
26.5
59.5
40.5
20.6
79.4
22.3
77.7
28.3
71.7
77.3
22.7
RETENTION
MEAN
(ug/1)
15.0
7.2
4.7
1.3
14.4
2.3
1.6
0.6
2.2
1.3
18.4
44.7
58.0
98.0
4.5
12.6
0.73
0.09
POND WATER2
PERCENT
OF TOTAL
67.6
32.4
78.3
21.7
86.2
13.8
72.7
27.3
62.9
37.1
29.2
70.8
37.1
62.9
26.3
73.7
89.0
11.0
PERCEWI1
REMOVAL
IN POND
54.5
95.1
88.3
96.2
49.7
77.0
36.0
33.0
12.0
23.5
76.4
85.1
53.6
77.5
-66.7*
-85. 3 J
57.1
82.0
STATE OF FLA.
CLASS III
VOTERS CRITERIA
(2/1/83)
30 (Total)
30 (Total)
30 (Total)
100 (Total)
50 (Total)
1000 (Total
None
None
0.8 (Total)
NOTES: 1. Number of observations = 150
2. Number of observations = 30
3. Denotes an increase
-------
Zn > Pb > Fe > Cd > Al > Cu > Ni > Cr » Mn
Dissolved forms of zinc were removed to the greatest degree with an
average removal of almost 90 percent. Dissolved iron was removed at an
efficiency of 75 percent, followed by lead, copper, aluminum, and cadmium
with removals of dissolved species ranging 50-60 percent. Removal
efficiencies for dissolved nickel and chromium were very poor, with removals
of only 36 and 12 percent respectively. The order for reduction of dissolved
runoff species upon entering the West Pond isi
Zn > Fe > Cd > Pb = Al > Cu > Ni > Cr » Mn
Studies by Yousef et al. (1985) as well as observations during this
research suggest that the removal of dissolved metal species is rapid with as
much as 90 percent removal occurring in four days. In the research by
Yousef, et al., isolation chambers were placed in a newly constructed
retention pond near Epcot Center, the isolation chambers were constructed of
inverted polyethylene 200-liter barrels placed on the sediments - which
isolated a 0.25 m area of the sediment and the overlying water column.
Chambers were constructed with both open and sealed bottoms to investigate
the effects of sediments on heavy metal concentrations. The chambers were
first installed then dosed with a solution of heavy metals. Periodic samples
were collected and analyzed for metal content. A summary of their work is
presented in Table 3.
Soluble concentrations of copper, zinc, iron, and lead were added to two
of the test chambers in concentrations between 0.5 and 1 mg/1 on 4/1/83.
However, when the next sample was collected on 4/4/83, concentrations of
copper, zinc, and lead had been substantially reduced by an average of 90
percent. By 4/18/83 (the next sample collection date) concentrations in the
closed chamber were indistinguishable from the control which received no
metal additions. No change was noted either with or without sediment contact
in these metal concentrations throughout the test period, even when anaerobic
conditions were established.
ACCUMULATION OF HEAVY METALS IN THE SEDIMENTS
HORIZONTAL DISTRIBUTION OF HEAVY METALS
Distributions of heavy metals in the top 1 cm of of the Maitland pond
sediments suggest that upon entering the receiving water body, the majority
of heavy metals associated with highway runoff settle out and are deposited
near the point of input for the runoff. The distributions of selected heavy
metals as a function of distance from the 45 cm RCP inlet are shown in Figure
2. This tendency was most obvious for lead and zinc which peaked in sediment
concentrations at a distance of only 15 m from the inlet followed by a rapid
decline in concentrations with increasing distance. Deposition patterns of
the other metals measured were much less pronounced than those observed for
lead and zinc. Chromium appeared to reach peak sediment concentrations at a
distance of 30 m from the inlet with increases and decreases much less rapid
than those observed for lead and zinc. Copper, nickel, and manganese did not
201
-------
TABLE 3. UPTAKE AND RELEASE OF HEAVY METALS INSIDE ISOLATION
CHAMBERS AT EPCOT POND
ro
o
ro
CHAMBER
TOTAL METALS CONCENTRATION BY DATE IN 1983 (Ug/1)
DESIGNATION
Sediment
Contact-
Control (no
metals added)
Sediment
Contact
(metals
added)
No Sediment
Contact
(metal s
added)
Pond
METAL
Cu
Zn
Fe
Pb
Cu
Zn
Fe
Pb
Cu
Zn
Fe
Pb
Cu
Zn
Fe
Pb
4-1
23
7
596
23
21
14
744
24
23
13
401
27
22
12
6U3
28
<
4-1*
AFTER
-
683
8b7
790
904
590
749
468
724
-
-_ni f f nc«
4-4
15
9
614
32
71
82
648
93
61
50
720
56
3b
4
855
52
4-18.
17
5
455
26
17
10
499
23
19
3
617
30
16
0
454
21
>
was Supplied
*
After addition of nutrient and heavy metal solution.
SOURCE: Yousef et al. (1985)
-------
250
ro
O
CO
30
45 60 75
DISTANCE FROM OUTFALL (m)
9O
105
120
Figure 2. Sediment Concentrations of Selected Heavy Metals In the Top 1 cm of the Maltland
West Pond as a Function of Distance from the Outfall.
-------
ro
20
30 40 50 60 70 80 90
SEDIMENT CONCENTRATION (ug/g DRY WT.)
100
110
Figure 3. Attenuation Of Heavy Metals in the Bottom Sediments of the Haitland Pond for
Samples Collected 10/15/82.
-------
TABLE 4. StMlARY OF BACKGROUND AND RUNOFF RELATED MEEAL CONCENTRATIONS
IN THE SEDIMENTS OF THE MAITLAND POND
u
1
3
b
H
SEDIMENT
DEPTH
- 1 cm:
Mean Cone.
Background
Added Cone.
- 3.5 cm:
Mean Cone.
Background
Added Cone.
.5 - 6 cm:
Mean Cone.
Background
Added Cone.
- b.S cm:
Mean Cone.
Background
Added Cone.
.5 - 13 cm:
Mean Cone.
Background
Added Cone.
MEAN SEDIMENT METAL
Cd
2.2U
U.4U
1.80
0.89
0.4U
0.49
U.56
0.40
0.16
U.45
0.40
U.US
U.66
0.40
U.26
Zn
45.4
L.b
43.9
10.7
1.5
9.2
6.49
1.50
4.99
4.64
1.50
3.14
3.16
l.bO
1.66
Cu
19.1
2.0
17.1
9.44
2.UU
7.44
7.50
2.00
b.iO
5.07
2.00
3.07
4.16
2.00
2.16
A1
49760
12058
37702
24574
1205B
12516
18256
12U5H
6198
1720b
12058
5147
12485
12058
427
CONCENTRATION (ug/g DRY NT.)
Fc
4554
654
3900
1761
654
1107
1303
654
649
874
6S4
22U
482
654
0
Pb
112.7
7.9
104.8
37.6
7.9
29.7
24.5
7.9
16.6
17.0
7.9
9.1
13.5
7.9
5.6
N1
16.5
1.98
14.5
6.49
1.9«
4.51
4.15
1. 98
2.17
4.03
1.98
2.05
3.31
1.98
1.33
Cr
44.7
7.63
37.1
19.3
7.63
11.7
15.4
7.63
7.77
10.8
7.63
3.17
4.01
7.63
0
Mn
43.9
1.6S
42.3
12.4
1.65
10.8
6.86
1.65
5,21
5.55
1.65
3.90
4.96
1.65
3.31
NO. OF
OBS.
138
6
138
6
138
6
138
6
138
6
-------
appear to exhibit pronounced peaks in sediment concentrations, but seemed to
settle out over a longer flow path length. However, in spite of the
differences in behavior, most of the metals in the runoff water entering the
Maitland pond were retained in the pond sediments within a distance of 60-90
m from the stormwater inlet.
Of the four metal species which exhibited the most rapid settling
characteristics (lead, zinc, iron, and aluminum), all but zinc had
particulate fractions in runoff which were near 80 percent of the total metal
measured. The remaining metal species (nickel, chromium, and cadmium) which
did not exhibit pronounced settling characteristics, were all present in
highway runoff at the Maitland site predominately in a dissolved form with a
small fraction of particulate species.
The results of- the horizontal analyses of heavy metals suggest important
design parameters for use in the design of retention basins to optimize
removal of heavy metals. Designs should provide physical configurations
where the flow velocity becomes very small to aid in sedimentation of
particles. The distance from points of input to the discharge point from the
pond should be maximized, and the design should minimize the possibility of
short circuiting and avoid hydraulically dead zones.
VERTICAL DISTRIBUTIONS OF HEAVY METALS
The vertical distribution of heavy metals in the sediments of the
Maitland West Pond was characterized by analysis of average sediment metal
concentrations by layer on each of the three sample dates. The vertical
distributions of selected heavy metals in the sediments of the Maitland pond
are given in Figure 3. Aluminum was the most abundant metal present in the
Maitland pond sediments at all depths, followed by iron. Lead was the third
most abundant heavy metal present^, followed by zinc and chromium, copper and
nickel, and finally cadmium. Concentrations of cadmium were generally very
small with many measured values, especially in the lower sediment depths,
approaching the limits of detection. Measured concentrations of total heavy
metals in the sediments of the Maitland pond exhibited highest concentrations
in the surface layer with a rapid decline in concentration with increasing
depth.
Background soil concentrations of heavy metals in the retention pond
area were estimated from mean soil metal concentrations in core samples
collected at depths of 3 m or greater during drilling of monitoring wells
beneath the pond. These background concentrations were subtracted from the
total sediment metal concentrations to provide an estimate of the added
accumulations as a result of inputs of highway runoff. A summary of
background and runoff related metal concentrations in the sediments of the
Maitland pond is given in Table 4.
Sediment concentrations of runoff related heavy metals were also
observed to decline rapidly with increasing depth. The rapid decline in
concentrations was found to observe an exponential decay relationship with
values of R-square in most cases in excess of 0.90 when fitted to the model:
In (C/C ) = -K x (depth). A summary of the regression of statistics for the
206
-------
serai-log model is given in Table 5. Values of K, which are a measure of the
rate of attenuation in sediment metal concentrations, indicated the following
order of attenuation of total heavy metal content in the sediment layers:
Most Rapid Least Rapid
Attenuation: Fe < Zn < Cd < Pb < Cr < Mn < Al < Ni < Cu: Attenuation
The calculated regression equations for runoff related metal
accumulations were used to estimate the extent of metal migration from runoff
related sources by estimation of the depths necessary to reduce runoff
accumulations by 90 percent and 99 percent to values which are 10 percent and
1 percent above estimated background levels. All runoff related
accumulations were reduced in concentration by 90 percent in the first 10 cm
or less.
Although the substantial majority of metal species were attenuated in
the first 5.0 cm of sediments, the depths necessary to achieve 99 percent
reductions in runoff accumulations, based on the calculated regression
equations, suggest that sediment concentrations of certain metals may be
slowly migrating to lower depths. However, the vertical extent of this
sediment-associated migration appears to be limited since all metal species
were reduced in concentration by 99 percent within 20 cm or less. These
calculations suggest a strong stability of the metal sediment associations
since, after eight years of metal accumulations in the Maitland pond, most
metals associated with sediments have remained in top 10 cm of the sediment
layer.
CURRENT STABILITY OF METAL-SEDIMENT ASSOCIATIONS
The stability of metal-sediment associations was evaluated from the
results of several different analyses. First, a sequential extraction
procedure was used to determine metal speciation in composite samples of each
of the five vertical core layers. Metal speciations were divided into
soluble, exchangeable, carbonate bound, bound to Fe/ton oxides, and organic
bound fractions. It is generally believed that the stability of the
metal-sediment associations increases in the following order: soluble <
exchangeable < bound to carbonates < bound to iron and manganese oxides <
bound to organic matter.
Fractional distributions of the total extracted heavy metals for each of
the five extracted species are presented in Table 6, Most of the metal
species tested, with the exceptions of lead, iron, and cadmium, appear to be
predominantly associated with only one major fraction. For most metals, the
dominant fraction is the one which is bound to Fe/Mn oxides. However,
cadmium is predominantly associated with the exchangeable fraction. Lead
also has a major association with this fraction. Alminum and iron appear to
have significant fractions with organic particles. Very few of the heavy
metals present in the sediments appear to be present in a soluble or
carbonate form although cadmium, zinc and nickel had soluble fractions of
approximately 10 percent. It appears certain that iron, manganese, and
207
-------
TABLE 5. SUMMARY OF REGRESSION STATISTICS FOR ATTENUATION OF
RUNOFF RELATED HEAVY METALS IN THE TOP 13 CM OF THE
THE MAITLAND POND FOR A SEMI-LOG RELATIONSHIP FOR
ALL THREE SAMPLE DATES COMBINED
HEAVY
METAL
Cd
Zn
Cu
AT
Fe
Pb
Ni
Cr
Mn
Organ i c
Content
NO. OF
OBS.
9
11
12
11
10
12
12
9
9
12
VALUE OF K FOR
"BEST- FIT" EQUATION OF
THE FORM:*
In (C/Co) = -KZ
In (Cd) =
In (Zn) =
In (Cu) =
In (Al) =
In (Fe) =
In (Pb) =
In (Ni) =
In (Cr) =
In (Mn) =
In (Org.)
-0.374 (Z)
-0.398 (Z)
-0.286 (Z)
-0.311 (Z)
-0.549 (Z)
-0.368 (Z)
-0.304 (Z)
-0.346 (Z)
-0.327 (Z)
= -0.241 (Z)
VALUE
OF
R-SQUARE
0.821
0.898
0.877
0.902
0.821
0.926
0.898
0.913
0.895
0.850
* Metal concentrations in units of ug/g; organic content in
percent; and depth (Z) in units of cm.
208
-------
TABLE 6. SPECIATION OF TOTAL HEAVY METAL CONCENTRATIONS IN THE SEDIMENTS
OP THE MAITLAND WEST POND AS A FRACTION OF THE TOTAL METAL PRESENT
HEAVY METAL
Cadmium
Zinc
ro
S Copper
Aluminum
Iron
Lead
Nickel
Chromium
Manganese
SOLUBLE
15
4
1
<1
<1
1
4
2
1
PERCENT OF TOTAL
EXCHANGABLE
52
1
3
<1
5
44
8
5
9
EXTRACTED METAL
CARBONATE
12
4
1
<1
<1
1
<1
1
1
CONCENTRATION
BOUND TO
Fe/Mn OX.
10
81
89
74
52
52
82
73
86
BOUND TO
ORGAN1CS
11
10
7
26
43
2
6
19
3
TOTAL
100
100
100
100
100
100
100
100
100
-------
organic content play dominant roles in regulating the mobility of metal
species in the sediments of the Maitland pond.
In addition to the speciation analyses/ an incubation apparatus was
constructed which allowed simultaneous control of pH and redox potential in
sediment suspensions to simulate metal adsorption or desorption under various
environmental conditions. Sediment suspensions were incubated at pH values
of 5.0, 6.5, and 7.5-8.0 which simulated current conditions of sediment pH in
the pond, and at redox potentials of -250 mv, 0.0 mv, 250 mv and 500 mv,
ranging from highly reduced to highly oxidized. A summary of metal release
under these conditions for selected heavy metals is given in Figures 4
through 6. In general, release of most heavy metals was less than a few
percent of the total sediment content under current conditions of pH
(7.5-8.0). The influence of pH was found to be much more important than
redox potential in regulating metal release for most metal species.
The results from the speciation and redox experiments combined with the
analyses of the sediment metal concentrations presents evidence that under
the current conditions of redox potential and pH within the sediments of the
Maitland pond, metal species, with the exceptions of cadmium, lead, and
manganese, are stable and exist in relatively immobile associations with
Fe/Mn oxides and organic matter. Lead and cadmium are apparently held to a
large degree in strong exchangeable associations. Changes in redox potential
from strongly oxidized to strongly reduced conditions did not appear to
affect the release of metals from the sediments under current pH values of
7.5-8.5. The release of most metals, except cadmium and manganese, from the
sediment phase to the water phase was substantially less than 1% of the total
metal present even after several weeks of incubation. However, cadmium and
manganese appear to be less tightly bound to sediments than other metals.
The release of both cadmium and manganese into solution from the sediment
phase during incubation is equal to approximately 5% of the total metal
present.
EFFECTS OF THE MAITLAND POND ON UNDERLYING GROUNDWATERS
A comparison of dissolved concentrations of heavy metals in the Maitland
pond water with groundwater collected beneath the pond, represented by wells
2 and 3 combined, is given in Table 7. In general, concentrations of all
heavy metals measured, except copper, were greater beneath the pond than
within the pond. For certain heavy metals such as zinc, manganese, aluminum,
and iron, measured concentrations in groundwaters were from 5 to 75 times as
great as measured concentrations in the pond water.
K
Analysis of variance procedures were used to estimate the vertical
extent of the migration of heavy metals in the aqueous phase. Zinc,
manganese, aluminum, and iron were significantly higher in groundwater
beneath the pond than in the pond at all depths tested. The extent of
significantly higher concentrations of lead extended to the 0.5-1.0 m range.
Copper was found to be significantly higher in the pond water than in
groundwater in all analyses. Average concentrations of zinc, manganese,
aluminum, and iron were found to be 4, 12, 8, and 50 times greater
210
-------
40
30-
20
O
UJ
10
***
§ o
UJ
cr
o
o
CADMIUM
±=^:
ZINC
pH-7.5-8.5
oioo
80
o
g
Q.
40
20
0
MANGANESE
pH-6.5
-100
I 1
100 200 300
REDOX POTENTIAL(mv)
pH=75-8.0
400
Figure 4. Fraction of Total Sediment Metal Concentrations of
Cadmium, Zinc, and Manganese Released at Various Values of Redox
Potential and pH.
500
211
-------
-100
0 100 200 300
REOOX POTENTIAL (mv)
4OO 500
Figure 5. Fractions of Total Sediment Concentrations of Copper,
Aluminum, and Iron Released at Various Values of Redox Potential
and pH.
212
-------
8
6
4
tu
Gl
UJ
I'
LEAD
pH= 6.5
pH=7.5-8.5
' ' ' '
' ' '
I I I
UJ 2
1
CO
UJ
CL
NICKEL
PH= 7.5-8.5
CHROMIUM
pH=6.5
= 5.0
PH = 75-8.5
-t-
-IOO 0 100 200 300 4OO 500
REDOX POTENTIAL (mv)
Figure 6. Fractions of Total Sediment Concentrations of Lead,
Nickel, and Chromium Released at Various Values of Redox Potential
and pH.
213
-------
TABLE 7. COMPARISON OF DISSOLVED CONCENTRATIONS OF HEAVY METALS IN THE POND WATER
WITH GRDUNDWATER COLLECTED BENEATH THE POND IN WELLS 2 AND 3
rv>
HEAVY
MFTAI
"ID 1 ML-
Cd
Zn
Mn
Cu
Al
Fe
Pb
Ni
Cr
pH
AVERAGE AVG. CONC.
DISSOLVED IN 0-1 cm
CONCENTRATION SEDIMENT
IN POND
(ug/D
0.73
4.82
4.47
14.4
57.9
18.4
15.0
1.62
2.18
7.46
LAYER
(ng/D*
1,015
20,938
20,246
8,809
22,949,100
2,100,285
51,977
7,610
20,615
0.1 m
1.48
22.9
53,3
9.40
709
1354
24.4
2.88
3.02
5.75
AVERAGE CONC. IN GROUNDWATER
BENEATH THE POND
0.5 m
2.06
18.5
44.2
10.3
742
834
24.8
2.63
4.15
5.92
1.0 m
1.50
18.8
23.3
9.12
192
766
20.5
1.98
2.60
5.17
3.0 m
1.17
19.1
78.9
11.7
89.2
797
11.2
1.94
1.29
4.86
6.0 m
1.00
19.5
70.2
13.1
543
1912
10.9
2.13
1.54
4.56
RATIO OF
G.W. CONC.
AT 0.1 m
TO AVG.
POND WATER
2.03
4.75
11.92
0.65
12.25
73.6
1.63
1.78
1.39
* Average sediment concentration per liter of sediment.
-------
respectively, in groundwater than in the pond water. Concentrations of
cadmium and iron in groundwaters exceeded water quality criteria for Class
III waters specified in Chapter 17-3 of the Florida Administrative Code.
Concentration ratios between the pond water and the groundwater indicate
that all metal species, except copper, have a greater affinity for the
groundwater phase than the pond phase and are leaching into groundwaters to
some degree. The order of release potential of heavy metals into groundwater
was estimated to be:
Least Most
Mobile: Cu < Cr < Pb < Ni < Cd < Zn < Mn = Al < Fe: Mobile
and was found to be inversely related to the order of attenuation for metal
species in the sediment phase. The magnitude of the release into
groundwaters was found to closely correspond to the order of release
predicted by the incubation experiments conducted under natural conditions of
pH (7.5-8.0). However, in spite of the increased metal concentrations
beneath the pond, the sediments are clearly the primary sink for heavy
metals.
TRANSPORT OF HEAVY METALS IN GROUNCWATER FLOW
One of the objectives of this research was to monitor groundwater
concentrations and flow patterns and to detect, if possible, the movement of
heavy metals which leach into groundwaters. To aid in this detection,
piezometers were installed at each well and a record of pie zone trie surface
was made approximately on a bi-weekly basis during 1983. The average
measurements are given in Table 8.
TABLE 8. AVERAGE MEASUREMENTS OF PIEZOMETRIC SURFACE AT MONITORING
WELLS AT THE MAITLAND INTERCHANGE DURING 1983
Location Piezometric Surface (m, MSL)
Well 1
Well 2
Well 3
Well 4
Wall 5
Pond
27.35
27.38
27.37
27.36
26.87
27.38
215
-------
As indicated in Table 8, little variation was measured in the
piezometric surface between each well. As a result of this very small
hydraulic gradient, horizontal movement of groundwater in the Maitland area
was calculated to be less than 10 m per year. It appears that vertical up
and down movement with changes in seasonal water table may be more important
than horizontal movement. As a result, metal contamination of groundwaters
appears to be very localized. Since the hydrologic conditions present at the
Maitland site are similar to conditions at many other sites in Central
Florida, it appears very unlikely that heavy metals from retention ponds
along highway systems in the Central Florida area will pose a pollution
hazard to nearby surface and groundwaters.
POTENTIAL FOR FUTURE MOBILIZATION OF HEAVY METALS
Natural aging processes within retention ponds as well as lakes result
in the increased deposition of organic matter to the bottom sediments
primarily as a result of the death and decay of both plant and animal matter.
As these processes occur, it has often been observed that sediments become
more reduced and decrease in pH. Although the incubation experiments
indicated that most metal species are stable and tightly bound to sediments
under current conditions of redox potential and pH, decreases in pH were
found to increase the solubility of all heavy metals tested. Changes in
redox potential produced no significant changes in release rates.
The results suggest that as the Maitland pond ages and accumulations of
organic matter in the sediments begin to cause sediment pH values to
decrease, mobilization of all metal species tested will increase and release
to groundwaters may occur. Although all metals were found to increase in
solubility with decreases in pH, the release was in general, only a small
fraction of the total sediment metals present. For zinc, iron, aluminum,
copper, and chromium, the maximum release was less than 3 percent of the
total acid-extracted metal in the sediments, even at the most extreme pH
value tested of 5.0. For nickel and lead, the release extended as high as
6-7 percent at a pH of 5.0. However, as seen in Figure 4, the release of
cadmium and manganese into groundwaters can be expected to increase as the
sediments become more acid. Manganese and cadmium were found to increase in
solubility substantially as sediment pH decreases with almost total release
of managanese and 25-35 percent release of cadmium at a pH of 5.0. Releases
of this magnitude may produce measurable increases in groundwater
concentrations beneath the pond. In the case of cadmium, a health hazard may
be present under these extreme conditions.
The results suggest that maintenance procedures may be necessary after a
period of time to remove the accumulated sediment deposits which may cause
conditions of low pH and release of metals.
CONCLUSIONS
From the results obtained in these investigations the following specific
conclusions were reached:
216
-------
1. Measured concentrations of heavy metals in highway runoff collected
at the Maitland Interchange during 1983-84 showed considerable variability
between storm events as well as during storm events.
2. With the exceptions of copper and cadmium, the majority of metal
species were present in a particulate form. The most common vehicle related
heavy metals found in highway runoff at the Maitland site were lead, zinc,
and copper which together accounted for approximately 91 percent of the
dissolved heavy metals and 94 percent of the total metal concentrations.
3. The Maitland pond was found to be very effective in removal of heavy
metal inputs from highway runoff. Particulate species of most metals were
removed in the range of 75-95 percent with most of this mass retained in the
pond sediments within a distance of 60-90 m from the stormwater inlet. In
general, dissolved forms of heavy metals were removed to a lesser degree than
particulate inputs with efficiencies near 50 percent for raost metals.
4. Mean concentrations of heavy metals within the pond were within
water quality criteria established in Chapter 17-3 of the Florida
Administrative Code (F.A.C.) for Class III (recreational) waters.
5. Measured concentrations of total heavy metals in the sediments of
the Maitland pond exhibited highest concentrations in the surface layer with
a rapid decline in concentration with increasing depth.
6. After eight years of accumulations in the Maitland pond, most metals
associated with sediments have remained in the top 10 cm of the sediment
layer.
7. Under current conditions of redox potential and pH within the
sediments of the Maitland pond, metal species, with the possible exceptions
of cadmium and manganese, are stable and exist in relatively immobile
associations with Fe/Mn oxides and organic matter.
8. In general, mean concentrations of all heavy metals measured, except
copper, were greater in groundwaters beneath the pond than within the pond.
Average concentrations of zinc, manganese, aluminum, and iron were found to
be 4, 12, 8, and 50 times grater, respectively, in shallow groundwater than
in the pond water. The extent of significantly higher groundwater
concentrations of nickel, cadmium, and chromium extended to depths of 0.5-1.0
m, lead extended to the 1-3 m range, while zinc, aluminum, manganese, and
iron were elevated in concentrations past the 6 m sample depth.
9. Violations of Class III water quality criteria were present for both
cadmium and iron in groundwaters beneath the pond.
