u
E
dtes
.nental Protection
Environmental Research
Laboratory
Athens GA 3061 3
EPA/600/9-87/016
August 1387
Research and Development
&EPA
Proceedings of
Stormwater and Water
Quality Model Users
Group Meeting
March 23-24, 1987
Denver, Colorado
-------
EPA/600/9-87/016
August 1987
PROCEEDINGS
OF
STORMWATER AND WATER QUALITY MODEL
USERS GROUP MEETING
March 23-24, 1987
Denver, Colorado
Edited by
William James
Cudworth Professor of Computational Hydrology
University of Alabama
Tuscaloosa, AL 35487
Thomas O. Barn well, Jr.
Center for Water Quality Modeling
Environmental Research Laboratory
Athens, GA 30613
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATHENS, GA 30613
Unv4 rr--.r./ji'-v,&l Projecti:
-------
DISCLAIMER
The information in this document has been funded in part by the United
States Environmental Protection Agency. Papers describing EPA-sponsored
research 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.
ii
-------
FOREWORD
A major function of research and development programs is the effective
and expeditious transfer of 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 the 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 23-24, 1987, in Denver, CO,
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
iii
-------
ABSTRACT
This proceedings includes 18 papers on topics related to the develop-
ment and application of computer-based mathematical models for water quality
and quality management. The papers were presented at the semi-annual meeting
of the joint US-Canadian Stormwater and Water Quality Model Users Group, held
on March 23-24, 1987, in Denver, Colorado.
Several papers deal with recent developments and adaptations of the
USEPA SWMM model itself. Its application in a variety of situations is
described in a nunber of additional papers. Other models covered include
UDSEWER, SEWERCADD, and RAFTS.
A number of papers provide a critical overview of hydrologic models and
modeling techniques, and a prediction of future development in stormwater
modeling, particularly on microcomputers. Other papers deal more specifically
with such topics as tidal flooding, corrective phosphorus removal, wasteload
allocations, and spreadsheet cost estimations for drainage design parameters.
iv
-------
CONTENTS
Page
FOREWORD iii
ABSTRACT iv
ACKNOWLEDGMENT vii
STORM SEWER SYSTEM DESIGN BY UDSEWER MODEL 1
Chwen—Yuan Guo and Ben Urbonas
MICROMOMPUTERS--THE COMPUTER STORMWATER MODELLING FUTURE 10
Geoff R. Thompson and Brett C. Phillips
ENHANCING SWMM3 FOR COMBINED SANITARY SEWERS 21
William James and T. Wayne Green
SEWERCADD 42
Michael H. Jackson and Jeff L. Lambert
CURRENT TRENDS IN AUSTRALIAN STORMWATER MANAGEMENT 42
Allan G. Goyen
A NEW GROUNDWATER SUBROUTINE IN THE STORM WATER MANAGEMENT MODEL 70
Brett A. Cunningham, Wayne C. Huber, and Victor A. Gagliardo
SWMM APPLICATIONS FOR MUNICIPAL STORMWATER MANAGEMENT: THE EXPERIENCE
OF VIRGINIA BEACH 105
John A. Aldrich and John E. Fowler
THE EFFECT OF SUBWATERSHED BASIN CHARACTERISTICS ON DOWNSTREAM
STORM-RUNOFF QUALITY AND QUANTITY 110
Rob G. Brown
SOME THOUGHTS ON THE SELECTION OF DESIGN RAINFALL INPUTS FOR URBAN
DRAINAGE SYSTEMS 119
Ivan Muzik
FIELD MEASUREMENT AND MATHEMATICAL MODELING OF COMBINED SEWER
OVERFLOWS TO FLUSHING BAY 125
Guy Apicella, Donald Distante, Michael J. Skelly, and Les Kloman
ACCOUNTING FOR TIDAL FLOODING IN DEVELOPING URBAN STORMWATER
MANAGEMENT MASTER PLANS 149
Stergious Dendrou and Kelly A. Cave
WASTELOAD ALLOCATION FOR CONSERVATIVE SUBSTANCES 171
Main Hutcheson
,THE USE OF DETAILED COST ESTIMATION FOR DRAINAGE DESIGN PARAMETER
ANALYSIS ON SPREADSHEETS 180
S. Wayne Miles, Thomas G. Potter, and James P. Heaney
-------
Page
CORRECTIVE PHOSPHORUS REMOVAL FOR URBAN STORM RUNOFF AT A RESIDENTIAL
DEVELOPMENT IN THE TOWN OF PARKER, COLORADO 194
William C. Taggart and Mary S. Wu
EVALUATION OF SEDIMENT EROSION AND POLLUTANT ASSOCIATIONS FOR URBAN
AREAS 205
Kim Irvine, William James, John Drake, Ian Droppo, and Steve
Vernette
UNCERTAINTY IN HYDROLOGIC MODELS: A REVIEW OF THE LITERATURE 217
T.V. Hromadka II
UNCERTAINTY IN FLOOD CONTROL DESIGN 228
T.V. Hromadka II
LIST OF ATTENDEES 248
VI
-------
ACKNOWLEDGMENT
The Stormwater and Water Quality Model Users Group appreciates the help
of interested members in making arrangements for certain of its meetings.