10. The horizontal movement of groundwaters in the study area was less
than 10 m/year and as a result, the influence of the pond on groundwaters
appears to be extremely localized.
11. As sediment accumulation occurs in retention ponds over time, the
corresponding decreases in pH and ORP of the sediments will increase the
2V
-------
release of metal ions into groundwaters. The effect of reductions in pH were
found to be more important than reductions in ORP in regulating the release
of metal species.
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
Shaheen, D.G. Contributions of Urban Roadway Usage to Water Pollution.
EPA-600/2-75-004, U.S. Environmental Protection Agency, Washington,
B.C., 1975.
Wilber, W.G. and Hunter, J.V. Aquatic Transport of Heavy Metals in the Urban
Environment. Water Resources Bulletin, 13:721, 1977.
Winters, G.L. and Gidley, J.L. Effects of Roadway Runoff on Algae. Report
#FHWA/CA/TL-80/24 Federal Highway Administration, Washington, D.C.,
1980.
Yousef, Y.A.; Harper, H.H.; Wiseman, L.P.; and Bateman, J.M. Consequential
Species of Heavy Metals in Highway Runoff. Final Report Nos. 99700-7255
and 99700-7272, Tallahassee, FL, 1985.
218
-------
SIMPLE TROEHEC STATE MODELS AND THEIR USE
IN WAgEELQAD AIIOCaTIONS IN FLORIDA.
by: R. W. Ogburn, P. L. Brezonik1 and B. W. Breedlove
Breedlove, Dennis & Associates Inc., Orlando, FL 32807
^University of Minnesota, Minneapolis, MN 55455
ABSTRACT
Florida's Department of Environmental Regulation (FDER) uses a simple
input-output trophic state model to establish wasteload allocation limits in
cases involving wastewater discharge with a lake as the impacted waterbody.
The model, SIMLAK, is based on in-lake chlorophyll a as predicted from
loadings of nitrogen and phosphorus. A wasteload allocation performed by
FDER for five potential dischargers to the Reedy Creek/lake Russell system in
Polk and Osceola Counties in central Florida resulted in phosphorus
concentration limits much lower than levels achievable with available
technology.
An independent analysis of the Reedy Creek/lake Russell system was
performed to examine the applicability of SIMLAK and other trophic state
models, and to explore the potential iinpacts of various nutrient loading
scenarios on Lake Russell. SIMLAK did not yield accurate predictions of
chlorophyll a for Lake Russell based on present loadings, possibly because of
the lake's highly colored water. The FDER allocation scenarios also required
the nitrogen-to-phosphorus ratio in the lake to remain constant at the
present ratio instead of controlling the limiting nutrient. Our analysis
showed that variation of nitrogen-to-phosphorus ratios in the model could
allow slight decreases in nitrogen concentration limits to offset increases
in phosphorus oancentrations (to achievable levels) with the same
chlorophyll a prediction.
Alternative trophic state models based on one limiting nutrient did not
yield better predictions of present conditions in Lake Russell. However,
they did indicate that wastewater discharges at achievable phosphorus
concentrations would not cause chlorophyll a in Lake Russell to exceed FDER's
allowable limit.
219
-------
The initial FDER allocation of 21.5 ragd would have required the five
potential dischargers to meet a concentration limit of 2.15 mg/L for total
nitrogen (IN), but total phosphorus (TP) limits ranged from 0.1 to 0.38 mg/L
depending on distance from lake Russell. A final allocation agreed to by all
parties allowed discharge of 22.5 mgd with concentration limits of 2.0 mg/L
for TN and 0.5 mg/L for TP.
INTRODUCTION
REEDY CREEK/LAKE RUSSELL SYSTEM
The Reedy Creek drainage basin in central Florida extends from the
southwest corner of Orange County, through Osceola County and into Polk
County, where it joins the Kissimmee River system (Figure 1). Reedy Creek
originates near the Disney World complex and flows southeastward through
Reedy Creek Swamp into Lake Russell. Below Lake Russell the stream splits
with approximately 30 % of the flow going to Cypress lake and 70 % to Lake
Hatchineha through the Dead River. Flow from these two lakes ultimately
reaches Lake Kissimmee. The upper third of Reedy Creek has been channelized
and modified by control structures, but the remainder of the creek flows
through a natural stream channel.
lake Russell is a highly colored lake with a surface area of about 300
ha, a mean depth of 1.83 m and a maximum depth of 2.6 m (Table 1). The
annual hydraulic residence time is 30 days, and the range of retention times
based on monthly flows is between 15 and 56 days. High levels of organic
color limit Secchi depth to about 0.5 m and also appear to limit levels of
chlorophyll a in -Uie lake. Color ranged from about 300 to 500 PCU between
1979 and 1983, and chlorophyll a generally was less than 1 ug/L during the
same period, except for one value of 16 ug/L in July, 1980 and one
measurement of 13.5 ug/L in July, 1981.
WASTELOAD ALLOCATION PARTIES
Irr 1982 the City of Kissimmee was ordered by the State to eliminate its
wastewater discharge to Lake Tohopekaliga; later in 1982 the city initiated
discussions with the Florida Department of Environmental Regulation (FDER)
concerning the possibility of discharge to Reedy Creek. At that time Reedy
Creek Utilities was the only discharger to the creek, but several additional
wastewater treatment plants also requested consideration from FDER for
discharge capacity to Reedy Creek. The five utilities requesting
consideration in the allocation included Reedy Creek Utilities, the City of
Kissimmee, Central Florida Utilities, Poinciana Utilities, and Osceola
Services Company. The FDER conducted a one-year water quality survey and
220
-------
WALT DISNEY
WORLD
REEDY
CREEK
JL. _QRANGECO_._
OSCEotA CO.
APPROXIMATE
PROJECT
AREA
LEGEND
MAJOR THOROUGHFARES
COUNTY LINES
LAKE AND CREEK BOUNDARIES
\
Figure 1. Reedy CreeVLake Russell study area.
221
-------
Russell/ estimate the assimilative capacity of the system, and allocate that
capacity among the five utilities.
WASTEIDAD ALLDCATICN RESUUTS
After FDER's examination of the Reedy Creek/Lake Russell system and
numerous modeling scenarios using the lake model SIMIAK, FDER published a
notice of its intent to adopt an agency order concerning a wasteload
allocation for the Reedy Creek basin (2). The proposed order included two
allocation scenarios with total point source discharges of 21.5 ragd and 11.0
mgd (Table 2). Both scenarios included equal TN concentration limits for all
five dischargers but they allowed higher TP concentration limits for
utilities with points of discharge (POD) farther from Lake Russell. In
addition, the allowable TP concentrations decreased every five years through
the year 2005, and the proposed order stated that design levels of treatment
for TN and TP removal were to be based on limits for the year 2005.
The TP limits established for the year 2005 were lower than levels
achievable with existing technology for all five dischargers in both
Table 1. Physical and chemical characteristics of Lake Russell.
Parameter SECRET^ FDER1
Conductivity (umho/cm) 95
pH 6.2 6.7
BODs (mg/L) 1.7 1.0 6.7
Secchi depth (m) 0.47 0.4 0.6
Color (PCU) 343 404 500
Ammonia N (mg/L) 0.15 0.36 0.07
Nitrate & Nitrite N (mg/L) 0.13 0.06 0.06
Total N (mg/L) 1.74 1.97 1.79
Total P (mg/L) 0.07 0.11 0.08
Chlorophyll a (ug/L) 5.5 0.02 0.2
Surface area (ha) 300
Mean depth (m) 1.83
Maximum depth (m) 2.7
Residence time (d) 30
1 Reference 1
2 Field and laboratory results form Breedlove, Dennis & Associates, Inc. 1983
- 1984.
222
-------
Table 2. Discharge limits of total phosphorus (mg/L) and total nitrogen
(mg/L) in FDER notice of Intent to Adopt Agency Order (2).
Flow TP
Discharger (mgd) TN 1985 1990 1995 2000 2005
ROJ 7.5 2.15 0.65 0.57 0.51 0.45 0.38
Kissimmee 6.0 2.15 0.17 0.15 0.13 0.12 0.10
CEU 3.0 2.15 0.29 0.25 0.23 0.20 0.17
Poinciana 3.0 2.15 0.17 0.15 0.13 0.12 0.10
OSC 2.0 2.15 0.41 0.36 0.32 0.28 0.24
21.5
ECU
Kissianmee
CFU
Poinciana
OSC
4.0
4.0
1.0
1.0
1.0
11.0
2.54
2.54
2.54
2.54
2.54
0.62
0.16
0.28
0.16
0.40
0.44
0.12
0.20
0.12
0.28
0.32
0.08
0.15
0.08
0.21
0.20
0.05
0.09
0.05
0.13
0.08
0.02
0.04
0.02
0.05
scenarios. Poinciana Utilities, Inc. subsequently filed a petition for an
Administrative Hearing before the Florida Department of Administration to
review FDER's proposed order and to request relief from the impacts of the
proposed order (3).
INDEPENDENT EVALUATION
Poinciana Utilities, Inc. contracted with the authors of this paper to
analyze the FDER wasteload allocation procedures as applied to Reedy Creek
and Lake Russell, and to conduct an independent analysis of the potential
impacts of wastewater discharge on lake Russell and downstream laloes. This
analysis included a thorough review of SIMLAK and the wasteload allocation
process, modifications of SIMLAK, alternative trophic state models, and
additional data related to potential impacts of wastewater discharge on
fisheries, water clarity and nuisance aquatic plants.
223
-------
FDER WRSTEIDAD ALLOC3VTICN
SIMLAK
SBHAK is a simple mathematical model that can be used to predict
concentrations of IN, IP and chlorophyll a in a lake based on Inputs of N and
P, and physical characteristics of the lake such as hydraulic retention and
basin morphometry (Figure 2). The S3MIAK equations were developed using
REEDY CREEK
r
i
i
i
~ .^^ - ~
N
LOADING
~4
LEO 2 |
"~
P
LOADING
f~-
1 EQ 3 1
TN
1
TP
I
CHLOROPHYLL A
LAKE RUSSELL
EQ 1: R = 0.482 - 0.112 In
EQ 2: TN=(1-R)
EQ 3: TP=(1-R)
N LOADING
FLOW
P LOADING
FLOW
EQ 4: CHLA=35.95
VOLUME
~+TP) J
LENGTH
(WIDTHXDEPTH)
Figure 2. Block diagram of SIMLAK model.
224
-------
regression analysis of data from 33 Florida lakes included in the National
Eutrophication Survey (4). Equation 1 is a nutrient retention factor that
predicts the fractions of IN and TP retained in a lake as a function of its
flushing rate (Figure 2). Equations 2 and 3 give predictions of in-lake TN
and TP based on loadings, the retention factor, and the water flow rate into
the lake. Equation 4 predicts chlorophyll a in the lake from the predicted
values of TN and TP, and a shape term derived from the length, width and
average depth of the lake.
WASTELCftD ALLOCATION PROCESS
The first step in the wasteload allocation process is to verify that
SIMIAK works for the particular lake in question by using data on present
inputs to predict existing concentrations of TN, TP and chlorophyll a in the
lake. However, FDER has not established criteria for determining whether
SIMIAK predictions are acceptable.
If the model produces "acceptable11 results, the next step is to
calculate allowable increases in nutrient loading to the lake. FDER does
this by estimating non-point source (NPS) flows and inputs of nitrogen and
phosphorus, and using SIMIAK to predict the amount of chlorophyll that should
be produced by NPS inputs alone. A maximum increase in chlorophyll a of 2
ug/L is allowed if the predicted NPS chlorophyll a is less than 60 ug/L. In-
lake concentrations of TN and TP are allowed to increase by the same
percentage as the increase in chlorophyll a.
Equations 2 and 3 are re-arranged to yield nitrogen and phosphorus loads
that would produce the TN and TP levels obtained in the previous step. New
values of flow (Q) and retention (R) are used to reflect the combined total
of NPS and design PS flows. When present NPS loads are subtracted from those
total loads, the remainders are the allowable point source loads.
Of this allowable increase, only 75% is allocated among the PS
dischargers, and 25% is set aside for increased NPS loads and as a safety
factor. The allocatable load to the lake is divided among potential
dischargers in the same proportion as their contributions to the total PS
flow rate. The procedure described thus far produces an allocation of
nitrogen, phosphorus and flow rates at the lake itself. FDER's next step is
to extend the allocation to obtain limitations on discharge flows and
nitrogen and phosphorus concentrations at the individual points of discharge
to the system.
In the Lake Russell wasteload allocation, FDER concluded that nitrogen
removal does not occur in Reedy Creek. Therefore the nitrogen concentration
and flow limits calculated at the entrance to the lake also were applied
without change to the points of discharge, and all potential dischargers
received the same effluent TN concentration limits (Table 2).
225
-------
However, PEER did conclude that phosphorus is removed during flow in
Reedy Creek, and they calculated an average rate of phosphorus removal per
mile of creek channel. This meant that higher phosphorus concentrations
could be discharged farther upstream of lake Russell and still theoretically
meet the loading limits at the lake. The last column in Table 2 shows that
using this method, potential dischargers with POD's far from Lake Russell
would receive less stringent phosphorus concentration limits than dischargers
whose assumed POD was closer to lake Russell.
The final step FDER performed in the Reedy Creek wasteload allocation
was an attempt to project NPS changes through the year 2005. Ihey used
information from the East Central Florida Regional Planning Council and from
Osceola County to estimate the effect of future growth on NPS inputs of
nitrogen and phosphorus to the Reedy Creek basin. FDER concluded that NPS
loads of nitrogen will increase by 7%, and that NPS phosphorus loads will
increase by 29% by the year 2005.
They then compared the projected NFS increases to the 25% reserve
loadings set aside in the WLA process. In the case of nitrogen, future
increases did not exceed the reserve amount. However, FDER's projected NPS
phosphorus increases did exceed the 25% phosphorus reserve amount. When that
occurred, FDER subtracted the additional NPS phosphorus increases from the
amount originally allocated to point sources, thus reducing phosphorus limits
as in the scenarios included in the FDER Notice of Intent.
INDEPENDENT EVALUATION
CRITICISMS OF SIMIAK
Our analysis of the Reedy Creek/Lake Russell WIA led to several
conclusions regarding the model itself and its use for Lake Russell. Most
nutrient loading models use different retention factors for TN and TP because
of observations that they are retained to different extents in lakes where
nutrient mass balances have been measured. Lakes generally retain more
phosphorus than nitrogen, and thus it is inappropriate to use the same
equation for both. In addition, the SIMIAK retention equation originally was
developed for phosphorus retention in northern lakes, (5) and it does not
give good predictions for Florida lakes.
SIMIAK also did not accurately predict chlorophyll a based on present
nutrient inputs to Lake Russell. The model was designed to predict mean
annual chlorophyll a, which has ranged from about 5 ug/L to <1 ug/L for Lake
Russell. SIMIAK predicted 20.7 ug/L of chlorophyll a, but FDER considered
that an acceptable fit because of the maximum values observed in 1980 (16
ug/L) and 1981 (13.5 ug/L).
226
-------
Ihe WIA precedure allows a maximum increase of 2 ug/L of chlorophyll a
over the NPS chlorophyll a level in a late, but SIMLAK is not capable of
predicating chlorophyll a values with that degree of accuracy (4). Ihe
average error of SIMIAK chlorophyll predictions for the data set used to
develop the model was +21 ug/L. In addition to the problem of model
accuracy, it would be very difficult to measure a 2 ug/L increase in
chlorophyll. For example, mean chlorophyll a in Lake Kissiiranee was 20.68
ug/L in 1979. Ihe winTimim statistically valid increase (at a 90% confidence
level) with a monthly sampling program would be 4.0 ug/L. Even with a daily
sampling schedule, 2.9 ug/L would be the minimum increase detectable at a 90%
confidence level.
SIMLAK uses both nitrogen and phosphorus to predict chlorophyll a.
Other nutrient loading models recognize that only one nutrient at a time
limits the production of chlorophyll and they therefore have separate
chlorophyll equations, one for phosphorus-limited lakes and another for
nitrogen-limited lakes. However, because SIMIAK predicts chlorophyll as an
additive function of IN and TP, the WIA procedure should allow for "trade-
offs" between the two nutrients such that an increase in one (above its
"allowable" values) could be compensated for in SIMLAK by a decrease in the
other (below its "allowable" value), the net effect of which would be to
keep chlorophyll a levels predicted by SIMIAK within the allowable increase.
Thus the WIA procedure does not allow the flexibility in computing nutrient
loads that is inherent in the assumptions of the SIMIAK model.
According to the SIMIAK model, a small decrease in nitrogen loading
could result in an increase in phosphorus loading to Lake Russell with no
change in predicted chlorophyll a for the lake. Ihe validity of the SIMIAK
results should not be affected as long as N to P ratios remained within the
range of N to P ratios found in the lake data used to develop the model
equations. Nitrogen to phosphorus ratios vary widely in Florida lakes and
they can change seasonally within a particular lake. There is thus no reason
to require that N to P ratios in point source loadings should be the sane as
N to P ratios in NFS loads.
FDER evaluated discharge scenarios by comparing nutrient loadings to
calculated loading limits rather than considering whether the chlorophyll a
limit would be exceeded. When FDER projected the Lake Russell allocation to
the year 2005, they added their predictions of future increases in NFS loads
to present and proposed loads. After the projected increases in NFS
phosphorus loads had equalled the 25% reserve phosphorus load, FDER
subtracted subsequent increases from the amount allocated to dischargers
in order to maintain the total phosphorus load within the WIA limit.
Instead of using that approach, we asked the question: If the PS
phosphorus load were not decreased, and the total phosphorus load exceeded
FDER's WIA limit, what would be the effect on Lake Pussell's chlorophyll a?
Using FDER's projections of future NFS loads and 24.5 ngd of PS discharge
with IN = 2.15 mg/L and TP = 0.5 mg/L, we calculated that allowable
chlorophyll a levels would not be exceeded through the year 2005 (Figure 3).
227
-------
WLA
LIMIT
Q
3i
if
II
V)
O
a.
I
c
O
u
O
w
o
O
111
DC
0.
1984
1984
1965 1990 1995 2000 2005
WLA
LIMIT
1985
1990 1995 2000 2005
WLA
LIMIT
1984
1985
1990 1995 2000 2005
Figure 3. Predicted chlorophyll a in Lake Russell using SIMIAK equations with
24.5 rogd point source flow (IN = 2.15 mg/L, TP = 0.5 mg/L) and DER
projections of future NPS loads.
22b
-------
The TP loading would exceed FDER's phosphorus limit throughout that time
frame, but it would be compensated for by the fact that IN loads would remain
less than the nitrogen limit.
In addition, FDER's estimate of future changes in NPS loading in the
Reedy Creek basin did not include the effect of surface water management
requirements on nutrient loadings. A smilar analysis was performed (Harper,
personal communication) using recent data on reduction of nutrient loadings
in stormwater management systems similar to those that will be required in
future developments. These results indicated that the land use changes
projected by FDER would not cause increases in NFS loadings of nitrogen or
phosphorus, and that both might actually decrease slightly. In such a case,
predicted chlorophyll a in lake Pussell in the year 2005 would be at least 1
ug/L lower than FDER's wasteload allocation limit.
ALTERNATIVE MDDEIS
Numerous models are available for evaluating potential effects of PS
discharges on lake trophic state and water quality. Some of the best-known
models are accepted as reasonable empirical approaches to predicting the
trophic state of temperate lakes (7,8) and several more recent models are
modifications of their basic approach. The models have been adapted to
Florida lakes by using a statistical approach to optimize equations that
relate nutrient loading rates to lake response data (6). When results from
those equations were compared to the SIMLAK equations we found that the
modified Dillon and Vollenweider equations generally gave better predictions
of nutrient and chlorophyll concentrations in Florida lakes than did the
SIMLAK equations (Table 3).
The alternative equations also yielded over-predictions of chlorophyll a
for lake Russell based on present loading rates (Table 4), but they all
predicted rather small increases in chlorophyll, N and P in the lake as a
result of increased PS flows. We did not identify the reason that all the
models predict higher chlorophyll levels than those observed in lake Russell,
but the high levels of color limit light penetration and may limit algal
productivity. Additional possibilities include a limiting micronutrient such
as molybdenum or zinc, the short water residence time, or high rates of
zooplankton grazing. Nevertheless, the models all indicated that achievable
concentrations of phosphorus in proposed discharges would have very small
impacts on lake Russell.
Additional regression models were used to evaluate the iiipact of
increased PS loads on user-perceived values in Lake Russell. One equation
relates Secchi depth in Florida lakes to chlorophyll and color (8):
In Secchi depth = 2.01 - 0.370 In Oil a - 0.278 In Color
The model indicated that predicted increases in chlorophyll a in lake Russell
would result in decreases in water clarity too small to be observable to lake
users.
229
-------
Table 3. Predictive equations from SIMLAK and other Florida nutrient loading
models.
Source
Equation
Coefficient of
Determination
Nutrient Retention
SIMLAK
Baker et al.
R = 0.482 - 0.112 In Q
V
(for nitrogen)
(for phosphorus)
= -0.010 + 0.597 log
Q
Rp = -0.056 + 1.40
1 + W
= 0.10
= 0.15
= 0.51
= 0.88
Chlorophyll Prediction
SIMAK
Baker, et al.
CHIA =f 35.95/1N ^rp\ JL
^3 J NWZ
CHLA = 8.30
CHIA = 73.4
71
/ \ ^
(o.65 Z + Q j
/ Lp \0.667
lo.65 Z + Q I
r2 - 0.52
r2 = 0.69
r2 = 0.60
230-
-------
Table 4. Summary of model predictions for Lake Russell with various
discharge scenarios, (TP = 0.5 mg/L, IN = 2.15 rag/L).
Predictions
Point Source Loadings (mgd)
Present 18.5
21.5
24.5
Total Nitrogen
(rog/L)
Observed
SIMLftK
Baker et al.
1.96
1.95
1.64
- -
2.02
1.63
2.02
1.63
2.02
1.63
Total Phosphorus
(mg/L)
Observed
SIMLAK
Baker et al.
0.110
0.068
0.073
0.070
0.078
0.074
0.078
0.077
0.079
Chlorophyll a
(ug/L)
Observed
SIMLAK
Baker
Baker
et
et
al.
al.
N
P
3
20
34
13
.0 (16.0
.7
.8
.7
Max)
21
35
14
^^B
.5
.3
.2
21.
35.
14.
6
3
2
^»«M4
21
35
14
^ft^M
.7
.2
.6
Nutrient
Retention Factor
SIMIAK
Baker et
Baker et
al.
al.
Rn
Rp
0.
0.
0.
1996
2496
2532
0
0
0
.1709
.2408
.2234
0
0
0
.1658
.2395
.2184
0.
0.
0.
1610
2383
2137
231
-------
Regression models also were constructed to relate nutrient
concentrations to late coverage by nuisance aquatic weeds from data contained
in the FIADAB late data base developed at the University of Florida (9). No
significant relation was found between nutrient concentrations and areal
coverage or percent coverage by water hyacinths or hydrilla. The lack of a
nutrient - aquatic weed relationship and the snail changes predicted in late
Russell nutrient levels suggested that proposed PS loadings were unlikely to
affect the status of aquatic weeds in Late Russell.
Finally, we used equations relating fish biomass and percent biomass of
sport fish, forage fish and rough fish to the chlorophyll trophic state index
in Florida lakes of varying trophic state (10). We found that sport fish
biomass expressed as a percent of total biomass decreases as trophic state
index increases (Figure 4). However, the predicted change is caused by an
increase in rough fish biomass, and the actual biomass of sport fish (kg/ha)
appears to be insensitive to changes in trophic state index.
SUMMARY AND CONCXIJSIONS
Hie wasteload allocation procedure applied by FDER resulted in overly
restrictive discharge limits to Reedy Creek because FDER did not take
advantage of the ability of SIMLAK to vary N to P ratios and because of the
25% reserves of N and P. Alternate models and the SIMLAK equations
indicated that point source flows of 24.5 mgd with nutrient concentration
limits of 0.5 mg/L for TP and 2.15 mg/L for TN should cause only slight
increases in nutrient concentrations in late Russell that would not be
noticeable to users of the late.
The Administrative Hearing was settled by a consent agreement that
allowed 22.5 mgd of PS flows to Reedy Creek with TP of 0.5 mg/L and 2.0 mg/L
of TN. The wasteload allocation procedures have been modified recently by
FDER to allow for more outside involvement, but the SIMLAK model has not been
changed.
Based on the Reedy Creek/Late Russell wasteload allocation, we recommend
that FDER develop several models better suited to different types of Florida
lakes (N limited, P limited, highly colored lakes). Their approach should
allow -reasonable variations in late conditions such as N to P ratios in order
to attain allocations with achievable nutrient concentration limits. FDER
also should develop procedures for model calibration and criteria for judging
the acceptability of model predictions. Their models should be capable of
predicting changes in late conditions that are equivalent to the increase
allowed by the WIA process.
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.
232
-------
20 -
10 -
50
300
o»
X
o
m
50
TOTAL
Figure 4. Fish-Trophic state relationship in Florida lakes (from Bays and
Crisman 1983; Dashed lines represent approximate range of TSI for
Lake Russell based on S3MLAK predictions of chlorophyll).
233
-------
1. Magley, W. Reed/ Creek/Lake Russell (Osceola County) Wasteload Study.
Water Qjaality Technical Series, 2 (82), Florida Department of
Environmental Regulation, 1984. 279 pp.
2. Florida Department of Environmental Regulation. Notice of Intent to
Adopt an Agency Position Concerning Wasteload Allocations in the Reedy
Creek Basin, July 11, 1984.
3. Poinciana Utilities, Inc. Petition for Administrative Proceeding,
August 24, 1984.
4. Hand, J. and IfcClelland, S. Ihe lake Model "SDCAK" Users Guide. Water
Quality Technical Series, 3 (3), Florida Department of Environmental
Regulation, 1979. 24 pp.
5. Kirchner, W. B. and Dillon, P. J. An empirical method of estimating
the retention of phosphorus in lakes. Water Resour. Res. 11: 182,
1975.
6. Baker, L. A., Brezonik, P. L., and Kratzer, C. R. Nutrient loading-
trophic state relationships in Florida lakes. Water Resources Res. Ctr.
Pub. No. 56, Univ. of Florida, 1981.