This particular meeting was arranged and organized by Dr. William James of
the Univesity of Alabama. It was the Group's first meeting in Colorado, a
particularly beautiful area. The meeting was locally sponsored by the Flood
and Drainage Control District 69, and local assistance was supplied by
Dr. James Guo of the University of Colorado at Denver.
vii
-------
STORM SEWER SYSTEM DESIGN BY UDSEWER MODEL
By
Chwen-Yuan Guo, Ph.D, P.E.
Civil Engineering Department
University of Colorado At Denver
Denver, Colorado.
and
Ben Urbonas, P.E.
Chief, Master Planning Program,
Urban Drainage and Flood Control District
Denver, Colorado
INTRODUCTION
Storm sewer system is a vital element in preserving an urban storm
drainage systems; the design of storm sewers involves surface runoff
hydrology and sewer hydraulics including surcharged and open channel flow.
Considerable effort has been devoted to the developments of methodology and
computer models for storm sewer system design. Despite of the existence of
many sophisticated hydrologic techniques for the design of storm sewers,
the most commonly used one is still the rational method.
Storm sewer design needs to meet the design requirements while
recognizing existing physical constrains such as slopes, depth of cover,
utility interferences, etc. Usually a. storm sewer design is achieved by a
series of trial and error calculations until flow conditions and
configurations satisfy all the design requirements and site constraints.
This iterative process is time consuming and manpower demanding.
In 1986, the Urban Drainage and Flood Control District in Denver,
Colorado sponsored the development of the personal computer software,
UDSEWER (1), for the design of storm sewer system. Although UDSEWER is
primarily programmed to follow the Urban Storm Drainage Criteria Manual (2)
at the District, it does provide the user options of implementing different
criteria to override default values.
This paper presents background information of UDSEWER, including its
capability, limitations and features. It is believed that the. use of
UDSEWER can improve the efficiency of storm sewer design.
-------
GENERAL INFORMATION ON UDSEWER
UDSEWER is a computer software developed for the use with the IBM
personal computer and compatibles. It is written in compiled BASIC
computer language and is menu driven with graphic displays, user
interaction on-screen editing and on-line help. It can be run on
machines that have a floppy or hard disk drives and can be printed using
standard dot matrix printer.
UDSEWER can handle a basin system having up to 100 manholes and up to
100 sewers. For larger basins, the user may input off-site runoff and in
effect link basins together. EAch manhole mode can have up to four
incoming sewers, but only one outgoing sewer. UDSEWER will check for
consistency of input data and for proper connections of the sewer-manhole
network.
The rational method is used to estimate the peak runoff and to size the
downstream sewer. Although UDSEWER uses open channel flow hydraulics to
size or evaluate sewer segments, it will also handle pressure flow in
existing sewers that are smaller than required. UDSEWER also calculates
surface water profiles throughout the sewer system to estimate the water
surface elevation at each manhole. Although the program will size only
circular pipes, it will also calculate pressure flow and hydraulic grade
lines in existing sewers of rectangular and arch shapes.
Final printout includes hydrology and depths of cover at each manhole
and flow conditions for each sewer segment. The latter including flow
velocity, surcharged length, possible hydraulic jump, etc. UDSEWER will
flag violations of design constraints set forth by the user.
REQUIRED INPUT DATA
There are four required groups of input data:
(1) design constraints
(2) rainfall intensity-duration-frequency
(3) manhole information and its surface hydrology and
(4) sewer hydraulic information
The programs user's manual provides input data summary tables designed
for the use with the data editor. The following describe in more detail
the specifies of each of the four data groups:
Design Constraints:
Design constraints includes minimum dept of coverage, minimum, sewer
size, and the range of permissible flow velocities in sewer (3). Design
-------
constraints do not affect hydraulic computations, but only serve as a basis
to flag any violations in criteria. The program has a pre-set default
values for these constraints; however, the user has an option to override
any of them.
Rainfall Intensity - Duration Frequency:
Rainfall intensity-duration-frequency data is needed for the Rational
Method. UDSEWER permits the user to provide either a rainfall intensity
formula or a rainfall intensity table for durations of 5, 10, 30, 40, 60,
and 120 minutes.
The default rainfall intensity formula in UDSEWER takes the format
developed for Denver area.
A • HI ,,,
i = _ (1)
(Td + B)
in which i - rainfall intensity in inches/hour, HI = one hour rainfall
depth in inches, Td - rainfall duration in minutes, and A, B, and C are
empirical constants.
For Denver area, it has been found that A = 28.5, B = 10 and C = 0.786.
When rainfall intensity formulate is not available or does not take the
above format, the user may enter the design rainfall intensity-duration
table. UDSEWER will use linear interpolation and extrapolation to find the
design rainfall intensity for any other rainfall duration. Due to the fact
that most rainfall statistics were developed from the data recorded with
the shortest time interval of five minutes, UDSEWER therefore uses the
intensity of five-minutes for any duration (time of concentration) that is
less than five minutes.
Manhole Information and Surface Hydrology:
The input data for each manhole includes the manhole identification
number, the identification numbers of incoming and outgoing sewers assigned
by the user and the ground elevation of the manhole.
When the local peak runoff is known, the user can input its valuation
with the contributing area and runoff coefficient for the design flood. If
the user wishes UDSEWER to calculate the flow, the user provides sub-basin
area, runoff coefficients for both the five-year flood and design flood,
overland flow length and its slope and gutter flow length and its velocity.
Program will calculate the time of concentration to the upstream manhole of
the basin using the following equations.
Tc = to + Tf (2)
-------
Tf - Lf / (Vf x 60) (3)
1.8 (1.1 - C5) Lo°-5
in which Tc - time of concentration in minutes . Tf - gutter flow time in
minutes, Lf - gutter flow length in feet, Vf = gutter velocity in
feet/second, Lo - overland flow length in feet, To = overland flow time in
minutes, So - overland flow slope, C5 - five-year overland runoff
coefficient.
In the computation of the time of concentration, the program defaults
to overland flow length of less than 500 feet for a rural area and 300 feet
for an urbanized area. According to the Denver Urban Drainage Design
Criteria, the time of concentration of the basin can not be shorter than 5
minutes for an urbanized area and 10 minutes for a rural area. These
design criteria have been programed into UDSEWER. The demarcation of
urbanization used in UDSEWER is the five-year runoff coefficient, C5. when
C5 > = 0.3, it is considered urbanized.
Assuming that the time of concentration is the critical design rainfall
duration, UDSEWER uses the rational method to estimate the peak runoff.
Qp - C i A (5)
in which Qp - peak runoff in cfs, C - runoff coefficient for design flood,
A — drainage area in acres.
As the runoff moves downstream in storm sewers, the times of
concentration at each manhole from the different parts of the basin
upstream are independently calculated. Often, in practice, the longest
time of concentration is used for design rainfall duration. However, it is
possible that a highly urbanized subbasin with a shorter time of
concentration may generate higher peak runoff than a larger composite
rural-urbanized subbasin with a longer time of concentration. Therefore,
the program calculates every possible combination of subbasins upstream and
use the highest peak runoff to size the immediate downstream sewer
dimensions.
Sewer Hydraulic Information:
The user needs to provide sewer identification number, length, slope,
Manning's roughness n, shape and upstream crown elevation. For an existing
sewer, the user needs to identify the sewer shape and dimensions and
UDSEWER will evaluate its capacity. For new sewers, round pipes will be
sized for the computed or given peak runoff rates. However, the user may
use the option of existing sewer to predetermine the sewer shape.