7. Vollenweider, R. 'A. Irqput-output models with specific reference to the
phosphorus loading concept in limnology. Schweiz, Z. Hydrol. 37: 53,
1975.
8. Dillon, P.J. The FC>4 budget of Cameron lake, Ontario: The importance
of flushing rate to the degree of eutrophy of lakes. Limnol. Oceanogr.
20: 28, 1975.
9. Canfield, D.E. and Hodgson, L.H. Prediction of Secchi disc depths in
Florida lakes: Impact of Algal biomass and organic color. Hydrobiologia
99:51, 1983.
10 Huber, W.C., Brezonik, P.L., Heaney, J.P., Dickinson, R.E., Preston,
S.D., Dwornik, D.S., and Demaio, M.A. A classification of Florida
lakes. Water Reources Research Center Pub. No. 72, University of
Florida, 1982
11. Bays, J.S. and Crisman, T.L. Zooplankton and trophic state
relationships in Florida lakes. Can. J. Fish Aquat. Sci. 40: 1813,
1983.
234
-------
MODEL COMPLEXITY FOR TROPHIC STATE SIMULATION IN RESERVOIRS
by: Raymond A. Ferrara
Assistant Professor, Department of Civil Engineering
Lafayette College
Easton, PA 18042
Thomas T. Griffin
Najarian and Associates, Inc.
Eatontown, NJ 07724
ABSTRACT
Frequently model users are faced with choosing an appropriate model
structure and level of complexity for a particular application. Increased
model complexity generally implies incorporation of more portions or a
larger portion of the real world, but also generally implies an increased
number of rate coefficients and input parameters, each of which has some
associated uncertainty. This paper examines this issue through mathematical
simulation of trophic state in reservoirs with two models of different levels
of complexity. Analysis and comparison of the models' output is accomplished
through various statistical measures and techniques including root mean
squared error, regression analysis with student's t test for slope and inter-
cept, and difference of means. This study revealed that for predicting
steady-state total phosphorus concentration, no advantage to using the more
complex model over the simpler model was observed. For time-variable simu-
lation, it was determined that the models were substantially in agreement
for long term averaging (i.e. greater than 10 years), but that such models
do not agree for short term simulation nor even for the frequency distribu-
tion of phosphorus concentrations within that simulation period. In less
biologically productive oligotrophic systems, the level of model complexity
appeared to b.e of lesser importance than for more biologically productive
eutrophic systems. The nonlinearity of the phosphorus uptake relationship
is demonstrated to be more pronounced in eutrophic than in oligotrophic
systems, and explains the discrepancy between the models' predictions.
INTRODUCTION
Frequently model users are faced with choosing an appropriate model for
their particular application. For trophic state determination in impound-
ments, the choices range from the steady state total phosphorus loading dia-
grams (1, 2) to the time-variable ecosystem models (3, 4, 5, 6, 7, 8). In
many cases a user may assume that the more complex the model, the more accu-
235
-------
rate or reliable the simulation. This may not always be true as even the
most complex models are never mathematical clones of real physical systems.
Thomaim (9) aptly states, "A mathematical model is simply an analytical ab-
straction of the real world. As such, it does not pretend to incorporate all
phenomena but rather abstracts only those portions of the real world that are
relevant to the problem under consideration."
Although increased complexity, generally implies incorporation of more
portions or a larger portion of the real world, it must be realized that in-
creased complexity also generally implies an increased number of rate co-
efficients and input parameters to utilize the model. Since the numerical
value of each of these has some associated uncertainty, then it is entirely
possible that uncertainty in the model output may also increase with complex-
ity. The parabolic curve of Figure 1 represents a feasible picture of the
relationship between model complexity and model uncertainty. Of course, as
the state of our knowledge of individual biological, chemical, and physical
processes improves throughout time, so too will our ability to mathematically
describe these processes and hence to build more certain complex models.
This paper does not attempt to resolve the entire dilemma, but rather to
examine it through mathematical simulation of trophic state in reservoirs by
comparing the output of two models of different complexity. The comparisons
are made under steady-state and time-variable conditions. For the latter,
two types of analyses are conducted: (1) Direct comparison of the models'
predicted concentrations at corresponding times, and (2) Comparison of the
frequency distribution of the models' predicted concentrations over the en-
tire simulation period. The methodology and conclusions are applicable to
modeling water quality constituents in general.
Model
Uncertainty
future
{e.g. 2000)
Model Complexity
Figure 1. A Relationship Between Model Complexity and Uncertainty
236
-------
METHODOLOGY
Two contrasting models of phosphorus dynamics in reservoirs are con-
sidered. Both models account for time-variable reservoir volume, influent
and effluent flow rate, and influent phosphorus concentrations. The first
model (10) represents a reservoir as a single fully mixed system and models
a single water quality variable, total phosphorus (TP). The model is de-
picted in Figure 2. The change in TP with time is modeled as the sum of the
influent load, minus the effluent load minus a net decay of phosphorus (e.g.
due to sedimentation), according to the following equation:
d(V'TP) vs
at = VTpi ~ VTP " T 'TP'V
(1)
where
V
t
Qe
TP±
TP
VS
H
reservoir volume
time
influent flow rate
effluent flow rate
influent phosphorus concentration
phosphorus concentration in the reservoir and
in the effluent
apparent settling velocity
mean depth of reservoir
Typical values for vg reported in the literature are 10 to 16 m/yr (1, 11,
12). The form of the model is similar to those previously presented by
Vollenweider (1) and Chapra (13). It will be referred to hereafter as the
Total Phosphorus Model, TPM.
The second model (14) represents a reservoir as having two fully mixed
layers, an epilimnion and a hypolimnion, during periods of thermal stratifi-
cation. When temperature gradients are small or absent, for example during
overturn, the reservoir is represented as a single fully mixed layer. Three
forms of phosphorus - organic phosphorus (OP), dissolved inorganic phosphorus
(DIP), and particulate inorganic phosphorus (PIP) - and dissolved oxygen are
modeled in each layer. The model is depicted in Figure 3. The model equa-
TPi
V, TP
1
Qe, TP
Figure 2. The Total Phosphorus Model (TPM)
237
-------
Stratified Model
Fully Mixed Model
Figure 3. The Multi-Component Phosphorus Model (MCPM)
238
-------
tions have been presented previously in Griffin and Ferrara (14) . The model
will hereafter be referred to as the Multi-Component Phosphorus Model (MCPM) .
TPM may be considered very nearly the simplest of model structures, i.e.
a low level of complexity. MCPM provides a higher level of complexity in-
corporating characteristics of nutrient cycling and biological transforma-
tions. However, it is not as complex as some of the ecosystem models cited
above. MCPM was developed to simulate seasonal changes in water quality,
yet still be simple and economical enough for long term simulation on the
order of decades. TPM was also derived for long term simulation, yet be-
cause of its simple structure, would not be expected to consistently predict
seasonal changes in water quality accurately.
TPM requires specification of one rate coefficient, i.e. vs. The quo-
tient ^ is sometimes reported as a constant, a, a first-order net sedimenta-
tion H rate coefficient. At steady-state, the solution to TPM is
TP . TP .
TP = - = - /2\
1 + ^s V 1 + at
3 Q
where t is the detention time, Q-. Vollenweider (15) has reported, based on
data from many lakes, that
Hence, knowledge of the detention time of a particular waterbody leads to an
estimate of the single reaction rate coefficient necessary to utilize the
total phosphorus model.
STATISTICAL MEASURES
Various statistical techniques exist for comparing data sets whether the
data are actual observations or the results of model simulation. Statistical
measures used in this paper include the following (16, 17).
1. Confidence Interval about Mean
For a limited number of data points, X, one can identify an interval
within which it can be said, with a certain amount of confidence, that the
true mean of X, yx, will reside. This interval is given as follows:
5 '
where X = the sample mean
S = the sample standard deviation
t = the percentile of the student t distribution
with N-l degrees of freedom and y equal to
100 minus the % confidence level.
239
-------
When comparing two data sets, if their intervals overlap, then the sample
means are not significantly different, and vice versa.
2. Root Mean Squared Error
A convenient method of comparing the difference between two data sets
is to compute the root mean squared error, R, where
R =
N
z <*ii - V
(5)
N
The subscripts 1 and 2 refer to data sets 1 and 2 respectively. The signif-
icance of R is pictured in Figure 4 where it is shown to represent the de-
viation from perfect agreement between two data sets. Since R has the same
units as X (e.g., concentration), it is useful for comparison to express a
relative error, RE, where
RE = -J- (6)
A
Note that either XT or X2 could be used in the denominator of equation (6).
3. Regression Analysis
A further test of agreement can be made by regressing the data as pre-
sented in Figure 4 and comparing the intercept, a, and slope, b, of the re-
gressed line with the theoretical values of a and $, respectively. A two-
tailed t test with N-2 degrees of freedom and the following statistics is
conducted,
'- <»
where Sfc and Sa are the standard deviation of the slope and intercept, re-
spectively. For t, leads to the following statistic
(8)
240
-------
ROOT MEAN SQUARED ERROR
T2
x2
R =
1 - Xj)2 + (Y2 - X2)2 + ... (YN - XM)
REGRESS[ON ANALYSIS
LINE OF PERFECT
AGREEMENT
LINE OF
BEST FIT
a=0
Figure A. Root Mean Squared Error and Regression Analysis
241
-------
where 6 = true difference between data set means
Sd = the standard deviation of d which is
given by a pooled estimate of the population
variance.
If the variance of the two data sets are assumed equal, then
d - 6 K*2 (N1+N2-2) I (9)
I 2 2~> « + N,
1(^-1) S^ + (N2-1)S2 ll 1 2
Note that for perfect agreement between data set means, 6=0. Then a value
of t{j
-------
TABLE 1, STEADY STATE SIMULATION RESULTS
Test
I
II
III
IV
V
VI
VII
VIII
IX
X
XI
XII
XIII
XIV
XV
XVI
XVII
XVIII
Mean
Standard Deviation
t(yr)
0.06
0.10
0.12
0.20
0.21
0.27
0.34
0.35
0.41
0.43
0.48
0.62
0.75
0.82
1.37
1.92
2.47
3.01
a(eq.3)
4.13
3.12
2.92
2.26
2.21
1.91
1.71
1.68
1.56
1.52
1.44
1.27
1.15
1.10
0.85
0.72
0.64
0.58
Coefficient -of Variation
*'TpM»"9' *
48.3
45.4
44.7
41.6
41.3
39.4
37.9
37.7
36.6
36.2
35.5
33.6
32.1
31.5
27.6
25.2
23.3
21.9
35.5
7.59
0.21
MCPM(ug/1)
43.0
39.1
41.1
41.1
37.6
41.2
37.7
41.6
30.1
42.1
37.9
38.5
39.0
29.0
29.0
29.3
29.9
30.5
36.5
5.26
0.14
243
-------
diet water quality changes from one year to the next, but rather to predict
the frequency distribution of water quality (as measured by total phosphorus
concentration) due to some reservoir management scenario. MCPM was intended
for simulation of seasonal changes in water quality, yet be economical enough
for long term simulation as well.
It might be expected from the steady-state analysis presented previously
in this paper and from the objectives of the two models that their predic-
tions for long term average concentrations should be in agreement. In such
cases it would be expedient to use TPM. However, it might further be ex-
pected that as the period of simulation decreases, the model predictions
would begin to diverge. At that stage, the model user must choose between
TPM and MCPM, and would likely accept the latter. The critical question
then becomes how to identify this point of divergence and how different the
model predictions really are. This may be accomplished through comparison
of the model predictions for simulations of varying time periods.
MCPM was run for a fifty-seven year simulation utilizing the rate co-
efficients previously defined (Figure 5). From_the model output a mean
total phosphorus concentration was calculated, TP^cpM = 27.2 ug/1. TPM was
then run for the same conditions. However, it was first necessary to choose
a value for the sedimentation rate coefficient. Equation (3) is inappro-
priate for two reasons, (1) it was derived for steady-state analysis, and
(2) the concept of detention time becomes meaningless under time-variable
simulation unless of course a temporally averaged flow rate and reservoir
volume are used to compute it. Values reported for vs in the literature
were similarly determined from steady-state data. Chapra (13) utilized a
to
=>
OL
O
X
Q.
in
o
50
45
4O
35
30
25
20
15
10
5
0
50
100 150 200 250 300 350 400 450 500 550 600 650 700
TIME I MONTHS)
Figure 5. Tine Variable Simulation - Caso 1
(dashed line represents TPM; solid line represents MCPM)
244
-------
value of v_ = 16 m/yr in a time-variable total phosphorus model of the Great
Lakes. Since there is no a priori method for choosing a value for vs and
a sufficiently large data base for model calibration did not exist_,_ it was
decided for this study to adapt a value which results in TP-ppM = TPjiCPM
for the 57 year simulation. This of- course forces the two models to agree
in the long term average but does not prejudice their relative agreement or
disagreement for shorter periods or for the frequency distribution of concen-
trations over the_long term. A value of vs = 9 m/yr results in TP~TPM approx-
imately equal to WMCPM (Table 2). This value is indeed close to the re-
ported steady-state range of 10 to 16 m/yr and the time variable value
utilized by Chapra (13).
The 57 year simulation may be broken down into various averaging periods
for comparison of the two models. For any averaging period of simulation
years there will be (57-x+l) values of TPjpM and TPMCpM for comparison. For
example, for a 57 year averaging period there is one value each for TF-pp^
and TPjicPM' i-e-» 27-5 and 27.2 ug/1 respectively. For a 25 year averaging
period, there will be 57-25+1 = 33 values for each model. For monthly
averaging periods of course there will be 57x12 = 684 total phosphorus con-
centrations for each model.
The root mean squared error for various averaging periods is provided
in Table 3. Note that the relative error or disagreement between the models
increases up to 19% when monthly average TP concentrations are compared,
whereas the disagreement is only 5% for the 25 year average concentrations.
The results of the regression analysis are also presented in Table 3. Slopes
and intercepts are significantly different than one and zero for the monthly
and 1 year groups but not for the 4, 10 and 25 year groups. This statistical
analysis verifies the hypothesis that the degree of agreement between the two
models does decrease significantly as the time period of simulation for com-
paring average total phosphorus concentrations is also decreased.
An analysis of the frequency distribution of predicted monthly average
total phosphorus concentrations for each model was also conducted. Although
it is shown above that the monthly average predictions in individual months
do not agree, it may be that the distributions of those predictions will
agree. This may be satisfactory for certain management decisions where for
TABLE 2. TP AS A FUNCTION OF v - CASE I
1 r M 5
Vs TPTpM(ug/])
5 34.5
7 30.6
8 29.0
9 27.5
10 26.2
12 23.9
15 21.1
245
-------
TABLE 3. STATISTICAL ANALYSIS OF TIME VARIABLE SIMULATION - CASE I
Averaging
Group
Monthly
1 year
4 year
10 year
25 year
57 year
R
5.05
4.66
3.84
2.19
1.48
0.30
RE
0.19
0.17
0.14
0.08
0.05
0.01
Slope
1.36
1.74
1.75*
1.23*
1.54*
-
Intercept
-9.32
-19.86
-20.11*
-5.18*
-13.51*
-
*Indicates value not significantly different than 1.0 for slope or 0.0
for intercept at 95% confidence level.
example it is only necessary to know the percent of time concentration ex-
ceeds a certain standard. The data were grouped into intervals of 0.5 ug/1,
(Figures 6.a and 6.b). Note that the data are approximately normally dis-
tributed but that the characteristics of the two models are interestingly
different. As in the steady-state simulations, although both models have
essentially the same mean value, TPM is characterized by a larger standard
deviation and hence coefficient of variation. Consequently, the distribution
of predicted monthly average total phosphorus concentrations appears to be
different. This was tested by conducting a regression analysis comparing
the frequency of occurrence in respective concentration intervals. The re-
sults of this analysis reveal a slope and intercept of 0.32 and 7.43 re-
spectively, values which were further determined to be indeed statistically
significantly different than the theoretical values of 1.0 and 0.0 re-
spectively. Therefore, in addition to the fact that the two models are pre-
dicting significantly different concentrations in corresponding individual
months, they are also predicting significantly different distributions of
concentrations over the 57 year simulation period.
TIME VARIABLE SIMULATION - CASE II
The above time-variable simulation was conducted on an impoundment which
might be "considered eutrophic, i.e. total phosphorus concentrations generally
greater than 20 ug/1. In order to compare the models in a less nutrient-rich
system, the same simulations were conducted with the influent phosphorus con-
centration reduced to one-fourth of its previous value. The model simula-
tions are presented in Figure 7. A value of TPtfcPM = iO-07 "S/1 (very nearly
oligotrophic conditions) was obtained. Interestingly, the calibrated value
for vs was found to be 3 m/yr (Table 4). Because of the reduced influent
phosphorus load, this system is far less biologically active. Since the role
of biological uptake and deposition has diminished, the system must be char-
acterized by a smaller net sedimentation/decay rate coefficient. In the pre-
vious system, a value of vs = 9.0 m/yr was obtained and was in close agree-
246-
-------
32
24-
o
Ul
O
Ul
£E
U.
8-
o
2
Ld
Z)
O
UJ
(E
84
63-
42-
21-
I
10 20 30
TOTAL PHOSPHORUS (ug/l)
|k
40
TPM
50
llVlt 1-1 r.
10 20 30 40
TOTAL PHOSPHORUS (ug/l)-MCPM
5O
Figure 6a. Frequency Distribution for Tine-Variable Simulation
Case I
247
-------
(Jj
700-t
600-
3 500-
400-
O
Q
2
Al
£ 300-
ZOO'
u.
0 100
MCPM
~PM
10 15 ZO Z5 30 35
TOTAL PHOSPHORUS (ug/1)
40
45
Figure 6b. Cumulative Frequency Distribution for Time Variable
Simulation - Case I
merit with the range of 10 - 16 m/yr, a range which was determined from data
on predominantly biologically productive eutrophic lakes. This study re-
veals that for a less productive oligotrophic/mesotrophic impoundment, vs
should be substantially reduced.
Results of the statistical analysis for this case are presented in
Table 5, and lead to conclusions similar to those drawn from Case I and
Table 3. The distribution of predicted monthly average concentrations is
presented in Figures 8.a and 8.V. Again the data are approximately normally
distributed, but TPM is characterized by a greater degree of variation as in
the previous case. The regression analysis again revealed that the models
are predicting significantly different distributions.
DISCUSSION
The analyses presented above raise an important question as to the ap-
nlicabilityof a simple total phosphorus model for predicting trophic state
in impoundments. Models which represent the influence of biological activity
on the cycling of phosphorus between organic and inorganic forms include
functional relationships for their reaction rate coefficients which express
a dependence on various internal or external system parameters. An example
248
-------
TABLE 4 - TPTpM AS A FUNCTION OF vs - CASE II
TP
TPM
1
2
3
5
8
10
11.2
10.6
10.0
8.7
7.3
6.6
of this in the multi-component phosphorus model is the organic phosphorus
growth rate term which is a function of nutrient availability. If nutrient
concentrations are high, biological uptake and deposition will also be high.
The converse is also true. The total phosphorus model with its single re-
action rate coefficient does not include this flexibility, and hence as
demonstrated^ two distinctly different values for vg were required to obtain
the appropriate model response under the two distinctly different phosphorus
loading scenarios.
This phenomenon is clearly demonstrated in Figure 9 which is a plot of
the time-variable nutrient limitation term. Note that in Case I, values are
typically greater than those in Case II corresponding to the reduction in vg
from 9 to 3 m/yr respectively. To further verify this, a third simulation
V)
D
tE
O
I
35
30-
25
20
15
10
5
0
MCPM
TPM
( 50 100 (50 20O 250 300 350 400 450 500 550 600 650 TOO
TIME {MONTHS)
Figure 7. Time Variable Simulation - Case II
(dashed line represents TPM; solid line represents MCPM)
249
-------
was conducted, utilizing conditions identical to Case I, however, the half-
saturation constant was increased by a factor of three for demonstration
purposes. TPMcpM was determined to be 37.4 ug/1, indicating fairly enriched
conditions, yet the appropriate value of vg was found to be 4 m/year. The
nutrient limitation term illustrated in Figure 9 provides the explanation.
Here again the net organic phosphorus growth rate terra is reduced preventing
132-
99-
O
UJ
tr.
u.
33-
1
IO 20 30 4O
TOTAL PHOSPHORUS (ug/l)-TPM
50
424-
318-
z
Ul 212-
r>
o
uj
tr
u.
IO6-
10 20 30 40
TOTAL PHOSPHORUS (ug/1)-MCPM
50
Figure 8a. Frequency Distribution for Time-Variable Simulation
Case II
250
-------
700
6OO-
UJ
i
Q
id
soo-
u
O 400
z
A|
Q.
300-
200-
100
MCPM
TPM
10
12
TOTAL PHOSPHORUS
(ug/l)
Figure 8b. Cumulative Frequency Distribution for Time Variable
Simulation - Case II
the cycling of phosphorus into the organic form with subsequent deposition to
the sediments. This study demonstrates the importance of including the
strong influence of biological activity to represent the net sedimentation of
phosphorus in a model for trophic state simulation. A simpler version does
not respond accordingly for time-variable simulation, unless of course vs is
adjusted as demonstrated to be required above.
The nutrient limitation term is also the key to the observed difference
between relative error and correlation coefficients for the two cases as re-
ported in Tables 3 and 5. There was significant improvement in Case II as a
result of the more constant nature of the nutrient limitation term. This re-
duces the nonlinearity of the multicomponent phosphorus model and provides
better agreement with the linear total phosphorus model.
Another difference which becomes apparent from this study is the con-
sistently higher coefficient of variation observed with the total phosphorus
251
-------
TABLE 5 - STATISTICAL ANALYSIS OF TIME-VARIABLE SIMULATION - CASE II
Averaging
Group
Monthly
1 year
4 year
10 year
25 year
57 year
R
0.979
0.949
0.776
0.374
0.197
0.05
RE
0.098
0.095
0.078
0.037
0.020
0.005
Slope
1.55
1.7
1.64*
1.54*
1.96
Intercept
-5.53
-6.9
-6.48*
-5.34*
-9.53
Indicates value not significantly different than 1.0 for slope or 0.0 for
intercept at 95% confidence level.
model in both steady-state and time-variable simulations. This results from
the fact that TPM because of its simple biochemical structure, is more
strongly dominated*by input-output forces, whereas MCPM represents an effec-
tive ecological buffering capacity which reacts to changes in external
forces. This difference in variance is most clearly demonstrated in the fre-
quency distribution histograms of Figure 6 and 8.
CONCLUSIONS
Several conclusions have been developed from this study. For steady-
state simulation, the two models do not differ appreciably in their predic-
tion of total phosphorus. However, for time-variable simulation, the models
0.6 -
o.o
LEGEND
CASE I
CASE 2
CASE 3
100 200 3OO 400 SCO
TIME (MONTHS)
600
70O
Figure 9, Time-Variable Nutrient Limitation Term
252
-------
do predict significantly different concentrations. Furthermore, the fre-
quency distribution of those concentrations even over very long term simula-
tion periods also differ. For a given simulation, the total phosphorus model
will predict a wider range of concentrations than the multi-component phos-
phorus model.
It is important to include a functional relationship describing the in-
fluence of biological activity in the cycling of phosphorus through its
various organic and inorganic forms thereby determining the net sedimentation
rate for total phosphorus. The value for vs is not constant across trophic
states. The nonlinearity of the nutrient limitation term is responsible for
the difference between model predictions. The relative error between the
models decreases as the degree of variation in this term decreases.
REFERENCES
1. Vollenweider, R.A. Input Output Models with Special Reference to the
Phosphorus Loading Concept in Limnology. Schweiz. A. Hydrol. 37:
1975.
2. Dillon, P.J. The Phosphorus Budget of Cameron Lake, Ontario: The
Importance of Flushing Rate to the Degree of Eutrophy in Lakes.
LjLmnol. and Oceanogr. 19:28, 1975
3. Jorgensen, S.E., Ecological Modeling of Lakes Chapter 9 In: G.T.
Orlob (ed.), Mathematical Modeling of Water Quality: Streams, Lakes,
and Reservoirs, Wiley-Interscience, N.Y., N.Y., 1983
4. Jorgensen, S.E. A Eutrophication Model for a Lake. Ecological
Modelling. 2: 147, 1976
5. Thomann, R.V., et al. Mathematical Modeling of Phytoplankton in Lake
Ontario, 1. Model Development and Verification. EPA-660/3-75-005,
U.S. Environmental Protection Agency, Corvallis, Oregon, 1975. 177 pp.
6. DiToro, D.M., and Matystik, W.F. Mathematical Models of Water Quality
in Large Lakes, Part I: Lake Huron and Saginaw Bay. EPA-600/3-80-059,
U.S. Environmental Protection Agency, 1980a.
7. DiToro, D.M. and Connolly, J.P. Mathematical Models of Water Quality
in Large Lakes, Part 2: Lake Erie. EPA-600/3-80-065, U.S. Environ-
mental Protection Agency, 1980b.
8. Chen, C.W. and Orlob, G.T. Ecologic Simulation for Aquatic Environ-
ments In; B.C. Patten (ed.) Systems Analysis and Simulation in Ecolo-
j»y_ Vol. 3, Academic Press, 1975.
9. Thomann, R.V., Systems Analysis and Water Quality Management, Environ-
mental Science Services Division, Environmental Research and Appli-
cations, Inc. N.Y., N.Y. 1972. 286 pp.
253
-------
10. Ferrara, R.A., T.S. Nadbielny, T.T. Griffin. Reservoir Management for
Water Supply and Water Quality Objectives. Current Practices in Envir-
onmental Science and Engineering. 2:1, 1986
11. Dillon, P.J. and Kirchner, W.B. Reply to Comment by S.C. Chapra Wa-
ter Resources Research. 11:6, 1035, 1975
12. Chapra, S.C. Comment on An Empirical Method of Estimating the Reten-
tion of Phosphorus in Lakes. Water Resources Research, 11:6, 1033, 1975
13. Chapra, S.C. Total Phosphorus Model for the Great Lakes. JEEP, ASCE
103: EE2, 142, 1977
14. Griffin, T.T. and Ferrara, R.A. A Multi-Component Model of
Phosphorus Dynamics in Reservoirs. Water Resources Bulletin,
20:5, 777, 1984
15. Vollenweider, R.A. Advances in Defining Critical Loading Le-
vels for Phosphorus in Lake Eutrophication. Mem. 1st Ital.