-------
Manning's equation is empldyed to compute the required sewer size and
UDSEWER will then suggest the next larger commercially available sewer size
for design.
When calculating hydraulic grade line, Benoulli energy equation is used
to balance the energy between two adjacent cross sections.
HI = H2 + friction loss + local loss (7)
HI = Yl + Zl
(Vl?2g) (8)
in which HI - Bernoulli sum at section 1, etc., Yl - flow depth in feet at
section 1, etc., Zl - elevation in feet at section 1, etc., and VI = cross
sectional average velocity in ft/sec at section 1, etc.
The friction loss is computer by the nonuniform open channel flow
equation.
r n2 v2 R(-*/O i
friction loss = I f^ I (9)
in which n - Manning's roughness, R - hydraulic radius in feet, Ls - sewer
length in feet.
The junction loss caused by the turbulence at each manhole is estimated
by UDSEWER to be 50% of the difference between the incoming and outgoing
velocity heads. The exit loss caused by the downstream surcharge or
submergence is assumed to be the entire velocity head at the exit.
CASE STUDY
The layout for the example storm sewer system is shown in Figure 1.
The user may utilize the input data forms to prepare the input data and
then use the UDSEWER data editor to input and edit data. UDSEWER produces
two reports: Report I, as shown in Table 1, tabulates input data and
Report II, as shown in Table 2, summarizes flow conditions in each sewer
and surface hydrology at each manhole.
You will note in Table 2, that for this example the overland slope is
manhole 3 is so flat that the overland flow time is 174 minutes. Instead
of taking the longest time for concentration as the design rainfall
duration, UDSEWER checks every possible combination of upstream subhasins.
As a result, the highest runoff rate at manhole 3 is determined to be the
combined runoff contributed from the upstream subbasins at manhole 7 and
manhole 5.
-------
In this example, the last sewer, ID number 992, the outlet is fully
submerged. As a result the entire sewer 992 and part of sewer 1499 are
surcharged.
A comparison between the predictions from UDSEWER and STORM1, developed
by the Construction Engineering Research Laboratory, Corp of Engineers (4),
is shown in Table 3. Although STORM1 uses Kirpitch formula to compute
overland flow time, for this example, both methods showed a good agreement
in sizing of storm sewers.
CONCLUSIONS
A personal computer software, UDSEWER, was developed for storm sewer
system design. Although UDSEWER follows the Denver Urban Storm Drainage
Criteria, it has the flexibility of input of any user defined criteria.
UDSEWER is capable of handling multiple basins with each basin having up to
100 manholes and up to 100 sewers. UDSEWER estimates surface hydrology at
each manhole and calculates flow conditions for each sewer segment.
UDSEWER is supported by the University of Colorado at Denver and Urban
Drainage and Flood Control District, Denver, Colorado.
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.
ACKNOWLEDGMENT
The development of UDSEWER model, a personal computer software for
storm sewer system design, was sponsored by the Urban Drainage and Flood
Control District, Denver, Colorado. The practical application of this
software is not, however, limited to the Denver region.
REFERENCES
1. Guo, C.Y. and Urbonas, B, "Storm Sewer System Simulator", the Fourth
National Conference on Microcomputer in Civil engineering held in
Orlando, Florida, Nov. 5-7, 1986, PP 312-316
2. "Urban Storm Drainage Criteria Manual", Vol 1, Runoff Section, Urban
Drainage and Flood Control District, Denver, Colorado, 1969
3. "Design and Construction of Sanitary and Storm Sewer", American Society
of Civil Engineers, New York, 1979.
4. "CESTORM: Storm Sewer Distribution Model"; Department of Army,
Construction Engineering Research laboratory, Champaign, Illinois.
-------
UDSEWER REPORT I: SUMMARY OF INPUT DATA
INPUT DATA FOR STORK mat SYSTEM DESIG*
USIMG UDtEVER-HODEL VERSION 1.1
DEVELOP CD
•r
INEE^I^'u.^^'^.''^ COLORADO AT DENVER
1* COOPEUATIOH vim DENVER
URBAN DRAINAGE AHD FLOOD CONTROL DISTRICT
DUIVEI. COLORADO
DEPARTMENT OF CIVIL
' PROJECT TITLE ;
CASE STUDY : EXAMPLE ONE
' DESIGN CONSTANTS A8E:
MIMMUH BUSIED DEPTH IH FT
HINInUH SEVER SIZE IK INCH
MAXIMUM FLOW VELOCITY IN FPS
MINIMUM FLOtf VELOCITY IN fps
* 2
• U
• 20
• 3
>•• DESIGN RAINFALL RETURN PERIOD IS 5 HAS
«•• DESIGN RAINFALL DURATION IMIN) VS. RAINFALL INTENSIFY (1W/I,RI
^UHKARr OF SUBBASIN INFORMATION
STORK SEVER SYSTEM DESIGN: NEV SEVERS AND EXISTING SEVERS
MANHOLE GROUND
IP
:
3
•.
NO. ELEVATION
FT
00 98.70
.00 100.30
.00 99.00
.00 99.20
.00 97, SO
.00 46.50
.00 97.50
2.00 93.00
MAW HOLE OVERLAND
10 NO. FLOV LHTH
FEET
J
.00 200.00
00 200.00
. C'O 200.00
.CO 100.00
.00 200.00
2.00 200.00
BASIN
AREA
ACRE* 5)
3 00
3.00
3 0
3 0
3 0
3 0
3 0
•3 0
DESIGN
RUNOFF C
<1.0
0.70
0.70
0.70
0.70
0.70
0.70
0.70
0.70
OVERLAND GUTTER
SLOPE FLOV LHTH
\ FEET
0.02
0.02
0.00
0.02
0.02
0.02
305.00
305.00
305.00
405.00
305.00
305 00
DESIGN
RUNOFF CS
<1.0
0. 0
0.
Q.
0.
0.
0 )
0.
Q.
CUTTER
VELOCITY
FFS
2.50
2.50
2.50
4.00
2.50
2.50
KNOWN
PEAK OP
CFS
0. 0
0. 0
0. 0
0, 0
0. 0
0. 0
0. Q
•• SUHMARt OF MANHOLE AND SEVER CONNECTIONS
MANHOLE OUTGOING IHCOMINO UW> ID HUMMUS
ID NO. SEWER It, 1ST SEVER 2ND IEVER 3RD (EVER «TH
5.0
7.0
3.0
6.0
10.0
11.0
95. 0
2.0
53.0
73.0
314.1
610.0
1014.0
1499.0
992.0
0,0
0.0
0.0
S3.0
0.0
610. 0
1014.0
1499.0
992.0
0.0
0.0
73.0
0.0
0.0
314.0
0.0
0.0 (
>.0 0.
.0 0.
• 0 0 .
.0 0.
.0 0.
.0 0.
.0 o.
.0 0:0
S EWE ft LENGTH
ID NO.
FT
3
10
14
9
3
4.
D.
4 .
»9.
2.