Idrobio. 33:1976
16. Thoraann, R.V. Measures of Verification In: R.V. Thomann and
T.O. Barnwell (ed.), Workshop on Verification of Water Quality
Models. EPA-600/9-80-016, U.S. Environmental Protection Agency,
Athens, Georgia, 1980. 258 pp.
17. Miller, I. and Freund, J.E. Probability and Statistics for
Engineers 2nd edition, Prentice Hall, Englewood Cliffs, N.J.
1977. 529 pp.
Disclaimer
The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessarily
reflect the views of tfie Agency and no official endorsement should be
inferred.
234
-------
F.D.O.T. Drainage Manual:Why "Drainage"1n an Age of "Stormwater Management"?
by: Edward G. Rings
Florida Department of Transportation
Tallahassee, Florida 32301
ABSTRACT
In 1963, the Florida State Road Department issued the first "Drainage
Manual". New policy and procedures were added by memorandum and design
directives. In the intervening 23 + years, enviromental awareness in general
and in Florida in particular increased dramatically. The intervening
development of Florida stormwater regulation is traced. When in 1985 a
replacement manual was begun, the decision was made to retain the title
"Drainage Manual" rather than changing the title to "Stormwater Management
Manual". This was not done to indicate that stormwater management was to be
disregarded, but to emphasize that drainage, which is critical to
transportation, must be provided. An overview of the draft Drainage Manual
is presented which demonstrates that proper drainage is good stormwater
management. Ineffective or inadequate drainage is inappropriate stormwater
management.
1963 FSRD DRAINAGE MANUAL
In Florida the "state of the art" for highway drainage as it existed 25
years ago can be seen by reviewing the 1963 (revised 1967) Florida State Road
Department Drainage Manual. A listing of the chapter headings is provided in
Table 1. The 1963 manual has only a few tables, equations and figures which
are identified in Table 2. The manual relied on the user having a good
working knowledge of hydrology and hydraulics and a fundamental
hydrologic/hydraulic library upon which to draw. The manual supplied the
necessary information to enable the user to utilize reference information to
accomplish highway drainage to Department standards.
What was the thrust of the manual? In two words it was runoff
conveyance. The methods were simple and straight forward, and not
surprisingly, they worked. This can be explained because the manual was
255
-------
TABLE 1. STATE ROAD DEPARTMENT DRAINAGE MANUAL 1963
TABLE OF CONTENTS
Chapter Chapter Heading Pages
1
2
3
4
5
6
7
8
9
10
11
12
13
Introduction
The Drainage Section
Engineering Law
Policies
Survey and Field Information
Rainfall and Runoff
Hydraulic Design of Culverts
Ditches and Canals
Hydraulic Design of Bridges (never completed)
Subgrade Drainage
High Water Clearance for Pavement Grades
Storm Sewers
Checking Plans
TOTAL
1
1
10
5
3
15
4
5
0
2
1
9
6
62
written by District and Central Office drainage engineers reflecting 35 +
year record of the highway drainage experience in Florida.
By 1963, Florida had begun to stir and start the transition from a
mostly rural state with a few predominately coastal urban centers into one of
the fastest growth states in the U.S. It is safe to categorize the 1960's
Florida drainage law as "common law" which permitted upper owners to increase
the rate (quantity) of discharge provided the increase did not cause damage.
The standard applied was one of reasonableness. This has been described as a
modified "look out below" drainage law. By the late 1970's most major urban
areas developed regulations and ordinances which limited and defined maximum
allowable peak discharges, since the cumulative effect of many "resonable"
increases is not necessarily reasonable. The special problem of total volume
amount in landlocked watersheds prevalent in north and central Florida's
karst topography was not addressed.
The first Drainage Manual required the highway drainage designer to
consider future development within a 20 year period but did not allow private
connections to the Department's drainage facilities. However, connections
were routinely granted provided the applicant demonstrated that the discharge
.does not come from divirted areas and that the Department's system was
designed for the level of discharge. The conveyance system below the
Department's outlet has always been evaluated with the extent of the
evaluation increasing in recent times. Quality was addressed in general
terms. Quality requirements were broad, requiring dischargers to meet State
water quality standards. A general interpretation of the earlier requirement
was "do not contaminate the stormwater".
256
-------
TABLE 2. STATE ROAD DEPARTMENT DRAINAGE MANUAL 1963
TABLES , EQUATIONS AND FIGURES
Tables Equations Figures
Storm Design Frequency Rational Rainfall Stations
Runoff Coefficient Talbot's Formula Intensity-Duration Curves
Minimum Culvert Sizes Time of Concentration for
Cleaning Velocities Overland flow
Base Clearances Talbot's Discharge Curve
Manning's "n" Values Storm Sewer Tabulation Form
Maximum Manhole spacing Highway Section Check List
THE INTERVENING PERIOD - 1963 to 1985
Prior to 1972, water management legislation had developed on a piecemeal
basis. In that year, a comprehensive law has enacted to provide extensive
protection and management of water resources throughout the state.
The Florida 1972 Water Resources Act, Chapter 373, Florida Statues,
provides a two-tiered administrative structure presently headed at the state
level by the Department of Environmental Regulation (DER). DER is the chief
pollution control agency in the state and was created from parts of several
agencies through Legislative reorganization in 1975. DER's jurisdiction over
water pollution control extends to "waters of the state". Practically all
dredge and fill activities conducted in areas either in or connected to
waters of the state are required to comply with water quality standards. DER
regulates discharge or untreated stormwater that could be a potential source
of pollution to the state. This regulatory scheme is predominantly
qualitative. All stormwater discharges must meet the water quality standards
of the class of water body the stormwater actually reaches. Additionally,
DER regulates stormwater by requiring retention or retention with filtration
systems that allow separation of polluting substances by percolating the
water into the ground. DER supervises five regional Water Management
Districts designed to provide the diverse types of regulation needed in
different areas of the state. These include the previously existing Central
and Southern Florida Flood Control District, renamed the South Florida Water
Management District, and the Southwest Florida Water Management District.
The three new districts established under the Act were the Suwannee River,
St. Johns River, and Northwest Florida Water Management Districts.
Most of the thrust of DERs and the Water Managment Districts efforts
have been towards water quality. Efforts to address stormwater quantity
control were mainly left to local government. Under the present law,
municipalities have authority to brovide for drainage of city streets and
reclamation of wet, low, or overflowed lands within their jurisdiction. They
may construct sewers and drains and may levy special assessments on benefited
property owners to pay all or part of the costs of such works. Additionally,
257
-------
municipalities have the power of eminent domain to condemn property for these
purposes. Thus, they have the means to deal directly with stormwater and
surface-water runoff problems. Municipalities may enact floodplain zoning
ordinances. Such ordinances may simply require compliance with special
building regulations or may exclude certain types of development in a
designated floodplain.
Most counties and municipalities have a drainage plan ordinance that
requires submittal of a drainage plan for proposed developments. In
addition, they commonly require that a drainage impact assessment be prepared
and submitted if there is to be a change in the development site. Many local
governments have ordinances restricting the amount of surface-water runoff
that may be carried by a particular drainage system, or the amount of
sediment transported by the runoff. Many local ordinances also incorporate a
flood plain regulation element or minimum elevations.
Subdivision regulations relating to surface-water runoff control tend to
be more detailed than local government ordinances, and often require
submittal of a comprehensive drainage plan, approval of which is often a
prerequisite for plat approval. Some regulations include runoff and rainfall
criteria to which the proposed drainage system must conform, while others
indicate permitted or preferred surface-water runoff control structures and
techniques. Other provisions found in subdivision regulations include: a
requirement that runoff from paved areas meet certain water quality
standards; the encouragement or requirement of onsite retention of runoff;
the regulation of grading and erosion control methods; and a monitoring
requirement for the discharge of surface-water runoff into lakes, streams,
and canals.
The Department of Community Affairs has recently been actively charged
with the responsiblity of coordinating growth management in the State, which
will reflect on drai/iage facilities and projected areas growth.
To provide adequate drainage and stormwater management, the regional
Water Management Districts work tn conjunction with the Departments of
Community Affairs and Environmental Regulation to obtain the necessary master
planning and overall coordination. The transportation facilities overseen by
the Department of Transportation must be recognized and included in the
planning and implementation strategy of these conjoining authorities.
1986 FOOT DRAINAGE MANUAL
The Florida Transportation Code establishes the responsibilities of the
State, counties, and municipalities for the planning and development of the
transportation systems with the objective of assuring development of an
integrated, balanced statewide system. The Code's purpose is to protect the
safety and general welfare of the people of the State and tq preserve and
improve all transportation facilities in Florida. The Code sets forth the
powers and duties of the Department of Transportation to develop and adopt
uniform minimum standards and criteria for the design, construction,
maintenance, and operation of public roads.
258
-------
The policy set forth In the manual provides guidelines for the planning
and design of drainage and stormwater management control features for the
Department. The guidelines, while positive and effective, are intentionally
broad, since no policy statement can cover all contingencies. Its primary
purpose is to establish uniform practices and to ensure that the best
information available is applied to project conditions. It neither replaces
the need for professional engineering judgement nor precludes the use of
information not presented in the manual.
Whether the Department must comply with local rules and programs for
stormwater management is a question that generates great doubt and confusion.
As a general rule, the Department should cooperate and comply with local
regulations where such compliance would not be detrimental to the
Department's interests or its ability to carry out its responsibilities.
Stormwater management must be a marriage of both quantity control and
quality control. This is the direction taken by the proposed Drainage
Manual. The manual is presented in three volumes: Policy, Procedures, and
Theory.
Volume One sets forth the prevailing policies as they pertain to
collecting, receiving and passing stormwater runoff from and through the
rights of way. The policies reflect the Department's responsibility to act
on the public's behalf in matters of safety, economics, effectiveness, and
environmental concerns. A listing of chapters is contained in Table 3(a).
Volume Two provides the most common or usually appropriate procedures
required to design the accordance with the policy. Volume Two includes
frequently used charts, equations, computational forms and figures extracted
from numerous technical publications. Textual material from these
publications is not duplicated in its entirety. Discussions on applying the
procedures focus on step-by-step calculations which assume that the reader
has a working knowledge of theoretical fundamentals and terminology. Desktop
technical procedures are presented in the manual. Computer programs of the
Department's mainframe computer are summarized. Tables and figures are
located at the end of each chapter. A listing of chapters for Volume Two is
presented in Table 3(b).
TABLE 3(a) 1986 FOOT DRAINAGE MANUAL VOLUME 1 - POLICY
Chapter Chapter Heading Pages
1 Introduction
2 Organization
3 Drainage Law
4 Policy Guidelines
Appendix
2
8
18
21
9
TOTAL "58
259
-------
TABLE 3(b)
Chapter
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
1986 FOOT DRAINAGE MANUAL VOLUME 2 - PROCEDURES
Chapter Heading
Introduction
Planning and Facility Location
Flood Plain Encroachment and Risk Evaluation
Data Collection
Stormwater Hydrology
Roadway Grades
Open Channel Hydraulics
Culvert Hydraulics
Bridge and Bridge Culvert Hydraulics
Storm Sewer
Culvert Material Selection
Gutter, Inlet and Pavement Hydraulics
Retention/Detention Hydraulics
Stormwater Pumping
Subsurface Drainage
Shore and Bank Protection
Erosion and Sediment
Computer Programs
Standard Indexes and Special Details
Permitting
TOTAL
Pages
3
9
25
39
68
5
40
76
25
40
23
32
20
12
32
21
30
14
30
39
573
Volume Three provides background information on drainage fundamentals,
supplementing the procedures presented in Volume Two. The focus of the
contents is on theoretical background. Not all chapters contained in Volume
Two are included. A listing of the chapters is presented in Table 3(c).
The manual has increased in size due mainly to an attempt to provide
most of the more widely utilized engineering materials. Stormwater quality
management requirements and recommended procedures are given chapters rather
than lines. There is also a difference in quantity requirements that must be
emphasized. In recoginzation of recent adverse court decisions the
Department has altered its design concepts significantly. Increases in peak
Stormwater discharge rates by upland owners will not be accepted or designed
for unless it is part of or provided for in a comprehensive local plan which
is substantially complete.
Volume One, Chapter 4 includes permitting both by and with the
Department. A discussion on the consideration of future development in the
design is presented.
Volume Two, Chapter 2 deals with floodplain encroachments, establishing
the level of impact, detailing agency coordination, planning and permitting
at the location planning phase. Chapter 3 identifies various highway
260
-------
TABLE 3(c) 1986 FOOT DRAINAGE MANUAL VOLUME 3 - THEORY
Chapter Chapter Title Pages
1
2
3
4
5
6
7
8
9
10
11
Introduction
Surface Hydrology
Subsurface Drainage
Open Channel Hydraulics
Culvert Hydraulics
Bridge and Bridge Culvert Hydraulics
Storm Sewer Hydraulics and Hydrology
Gutter, Inlet and Pavement Hydraulics
Retention/Detention Hydraulics
Erosion and Sediment Control
Shore and Bank Protection
References
Glossary
TOTAL
3
45
18
23
14
24
8
20
12
17
19
11
25
239
activities in floodplains givinr tht level of studies required by Drainage
for each activity. Chapter 5 discusses various hydrologic procedures, often
required to satisfy permit requirements of other agencies as well as
Department design needs. Chapters 6,7 and 8 are hydraulic procedures used to
establish highway grades, open channel flow and culvert requirements.
Chapter 9 provides information required for bridges and bridge culverts,
including coordination requirements with the local communities when Federal
Emergency Management Agency (FEMA) regulated floodways are involved. Chapter
13 provides design procedures for. retention and detention design both to
reduce outfall requirements and to conform with permit requirements. Flood
routing procedures, filtration calculations for detention with filtration and
landlocked basin retention requirements are presented. Chapter 15 provides
procedures for subsurface drainage, including exfiltration (french) drains
and drainage well designs. Chapter 16 provides information for shore and
bank protection. Chapter 17 contains erosion and sediment predictive methods
and control procedures both temporary and permanent. Standard details and
indexes used by Drainage are explained in Chapter 19. Permitting activities
with regulatory agencies as well as permits issued by the Department in
drainage or stormwater related matters are explained in Chapter 20.
Most of the procedures in the manual are obtained from widely available
sources, particularity the U.S. Department of Transportation, Federal Highway
Adminstration (FHWA).
The manual will be updated annually. Updated material will be furnished
to purchasers of the manual at no charge provided that the manual owner's
address is current. The manual is being given a final internal review prior
261
-------
to printing. It is anticipated to be available by Fall 1986. Its purchase
price has not been established. If you would like to get on a "reserved"
pre-publication list you may do so by either writing or telephoning and
requesting to be put on the list. Requests should be directed to the State
Drainage Engineer, Florida Department of Transportation, Mail Station 32, 605
Suwannee Street, Tallahassee, Florida 32301; Telephone (904) 487-1700.
Persons who are on the list will be advised of the price and availability and
would be guaranteed a copy from the initial printing.
I hope the preceeding presentation has provided insight into the
changing needs which the new Drainage Manual is addressing. The need for
safe, well-drained and available transportation facilities has not changed,
but the technique required to meet a new awareness of the need for quanity
and quality stormwater management has changed significantly. The Department
is dedicated to preserving and enhancing the unique environmental
geographical area known as Florida while providing for its transportation
needs.
The Department should be considered as a "developer" of transportation
facilities much the same as a residential site developer or a commercial site
developer. As such the Department should support and conform to regional and
local comprehensive stormwater plans rather than establish such plans. The
primary need of the Department remains adequate drainage. Required stormwater
management practices should reflect the procedures developed and required by
those agencies primarily responsible for the entire basin.
The work described is 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.
262
-------
APPLICATION OF THE OTTSBMM MODEL
FOR A RELIEF SEWER STUDY IN LAVAL, QUEBEC
by
R. Roussel ( 1)
J.C. Pigeon (2)
J.R. Noiseux (3)
ABSTRACT
The OTTWSMM (Ottawa University Storm Water Management Model) model is
an expanded version of EPA - SWMM which accounts for overland flow, inlet
restriction and surface ponding. The model was recently applied in the re-
lief sewer study on an area with an old combined sewer in Laval, a municipa-
lity in the Montreal Urban Conmunity.
The first part of the paper describes briefly the area, flooding pro-
blems and constraints related to the interceptor and high levels in the re-
ceiving water body.
The second part of the paper describes the solution developed by means
of the inlet control analysis and gives an economic comparison with more ad-
ditional alternatives.
The last part of the paper presents some considerations on the use of
the model and its interface with a simple screening technique.
INTRODUCTION
Severe flooding problems in the city of Laval prompted the municipal
authorities to request a full scale investigation. This was undertaken by
Les Consultants Dessau Inc. and the proposed design is presented in this pa-
cer.
(1) Drainage Engineer, Les Consultants Dessau Inc.,
1200 ouest, boul. St-Martin, Laval (Quebec) H7S 2E4
(2) Director, Drainage Dept. City of Laval,
3, Place Laval, suite 300, Laval (Quebec) H7N 1A2
(3) Manager, Environmental Division, Les Consultants Dessau Inc.,
1200 ouest, boul. St-Martin, Laval (Quebec) H7S 2E4
263
-------
The city of Laval is the second largest city in the province of Quebec/
Canada and has a population of 300,000. It is situated on an island sur-
rounded by Des Mille-Iles and Des Prairies rivers. The undertaken investi-
gation deal with a particular watershed currently served by a combined sys-
tem.
DEFINITION OF THE PROBLEM
FLOODING PROBLEMS
The Montmorency-Brien watershed, located in Laval, Quebec, has a drai-
nage area of 102 hectares. A single family housing development has led to
an imperviousness of 35%. This urbanized area is currently drained by a
combined sewer system that appears to be insufficient to handle a 10 year
flow.
In order to determine the extent of flooding, a study was performed in
which citizen complaints were evaluated. The following trends were evident:
Flooding occurs all over the watershed as a result of insufficient
pipe capacity (collector and laterals)
Spring and Summer are the critical flooding periods with lower ele-
vation areas being the most affected. The frequency of flooding has been
known to range from none to the three times per year. The average return
period of flooding for the whole watershed is around two years.
Most complaints were for basement flooding as a result of combined
sewer surcharge backing up into the houses.
It should be pointed out that the study was based on actual complaints
registered at the City Hall. In reality, the number of flooding incidents
is somewhat higher.
PRELIMINARY INVESTIGATIONS
A computer simulation of the existing pipe network indicated that 90%
of the existing collector pipes as well as 60% of the lateral pipes do not
meet the up-dated design criteria required by the City of Laval. The
collector average capacity corresponds to a two year storm while the lateral
pipes indicated a very diverse capacity to handle storm generated flow (3
times in a year to one in 25 years). These are definitely not within the 10
year return period design restraints as specified by the city.
In an attempt to better comprehend the flooding problems, Spring and
summer conditions were individually simulated. Spring conditions are cha-
racterized by a lower rainfall (3 times less than summer conditions) and a
high water level of the receiving body as a result of snowmelt. It is
indicative that the outfall pipe is found to be completely submerged. As a
result of the higher water level in the receiving water, flooding conditions
are more severe in the depressed areas, which have a lower potential ener-
gy. This was reflected in the citizen complaints. During the summer
264
-------
months, the water level is at the half-point mark in the outfall pipe even
though the combined sewer is subject to intense rainfall.
As a result of the preliminary analysis, it was felt that due to the
existing circumstances, all pipes should be individually analysed. For the-
se purposes, the OTTSWMM model (Ottawa storm water management model) was
used in connection with the surcharge sub-model EXTRAN (Extended trans-
port) . In this way, the influence of the variations of the river level and
the very low slope of the existing combined collector were accounted for.
OPTIMUM SOLUTION
DESCRIPTION
The traditional approach to rectify the flooding problems of the Mont-
morency-Brien watershed would have called for the twinning of 60% of the
lateral pipes. Due to space limitation twinning could not be applied to the
collector pipes hence almost 90% of them would have to be replaced by higher
capacity pipes. The implementation of this programm would have led to ex-
tensive excavations in the majority of the existing roads. Furthermore, due
to topographic peculiarities, a combined sewer of 3,000 mm diameter at a
depth of 9 meters would have to be installed at a location. Apart from the
obvious economic and severe inconvenience repercussions this approach would
involve, it would provide only a limited protection against flooding. (1 in
10 years)
A solution was developed in order to reduce the twinning of the exis-
ting sewer system and also provides for a higher protection against floo-
ding. This system utilizes inlet control techniques to reduce the flow en-
tering in the existing system (as indicated in figure 1). Calibrated orifi-
ces are installed in the catchbasins in order to restrict the water flow en-
tering in the combined sewer. This keeps the hydraulic grade line below ba-
sement levels and consequently, avoids the possibility of basement floo-
ding. This solution uses the streets to carry the water flow exceeding the
pipe capacity. Hence the streets are actually acting as open channels, in
rare storm events, transporting the excess flow which is subsequently captu-
red by new storm sewers. Figures 2 to 5 outline typical cases of actual
conditions and how these will be altered by the implementation of the propo-
sed solution.
The present lateral pipe and collector could not accomodate a 10 year flow.
In the proposed system (figure 2b), use of inlet control devices allow for
surface runoff to be collected at a lower point by a proposed storm sewer
collector. This method avoids twinning of the lateral pipe.
In figure 3a, a low point is located upstream of the existing collector.
Street ponding was not a suitable solution due to excessive volume of flow
and traffic restrictions. In the proposed system, inlet restrictions are
installed in the latter part of the street. A storm collector is installed
to accomodate the increased flow to the low point. This eliminates twinning
of the latter part of the lateral pipe.
26b
-------
CATCH8A5IN ORATING
COU-tCTlHO C4TCHB01N
INLET CONTROL
FIGURE'!
A similar problem as the one shown in figure 3. The solution (figure 4b)
involved the installation of an additional lateral pipe and collector nearer
to the ground surface. Thus by this technique overloading of the downstream
system was avoided.
In a particular location, it was deemed necessary to install restrictions in
the catchbasins to direct the flow in a shallow storage. Storm water was
stored and gradually released into the system at a rate which allowed for
its accomodation by the existing network. While this approach allowed for
keeping the lateral pipe and collector intact/ it was difficult to install
the storage due to espace limitations.
266
-------
1
LOW POINT
FIGURE 2o
EXISTING SYSTEM
EXISTING
COMBINED
COLLECTOR
|«mrol
FIGURE '2b
PROPOSED SYSTEM
LOW POINT
EXISTING I
COMBINED j-
COUECTOR
DATA INPUT AND ANALYSIS
In order to implement the proposed solution, the following input para-
meters are required:
Tributary area
Elevation of high and low points
- Street longitudinal inclination
267
-------
- Curb or sidewalk elevation
Catchbasin grating specification (figure 6)
Actual pipe capacity
It is pointed out that flow generated by impervious areas directly con-
nected to the system will not flow on street. This was incorporated in all
subsequent analysis.
HIOH POINT
LOW POINT
::=^:
1
-
I
L-
\
<
!
I
LOW POINT
\
rr^
r
EXISTING
COMBINED
COLLECTOR
S*\
1 1-r -,
PRQTFCT.^ |QYFJPe -»-
1 ^1^ PROTECTION f invr... .
/
/
FIGURE-3o
EXISTING SYSTEM
HIGH POINT
LOW POINT
LOW POINT
PROPOSED
STORM
COLLECTOR]
EXISTIN6
COMBINED
.COLLECTOR
FIGURE = 3b
PROPOSED SYSTEM
268
-------
LOW POINT
HIGH POINT
EXISTING
COMBWEO
COLLECTOR
FIGURE 4o
EXISTING SYSTEM
PROPOSED
STORM
COLLECTOR
FIGURE ' 4b
PROPOSED SYSTEM
A detail analysis was undertaken to predict the extent of necessary
measures to alleviate the flooding problem in the Montmorency-Brien water-
shed. It was necessary to perform this analysis pipe by pipe, street seg-
ment by street segment and catchbasin by catchbasin. The following were
evaluated:
Flow in the existing combined sewer
Flow in the proposed storm sewer
- Flow on all the street segments and corresponding depths of flow in
order to avoid flooding in shallow curb areas.
269
-------
The purpose of the above-mentioned detailed hand-analysis was the de-
termination of the optimum inlet control location and the development of a
preliminary "sizing" of the pipes. We used the OTTSWMM model (Ottawa Uni-
versity storm water management model) with the results of the detailed
hand-analysis as input parameters. The model was manipulated to generated
and "cut" hydrographs which correspond to existing and proposed systems (as
showm on figure 7). In order to compute the surcharge imposed on the sys-
tems, the model EXTRAN (Extended transport) was used in connection with
OTTSWMM. This was necessary so as to accomodate the fluctuations of the ri-
ver flow as well as low gradient of the existing combined collector.
1
LOW POINT
HIGH POINT
EXISTING
COMBINED
COLLECTOR
FIGURE'So
EXISTING SYSTEM
LOW POMT
PROPOSED
STORAGE
~l
HIGH POINT
UH8T
EXISTING
COMBINED
COLLECTOR
PROTECTION 5 IOYPAP
-------
°DXX) OX)1
O.O2 OX» O.O4 OJ05 O.06 O.O7 ODB OO9 0.1O
0 APPROACH (cmi)
FIGURE 6a
A ^^ ^IAPPROACH FLOW
T_ / \ ^ICARRY-OVER
OJ2
CAPTURED
BY ORATING
WITH ORIFICE
RESTRICTION
TIME
FIGURE'-6b
HYDRAULIC PERFORMANCE OF CATCHBASIN
271
-------
It is important to follow up this analysis by monitoring of the field
implementation of the recommended alterations to the system.
GLOBAL IMPLEMENTATION
The global implementation of the suggested solution required several
alterations to the existing combined system. A shallow storm sewer collec-
tor of 1,650 mm diameter at the outlet was proposed. For space limitations,
some wide rectangular sections were used. It was thought appropriate to se-
lect a shallow depth so as to minimize excavation costs and disruption of
other utilities systems.
It was necessary to completely block 88 catchbasins and to install in-
let controls in 40 of them. In the City of Laval, catchbasins are located
at almost twice the frequency compared to a neighbouring municipality. Hen-
ce , it was possible to block as many without serious repercussions. Two
hundred and fifty-four (254) catchbasins of the existing system had to be
cut-off and reconnected to the proposed storm system. The reason for the
large number of reconnected catchbasins is that the proposed laterals and
collector pipes intersect low points where a large number of these catch-
basins were situated. A larger than usual special inlet was installed at
low points locations in order to discourage surface ponding. The new gra-
ting was directly connected to the new storm system.
FIGURE:/
HYDROGRAPHS SIMULATED
272
-------
As a result of the detailed analysis, four street intersections had to
be reprofiled and sidewalk lengths totaling about 60 m had to be built. In
this way, runoff flow was prevented from intruding into garages located be-
low the street levels. Furthermore, three (3) new combined lateral piped
had to be installed so as to accomodate the flow generated by the impervious
areas directly connected.
MODEL SELECTION
It is evident that there is a need for selecting the most appropriate
model for a given set of conditions. The opinion of the authors is that
simple models should be used in order to better evaluate the prevalent con-
ditions and allow for the selection of the most appropriate sophisticated
model for the final design. This approach carries the benefit of reducing
the necessary man-hour and computer time.
In this study, the IMPRAM model (Improved rational method) was employed
to define the actual problems. ^Preliminary results prompted the use of
OTTSWMM and EXTRAN models for the final analysis.
DISCUSSION AND CONCLUSION
A need to improve the current combined system serving the Laval munici-
pality prompted this study. A novel technique-was used by which the exis-
ting system was relieved by means of storm sewers and catchbasin inlet con-
trols. This led to the development of a mostly separated system with consi-
derable advantages:
a) Reduction of the length of proposed new pipes from 10 km projected
by traditional design techniques to 2,8 km.
b) The installation of the shallow storm sewer will not interfere with
other utility connections and will be less prone to river fluctuation in-
fluences .
c) As the proposed system is not directly connected to the residences,
flooding by surcharge will not be a problem even if the hydraulic grade line
reaches the ground surface.
d) Inlet controls keep the hydraulic grade line (of the existing sys-
tem) below basement levels while still use the existing collector to its
full capacity.
e) Implementation of the proposed system will restrict the volume as
well as the frequency of combined overflow into the river by about 65%.
f) A traditional approach to this problem was estimated to cost around
7.5 million $. This proposed solution is projected at 4 million $.
273
-------
ACKNOWLEDGEMENTS
The work presented in this paper was performed by Les Consultants Des-
sau Inc. The investigation was commissioned by the City of Laval, Quebec,
which the authors wish to thank for permission to publish this paper.
Note: The work described in this paper was not funded by the U.S. Environ-
mental Protection Agency and therefore the contents do not necessa-
rily reflect the views of the Agency and no official endorsement
should be inferred.
274
-------
APPLICATION OF INLET CONTROL
DEVICES AND DUAL DRAINAGE MODELLING
FOR NEW SUBDIVISIONS
by
Paul Wisner, PHD, P. Eng.
Dept. of Civil Engineering
University of Ottawa
Ottawa, Ontario
C. Kochar, MASc, P. Eng.
Canada Mortgage &
Housing Corporation
Ottawa, Ontario
Hugh Fraser, MASc, P. Eng.*
Novatech Engineering
Ottawa, Ontario
C. Rampersad, P. Eng.
Andrew Brodie Associates
Thornhill, Ontario
ABSTRACT
Concern over basement flooding has led to the application of a new technique
in stormwater management practice, known as inlet control. Major storms have
often caused basement flooding when the sewer pipes have surcharged. A
traditional, but costly remedy to this problem involves resizing the sewer
pipes to handle the higher flows. A new method is to control the flow into
the sewer system and thus reduce the occurrence of basement flooding. This is
achieved by applying the dual drainage concept and installing inlet control
devices in the catchbasins.
This paper summarizes the results of a study conducted for the Canada Mortgage
and Housing Corporation that evaluated the performance of three commercially
avaialable inlet control devices. The hydraulic operation of the units was
tested with a physical model using debris laden water. The study recommended
that a minimum size for the orifice type flow regulators be between 14 and 20
Ips.
Numerous applications of this concept have been made in the United States and
Canada in connection with relief sewer studies. In addition, the concept is
being applied in new subdivisions where significant cost savings may be
obtained by designing the pipes to be compatible with the level of inlet
control.
*presently with CCL Consultants Inc.
275
-------
INLET CONTROL DEVICES AGAINST SEWER SURCHARGES
The initial objective of storm sewers, traffic convenience, has been
replaced by property protection against flooding. This was assured by a
judgemental selection of design storm return periods that vary from 5 years
(Metro-Toronto) to 100 years (Dallas). After a major storm that caused
damage by basement flooding, the traditional solution was to Increase the
design return period or modify the inlet time in the rational method.
The choice of the municipal engineer used to be between an expensive
increase of storm sewer sizes or risk flooding during the life span of a
dwelling. In the recent past the problem has become more critical with
basements being used for family or-recreation rooms in modern homes, damage
from flooding is more severe.
One of the most promising solutions to this dilemma was found to be the
application of the dual drainage principle in conjunction with inlet control
devices (Wisner and Hawdur, 1984). The principle of this solution, which
eliminates surcharge without pipe size increases, is shown in Figs. 1 and 2
for new developments and Fig. 3 for old developments with points. This
solution is now mandatory in many Canadian municipality where it is designed
with the OTTSWMM model. OTTSWMM (The Ottawa University SWM Model) is a
modified version of the EPA SWMM-EXTRAN model which accounts not only for
conduit but also for overland flow, ponding and inlet controls. A schematic
of the models operation is shown in Fig. 4 (Kassera, 1983, Wisner, 1982). The
Appendix gives a brief description of OTTSWMM which is available for mainframe
and microcomputers. Development of OTTSWMM was made by the IMPSWM
(Implementation of Storm Water Management) program at the University of
Ottawa. This program is based on a co-operation between IMPSWM researchers
and participants from engineering firms and municipalities. The research
varies from modelling and experimental studies to review of policies, public
attitudes and economics of storrawater management.
Many projects in Southern Ontario have used the dual drainge-inlet
control principle in new subdivisions. In these cases the maximum flow
accepted in the storm sewer and controlled by ICD's (inlet control devices)
varies from 28 Ips to 40 Ips per catchbasin (Table 1). These limits were
selected to eliminate sewer surcharge for a 100 year storm with sewer sizes
designed for free surface flow at a 5-year design storm. One of the main
reasons for this ICD limit was the lack of information on the performance of
various available types of ICD's and concerns regarding clogging. Relief
sewer projects in the U.S.A. for older systems used lower limits such as 10
Ips per catchbasin or even less. Some of these designs were done with a
proprietory model (Donahue 1982), others have been developed without a
detailed analysis of street flow depths (Pisano, 1982).
A study co-ordinated by NOVATECH ENGINEERING with co-operation from
ANDREW BRODIE ASSOCIATES and IMPSWM in 1985, examined the operation of ICD's
and the economic and operational implications of various levels of ICD
control. The study was sponsored by the Canadian Housing and Mortgage
Corporation. This paper presents briefly some of its findings and
conclusions.
276
-------
ORIFICE
PLATE
O
J
TIME
TO STORM
SEWER
APPROACH FLOW
CARRY-OVER
CAPTURED BY
GRATING
WITH ORIFICE
RESTRICTION
APPROACH FLOW
Severe Storm
STREET PROFILE
.CARRY OVER FLOW
» ^
STREET SECTION
MAXO
K*^ 1
15 mj
SIDEWALK
INLET CONTROL DEVICE LIMITS FLOW
STORM SEWER
Figure 1 Principle of Inlet Control
277
-------
FIG. 2a
LOCATION OF LOW POINTS AND LANDSCAPING OF (DEPRESSED)
PARK AREA FOR RUNOFF DETENTION.
FISHBONE COVER
Inlet Control t.Scfs
Inlat Control tdi
EDMONTON COVER
8 10 12
0 APPROACH (Cfs)
14
IB
18
RG. 2b HYDRAULIC PERFORMANCE OF CATCHBASIN COVERS
278
-------
EXPERIMENTAL RESULTS
A comparison of several commercial types of ICD's was conducted in the
field in Skokie, Illinois (Donahue, 1984). Field observations may confirm if
an operation is acceptable or not, but will not explain the hydraulic
performance and clogging mechanism. Critical rainfall events are rare, and
difficult to monitor. Therefore, some field observations used an artificial
catchbasin loading from fire hydrants (Pisano, 1985). For this reason the
present study compared four ICD's on a hydraulic model at 1:1 scale in the
hydraulic laboratory at the University of Ottawa (Fig. 5). ICD performance
was observed in a plexiglass catchbasin and flow measurements were made with a
triangular weir.
The three ICD types commercially available and tested at the University
of Ottawa are:
A. SCEPTER - an orifice type of ICD with a self-cleaning notch (Townsend,
1984).
B. CROMAC - an orifice type with a vcariable area slot ICD.
Figure 3a Inlet Control to reduce the Sewer Reconstruction Cost
PARKING LOT
STREET
Figure 3b Parking Lot Inlet Control
279
-------
C. HYUROVEX - a vortex ICD, representing an improved German version of
HYDROBRAKE orifice.
The forth type called the HANGING TRAP is a self made ICD proposed in the
Skokie studies. The schematics of different devices are shown in Fig. 6.
Head-Discharge curves were determined and found to be practically the
same as those indicated by the manufacturers. The discharge coefficients are
relatively the same for the first two devices, but are much small for
HYDROVEX.
RAINFALL
RUNOFF
TIME-
MAJOR
SYSTEM
SUB-MODEL
SURFACE
RUNOFF
SUB. MODEL
INLET
SUB.MODEL
A»*«3ACMnav
CMHT-CVCK
awnmeo IT
xrtTH oNmci
~~iTir "
*«
T1MC
MINOR
SYSTEM
SUB.MODEL
MAJOR SYSTEM
STORAGE
SU8-MOOEL
MINOR SYSTEM
STORAGE
SUB-MODEL
OUTPUT
FIG. 4 : AN OVERVIEW OF THE UNIVERSITY OF OTTAWA
STORMWATER MANAGEMENT MODEL (OTTSWMM)
280
-------
TABLE 1 - EXAMPLES OF CASE STUDIES USING INLET CONTROL DEVICES
CASE
STUDY
A
B
ro
co
1-1 c
D
^
AREA
HA.
465
117
10
32
44
LAND
USE
RESID
RESID
RESID
RESID
RESID
AV.IMP.
00
31
30
25
40
35
TOTAL NO. INLET Z INLETS
OF INLETS DENSITY WITH
(CB/ha) RESTRICTORS
MDP
320
32
100
166
STUDY
2.73
3.2
3.1
3.7
25
75
50
51
LEVEL OF SURCHARGE
INLET ELIMINATED
CONTROL (LPS)
5 YEAR
42
42
34 9 28
16 0 42
28
CONTROL AVOIDED
Limited &
Acceptable
Negligible
Ipa Limited &
Ips Acceptable
Limited 6
Acceptable
MAX. STREET
PLOW DEPTH
(cm)
28
20
24
26
PARK STORAGE
USED
YES
YES
NO
NO
YES
EXTERNAL COMMENTS
AREAS
(ha)
- A detailed inlet
control analysis
was not conducted.
ICD's used mainly
to avoid surcharg-
ing the sewers
during rare storms.
15.4
118 RESID 35
100
28
ELIMINATED 37
YES
90.8 - Severe Inlet con-
trols were required
because of the
status of the pro-
ject at time of
analysis
12 RESID 25 50
42
68
28 Limited 4
Acceptable
12
NO
-------
MULTIPORT OIFFUSER
DIFFUSING TANK
CATCH BASIN TANK
L_T1_
(UDNGITUDINAL STREET SLOPE I -5 %)
ENTRANCE BAFFLE
HEADER TANK
GUTTER
|
CURB
4-27m
PLAN
J
GRATE
8% GUTTER SLOPE
>/0 STREET SLOPE
PLEXIGLASS CATCHBAStN
400mm PLEXIGLASS LEADER PIPE
DEBRIS TRAP
RECTANGULAR NOTCH
CATCH BASIN TANK
SECTION A-A
Figure s The Physical Model
282
-------
SIZE
CONSTANT
A) Schematic of the Scepter
Inlet Control Device
Low flow
free discharge
(D.W.F.)
Controlled flow
B) Schematic of the Hydrovex linit in Operation
ORIFICE
C) Schematic of the Cromac ICD
D) Banging Trap ICD
Figures Inlet Control Devices
283
-------
Clogging experiments were in general conducted in a conservative mode as
compared to natural occurences. The limit found for operation without
permanent clogging was 20 Ips for CROMAC and 14 Ips for SCEPTER. HYDROVEX was
tested and operated well down to 8.5 Ips, although it may function adequately
for even lower flows than in the test. For all ICD's it was found that
certain combinations of leaves and branches loading may temporarily plug the
orifice. Consequently, it is considered that periodic visual inspection and
cleaning is necessary.
Hydraulic laboratory findings were compared with field experience in
Markham and Scarborough, municipalities near Toronto, where ICD's have been
successfully used for several years. A group of municipal engineers attended
demonstration tests and presented useful comments. As a result of these
discussions, another factor introduced in the selection process is the effect
of protruding ICD's on current catchbasin cleaning practice. Based on these
considerations, for typical Canadian separated sewer catchbasins the HANGING
TRAP device was not recommended. The cost of HYDROVEX is at present higher
than that of the simpler orifice type ICD's and it is therefore useful only if
a low level of control is required. This, however, has other implications,
which can be examined with OTTSWMM.
SELECTION OF AN INLET CONTROL DEVICE
As indicated above, the present practice in new Canadian subdivisions is
to use a minimum control level of 28 Ips. This limit can be reduced as low as
14 Ips and orifice ICD's (SCEPTER) could still be used. For a lower level,
the HYDROVEX is the only recommended ICD. Operations and cost implications
were analyzed with the OTTSWMM model for five levels of control in a 42 ha
typical subdivision, shown in Fig. 7. The OTTSWMM model sized conduits to
avoid surcharge for the 100-year storm (Keifer and Chu distribution in
Metro-Toronto).
Reducing the control level of ICD's gives pipe flow equivalent to a
traditional design for a more frequent storm. As an example a 2-year storm,
corresponds for Metro-Toronto conditions to a 20 Ips ICD.
Results are summarized in Table 2. It was found for example that for a
20 Ips ICD, park storages for overland flow would operate about twice per year
instead of once every five years as for the 42 Ips ICD currently used. The
depth in the parks for this more frequent flooding would, however, be
relatively small. The increase in maximum street flow depth at the gutters
would be very small as compared to the present practice (1-4 cm for a 100-year
storm).
Table 2 is an example of information available to a decision maker from
the OTTSWMM model. It is found that if the level of control is lowered to
28 from 42 Ips currently used in Ontario for ICD's In new subdivision, the
savings may be significant. For example, in a typical residential area, inlet
control at 20 Ips per catchbasin can be achieved with the less expensive
orifice type ICD's. The corresponding saving is $5,500/ha. Of course
municipalities now using a 10-year or larger storms can achieve greater cost
reductions.
284
-------
r\j
00
MINOR SYSTEM
MAJOR SYSTEM
CATCH BASINS
Figure 7 Studied System
-------
Table 2
COMPARISON OF ECONOMIC AND OPERATIONAL CHARACTERISTICS
Area of
Park
Approximative Storage
Level Savings in Maximum Gutter % of
of $l,000/ha Depth (cm) Total Area
Control 100-Year 5-Year 100-Year 100-year
(Ips) 7.3 ha .42 ha 7 ha 42 ha 7 ha 42 ha 42 ha
42 0 0 7.4 14 13.7 27 1.1
28 2.9 3.2 8.4 16 15.0 30 1.4
20 4.6 5.5 9.6 18 15.7 31 1.8
14.0 4.7 6.6 10.2 20 16.2 33 2.2
8.5 6.4 8.6 10.9 22 16.5 34 2.7
Average
Depth
in Park
for a
5-Year
Storm
(cm)
18
17
18
20
23
The main limitation in achieving higher savings with inlet controls of
HYDROVEX type, is the concern related to higher street flow depth during major
storms and increased frequency of park flooding. These effects vary with
contributing areas, imperviousness, slopes, -etc. and can be assessed using
detailed OTTSWMM computations for the specific conditions of each project.
The sewer network or minor system can follow either a dendritic or looped
pattern. When a new system is being designed, the pipes should follow a
dendritic pattern. More sophisticated analysis of looped or surcharged pipes
is possible with the EXTRAN (Extended Transport) submodel. The pipe slopes
control the flow direction in the minor system. Water enters the minor system
through storm inlets which are connected to manholes at sewer junctions. At
286
-------
each junction the flow from upstream pipes is added to the street inlet flow
giving the total flow to be routed to the next segment. Pipes are sized for
the peak flow at each junction.
MODEL OPERATION
The model is composed of four main submodels, a surface runoff submodel,
inlet submodel, minor system submodel and the major system submodel.
The input data consisting of subareas, street segments, pipe segments,
and storages are read and connectivity matrices set up. The input data order
is shown in Figure C.I. Computations are done in a number of steps. First
the runoff for each subarea is computed using the Runoff Block routine
borrowed from EPA-SWM Model. This is followed by the major system routing.
Starting at an upstream major system segment subarea runoff is routed down the
street segment with any upsteam carryover flow. In conjunction with the major
system routing, inlet flows are determined. Any flow captured by the inlets
are stored for minor system analysis. Excess flow not caputred by the inlets
form carryover flow to be routed down the following major system segment. The
street segment routing and inlet capture are continued until all of the street
segments have been considered.
The computations for the minor system components are performed next. The
user selected input determines how the pipe system is analyzed. Design of a
new system starts with the most upstream sewer segment and proceeds
downstream. The pipe flow is the total of inlet capture flow and the upstream
pipe flow. Design or resizing of individual pipes is performed with the MINOR
submodel. When the pipes are sized, free surface and a dendritic sewer system
is assumed. Pipes are selected based on the input slopes, pipe roughness and
computed peak flow, using the Manning equation. The model selects the
smallest commercial sewer size that will maintain free surface flow.
If the pipe surcharge analysis option is selected the inlet flows are
saved in a separate file, to be used by EXTRAN subroutine. Analysis with
EXTRAN can be conducted in looped pipe systems and in systems where pipes have
insufficient capacity for free surface flow. Some small surcharging may be
desirable during major storm events, to prevent resizing of surcharged pipes.
Surcharge levels must be kept below foundations, to prevent basement flooding
damage. The EXTRAN model must be used to determine if the surcharge levels
are acceptable.
ARE MODELS SAVING MONEY?
The change of a well entrenched methodology requires not only studies but
also pilot projects and innovative municipal engineers. The first ICD-dual
drainage project in Markhara, Ontario was proposed in 1978. At present the
system is used in other Metro-Toronto municipalities (Vaughan, Scarborough)
and is recommended by criteria in other smaller cities (Oakville, Barrie). It
was also applied in various projects in Nepean, Cumberland, etc. Some initial
applications were implemented after flooding experiences.
287
-------
The proposed reduction of pipe size's, is a next step and may require a
dialogue with decision makers and developers. The use of OTTSWMM for new
subdivisions showns that modelling can be used for a more economic drainage
design.
A previous IMPSWM study (Wisner, Cheung 1982) developed an improved
rational method in which the runoff coefficient is a function of
iraperviousness and the inlet time varies with rainfall intensity. This
method, available for microcomputers under the name IMPRAM gives the peak
flows in close agreement with SWMM model (Fig. 8).
It follows that for a conventional design the use of a more sophisticated
model does not offer an advantage as compared to a slightly improved
traditional computation. In the case however, of the sophisticated dual
drainage-inlet control-park storage design, modelling is justified by
significant savings and evaluation of a more realistic level of protection.
Experience shows that the OTTSWMM model is very simple to use, even if
consultants are not familiar with the SWMM model. Huitt-Zollars, a consulting
firm in Dallas, used OTTSWMM without previous SWMM experience on a relatively
complex relief sewer study for a large downtown area in Dallas (Urable 1986).
The study found that the role of gratings even without ICD's is very
important. OTTSWMM users have support through the IMPSWM (Implementation of
Storm Water Management) program at the University of Ottawa which operates a
hot-line (613-564-39U or 613-564-7022).
The main difficulties are of a different type:
0 acceptance of the reality of overland flow by decision makers.
0 integration of shallow storage parks for overland flows at strategic
locations by planners.
0 convincing consultants for developers to pay more attention to street
grading and avoidance of low points.
If a minimum $2,000/acre saving is not an adequate incentive to overcome
these difficulties, one should also consider the considerable damages caused
by basement flooding in areas where foundation drainage is connected with
storm sewers (Wisner, Hawdur, 1983).
ACKNOWLEDGEMENTS
Mr. D. Keliar, Director of Engineering, Town of Markham, Mr. Jean Claude
Pigeon, Drainage-Director in Laval, Quebec and other engineers in Canada made
a significant contribution to the implementation of OTTSWMM based
applications of ICD's and dual drainage. This was seminal in generating
further applications and particularly this study. The Ontario Ministry of
Environment included the ICD-dual drainage concepts in its recent criteria and
provided assistance for OTTSWMM.
288
-------
AREA Ma.
SUB AREA BOUNDARY
UINOM SYSTEM
Figure 8 Comparison of Peak Flows done with
SWMM and IMPRAM* for the example area
Cumulative area
(fla)
3; 27
4.52
5.90
8.94
10.52
12.42
13.96
15.72
13.37
(ti «
SWMM
(cms)
0.32
0.44
0.55
0.83
1.00
1.19
1.34
1.34
1.78
10 min.)
IMPRAM
(cms)
0.3
0.4
0.5
0.7
0.8
0.9
1.1
1.5
2.0
* IMPRAM (Improved Rational Method) by IMPSWM available
for microcomputers.
289
-------
Mr. John Meunier, distributor of HYDROVEX, Mr. Bob Crosmas, developer of
the CROMAC device attended the experiments and provided ICD devices. The
SCEPTER ICD was also made available by the Manufacturerers. Drs. Stuart
Welsh from Donahue Associates and W. Pisano provided documentation on ICD
applications on existing sewer systems.
The OTTSWMM model was used in a microcomputer version of the IMPSWM
mainframe program, developed for Andrew Brodie Associates by Mr. Kassera.
Review of the study and comments from Mr. Martin Hawdur, P. Eng., President,
Novatech Engineering and advice and discussion with Dr. Ron Townsend,
Professor at the University of Ottawa, on the methodology of experiments and
data from Mr. Andrew Brodie on his numerous applications of ICD's were very
useful. All these contributions are gratefully acknowledged.
REFERENCES
1. Donahue and Assoc. Inc., "Flow Regulator Pilot Study Runoff Control
Program Howard Street Sewer District", Village of Skokie, Illinois, March
1984.
2. Kassem, A., "Development and Application of Simultaneous Routing Model
for Dual Drainage", Ph.D. thesis, Department of Civil Engineering,
University of Ottawa, 1982.
3. Ontario Ministry of the Environment, "Ontario Drainage Design
Guidelines", Draft 1983.
4. Pisano W.C., "An Overview of Four Inlet Control Studies for Mitigating
Basement Street Flooding in Cleveland and Chicago Areas", Presented to
ASCE Luncheon Meeting, Cleveland, November 1982.
5. Townsend R.D. Wisner P.,. and Moss D., "Inlet Control Devices for
Stormsewer Catchbasins: A Laboratory Study", Canadian Hydrology
Symposium, Toronto, May 1980.
6. Townsend R.D., "A Novel Inlet Control Device for Storm Sewer Systems",
Proceedings, International Symposium on Urban Hydrology, Hydraulics and
Sediment Control, Lexington, Kentucky, July 1984.
7. U.S. Department of Transportion, "Design of Urban Highway Drainage - The
State of the Art", Federal Highway Admninistration, Office of Research
and Development, (FHWS-TS-79-226), August 1979.
8. Wisner P., "IMPSWM - Implementation of Storrawater Management Procedures
for Urban Drainage Modelling - 2nd Edition", University of Ottawa,
February 1983.
9. Wisner P. and Hawdur M., "Evaluation of Urban Drainage Methods for
Basement Flooding Proofing". A report prepared by Novatech for Canada
Mortgage and Housing Corporation, October 1983.
290
-------
10. Wright-McLaughlin Engineers, "Urban Storm Drainage Criteria Manual",
Denver Regional Council of Governments, Denver, Colorado, 1968.
11. Wisner P. Eraser H, Hawdur M and Rarapersad C "Evaluation of Inlet Control
in Dual Drainage Systems" A report by Novatech for Canada Mortgage and
Housing Corporation, 1985.
APPENDIX
DUAL DRAINGE COMPUTER MODEL - OTTSWMM
Most urban storm drainge models assume that all the catchment runoff is
transferred directly into the minor system. They have been developed for
designing or analyzing systems with low return period rainfall events.
However, if the dual drainage conept is to be employed, drainage systems must
be designed for events with a high return period. In this case, all the storm
water runoff is not captured by catchbasins and transferred into the pipe
system. A portion of the street flow is captured and transferred to the pipe
system while the remaining carry over flow is transported by the streets.
The OTTSWMM model was designed specifically for analyzing dual drainage
systems. The program has the capability of determining the surface flow, the
hydraulic capture by catchbasin inlets and the pipe flow. It can be used in
four modes:
i) to determine pipe sizes for free surface flow;
ii) to analyze an existing system or proposed design and resize pipes to
maintain free surface flow;
iii) to determine the level of inlet restriction to maintain free surface flow
in pipes;
iv) to conduct a pipe surcharge analysis.
Whatever mode of operation is being used, the basic assumptions of the
model remain the same. The model conducts an analysis of two interconnected
systems, the surface or major system and the pipe or minor system. Since the
computations are done on two levels, the surface and sewer network flows do
not necessarily have to be in the same direction.
The major system is formed by the street network and must follow a
dendritic pattern converging to a downstream outlet. The major system should
be continuous, no water ponding is allowed except at storage locations. In
new subdivision, streets can be designed so that low points are avoided.