0
Ci
0
0
0 1
0 1
00.00
00.00
20.00
00.00
300.00
10.00
SLOPE
\
0.47
0.01
0.80
0.02
0 15
0 02
UPSTREAM
CROWN
ELEV (FT)
95. AS
93.06
95. t«
93.12
92.96
91.46
HANNINC'M
ROUGHNESS
0.013
0.013
0.014
0.013
0.013
0.014
SHAPE EJtI£
DIM
(IN)
BOX
ROUND
ARCH 1
ROUND
ROUND
ROUND
INC LTOKDITION
1CHI WIDTH
FT) (IN) IFTt
.5
.0
.0
.0
.0
.0
1.50
0.00
24.00
0.00
Q.OD
) 0.00
(1) DIMENSION UNITS FOR ROUND AND AfiCH 5EVEA ARE INCHES
DIMENSION UMTS FOR BOX SEVER ARE FEET
(2) WHEN DIMENSION OF SEWER IS NOT GIVEN (-0), PROGRAM WILL
-------
UDSEWER REPORT II: SUMMARY OF SEWER SYSTEM DESIGN
SUMMARY or COMBINED TIMII OF CONCENTRATIOH AT MANHOLES
REPORT OP STORH SEWER SYSTEK DESIGN
USING UDSEVER-MODEL VERSION 1.1
MANHOLE LOCAL INCOMING Tc AT TRAVEL AREA'C TIME Or
ID NUMBER BASIN Tc (EVER ID UPST NNHL TINE CNCNTRTION
MINUTES MINUTES (MINI
•I 9.00
JAMES C.». GUO, PHD, PE
DEPARTMENT OP CIVIL ENGINEERING, UNIVERSITY OF COLORADO AT DENVER 7 00
IN COOPERATION WITH
URBAN DRAINAGE AND FLOOD CONTROL DISTRICT 3 oo
DENVER, COLORADO
6.00
10.00
*•• PROJECT TITLE :
CASE STUDY : EXAMPLE ONE 14.00
96.00
RAINFALL INTENSITY TABLE II GIVEN
2.00
••* SUMMARY OF SUBBASIN RUNOFF PREDICTIONS
12
12
176
12
19
19
44
39
.91
.(1
.14
.81
.54
.94
.06
.94
51
71
(10,
1014
314
1199
992,
.00
.00
.00
.00
.00
•to
.00
12
12
13
19
14
46
10
.91
.11
.11
.94
.91
.70
.95
1
2
0
1
9
3
12
.99
.11
.91
.16
.SB
.9!
.91
2
3
2
2
2
2
2
2
2
4
6
2
12
1
14
.10
.10
.10
.10
.10
.10
.10
.10
.10
.20
.10
.10
.(0
.10
.10
13.11
12.11
116.14
14.40
14.91
12.91
19.54
13.'-,
19. 54
46.10
34.il
44. Of
SO. SS
19.54
(1.46
TIME OF CONCENTRATION
OVERLAND GUTTER BASIN
10
NUMBER AREA
5.00
7.00
3.00
(.00
10.00
14.00
91.00
2.00
« c
2.10
2. ID
2.10
2.10
2,10
2.10
2.10
2.10
To (Kim
37.51
37.51
174.11
26.52
17. St
37.51
41.94
37.51
Tf iniK)
2.03
2.03
2.03
1.C9
2.01
2.03
2.13
2.03
TC (HIN) INCH/HR CF5
12.81 4
12. S
176.1
12. S
)9.5
39.5
4
0
4
2
2
44.06 2
39.54 2
04
04
3S
04
39
39
26
39
.49
.49
.79
.49
.02
.02
.75
.02
** SUMMARY OF HYDRAULICS AT MANHOLES
HANHOLE SUM OF RAINFALL RAINFALL PREDICTED GROUND WATER COMMENTS
ID NUMBER AREA • C DURATION INTENSITY PEAK FLOW ELEVATION ELEVATION
HINUTES INCH/HR CFS FEET FEET
5.00
7.00
3.00
(.00
10.00
14.00
9S.OO
2.00
2.10
2.10
6.30
2.10
4.20
12. (0
14.70
0.00
12. SI
12. SI
14.93
12. SI
19.94
41.70 •
SO.S9
0.00
4.04
4.04
1.99
4.04
2.19
2.19
2.09
0.00
9.49
S.49
It. 15
9.49
10.01
11.54
10. (S
0.00
9S.70
100.90
99.00
99.20
91.50
94.50
97.90
91.00
• 94. 7C
M.73
92.92
95.51
92.9-5
92.95
92.22
92.00
OK
OK
OK
OK
OX
OK
OK
OK
FOR RURAL AREA, BASIN TIHE or CONCENTRATION >10 MINUTES
FOR URBAN AREA. 1AE1N TIME OF CONCENTRATION >5 MINUTES AND
AT THE 1ST DESIGN POINT, TC < (lO'TOTAL LENGTH/1t0) IN MINUTES
VHEN WEIGHTED RUNOFF COEFF >0.10, THE RASIN IS CONSIDERED TO BE URBANIZED
NOTICEl WHEN TO>TF <> TC, IT INDICATES THAT THE ABOVE OES1CH CRITERIA SUPERCEDE COMMENTS ARE OK WHEN WATER ELEVATION l< LOWE* THAN THE GROUND ELEVATION AT
COMPUTATIONS MANHOLE
SUMMARY OF SEWER HYDRAULICS
SUMMARY OF HYDRAULIC GRADIENT LINE ALONG SEWERS
SEWER
ID NUMBER
53.00
13.00
114.00
610.00
1014.00
992.00
HAMHOLE NUMBER
UPSTREAM DNSTREAM
ID NO. ID NO.