Existing development often have low points. In recognizing this the model
permits two types of inlets. Normal inlets are those where flow partly enters
the minor system and is partly passed down the major system. Storage inlets
are located at low points, all of the water enters the minor system at these
inlets.
291
-------
The sewer network or minor system can follow either a dendritic or looped
pattern. When a new system is being designed, the pipes should follow a
dendritic pattern. More sophisticated analysis of looped or surcharged pipes
is possible with the EXTRAN (Extended Transport) submodel. The pipe slopes
control the flow direction in the minor system. Water enters the minor system
through storm inlets which are connected to manholes at sewer junctions. At
each junction the flow from upstream pipes is added to the street inlet flow
giving the total flow to be routed to the next segment. Pipes are sized for
the peak flow at each junction.
MODEL OPERATION
The model is composed of four main submodels, a surface runoff submodel,
inlet submodel, minor system submodel and the major system submodel.
The input data consisting of subareas, street segments, pipe segments,
and storages are read and connectivity matrices set up. The input data order
is shown in Figure C.i. Computations are done in a number of steps. First
the runoff for each subarea is computed using the Runoff Block routine
borrowed from EPA-SWM Model. This is followed by the major system routing.
Starting at an upstream major system segment subarea runoff is routed down the
street segment with any upsteam carryover flow. In conjunction with the major
system routing, inlet flows are determined. Any flow captured by the inlets
are stored for minor system analysis. Excess flow not caputred by the inlets
form carryover flow to be routed down the following major system segment. The
street segment routing and inlet capture are continued until all of the street
segments have been considered.
The computations for the minor system components are performed next. The
user selected input determines how the pipe system is analyzed. Design of a
new system starts with the most upstream sewer segment and proceeds
downstream. The pipe flow is the total of inlet capture flow and the upstream
pipe flow. Design or resizing of individual pipes is performed with the MINOR
submodel. When the pipes are sized, free surface and a dendritic sewer system
is assumed. Pipes are selected based on the input slopes, pipe roughness and
computed peak flow, using the Manning equation. The model selects the
smallest commercial sewer size that will maintain free surface flow.
If the pipe surcharge analysis option is selected the inlet flows are
saved in a separate file, to be used by EXTRAN subroutine. Analysis with
EXTRAN can be conducted in looped pipe systems and in systems where pipes have
insufficient capacity for free surface flow. Some small surcharging may be
desirable during major storm events, to prevent resizing of surcharged pipes.
Surcharge levels must be kept below foundations, to prevent basement flooding
damage. The EXTRAN model must be used to determine if the surcharge levels
are acceptable.
Storage can be provided to control either minor or major system runoff.
The minor system storage may be provided by a superpipe, or other in system
detention facility. For the major system, street flow is typically directed
292
-------
Subcatchments/Surface Runoff
(9)
| Storage (Major System)
(8)
Major System Network & Inlets
(7)
Inlet Capture
(6)
Major System Characteristics
(5)
Storage (Minor System)
(4)
Minor System Network
(3)
f
Rainfall
(2)
Run Control Parameters
(1)
Fig. Cl Main Card Groups For Input Data to QTTSWMM
to a surface storage facility such as a depressed park or detention pond.
Minor system storage is provided to obtain economies in pipe sizing or to
ensure that peak flows do not exceed predevelopraents levels. Major system
r^r?nr1%rCeSSary " ^^^ Street fiow. «* release it at a controlled
rate into the minor system or receiving stream. The designer inputs the
stage-discharge curve to provide the desired control, the computer output
gives the required storage volumes. !f a storage volume was input and it is
exceeded the overflow volume is given in the output. a a in is
The basic computer output of OTTSWMM provides the following information:
1. a print-out of the input data as well as a summary statistics of the
watershed: number of subareas, sewers, storage units, total drainaee
area, etc density of inlets (number of inlets per unit area^avefLe
distance between inlets, etc.; <"«a/, average
2. required sizes of sewers for free surface flow conditions;
293
-------
3. inlet control requirements, that is locations of inlets which may need
flow constricting devices, and limiting capacities if the latter is not
specified;
4. detailed simulation results for specified elements, in printed and
plotted forms:
a) time history of surface runoff;
b) time histroy of major system flows and depths;
c) time history of sewer flows;
5. a summary of simulation results including maximum flows and depths at
various locations as requested.
6. Storages:
a) required storage volume for major system flow;
b) required storage volume for minor system flow, or volume of overflow
if a given storage volume is exceeded.
7. with EXTRAN for sewer flow routing the following information is obtained:
a) printout summary for water depth at junctions;
b) printout summary for conduits showing design flows.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement whould be inferred.
294
-------
USE OF CONTINUOUS .SWMM FOR SELECTION OF HISTORIC RAINFALL
DESIGN EVENTS IN TALLAHASSEE
by: Wayne C. Huber, Brett A. Cunningham and Kevin A. Cavender
Department of Environmental Engineering Sciences
University of Florida
Gainesville, Florida 32611
ABSTRACT
Conventional design methods often utilize a "design storm" synthesized
from a total depth taken from an intensity-duration-frequency curve for an
arbitrary duration,, coupled with an assumed temporal distribution using any of
several available shapes for the hyetograph. The synthetic design storm is
then used as input to a rainfall-runoff model. Unfortunately, the true return
period of runoff parameters such as peak flow or volume is unknown since
antecedent conditions must be arbitrarily assumed when using such a method.
In addition, there is seldom a firm basis for the choice of the storm duration
or temporal distribution. Nonetheless, the use of this method is very common.
As an alternative, a model may be calibrated and verified for the drain-
age basin and then used in a continuous simulation for as many years as there
are available rainfall data. A frequency analysis of the predicted runoff
events may then be used to select historic rainfall events for use in design
based on desired return periods on the parameter of interest, such as peak
flow, runoff volume, flood stage, pollutant load, or pollutant concentration.
The historic rainfall events so identified may then be simulated in more
detail for design purposes.
The results of such an analysis for the Megginnis Arm Catchment in Talla-
hassee, Florida are described in the paper. SWMM simulations with historic
and synthetic design storms are compared with respect to a frequency analysis
of peak flows. Contrary to what is often assumed, it is found that the
synthetic design storms are not necessarily conservative (i.e., they do not
necessarily produce higher peak flows) over the range of return periods consi-
dered (up to 25 years).
295
-------
INTRODUCTION
ANALYSIS OF URBAN STORMWATER
Urban drainage studies may be conducted for several purposes, including
flood-control and water quality control. "Flood control" may simply mean
conventional drainage of urban streets or be more comprehensive in nature,
e.g., for a whole basin. The usual design criterion is peak flow, although
stage (or hydraulic grade line) is likely more important. Runoff volumes are
also of interest when storage is used as a control measure. Quality control
may be applied to limit pollutant loads to downstream receiving waters and may
or may not necessitate a detailed simulation of a pollutograph (concentration
versus time) during a storm. In most cases, receiving waters are sensitive
only to the total load and not detailed variations within a storm (Drlscoll,
1979; Hydroscience, 1979); hence, in roost cases, only total loads need to be
predicted.
For both quantity and quality control, design conditions must be speci-
fied. In some instances, the design engineer has little choice. For in-
stance, in Florida, the Department of Environmental Regulation (DER) specifies
in part that stormwater quality criteria can b'e met by providing for the
retention of the runoff resulting from the first one inch or rainfall or of
the first half-inch of runoff, for projects of less than 100 ac. This leaves
few design options to the engineer, although it is easy to administer.
In the area of drainage design, another common guideline promulgated by
agencies is to specify the use of a synthetic design storm. This is usually
constructed by obtaining the depth for a given frequency and duration (e.g.,
25-yr, 24-hr) from an intensity-duration-frequency (IDF) curve. The depth Is
then distributed In time by using an assumed temporal distribution of the
storm, e.g., the Soil Conservation Service (SCS) Type II distribution for the
hyetograph. What is usually unknown and ignored in such decisions is the true
return period of the design. What level of protection has really been af-
forded by this design? If the desired level of protection is, say, 25 years,
based on rainfall statistics, are the statistics (frequencies) of peak flows
or volumes consistent with this value? It is easy to comprehend that the
return period of a rainfall event is unlikely to be the same as the return
period of the flow peak or volume caused by the event because of variable
antecedent conditions and the inherent nonlinearity of the catchment. Is the
resulting design conservative (over designed) and therefore uneconomic, or the
reverse? It is difficult to say a priori, although comparisons have been made
(Marsalek, 1978; Arnell, 1978,1982). However, most conventional wisdom
assumes that synthetic design storms are conservative, a result that will be
shown not always to be true. For quality, Geiger (1984) presents convincing
combined sewer overflow data showing that return periods of event mean concen-
trations can be quite different from the rainfall and runoff that caused them.
The choice of rainfall input for models has been widely studied, e.g., by
Adams and Howard (1985), Arnell (1978, 1982), Harreraoes (1983), Huber et al.
(1981), James and Robinson (1982), Marsalek (1978), McPherson (1978), Patry
and McPherson (1979), Wenzel and Voorhees (1981). Adams and Howard (1985)
make a particularly strong argument against the use of design storms, synthe-
296
-------
tic or historic, and recommend continuous simulation or derived frequency
distributions.
This paper will illustrate the use of continuous simulation for selection
of historic design rainfall events for use in detailed design simulations. In
this method, a model is first calibrated and verified for the catchment. A
continuous simulation is then performed using as long a record as possible of
hourly or shorter increment rainfall data (about 25 years of computerized
hourly values are typically available from the U.S. National Weather Service).
A frequency analysis is then performed on the parameters of interest,'such as
peak flow, runoff volume, or quality loads and concentrations, from which
historic rainfall events that produce a runoff event of the desired return
period are identified. Finally, the storms may be input again to the model in
a more detailed design simulation. By this means, questions of antecedent
conditions, storm duration, and storm shape are avoided, and a better idea of
the (unknown) "true" return period for the design is obtained.
Why not use measured flows for this purpose? Although this indeed might
resolve the question of the true design frequency, it is seldom possible in
urban areas because of lack of a gage altogether, a short or incomplete
record, or changing land use during the time of the gaging. Simulation offers
a less exact option to bypass these difficulties.
OBJECTIVES
Ultimately, it is desirable to determine the relative effect of using
various methods for drainage design on the design condition. This paper will
partially address this question for peak flows using one catchment as a case
study. Eventually it may be even more important to determine the economic
impact of alternative design methodologies, both on the cost of the project
and on the cost of the engineering design. In other words, if the use of more
sophisticated and complex methods only results in a marginally different
answer from that obtained by conventional means then it may be constructed for
almost the same cost either way, but be much cheaper to design using the
conventional (simpler) methods. This latter question of costs is currently
under study at the University of Florida and will not be addressed in this
paper.
The immediate objectives of this paper are thus to:
1. Determine design storms from historic rainfall series using continuous
simulation, and
2. Compare peak flow results obtained by simulation with historic design
storras, synthetic design storms and other methods.
METHODS
The options will be compared through a case study using the Megginnis
Arm Catchment in Tallahassee, Florida. The steps are enumerated below.
1. Calibrate and verify the SWMM model on the catchment.
297
-------
2. Perform a continuous simulation using the 22 year historic record of
hourly rainfalls (1958-79).
3. Perform a frequency analysis on predicted peak flows and runoff
volumes.
4. Select historic rainfall events that give peaks or volumes of desired
return periods.
5. Run these historic storms through the SWMM model again using a more
detailed simulation and compare with the results using synthetic design storms
and other procedures.
6. Show the final comparisons on a plot of peak flow versus return
period for the different methods.
SWMM SIMULATIONS
WHY SWMM?
The EPA Storm Water Management Model (Huber et al., 1981; Roesner et al.,
1981) was chosen because of its many applications to urban areas (Huber et
al., 1985) and because it may conveniently be used for both continuous and
single event simulation. Alternative choices include STORM (Roesner et al.,
1974; HEC, 1977) and HSPF (Johanson et al., 1980). However, STORM is not well
suited for single event simulation and generally has only simple hydrologic
and water quality routines (which do not detract from its usefulness as a
planning tool). HSPF might be a viable alternative, but it is less well
suited to urban areas than is SWMM.
THE CATCHMENT
The 2230 acre Megginnis Arm^Catchment in Tallahassee, Florida (Figure 1)
discharges to Lake Jackson north of the city and has experienced both quantity
and quality problems in recent years due to increased urbanization. A runoff
data base exists since 1973, with rainfall-runoff data available since 1979
(both collected by the USGS). It is the site of a multi-million dollar exper-
imental water quality control facility consisting of a detention basin and
artificial marsh, and has been included by the USGS in th&ir urban runoff
studies (Franklin and Losey, 1984). A summary of reports available for the
catchment is given by Esry and Bowman (1984). Preliminary applications of the
Runoff Block of SWMM have been reported by Huber et al. (1986) for purposes
of predicting 5-year peak flows. The more comprehensive results of the con-
tinuous and single event simulations will be illustrated here .and compared
with alternative engineering analyses of the basin for prediction of peak
flows. The various methodologies will be compared over a rango of return
periods, although there will be considerable uncertainty in the peak flow
estimates at the end of the 22 year rainfall record. The 5-yeair results are
more reliable since they fall in the middle of the range of refturn periods and
will be singled out for special consideration of the historic storms involved.
298
-------
Land to be
Purchased
1 mi le
1 mile
SCALE
Figure 1. Megginnis Arm Catchment in Tallahassee, Florida.
Rainfall and flow measurements are made at the
entrance to Pond I.
299
-------
SWMM CALIBRATION
Calibration of SWMM first required estimates for the Runoff Block parame-
ters listed in Table 1. The average slope was found by selecting eight points
at the edge of the catchment, calculating the path length of each point to the
inlet, dividing the path lengths by the change in elevation, and taking a
weighted average of the eight slopes. Percent imperviousness was obtained
from an earlier USGS modeling study by Franklin and Losey (1984). Green-Ampt
infiltration parameters were estimated by identifying the soils in the catch-
ment (primarily sandy) from a county soil survey map, finding the hydraulic
conductivity and capillary suction for each soil from data published by Car-
lisle et al. (1981), and then taking a weighted average over the soil types.
Manning's n values were selected from charts based on average type of ground
cover. Average monthly pan evaporation data were obtained from National
Weather Service (NWS) values published by Farnsworth and Thompson (1982), from
which actual evapotranspiration (ET) estimates were calculated by multiplying
by a pan coefficient of 0.7. Final parameter estimates are shown in Table 1.
The overall catchment was schematized using only one subcatchment and no
channel routing in order to maintain a reasonable computation time for the
continuous simulation. Later, a five subcatchment schematization was used,
but the there was only a minimal improvement in the predictions. Hence, de-
tailed simulation results are reported only for the single subcatchment sche-
matization.
TABLE 1. RUNOFF BLOCK PARAMETERS FOR THE MEGGINNIS ARM CATCHMENT
Parameter Value
Area 2230 ac (903 ha)
Width 6000 ft (1830 m)
Percent Imperviousness 28.3
Slope 0.0216
Manning's Roughness
Impervious 0.015
Pervious 0.35
Depression Storage
Impervious 0.02 in (0.5 mm)
Pervious 0.50 in (13 mm)
Green-Ampt Parameters
Suction 18.13 in (461 mm)
Hydraulic Conductivity 5.76 in/hr (146 mm/hr)
Initial Moisture Deficit 0.15
Rainfall-runoff data for calibration and verification were obtained from
available USGS records for the catchment. Ten of the largest storms were
selected from the period 1979-81 and randomly divided into two sets of five
storms each, one for calibration and the other for verification. The larger
storms were chosen since the ultimate use of the modeling was to be drainage
300
-------
design, for which calibration for large storms is preferable. The model was
calibrated for the five storms simultaneously, that is, while maintaining the
same parameter values for each (Maalel and Huber, 1984). Calibration of
runoff volumes was sufficient using the assumed value of catchment imper-
viousness (Figure 2); thus, calibration for peak flows was achieved only by
varying the subcatchment width (a parameter in SWMM that is equivalent to
making changes in the slope or roughness). The results for peak flows are
shown in Figure 3.
Verification was accomplished by running five different storms using the
same parameters that were used in the final calibration runs. Results of the
verification runs were comparable to the calibration runs and are also shown
in Figures 2 and 3. Individual storms exhibit varying goodness of fit as shown
in Figures 4-7. Figure 4 illustrates the best of the ten fits and Figure 5
the worst, while Figures 6 and 7 illustrate typical intermediate results. It
should be emphasized that antecedent condtions were not altered for individual
storms which would have aided in the fits. (Continuous simulation eliminates
this problem.) However, a robust calibration was desired, which would produce
a good fit, on the average, for many storms (Maalel and Huber, 1984). Hence,
the agreement between measured and predicted volumes and peaks shown in Fi-
gures 2 and 3 was considered adequate for this study.
FREQUENCY ANALYSIS
Upon completion of the calibration and verification exercises, a contin-
uous run of 21.6 years (259 months, June 1958 to December 1979) was made using
hourly rainfall data from the Tallahassee Airport NWS rain gage, located
approximately 7 miles southwest of the study area. Statistical analysis of
the predicted flows was performed using the Statistics Block of SWMM. The
time series of hourly runoff values was separated into 1485 independent storm
events by varying the minimum interevent time, MIT, until the coefficient of
variation of interevent times equalled 1.0, yielding a value of MIT = 19
hours. This method of delineating independent events (Hydroscience, 1979;
Restrepo-Posada and Eagelson, 1980) is based on the fact that the exponential
distribution is often fit to interevent times, and it has a coefficient of
variation (standard deviation divided by the mean) equal to 1.0.
The SWMM Statistics Block performs a frequency analysis on any or all of
the following parameters: runoff volume, average flow, peak flow, event dura-
tion, and interevent duration. (If pollutants were being simulated, frequency
analyses could also be performed on: total load, average load, peak load, flow
weighted average concentration, peak concentration.) Storm events are sorted
and ranked by magnitude for each parameter of interest, and assigned an empir-
ical return period in months according the the Weibull formula (T = (n+l)/m,
where n Is the total number of months and m is the rank of an event). The
largest magnitude event for this 259-month simulation was thus assigned a
return period of 260 months. Hence, for this simulation, a five year event is
bracketed by return periods of 52 and 65 months (fifth and fourth largest
events, respectively), both of which were selected as "design events."
On this basis, the historic storms producing the nine highest peak flows
301
-------
1 5
O
O
ui
o
Ui
ct
a
0.2
1.2
0.6 0.8 1
MEASURED VOL.. IN
OCALIBRAT1ON VERIFICATION
Figure 2. Goodness of fit of runoff volumes.
1.4
0.2
-i 1 r
0.4 O.6 0.8
(Thousands)
MEASURED PEAK FLOW, CFS
I VERIFICATION
OC A LI BRAT I ON
Figure 3. Goodness of fit of runoff peaks.
302
-------
SVU. UUUUU
4OO. OOOOO
-
3OO. OOOOO
FLOW
IN
CF8
200. OOOOO
100. OOOOO
O. 0
8.
*
»*
#»
**
**
**+
«» **+
f\ 3 *»»+
*+ *
*+ *
*+ *
*+ »
»-»
t *£
1 i 1 i 1 1 1
I
I
+
+
+
+
+
j.
*+ * +
+
»* »+
»» *» +
» «* +
-------
430. OOOOO
360. OOOOO
_
270. OOOOO
FLOW
IN
CFS
1BO. OOOOO
OJ
O 90. OOOOO
-^
Of*
. w
6
1 1 1 _- i 1 1_ _ 1 1
4> 4.
* 4- +-f 4-4.
+ 4- 4-4- 4- + 4-
4- 4- + +
4- +44.
~10m3/s / I I \
' * % + t
4« 4-4- 4-
4-« > + 4-
+ + + 4-
4- 4-4- 4>
4- 4- 4> '+
4- * +4- 4-
+ ++ +
4- * * 4-
4- ** « +
+ « * 4. +4-f
*+ 4- »* » » 4>
4-4- 4- » » » * +
* * 4- » » *
+ + *» » * 4-
* 4- 4- * »* 4-
4>4-4- » « 4-
4- + * * ** +
+ 4- * 4- » «* 4-
+ + « + » * » 4-4
4- 4> *+ * * » 4-4-
4- 4>4> « « 4-M-
4-** * 4-4-f
4- » * * «* 4-M-
4- «»** «» !>>» +
4- « * 1 > 1 1 14-
*4- «»« 4-4-4-t-4-4-+4-*-»+
*+ »***
4-4- «»**»«-»*»*«»«-»** I
O 7.*O 8."o 9.*O 10*0 ll.*0 12*0 13*0 14*0 19*0 16.
TIME OF DAY. IN HOURS
STORM 4 MARCH 1O 198O
PREDICTED-*. MEASURED-*
LOCATION 98
HYDROS*APH STATISTICS FOR LOCATION «78
VOLUME
CUBIC FEET
PREDICTED.
TOTAL TIME
MEASURED,
TOTAL TIME
PREDICTED,
OVERLAPPING
TIME
MEASURED.
OVERLAPPING
TIME
DIFFERENCES,
ABSOLUTE
X OF MEAS
0.
0.
0.
0.
O.
2B463E+07
3B339E+07
2B430E+07
9B339E4-O7
29889E+07
INCHES
0.
o.
o.
o.
0.
51.
392
721
391
721
369
233
PEAK
TIME. HR
8. 983
8. 667
8. 9S3
8. 667
0. OB3
FLOW
DURATION
FLOW. CFS
321.
414.
321.
414.
92.
22.
699
OOO
699
000
341
309
START, HR
6. 667
6. 667
6. 667
6. 667
Figure
END, HR
19. 833
19.790
1 9. 73O
19. 73O
LENOTH, HR
9. 167
9.083
9. OB3
9.083
5. Predicted
March 10,
NO.
POINTS
111
110
110
110
and meas
1980 (ve
-------
^9U. WUUU
2OO OOOOO
150. OOOOO
-
FLOW
IN
CFS
100. OOOOO
- -.
10
O 30. OOOOO
(Jl
Oft
, Q
2.
1 1 i 1 ;l 1 1 1 1 1
4- 4-
4- 4-
4- »
4 4-
4 4-
4- 4-
+ *
4- 4
4- 4-
+ 4-
4- 4-
4- +
4 4.
»4 +
»4> 4-
. *4-« «
- 4 «4-. 4-
3 / *+* +
«n /s ::% :
4 * 4>
4- * 4- 4-
4- » 4- 4-4-
4- » 4- 4- + 4-
4- #«4- +-* 4-4-
+ * 4- 4- + + 4-
4- * 4- » 4- »4 4
4- »4- 4- 4- 4-« 4
4- * 4- +« + *4-«* +
+ * 4- H-» 4- »4- » 4-
4. ^4. 4.4^ ^ 4.4-94. 4.
-» » + «**++» +
»+ «4 4- 4- » 4-4- * 4>
4- »4- » 4- 4-4- 4-4. » * * 4-
4.4. «+ **4-44- 4-4-4-4- «« » *+
4-4- *+ »+«» 4- » 4-4-
4-4-4- » » » 4-
4- 4- 4- » »* 4-4-
4- 44- »» 4-4.4-
4-*»»»4-4- »»
*+ * **
*4- »»*
4-4-
T T T T . _^__ t _ -i«T « _^«_ f -i w i-. T ^ T ^_^
0 40 6. 0 80 10.0 UO 140 Ife. O 18. O 20. O 22.
i
TIME OF DAY, IN HOURS PREDICTED**. MEASURED-*
STORM2 MARCH 9 198O LOCATION 98
HYDROCftAPH STATISTICS FOR LOCATION 98
VOLUME PEAK FLOW DURATION NO.
CUBIC FEET INCHES TIME. HR FLOW, CFS START. HR END. HR LENGTH, HR POINTS
PREDICTED. 0 18009E+O7 0 222
TOTAL TIME
MEASURED. O. 23986E+07 0. 296
TOTAL TIME
PREDICTED. 0. 17979E+07 0. 222
OVERLAPPING
TIME
MEASURED. 0 239B6E+O7 0 296
OVERLAP?INQ
TIME
DIFFERENCES.
ABSOLUTE O 6OO63E+O6 O 074
V. OF MFAS 25 041
S. 500 158.767
5. 730 234 000
3 500 138. 767
3 750 234 000
0 230 75 233
32 151
3. 500 13. 300 10. OOO
3. 300 13. 417 9. 917
3.300 13.417 9.917
3 300 13 417
9 917
121
120
120
120
Figure 6. Predicted and measured hydrograph,
March 9, 1980 (calibration run).
-------
190. OOOOO
120. OOOOO
9O. OOOOO
FLOW
IN
CFS
60. OOOOO
CO
g 30. OOOOO
Of\
-4
11.2
STORM 1 SEPT 27
1 ! i _ !_ ! j
1.1 /S
+ ***
+ * * *
+ +» *
+ *
»+ » + *
* + + *
*+ + »
+ *
+ + * »
+ ++ » «»
*+ + * *
+ + »
* + + » »
«*+ + « *
+ 4- *** *
* + + «»
» » + ++*++» »
» * »«*+*
» + ++ + * + » *
+ «+* * «
+ + «»»«
* + +++++* »
* + +++ » * * *
+ +4- * »
+ »+«+'»»«.»+ «»*»
+ * -M-t-l-
* * »+
+ *
++ #*#
* »*»
*+
11.9 12.6 13.3 14.0 14.7 19.4 16.1 16.8 17.9 IB.
TIME OF DAY. IN HOURS PREDICTED-*, MEASURED-*
1979 LOCATION 9S
HYDROORAPH STATISTICS FOR LOCATION 98
VOLUME PEAK FLOW DURATION NO.
CUBIC FEET INCHES TIME.HR FLOW, CFS START, HR END. HR LENGTH, HR POINTS
PREDICTED. 0. 136S2E+07
TOTAL TIME
MEASURED. 0. 1197&E+07
TOTAL TIME
PREDICTED. O. 13636E+O7
OVERLAPPING
TIME
MEASURED. O. 11976E+07
OVERLAPPING
TIME
DIFFERENCES.