5.00 1.00
1.00 1.00
1.00 14.00
(.00 10.00
10.00 14.00
96.00 2.00
SEWER
SHAPE
ROUND
BOX
ROUND
ARCH
ROUND
ROUND
REQUIRED SUGGESTED EXISTING SEWER SEWER SURCHARGED CROWN ELEVATION WATER ELEVATION PLOW
DlAUICHl DIA(KIGK) DIAtHIGH) WIDTH ID NUMBER LENGTH LENGTH UPSTREAM DNSTREAM UPSTREAM DNSTREAH CONDITION
1-30 • 1.50 1.50 l.SO 11.00 600.00 .00 95.86 91.06 91.11 92.93 UBCR
50.17 54.00 0.00 0.00 114.00 800.00 .00 91.06 . 92.91 91.92 92.85 UBCR
17.58 18.00 12.00 24.00 (10.00 120.00 .00 95.66 91.12 99.91 52.15 UiCR
36.96 43.00 0.00 0.00 1014.00 700.00 .00 91.12 92.91 92.95 92.99 UBCR
59.57 60.00 0.00 0.00 992.00 1210.00 1210.00 91.49 91.29 13.22 92.00 PR S'ED
DIMENSION UNIT* FOR ROUND AND ARCH SEWER AR.C
DIMENSION UNITS FOR BOX SEVER ARC FEET
REQUIRED DIAMETER - HYDBAUUCALLY DETERMINED;
"AVAILABLE
TOU A NEW SEVER, FLOW If ANALYZED BY THE SUGGESTED SEWER SIZE; OTHERWISE,
EXISTING SIZE IS USED
111 CHECK THE ADEQUACY OF EXISTING SEWER SIZE II
KCH£S '
SUBCR-SUBCRITICAL FLOW; PRSS'ED-PRISSUIIID-FLOWj JUHP'POSilBLB OCCURCNCB OF KYDftA
SUGGESTED DIAMETER • COHMVHCIALL' ULIC JUH?
SEWER DESIGN g P-FULL 0
ID NUMBER CFS CFS
53.00
73.00
314.00
610.00
1014.00
1499.00
S92.00
6.49
8.49
16.15
8.49
10.03
27.5S
10.65
10.02
9.12
19.72
9.07
14.16
37.90
11.34
DEPTH FLOW AREA VELOCITY
FCET SO FT FPS
1.23
1.20
3.08
1.16
2.14
2.20
4.99
1.80
1.81
11.61
1.46
6.16
6.37
19.62
4.70
4.70
1.19
5.80
1.63
4.33
1.56
FROUDE COMMENTS
NUMBER
0.7S
0.26
0.15
0.95
0.21
0.56
0.05
-OK
-OK
-LOW
-OK
-LOW
-OK
-LOW
FROUDE NUMBER-0 INDICATES A PKIfflURED FLOW OCCURS
SEVER SLOPE INVERT ELBVATIOM BURIED DEPTH COMMENTS
ID NUMBER UPSTREAM DNSTREAM UPSTREAM ONSTREAK
* (FT) (FT) (FT) (FT)
53
73
314
610
1014
1495
952
00 <
00
00
00
00
00
00
.41
.47
.01
.80
.03
.15
.02
93.16
94.36
66.56
94.18
89.62
89.46
86.48
91
91
86
91
99
87
66
32
56
44
62 '
46
96
29
1.79 5.
.42 i.
.94 1.
.52 4.
.38 1.
.52 6.0
.02 1.7
OK
OK
OK
«
OK
2 OK
1 NO
COMMENTS ARE OK WHEN THE BURIED DEPTH IS GREATER THAN THE BEOUIRED BURIED DEPTH
OF 2 FEET
-------
COMPARISON BETWEEN PREDICTIONS FROM UDSEWER AND STORM 1.