ABSOLUTE -0. 166O7E+06
X OF MEAS
0.169 13.000 122.836 11.730 17.730 6. OOO 73
0.148 12.667 119.000 11.730 17.667 3.917 72
0.168 13. OOO 122. B3B 11.730 17.667 3.917 72
0.148 12.667 119. OOO 11.730 17.667 3.917 72
-13 868 -0-333 -g- |3B Figure 7. Predicted and measured hydrograp
September 27, 1979 (Verification run)
-------
and.the nine highest runoff volumes (total flows) are indicated in Tables 2
and 3. The hyetographs of 52 and 65 month storms for each case are shown in
Figures 8 and 9 and their characteristics are listed in Table 4.
TABLE 2. STORMS RANKED BY PEAK FLOW
Rank
1
2
3
4
5
6
7
8
9
Return Period
months
260
130
86.7
65
52
43.3
37.1
26
23.6
Date
9/ 8/68
7/21/70
7/16/64
7/21/69
9/ 3/65
9/20/69
12/ 3/64
11 9/65
6/30/64
Duration
hrs
6
34
73
58
11
88
61
81
48
Depth
in
6.52
8.21
10.16
6.11
4.35
13.79
9.92
6.01
4.67
Peak Flow
cfs
1907
1605
1505
1131
1104
1070
1028
807
805
TABLE 3. STORMS RANKED BY RUNOFF VOLUME (TOTAL FLOW)
Rank Return Period
months
1
2
3
4
5
6
7
8
9
260
130
86.7
65
52
43.3
37.1
26
23.6
Date
9/20/69
3/28/73
12/ 3/64
7/16/64
7/26/75
7/21/70
8/ 2/66
3/31/62
10/ 6/59
Duration
hrs
84
108
61
73
97
34
171
24
87
Depth Runoff Volume
in in
13.79
10.91
9.92
10.16
9.30
8.21
7.72
7.78
7.35
3.69
2.86
2.74
2.69
2.39
2.24
2.09
1.96
1.95
Tables 2 and 3 show that the return periods of the individual storms are
quite different when ranked by another parameter. This is illustrated further
for the 5-year storms in Table 5. Storm V-65 is rare by both flow measures;
in fact, it is the third largest storm of record on the basis of peak flow
(and the second largest rainfall volume event). However, when storm V-52 is
ranked by peak flow, and when storm P-52 is ranked by volume, they are seen to
be not especially rare, with return periods of 7.4 and 6.8 months, respec-
tively. Return periods for rainfall volumes shown in Table 5 were obtained
from the SYNOP program for rainfall frequency analysis (EPA, 1976; Hydrosci-
307
-------
7/21/69
OE
I
I/)
a
i
t
to
UJ
Z
-^ -r
32 -
3 -
28-
2 6 -
24 -
22-
2 -
1 8 -
1 6 -
1 4- -
1 2 -
1 -
08 -
O 6 -
O 4 -
02 -
o
-50
mm/hr
F'
r '
1 1 1
EP
rk.n
21 31 41 51 61
TIME. HOURS
Figure 8a. Hyetograph for storm P-65, July 21, 1969.
65-month return period based on peak flow.
9/3/65
.> 1 -
3 2 -
3 -
2 6 -
2 6 -
2 4 -
2 2 -
2 -
1 8 -
1 6 -
1 4 -
1 2 -
1 -
O 8 -
06-
04 -
02-
50 mm/hr
// _
//
-------
7/16/64
cc
x
in
r
3.4 -i
3.2 -
3 -
2.8 -
2.6 -
2.4 -
2.2 -
2 -
1 .8 -
1 .6 -
1 .4 -
1 .2 -
1 -
O.8 -
0,6 -
0,4 -
0.2 -
'
-50
mm/hr
1 ru r
i i
1
'
t
'
Ua
U n Lfi
M rlLJIrt! n ₯_
1 21 31 41 51 61 71
1 11 21 31 41 51 61 71
TIME. HOURS
Figure 9a. Hyetograph for storm V-65, July 16, 1964.
65-month return period based on runoff volume
(t
r
\
z
£
t/>
2
5
3 4 -,
3.2 -
3 -
2 8 -
26-
2.4 -
2 2 -
2 -
1 6 -
1 6 -
1 4 -
1 2 -
1 -
08 -
06 -
04 -
02 -
0 -
7/26/75
-Ml in 111 /lu
p
I
f n ?H(fl (l^Ui|imjr, Ik jWlH
L
1 11 21 31 41 51 61
1 « |.
I-,'.?M JS^.TI 1.1 1,,. i M'Hll llfl, ,WH
71 61 91
TIME, HOURS
Figure 9b. Hyetograph for storm V-52, July 26, 1975.
52-month return period based on runoff volume,
309
-------
ence, 1979). Thus, although the return periods were computed slightly differ-
ently, Table 5 further illustrates that return periods of the same event
ranked by different parameters are rarely the same.
TABLE 4. CHARACTERISTCS OF 5-YEAR HISTORIC DESIGN STORMS
No.* Date
Runoff Runoff Rainfall Peak Time Since Rainfall
Duration Volume Volume Flow Last Event Last Event
hr in in cfs hr in
V-65
V-52
P-65
P-52
7/16/64
7/26/75
7/21/69
9/ 3/65
73
97
58
11
2.69
2.39
1.61
1.19
10.16
9.30
6.11
4.35
1670
685
1131
1104
74
31
32
67
0.27
0.07
1.35
0.11
* V means ranked by volume, P means ranked by peak, and numbers are return
periods in months.
TABLE 5. RETURN PERIODS (MONTHS) OF "5-YEAR" STORMS
BY VOLUME, PEAK FLOW ANT) RAINFALL
No. Date Return Period
by Volume
Return Period
by Peak Flow
Return Period
by Rainfall Vol.
V-65
V-52
P-65
P-52
7/16/64
7/26/75
7/21/69
9/ 3/65
65
52
14
6.8
87
7.4
65
52
156
78
12
7.8
Seven of the historic storms defined above were then run again through
the model (storms for return periods of 37.1 and 43.3 months were omitted)
using hourly rainfall inputs but a 5-min time step, instead of the hourly time
step used in the continuous simulation. In lieu of continuous simulation for
these detailed analyses, 4 to 6 days of prior rainfall were run through the
model prior to the beginning of each design event. This "pseudo-continuous
simulation" thus accounts for antecedent conditions without the need for
running the entire rainfall time series at a short time step. Results will be
described following a discussion of the generation of synthetic design storms.
SYNTHETIC DESIGN STORMS
Following the techniques of Arnell (1982), five synthetic design storms
310
-------
were constructed for comparision with the historic storms in Tallahassee: Soil
Conservation Service (SCS, 1964), Chicago (Keifer and Chu, 1957), Illinois
State Water Survey (Huff, 1967), Sifalda (1973), developed in Czechoslovakia,
and Flood Studies Report, FSR, (Natural Environment Research Council, 1975),
developed in Great Britain. The choice of a 24-hour duration for the storms
was made for two reasons: 1) it is commonly used in engineering practice in
the'Tallahassee area, and 2) approximate calculations of the time of concen-
tration of the basin using the kinematic wave equation (Eagleson, 1970)
yielded estimates ranging from 13 to 60 hours, depending on the choice of
rainfall excess. Thus, 24 hours is at least in the range of possible times of
concentration. But the very idea of having to arbitrarily select a storm
duration in the first place illustrates one of the major difficulties in using
synthetic design storms.
Five-year storms are considered as an example. From the Tallahassee
region IDF curves (Weldon, 1985), the 5-year, 24-hour average intensity is
0.305 in/hr, giving a 5-year, 24-hour depth of 7.32 in. The five synthetic
design storms were then scaled to produce this total storm depth. Various 24-
hour depths frora the IDF curves are compared with historic storms in Table 6.
These were compiled from the SYNOP program (EPA, 1976) with a minimum intere-
vent time of 5 hours and thus differ In magnitude slightly from the rainfall
volumes identified from the frequency analysis of runoff conducted using SWMM.
It may be seen that although the IDF and historic depths are comparable, the
actual durations of the historic storms are certainly not 24 hours.
TABLE 6. COMPARISON OF 24-HOUR IDF DEPTHS WITH SYNOP
FREQUENCY ANALYSIS OF HISTORICAL RAINFALL EVENTS
Storm
Date
9/20/69
IDF
7/17/64
IDF
12/13/64
7/28/75
7/21/70
IDF
8/30/50
6/18/72
Return
Period
yr
26.0
25
13.0
10
8.7
6.5
5.2
5
4.3
3.7
Duration
hr
54
24
33
24
36
53
20
24
36
46
Depth
in
13.41
10.08
9.76
8.64
9.73
8.84
8.18
7.44
7.34
7.17
The final group of synthetic hyetographs is shown in Figures 10 and 11.
It is clear that they bear no resemblance to the actual historic storms shown
in Figures 8 and 9.
311
-------
E
M
Z
DESIGN STORMS FOR A 24 HR. DURATION
AND A 5 YEAR RETURN PERIOD
3.2
1 -
O.B
0.6 -
0.4 -
0.2 -
3 -
2.8 -
2.6 -
2.4 -
2.2 -
2 --50mm/hr
1.8 -
1.4 -
ILLINOIS
CHICAGO
JTSR
16
2O
O 4 6 12
" TIME (hr.)
Figure 10. Chicago, Flood Studies Report, and
Illinois design storms.
24
24
01
z
DESIGN STORMS FOR A 24 HR. DURATION
AND A 5 YEAR KLTUKN PERIOD
3 -
2.B -
2.6 -
2.4 -
2-
l.B -
1.6 -
1.4 -
1.2-
1 -
O.B -
0.6 -
O.4 -
0.2 -
-5O mm/ hr
r ^
0 4
0
i
SCS
^A
' i
^-^ -^=^=^= _j
5 12 16 2O 2.
2'
TIME (hr.)
Figure 11. Soil Conservation Service and
Sifalda design storms.
312
-------
OTHER DESIGN TECHNIQUES
The USGS has performed a flood frequency analysis on 15 basins In the
Tallahassee area using the combination of continuous simulation and regression
discussed earlier (Franklin and Losey, 1984). Their predicted 5-yr peak for
Megginnis Arm is 1570 cfs.
Finally, the venerable Rational Method could be applied. The results for
this method are also shown below, using a runoff coefficient of 0.27 (from the
SWMM modeling) and Weldon's (1985) IDF data.
RESULTS
The main objective is a comparison of peak flows developed using alterna-
tive techniques. For this purpose, seven of the historic storms listed in
Table 2 (with antecedent conditions included) and five synthetic storms were
run on a single event basis using the calibrated version of SWMM described
earlier. Seven historic storms from Table 3 (for total flow) were also simu-
lated. (Storms for return periods of 43.3 and 37.1 months were not run.) For
all runs, a time step of 5 minutes was used, but with hourly hyetograph
values, since historic rainfall data at finer time increments were not avail-
able in time for this analysis. (Copies of the microfilmed weighing-bucket
charts must be ordered from the NWS National Climatic Data Center in Ashe-
ville, NC in order to obtain, say, 15 minute hyetograph increments.) Peak
flows calculated by the model in this way are somewhat higher than for the
identical storms during the continuous simulation because SWMM calculates
average flows over the hourly time step during continuous simulation (in order
to avoid large continuity errors), whereas instantaneous flows at the end of a
time step are calculated during single event simulation. Results for the 5-
year storms are given in Table 7 along with results for the other methods that
have been applied to this basin. Results for all storms are shown in Figure
12. (The peculiar shape of the curve for runoff volumes is because the
volumes are plotted at return periods corresponding to a volume ranking, not a
peak flow ranking.)
Before discussing the results, it is interesting to note that when run on
a single event basis, the P-65 storm produces a lower peak than does the P-52
storm, in contrast to the results for the same storms during the continuous
simulation (Tables 2 and 4). This is an artifact of the numerical methods
used in the SWMM Runoff Block. For continuous simulation, average flows over
the time step (usually one hour) are computed in order to avoid continuity
errors when using long time steps. For single event simulation, instantaneous
flows at the end of the time step are computed. The peak instantaneous values
are higher than the averages and respond more directly to peak rainfall inten-
sities. This factor, coupled with the shorter time step used for the detailed
simulations were enough to alter the rankings of these two storms. In gen-
eral, all the peak flows predicted during the single event simulations of the
historic storms were higher by as much as a few hundred cfs than those pre-
dicted during the continuous simulation, but the only change in rankings is
the one discussed above.
313
-------
U)
v>
-P.
(ft
LL
3.2
3 -
2.8 -
2.6 -
2.4 -
2.2 -
2 -
1.8 -
1.6 -
1 .4 -
1.2 -
1 -
0.8 -
0.6 -
O.4 -
0.2
0
-60m3/.s
HIST. PEAK
OCHICAGO
HlST.jrO'ML FLOW
1 1
0 4
+ SCS
OUSGS
T
I
12
T
8 12 16
RETURN PERIOD, YEARS
20
24
o ILLINIOS
OPEAK PLOW
A SIFALDA
OTOTAL FLOW
F.S.R
Figure 12. Peak flows versus return period predicted using historical storms,
synthetic design storms and the USGS method. Historical storms
based on total flow are plotted at return periods corresponding
to runoff volume, not peak flow.
-------
Considering a 5-year return period, for the Rational Method to predict a
peak flow of 1748 cfs would require a duration on the IDF curve of approxi-
mately 50 min, much less than the actual (but unknown!) time of concentration.
Hence, it does not appear to be applicable. Furthermore, the return period
for the method is assigned on the basis of conditional frequency analysis of
rainfall, as oppposed to the storm event analysis of runoff peaks or volumes
of the other methods.
TABLE 7. COMPARISON OF 5-YEAR PEAK FLOWS USING SEVERAL METHODS
Storm
V-65
V-52
P-65
P-52
SCS
Chicago
Illinois
Sifalda
F.S.R.
Peak Flow
cfs
1996
983
1384
1748
1926
1677
511
441
1235
Method Peak Flow
cfs
uses
Rational Method
tc = 24 hours
t_ = 13 hours
tc = 50 min
1570
187
289
1764
Somewhat unexpectedly, the results for the synthetic design storms fall
on the low end of the estimates; only the SCS storm is consistently high. For
example, a similar study by Marsalek (1978) showed higher peaks using the
synthetic storms, but his basins were much smaller (64 acre and 321 acres).
The difference may be partly due to the choice of a hyetograph interval of 1
hour; synthetic hyetographs are sensitive to this choice since as the interval
becomes smaller, the peaks for shapes such as the Chicago and SCS storms tend
to become very high (and thus produce higher runoff peaks). This is another
complication in the use of such methods. On the other hand, the historic
rainfall was also discretized at 1-hour intervals and might also be expected
to have higher peaks at shorter intervals (but the shorter interval data were
not available for this study). Furthermore, the Megginnis Arm basin is large
with a long time of concentration; in principle, short time increment fluctua-
tions should be well damped by the basin response.
Another reason why the peaks are higher with the historic storms lies in
the differences in storm volumes and durations. For instance, the two "5-
year" historic storms chosen on the basis of volume are both larger than the
24-hr IDF value of 7.32 inches. On the other hand, historic storm P-52 has a
duration of only 11 hours in which to concentrate the rainfall of 3.93 inches.
The anomaly is historic storm P-65, which has a lower rainfall volume and a
longer duration than do the synthetic storms but it has a pocket of high
intensity rainfall (Figure 8a) coupled with wet antecedent conditions.
315
-------
The main difference between the results of this study and those of Marsa-
lek (1978) is in the manner of choosing the historic storms. Marsalek iso-
lated storms on the basis of rainfall criteria only, including a requirement
that storms have a total depth greater than 4.9 in/hr and/or contain a 10-
minute intensity greater than 5.9 In/hr. His historic storms were further
screened by seeking maximum depths for durations of 10, 15, 30 and 60 minutes.
In this study, the storms were screened solely on the basis of the- continuous
simulation, allowing the catchment to filter out the important elements of the
rainfall-runoff response (to the extent that it is properly modeled in SWMM).
Arnell (1982) found that the Sifalda storm gave the best agreement with
historic storms, in contrast to this example in which it gives a low estimate
for peaks. In general, this catchment produces higher peaks for storms with
late and high peak rainfall intensities, appearing after the infiltration
capacity of the basin is satisfied.
The USGS estimates (based on a log-Pearson Type 3 frequency analysis of
simulated flows) do not differ widely from those of the historic SWMM runs at
low return periods, but all methods give lower peak flows at high return
periods than that predicted by SWMM for the 260-month peak-flow storm. As can
be seen in Figure 12, the SWMM simulations of the historic storms produce
lower peaks than those of the USGS and the Chicago and SCS design storms at
low return periods. Hence, it cannot be said unequivocally that synthetic
design storms produce conservative results. On the basis of this single study
on a single catchment, the SCS and Chicago storms are conservative at low
return periods but not at high ones.
However, there is considerable uncertainty as to the true return period
of the largest storm runoff peak in the 22 year period. Assuming on the basis
of order statistics that the empirical frequencies assigned to the historic
storms have a beta distribution (Gumbel, 1958, Chapter 2), then +/- 25% confi-
dence intervals about the expected value of the return period of the largest
peak (23 years) are 11.1 and 47.5 years. That is, the probability is 50% that
the true return period of the largest simulated peak in the 22 year record
lies between 11.1 and 47.5 years. These confidence intervals narrow consider-
ably at lower return periods. Hence, it can be said with more confidence that
the SCS and Chicago storms are conservative at low return periods than that
they are not conservative at high return periods. A much longer rainfall
record would resolve the question at the 22 year return period, but there
would always be doubt at the end point of the record as to the proper fre-
quency (even if a frequency distribution, such as the log-Pearson Type III,
were fit to the data). A main result of this study is simply that there is no
guarantee that synthetic storms are conservative over the whole range of
return periods.
Ideally, the measured runoff records should be analyzed to isolate storm
events and resolve the question of which of the several curves shown on Figure
12 is correct. But the gage for the Megginnis Arm Catchment has only been in
operation since 1973, making it difficult to identify storms for return per-
iods greater than about 10 years. Another difficulty is the changing land use
within the catchment, with rapid urbanization altering the nature of the
rainfall-runoff response. This factor tends to direct the analysis back to a
316
-------
properly calibrated model. However, for the record, the largest peak observed
during the intermittent sampling since 1973 is 724 cfs on May 23, 1980 from a
storm of 1.43 inches. It is not suprising that the much higher rainfalls used
in the simulations would produce peaks on the order of 2000 cfs (Figure 12).
In general, maximum rainfalls after 1973 have been much less than before (see
Table 6). Thus, peak flows simulated using 1958-79 rainfall data are quite
likely to be higher than those measured since 1973.
HISTORIC VERSUS SYNTHETIC STORMS
The reader will probably not be surprised to learn that the authors favor
the historic storm values for the following reasons:
1. They are based on a frequency analysis of peak flows (or runoff volumes),
not on a conditional frequency analysis of rainfall depths- That is, the
frequency analysis is on the parameter of interest.
2. The frequency analysis of the continuous time series of flows includes all
effects of antecedent conditions whereas the synthetic design storms do not.
Moreover, the detailed single-event simulations of historic storms may be
conducted in a "pseudo-continuous manner," that is, by simulation of several
days of antecedent weather prior to the beginnning of the storm of interest.
3. The historic storms avoid the vexing questions of storm duration, shape and
hyetograph discretization. The duration is especially critical, since peak
flows may arise out of a storm that lasts several days (e.g., storms V-65, V-
52, P-65) or out of a short, intense one (e.g., storm P-52). There is simply
no basis for establishing a standard duration, such as 24 hours, for all
design work, as seems to be the unfortunate tendency in Florida.
4. Historic storms can also be used for analysis of volumes for design of
basins for detention or retention. The volume of synthetic storms is arbi-
trarily linked to the assumed rainfall duration. A given volume can result
from an infinite number of combinations of Intensity and duration, with a
corresponding range of return periods.
5. If historic storms are used for design, the local citizenry can be confi-
dent that the design will withstand a real storm that may be remembered for
its flooding by many, as opposed to an "unreal" synthetic storm. Thus, .the
engineers or agency may make a statement such as, "our design will avoid the
flooding that resulted from the storm of September 3, 1965."
PROBLEMS WITH USE OF HISTORIC STORMS FOR DESIGN
The analysis outlined in this paper is neither brief nor simple. A
continuous simulation model must be calibrated and used, first to identify
storm events and second to perform the more detailed design simulations. This
means more work both for the design engineers and for the reviewing agencies
(leading to the common arguments in favor of synthetic design storms they
are quick, easy, and consistent from application to application). A possible
317
-------
compromise would be to compile an atlas of historic storms and their antece-
dent conditions for use in design, based on generic runs of a continuous model
for representative catchment sizes, land uses, soil types, climatic regions
etc. Research on this possibility is underway at the University of Florida.
Another consideration is the high spatial variability of real storms and
the sensitivity of runoff to storm direction (James and Shtifter, 1981;
Surkan, 1974). Historic storms must still be derived from point rainfall
data. The catchment may be more sensitive to the spatial dynamics of storms
than to the relative rankings within a time series. Of course, the same
remarks apply to synthetic storms. However, the analysis of historic storms
may be carried further and dynamic storms constructed from historical data
(James and Scheckenberger, 1984). Spatial dynamics are more important as the
size of the catchment increases.
Finally, the question of economics versus safety factors raised early in
this paper is unanswered. Is there a savings in construction costs by using
the type of analysis outlined in this paper (with the implication that the
synthetic design storms commonly used in design are too conservative)? Or are
the synthetic design storms not conservative and therefore providing less than
the level of protection assumed by design engineers? What are the implica-
tions on flood stage as opposed to peak flows? Stage may only be affected in
a minor way by differences of several hundred cfs in flow for a wide flood
plain. In such cases the design may be insensitive to the design methodology
used, and the quickest and fastest method may suffice. These questions are
also under investigation at UF.
CONCLUSIONS
Given a calibrated and verified model, selection of historic design storm
events from a continuous simulation offers the advantages of a meaningful
frequency analysis on the runoff parameter of interest, such as peak flow or
runoff volume; the frequency analysis need not be tied to rainfall frequencies
as with the use of synthetic desfgn storms. For this study, historic design
storms were selected from a frequency analysis of peak flows and runoff vol-
umes on the basis of a continuous SWMM simulation using the 22 year period of
available rainfall. The model had previously been calibrated and verified
using local rainfall-runoff data for the Megginnis Arm Catchment in Tallahas-
see. The historic storms were then simulated and the results compared to
those obtained using five different synthetic design storms, including the
widely used SCS Type II storm. In general, the SCS and Chicago storms pro-
duced higher peaks at low return periods than produced by the historic storms,
while the historic storms produced a higher peak at the highest return period.
Thus, it cannot be assumed that the synthetic storms are conservative over the
entire range of return periods.
ACKNOWLEDGEMENTS
This research is supported by EPA Cooperative Agreement CR-811607 and by
USGS project USDI-14-08-0001-G1010.
318
-------
REFERENCES
1. Adams, B.J. and C.D.D. Howard. Pathology of Design Storms. Publica-
tion 85-03, University of Toronto, Dept. of Civil Engineering, Toronto,
Ontario, 1985.
2. Arnell, V. Analysis of rainfall data for use in design of storm sewer
systems. Proc. First Int. Conf. on Urban Storm Drainage, University of
Southampton, Pentech Press, London, 71-86, April 1978.
3. Arnell, V. Rainfall Data for the Design of Sewer Pipe Systems. Report
Series A:8, Chalmers University, Dept. of Hydraulics, Goteborg, Sweden,
1982.
4. Carlisle, V.W., Hallmark, C.T., Sodek, F., Ill, Caldwell, R.E., Hammond,
L.C and V.E. Berkheiser. Characterization Data for Selected Florida
joils. Soil Science Research Report No. 81-1, Soil Science Department,
University of Florida, Gainesville, Florida, June 1981.
5 Driscoll, E.D. in Benefit analysis for Combined Sewer Overflow Control.
Seminar Publication, EPA-625/4-79-013, Environmental Protection Agency,
Cincinnati, Ohio, April 1979.
6. Eagleson, P.S. Dynamic Hydrology. McGraw-Hill, New York, 1970.
7. Environmental Protection Agency. Areawide Assessment Procedures Manual.
EPA-600/9-76-014, Cincinnati, Ohio. Il:E-l - E-115, 1976.
8. Esry, D.H. and J.E. Bowman. Final Construction Report, Lake Jackson
Clean Lakes Restoration Project. Northwest Florida Water Management
District, Havana, Florida, 1984.
9. Farnsworth, R.K. and E.S. Thompson. Mean Monthly, Seasonal, and Annual
Pan Evaporation for the United States. NOAA Technical Report NWS 34,
Office of Hydrology, National Weather Service, Washington, DC, December
1982.
10. Franklin, M.A. and G.T. Losey. Magnitude and Frequency of Floods from
Urban Streams in Leon County^ Florida. USGS Water Resources Investi-
gations Report~54-4004, Tallahassee, Florida, 1984.
11. Geiger, W.F. Characteristics of combined sewer runoff. Proc. Third Int.
Conf. on Urban Storm Drainage, Chalmers University, Goteborg, Sweden!
2:851-860, June 1984.
12. Gumbel, E.J. Statistics of Extremes. Columbia University Press, New York
1958.
13. Harremoes, P. (Ed.). Rainfall as the Basis for Urban Runoff Design and
Analysis. Pergamon Press, New York, 1983.
319
-------
14. Huber, W.C., Cunningham, B,A. and K.A. Cavender. Continuous SWMM modeling
for selection of design events. Urban Drainage Modeling, Proc. Int. Symp.
on Comparison of Urban Prainge Models with Real Catchment Data, Dubrov-
nik, Yugoslavia, Pergamon Press, New York, 379-390, April 1986.
15. Huber, W.C., Heaney, J.P. and B.A. Cunningham. Storm Water Management
Model (SWMM) Bibliography. EPA/600/3-85/077 (NTIS PB86-136041/AS), Envi-
ronmental Protection Agency, Athens, Georgia, September 1985.
16. Huber, W.C., Heaney, J.P., Nix, S.J., Dickinson, R.E. and D.J. Polmann.
Storm Water Management Model User's Manual, Version III. EPA-600/2-84-
109a (NTIS PB84-198423), Environmental Protection Agency, Cincinnati,
Ohio, November 1981.
17. Huff, F.A. Time distribution of rainfall in heavy storms. Water Resour-
ces Research. 3(4); 1007-1019, 1967.
18. Hydrologic Engineering Center. Storage, Treatment, Overflow, Runoff Mo-
del, STORM, User's Manual. Generalized Computer Program 723-S8-L7520,
U.S. Army Corps of Engineers, Davis, California, 1977.
19. Hydroscience, Inc. A Statistical Method for Assessment of Urban Stormwa-
ter Loads - Impacts - Controls. EPA-440/3-79-023, Environmental Protec-
tion Agency, Washington, DC, 1979.
20. James, W. and M.A. Robinson. Continuous models essential for detention
design. Proc., Engineering Foundation Conference on Stormwater Detention
Facilities Planning Design Operation and Maintenance, Henniker, New Hamp-
shire, American Society of Civil Engineers, New York, 163-175, August
1982.
21. James, W. and R. Scheckenberger. RAINPAC - a program for analysis of
rainfall inputs in computing storm dynamics. Proc. Stormwater and Water
Quality Meeting, Detroit, Michigan, EPA-600/9-85-003 (NTIS PB85-
168003/AS), Environmental Protection Agency, Athens, Georgia, 81-100,
April 1984.
22. James, W. and Z. Shtlfter. Implications of storm dynamics on design storm
inputs. Proc. Stormwater and Water Quality Managment Modeling and SWMM
Users^ Group Meeting, Niagara Falls, Ontario, Report CHI-81, Dept. of
Civil Engineering, McMaster University, Hamilton, Ontario, 55-78, Sep-
tember 1981.
23. Johanson, R.C., Imhoff, J. C. and H.H. Davis. User's .Manual for Hydrolo-
gical Simulation Program ^ Fortran (HSPF). EPA-600/9-80-015, Environmen-
tal Protection Agency, Athens, Georgia, 1980.
24. Keifer, C.J. and H.H. Chu. Synthetic storm pattern for drainage design.
J. Hyd. Div., Proc. ASCE, jtt:(HY4), July 1957.
25. Maalel, K. and W.C. Huber. SWMM calibration using continuous and multiple
event simulation. Proc. Third Int. Conf. on Urban Storm Drainage, Chal-
320
-------
mers University, Goteborg, Sweden. 2i595-604, June 1984.
26. Marsalek, J. Research on the Design Storm Concept. ASCE Urban Water
Resources Research Program Tech. Memorandum No. 33, (NTIS PB-291936),
American Society of Civil Engineers, New York, 1978.
27. McPherson, M.B. Urban Runoff Control Planning. EPA-600/9-78-035, Environ-
mental Protection Agency, Washington, DC, October 1978.
28. Natural Environment Research Council. Flood Studies Report. Five Vol-
umes. Institute of Hydrology, Wallingford, UK, 1975.
29. Patry, G. and M.B. McPherson (Eds.). The Design Storm Concept. EP80-R-8,
GREMU-79/02, Civil Engineering Dept., Ecole Polytechnique de Montreal,
Montreal, Quebec, December 1979.
30. Restrepo-Posada, P.J. and P.S. Eagleson. Identification of independent
rainstorms. J^ Hydrology. ^5j309-3l9, 1982.
31. Roesner, L.A., Nichandros, H.M., Shubinski, R.P., Feldman, A.D., Abbott,
J.W. and A.O. Friedland. A_ Model for Evaluating Runoff-Quality J£ Metro-
politan Master Planning. ASCE Urban Water Resources Research Program
Technical Memorandum No. 23 (NTIS PB-234312), American Society of Civil
Engineers, New York, April 1974.
32. Roesner, L.A-, Shubinski, R.P. and J.A. Aldrich. Storm Water Management
Model User's Manual, Version III: Addendum I± EXTRAN. EPA-600/2-84-109b
(NTIS PB84-198431), Environmental Protection Agency, Cincinnati, Ohio,
November 1981.
33. Sifalda, V. Entwicklung eines berechunungsregens fur die beraessung von
kanalnetzen. Gwf - Wasser/Abwasser. 114(H9), 1973.
34. Soil Conservation Service. National Engineering Handbook, Section 4,
Hydrology. U.S. Dept. of Agriculture, Washington, DC, 1964.
35. Surkan, A.J. Simulation of storm velocity effects on flow from distri-
buted channel networks. Water Resources Research. 10(6):1149-1160, Decem-
ber 1974.
36. Weldon, K.E. F.D.O.T. rainfall intensity-duration-frequency curve genera-
tion. In M.P. Wanielista and Y.A. Yousef (Eds.), Stormwater Management,
_An Update, Publication 85-1, University of Central Florida, Environmental
Systems Engineering Institute, Orlando, Florida. 11-31, July 1985.
37. Wenzel, H.G., Jr. and M.L. Voorhees. An Evaluation of the Urban Design
Storm Concept. Research Report 164, Water Resources~~Center, University^ of
Illinois, Urbana-Champaign, Illinois, August 1981.
The work described in this paper was partially funded by the U.S. Environmental
Protection Agency and has been subject to the Agency's peer and administrative
review. It has been approved for publication as an EPA document.
321
-------
An Expert System Prototype for RECEIV-II Using TURBO PASCAL
by
Robert E. Dickinson
Department of Environmental Engineering Sciences
University of Florida
Gainesville, Florida 32611
Ivan B. Chou
Applied Technology & Management, Inc.
Gainesville, Florida 32602
Fred V. Ramsey
Applied Technology & Management, Inc.
Gainesville, Florida 32602
Abstract
A pre-processor written In TURBO PASCAL was created for the RECEIV-II
model. The graphics based pre-processor enables trained engineers and
scientists familiar with estaurine hydrodynamics to easily run the
receiving water model. The pre-processor acts similarly to programs such
as LOTUS-123 in enabling a higher percentage of people to become competent
modelers.
Introduction
The use of the personal computer by the engineer and scientist Is a
landmark advance that greatly increases their analytical tools.
Spreadsheets, Word Processing programs, and Data Base Management programs
increase productivity by (1) compressing the time between the initial
conceptualization and the final product; and (2) eliminating some of the
mechanical impediments to original creation. The PC aids both in the
creation and production of engineering and scientific products.
The primary usage of the PC, however, is restricted to programs equally
applicable to any profession, trade, or occupation. The scientific and
322
-------
engineering community has only begun to realize the potential of the PC as
design aids and expert systems.
The purpose of this paper is to introduce a prototype expert system for
the RECEIV-II model (Raytheon, 1974) constructed using the Borland
International program TURBO PASCAL (Borland, 1985).
Expert Systems
An expert system incorporates the human expert's knowledge into a
computer program so that others can solve the same type of problem (Van
Horn, 1986). Some of the characteristics of "real" expert systems as
described by Van Horn(1986) are: (1) the program contains the heuristic
knowledge of an expert; (2) the program is able to interface with the user
and developers in natural language; (3) the program asks questions to
obtain needed data; (4) the program is easily refined and upgraded without
extensive reprogramming; (5) the program can explain its conclusions; (6)
the program can accept uncertain input and assign it a certainty factor;
(7) the program gives answers with a certain level of confidence; and (8)
the program learns from its own performance.
A combination pre-processor and water quality model on the PC addresses
most of the requirements of an expert system using a liberal interpretation
of the definition of an expert system. The model itself functions as the
rule base of the expert system. The inference engine is built into the
structure of the program, and the pre-processor acts as a quasi natural
language interface.
The key gain in using a pre-processor and model in the PC environment
is the increase in modeling competence, performance, and productivity.
The model can now be used by any engineer or scientist with a knowledge of
coastal hydrodynamics and limited computer experience.
Pre-Processor
This section will explain the actual construction of a menu driven
pre-processor in TURBO PASCAL. The tools used to construct the pre-
processor include TURBO SCREEN (Pascom Computing, 1985), TURBO PASCAL
(Borland International, 1985), and TURBO Toolbox (Borland International,
1984).
TURBO SCREEN is a program used to construct graphical menus for data
input. The menus are made using a menu driven text editor with WORDSTAR
commands. The program generates TURBO PASCAL code that will mimic in a
compiled TURBO PASCAL program the menus created by the programmer. The
menus, which consist of colored or shaded foreground text on colored or
shaded backgrounds, are restricted to 24 row by 80 columns. The text
explains the data input requirements for each data field and function as
323
-------
built in help screens. The pre-processor data fields, programmer defined,
are either real, integer, string or character.
A graphical menu approach to the pre-processor was chosen over a
simple question and answer approach for aesthetics, program modularity, and
the ability to use window management. Each menu is defined as a picture or
window and is manipulated as a record in PASCAL. PASCAL records are
similar to arrays in FORTRAN except records can have mixed types, e.g.
integer, real, and string components in the same record. Each menu is
easily manipulated by the programmer by simply specifying the picture
number. The beauty of PASCAL over BASIC and FORTRAN is this ability to use
records.
The pre-processor either creates a new input file or edits an existing
file. The user of the RECEIV-II model does not have to learn the FORTRAN
field codes for data input; the pre-processor handles the actual ASCII file
manipulation. The pre-processor allows movement between fields in the
screen by reading the keyboard buffer. The data input stream does not have
to be sequential. It can be entered vertically or horizontally in the
menu. As an example entering channel data can be done channel by channel
or field parameter by field parameter.
The creation of the pre-processor (1950 PASCAL statements) took
approximately three days from start to working product using the advanced
programming tools available from Borland International. The program
creation was greatly aided by the sequential nature of the menus built
using TURBO SCREEN and the editing environment of TURBO PASCAL.
Discussion
Compilers constructed specifically for the PC such as TURBO PASCAL and
TURBO PROLOG (Borland, 1986) offer an enhanced editing environment, faster
compilation, and the advantages of a higher level language. The
programmer's productivity is increased because: (1) low level language
errors such as undefined variables are eliminated; (2) there is better
error checking during compilation; and (3) faster compilation speeds (less
than 1 minute for a 2000 line program). The structure of PASCAL is similar
to FORTRAN. A FORTRAN programmer can easily learn the ancillary techniques
to program in PASCAL.
The most expensive component of software development is the
programmer's time in writing and debugging source code (Brooks, 1975).
Languages and compilation systems that enhance the creation and
implementation of programs will: (1) increase programming productivity; and
(2) enable a higher percentage of the population to program. As
experienced FORTRAN programmer's we know the frustration of late night
sessions hunting for a bug in the program. A language that eliminates most
bugs in the compilation stage has a tremendous evolutionary advantage over
lower level languages.
324
-------
Over the last 40 years as computer languages have advanced from
machine code to assembly language to BASIC and FORTRAN to PASCAL to MODULA-
2 the ease of programming has increased and the percentage of the
population capable of programming has Increased. The popularity of
programs such as LOTUS-123 and DBASE-III Is partly due to the need of
everyone to appear as competent programmers and computer "literate". These
programs allows a novice programmer the almost Instantaneous ability to
generate programs.
A pre-processor linked to a model on the PC allows not just the
experienced programmer but the trained engineer and scientist to be a
competent modeler. The "natural language" interface is similar to LOTUS-
123 in permitting a higher percentage of the population of engineers and
scientists to become competent RECEIV modelers.
References
Borland International, TURBO PASCAL, Borland International, Scotts Valley,
Ca., 1985.
Borland International, TURBO TOOLBOX, Borland International, Scotts Valley,
Ca., 1984.
Brooks, F.P, The mythical man-month, Addison-Wesley, Reading Massachusetts,
1975.
Van Horn, M., Understanding Expert Systems, Bantam Books, Toronto, 1986.
Pascotn Computing, TURBO SCREEN Programmers Manual, Pascom Computing,
Cleveland, Ohio, 1985.
Raytheon Company, Part 1 - RECEIV-II Water Quantity and Quality Model, U.S.
Environmental Protection Agency, Washington, D.C., 1974.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
325
-------
MODELING FLOOD HYDROLOGY USING HYMO
by: James E. Scholl, P.E.
CH2M HILL
Gainesville, Florida 32602
ABSTRACT
HYMO is a computer program well suited to model flood hydrology for a
wide range of watershed conditions. Any design storm duration and
distribution can be input along with site-specific unit hydrograph shape
parameters. Stream or channel routing is accomplished using the variable
storage coefficient method; the storage-indication method is used for
reservoir routing. The original program was developed for mainframe
equipment, but can easily be adapted for use on microcomputers.
After a brief summary of HYMO commands, formatting, and storage, the
paper describes hydrograph analyses performed in the U.S. Virgin Islands.
Results of the analyses were used to develop a basis for assigning unit
hydrograph parameters to model flpod hydrology.
HYMO DESCRIPTION
HYMO, derived from the words hydrologic model, is a computer language
developed by the Agricultural Research Service, U.S. Department of Agricul-
ture, in cooperation with the Texas Agricultural Experiment Station, Texas
A&M University (1). The original program was developed for mainframe equip-
ment, but can easily be adapted for use on microcomputers. HYMO commands
are simple to use and offer a great deal of flexibility. For example, any
design storm duration and distribution can be input along with site-specific
unit hydrograph shape parameters. Flood hydrographs are developed using
unit hydrograph theory and the SCS rainfall-runoff relationship.
The variable storage coefficient {VSC) flood-routing method is used for
stream routing. The VSC method accounts for changes in channel velocity or
reach travel time with flood stage. Although an iterative solution is used,
326
-------
the VSC method requires minimal computer time and is free of convergence
problems. Channel section rating curves used for VSC flood-routing calcula-
tions can be input directly or computed internally by HYMO for uniform flow
using Manning's Equation. Channel section data for internal calculations
can be segmented into flood plain and main channel components with appro-
priate slope and n values for each segment.
The storage-indication method as documented in the SCS National
Engineering Handbook, Section 4, Hydrology (2) is used by HYMO for reservoir
routing computations.
HYMO COMMANDS
The HYMO language includes 17 commands for performing flood hydrology
and sediment yield calculations. Commands are expressed in the first 20
columns of the data card, with columns 21 through 79 used for numeric data
and keywords. Column 80 is reserved for a page change code (an asterisk in
column 80 causes the printer to advance to a new page). Continuation cards
are allowed when 59 characters are insufficient to present the data.
The data can be written in any format, but at least one blank space
must be left between data items. A decimal is required for numbers
containing fractions, but not for whole numbers. Keywords can be written
with the data to describe individual data items. Comment cards may be used
at any point in a HYMO program by punching an asterisk in column 1 and the
comment in columns 2 through 79.
Six hydrographs can be stored in a HYMO program at a time, identified
by storage location numbers 1 through 6. The storage location numbers must
be.repeated for watersheds requiring more than 6 hydrographs. However, no
more than six hydrographs are ever needed at one time because HYMO programs
begin at the head of a watershed and work downstream through one reach at a
time. Because the first hydrograph is lost when a storage location number
is used to store or compute another hydrograph, the user should be sure that
the hydrograph will not be referred to again before using the storage
location number for another command.
Details of rules for HYMO commands are presented in the Users Manual
(3) along with an example problem.
HYDROGRAPH ANALYSIS
To establish a site-specific basis for using HYMO in the U.S. Virgin
Islands, published and unpublished streamflow data were obtained from the
U.S. Geological Survey (4)(5). Because long unbroken periods of record were
not available, observed data did not provide an adequate basis to develop an
annual maximum flood series. The data did, however, contain several
isolated single-event runoff hydrographs which could be analyzed to
determine site-specific flood hydrograph parameters. These included the
time to peak, the total runoff volume, the peak flow rate, and the SCS unit
hydrograph shape factor.
327
-------
Streamflow data from four gaged watersheds on St. Croix were used to
establish a basis for assigning unit hydrograph parameters for ungaged
watershed areas. Streamflow records considered in this study are summarized
in Table 1.
TABLE 1. USGS STREAMFLOW DATA AVAILABLE FOR ST. CROIX, U.S. VIRGIN ISLANDS
Station Name
River Gut at River
River Gut at Golden Grove
Jolly Hill Gut at Jolly Hill
Station
Number
3320
3330
3450
Drainage
Area (mi2)
1.42
5.12
2.10
Period of
Record Examined
1963
1963
1963
- 1967 (5 years)
- 1971 (9 years)
- 1968 (6 years)
Creque Gut above Mount
Washington Reservoir
3470
0.50
1965 - 1967 (3 years)
The first step in the hydrograph analysis was to obtain the original
stage hydrographs {strip charts) from the USGS files in Puerto Rico (6).
Strip charts were obtained for selected days for all four stream gaging
stations listed in Table 1. These strip charts were then screened for
single-peak events which resulted from short duration rainstorms. Since
synchronized rainfall records were not available at these gages, this
selection process was largely a matter of judgment.
Once the stage hydrographs were selected for analysis, they were
converted to discharge hydrographs by application of the appropriate rating
table, and the following hydrograph parameters were measured:
1. Time to peak {T ), in hours, defined as the time from the
beginning of rile to the peak of the hydrograph
2. Peak flow rate (q ) of the runoff hydrograph, in cfs
3. Total volume of runoff (Q), in inches
Using these measured parameters, the hydrograph shape factor, B, was
computed for each event as follows:
B =
(1)
where:
B = Hydrograph shape factor
A = Watershed drainage area, in square miles
{all other terms are as previously defined)
These parameters are reported in Table 2 for each event analyzed.
328
-------
TABLE 2. HYDROGRAPH ANALYSIS RESULTS, BASED ON USGS DATA
Location
River Gut at River,
No. 3320
Mean
River Gut at Golden
Grove, No. 3330
Mean
Jolly Hill at Jolly
Hill, No. 3450
Mean
Creque Gut above
Mt. Washington
Reservoir, No. 3470
Mean
T
Event Date p
4/06/65
5/30/65
11/09/65
8/01/63
11/02/63
11/21/63
10/24/69
10/30/69
12/02/69
1/03/63
1/04/63
12/12/65
12/12/65
10/13/65
11/12/65
11/09/65
5/17/65
(hours)
1.7
3.0
1.0
1.9
4.00
1.30
1.25
1.50
5.50
2.75
2.72
0.60
0.50
0.60
0.70
0.60
1.0
1.0
0.5
0.5
0.75
Q (inches)
0.033
0.064
0.023
0.240
0.100
0.021
0.112
0.062
0.070
0.024
0.097
0.009
0.023
0.297
0.250
0.174
S (cfs)
11.44
11.78
18.91
144.0
123.0
28.1
169.0
26.5
41.5
39.5
177.0
17.5
39.5
91.1
79.8
83.7
B
415
389
579
461
467
311
325
440
457
317
386
470
434
556
572
508
613
638
481
462
549
Equation 1 is an algebraic transformation of the hydrograph peak rate
equation employed by the Soil Conservation Service (SCS) (7). The standard
value for B used by the SCS in the majority of hydrologic design applica-
tions is 484. However, Mockus (1972) reported that the hydrograph shape
factor has been known to vary from about 600 in steep terrain to 300 in very
flat swampy country. From equation 1, it can be seen that the larger the
hydrograph shape factor, B, the larger the peak rate of runoff generated by
a given volume of runoff, Q. Thus, the hydrograph shape, which is generally
related to topography, is also an important factor influencing flood
potential.
The hydrograph shape factor is a watershed characteristic and should be
a constant for each watershed analyzed. However, considerable variation is
reported in Table 2 for individual event B values at each gaging station.
These variations are primarily due to the fact that all errors present in
each measured hydrograph parameter (i.e., T , Q, and q ) are combined in the
computation of B. The time to peak, T , is^a particularly difficult para-
meter to measure accurately due to the time scale of the original strip
charts (1 inch = 10 hours) and the short times of observation. Therefore,
329
-------
the mean of the event B values is probably the best estimate of the hydro-
graph shape factor for each watershed analyzed.
Plotting the results from Table 2 as shown in Figure 1, the relation-
ship between the unit hydrograph shape factor, B, and drainage area provides
a basis for modeling ungaged watersheds. A total of 13 watershed areas in
the U.S. Virgin Islands ranging in size from 365 acres to 3,396 acres have
been modeled using this procedure. Nine of these watersheds were on St.
Croix and four on St. Thomas. The results of this modeling work are
contained in three technical reports by CH2M HILL (8){9}(10).
CONCLUSIONS
HYMO is a practical tool for evaluating flood hydrology for a wide
range of conditions and project requirements. The analysis of observed
streamflow hydrographs, as demonstrated using data for the U.S. Virgin
Islands, can provide a site-specific basis for assigning unit hydrograph
parameters for ungaged 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
1. Williams, J.R. and Hann, R.W. HYMO: Problem-Oriented Computer
Language for Hydrologic ModelingUsers Manual. ARS-S-9, Southern
Region Agricultural Research Service, U.S. Department of Agriculture,
1973.
2. Mockus, V. Section 4, Hydrology. In: SCS National Engineering
Handbook (NEH-4). U.S. Department of Agriculture, Soil Conservation
Service, Washington, D.C., 1972.
3. Williams, J.R. and Hann, R.W. HYMO: Problem-Oriented Computer
Language for Hydrologic ModelingUsers Manual. ARS-S-9, Southern
Region Agricultural Research Service, U.S. Department of Agriculture,
1973.
4. Robison, T. M. et al. Water Records of the U.S. Virgin Islands,
1962-69. U.S. Geological Survey, San Juan, Puerto Rico, 1972.
5. McCoy, J. Personal Communications. U.S. Geological Survey, San Juan,
Puerto Rico, 1979.
6. McCoy, J. Personal Communications. U.S. Geological Survey, San Juan,
Puerto Rico, 1979.
330
-------
Hydrograph Shape Factor B
8 8 § 1 § 1
Crequc
B = 4
h-
> Gut Above ML W
1
1 J
M Slandar
dSCS
-L.
ashinglon
Reservoir
telue
**-,
^^M
^
^M
'-
^*"*-
River Gut at River
Rh
Jolly Hil
er Gut at (
Gul at
-^
iolden <
Jolly 1
^
5rove
Hill
«_
"*
**
*
0,2 0.3 0.4 0.5 0.6 0.6 1.0 2.0
Drainage Area (square miles)
3.0
4.0 5.0 6.0
8.0
FIGURE 1. Hydrograph Analysis Results for the U.S. Virgin Islands.
7. Mockus, V. Section 4, Hydrology. In: SCS National Engineering
Handbook (NEH-4). U.S. Department of Agriculture, Soil Conservation
Service, Washington, D.C., 1972.
8. CH2M HILL. A Flood Damage Mitigation Plan for the U.S. Virgin Islands.
Prepared for the Disaster Programs Office, Office of the Governor,
Government of the U.S. Virgin Islands, 1979.
9. CH2M HILL. Planned Drainage Basin Studies for the Protection of Roads
from Flood Damage in the U.S. Virgin Islands, Volume 1St. Thomas.
Prepared for the Public Works Department, Government of the U.S. Virgin
Islands, 1982.
10. CH2M HILL. Planned Drainage Basin Studies for the Protection of Roads
from Flood Damage in the U.S. Virgin Islands, volume 2St. Croix.
Prepared for the Public Works Department, Government of the U.S. Virgin
Islands, 1982.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
331
-------
ATTENDEES
Name
John Aldrich
Richard Baker
Tom Barnwell
Christina Barrett
Vern Bonner
John M. Grouse
Brett Cunningham
Geoffrey Dendy
Roy R. Detweiler
Robert E. Dickinson
Forrest Dierberg
Jon Dobson
Eugene Driscoll
J. D. Edgman
Jeff Einhouse
Andrew C. Eversull
Raymond Ferrara
Huqh Fraser
David R. Gaboury
Victor Gaghorado
Richard Gietz
Lou Grant
Thomas T. Griffin
Judy Grim
Representing
COM
Metcalf & Eddy
EPA
Greiner Engineering, Inc.
Hydrologic Engineering Center
Greenhorne & O'Mara, Inc.
University of Florida
Greinger Engineering, Inc.
Chados Ford Enterprises, inc.
University of Florida
Florida Institute of Technology
Gainesville, FL
Oakland, NJ
Dallas Water Utilities
Miller, Miller, Sellon, Einhouse
Louisiana State University
Lafayette College
Cumming-Cockburn Associates, Ltd,
Woodward-Clyde Consultants
Poland, FL
Regional Municipality of Ottawa-
Carleton
Bio Environmental Services
Najarian & Associates, Inc.
Briley, Wild & Associates
332
-------
ATTENDEES
Name
Ji Han
Michael C. Hancock
James P. Kearney
Jeffrey D. Holler
Wayne Huber
William James
Howard A. Jongedyk
Roger T. Kilgore
Stephen Kintner
Tinny H. Lee
Agustin E. Maristany
Robert McCarthy
Joseph M. McGinn
S. Wayne Miles
Michael Morrison
Ananta K. Nath
J-Rene' Noiseux
J. Robert Owen
Richard J. Pfevfrer
Larry A. Roesner
Mark Robinson
Robert Roussel
Representing
Post, Buckley, Schuh, & Jernigan Inc.
University of Florida
University of Florida
South Florida Water Management District
University of Florida
McMaster University
Federal Highway Administration
GKY 6 Associates
Brevard County Water Resources Dept.
Briley, Wild & Associates, Inc.
N.W. Florida Water Management District
Dallas Water Utilities
CB MacGuire Inc.
Raleigh, NC
Hayes, Seay, Mattern & Mattern
Nebraska Natural Resources Commission
Les Consultants Dessan, inc.
Colorado Dept. of Health
South Florida Management Division
Camp, Dresser, & McKee
Computational Hydraulic Group
Les Consultants Dessan, Inc.
333
-------
ATTENDEES
Name
A. Charles Rowney
Marty Sanders
James E. Scholl
Linda Seigle
Phil Shelly
Himat Solanki
Wing H. Tang
Wilson Timmens, Jr.
Ben Urbonas
Michael G. Waldon
Raymond E. Wiles
Dr. P. Wisner
Yousef A. Yousef
Claudia L. Zahorcak
Representing
Queen's University
Lindahl, Browning, Ferrari & Hellstrom
CH2ZM-Hill
Florida Dept. of Transportation
Washington Analytical Sers. Center Inc.
Smally, Wellford & Nalven, Inc.
Univ. of Puerto Rico
Brevard County Water Resources
Urban Drainage & Flood Control District
LSU - Louisiana State Univ.
GeoScience, Inc.
Univ. of Ottawa
Univ. of Central Florida
Brown and Caldwell
*U.S. GOVERNMENT PRINTING OFFICE: 1986-646-116-40655
334
------- |