Predicted Peak Runoff
Sewer
Segment
53
73
314
610
1014
1499
992
in CFS
UDSEWER
8.49
8.49
16.21
8.49
10.03
27.58
32.65
STORM1
6.42
6.42
18.28
6.42
12.41
30.76
33.89
Pipe Size
in Inches
UDSEWER
Round Arch Box
21
18x18
54
18x24
42
42
63
STORM1
Round
21
18
54
21
42
42
66
FIGURE I
LAYOUT OF STORM SEWER SYSTEM IN CASE STUDY
- Manhole with 10 nu»b«r of 5
- Sever with ID nu«b«r of 53.
-------
MICROCOMPUTERS - THE STORMWATER MQDELLTNG FUTURE
by
G. R. THOMPSON, MIEAust
Senior Systems Engineer
Willing & Partners Pty Ltd
Canberra, A.C.T., 2605, Australia
B. C. PHILLIPS, Ph.D., MIEAust
Senior Engineer
Willing & Partners Pty Ltd
Canberra, A.C.T., 2605, Australia
ABSTRACT
The rapid development in recent years of increasingly powerful microcomputers is radically
changing the availibility and application of mathematical models. By 1985, former mainframe
computer stormwater modes including HEC-1, HEC-2, SWMM, DAMBRK, DAMS2,
ILLUDAS, RAFTS, RATHGL and CELLS were available on the IBM PC family of
microcomputers. In 1987 an increasing number of stormwater models for the Apple Macintosh
family, including HEC-2, SWMM, DAMBRK, RAFTS, RATHGL and CELLS are being
released. Clearly, the future of stormwater modelling, its widespread use and the manner in
which the profession responds to the challenge of responsibily implementing the power of
microcomputers is being governed by the rapid development of new generations of powerful and
inexpensive microcomputers.
The historical development of microcomputers is briefly reviewed and the processing
power of current microcomputers is presented. The results of three benchmark tests of the
SWMM, HEC-2 and RAFTS models are presented and conclusions are drawn. Future
developments in the microcomputer field are speculated upon and the implications for stormwater
modelling are discussed. It is concluded that the current implementation of stormwater models on
microcomputers heralds the future direction of stormwater modelling.
The work described in this paper was not funded by the U.S. Environmental Protection
Agency and therefore the contents do not neccessarily reflect the views of the Agency and no
official endorsement should be inferred.
10
-------
INTRODUCTION
Mathematical models are becoming increasingly important and more frequently used in the
analysis of complex hydrological processes. With the increase in microprocessor power former
mainframe computer stormwater models including HEC-1, HEC-2, SWMM, DAMBRK,
DAMS2, ILLUDAS, RAFTS, RATHGL and CELLS are now being routinely executed on
microcomputers.
It is no coincidence that the upsurge in interest and usage of mathematical models has been
directly linked to the development of mainframe computers and more recently to the development
of powerful and inexpensive microcomputers. Such a linkage is demonstrated in the history of
the SWMM model. This model was first developed in 1971 to analyse water quantity and quality
problems resulting from urban storm water runoff and combined sewer overflows and was
available solely as a mainframe computer model. Until recently, it was still perceived that
SWMM could only be executed on large computers. This perception is evidenced by the
following statement, issued as recently as 1984, on computer needs for SWMM (Huber et al (1)):
"A large high speed computer is required for operation of the SWMM, such
as an IBM370, Amdahl 470, UNIVAC 1108 or CDC 6600 Through
considerable efforts, users have been able to adapt different blocks of the
program to various mini computers, but only with extensive use of off-line
storage and increase in execution time."
The rapid development of powerful and inexpensive microcomputers from the mid 1970's
onwards culminated in the release of a microcomputer version of SWMM3 for IBM PC
compatibles (PCSWMM3) in early 1984. The implementaion of SWMM3 on microcomputers
has recently been further enhanced by the release of a version of SWMM3 for the Apple
Macintosh computers (MACSWMM) which utilizes recent advances in microcomputers including
the window environment, mouse control and expanded memory capabilities.
The implementation of the SWMM model on microcomputers is but a single example of
the ability of today's microcomputer to implement mathematical models which until recent years
were considered to be solely in the domain of mainframe computers. Clearly, the future of
stormwater modelling, its widespread use and the manner in which the profession responds to the
challenge of responsibily implementing the power of microcomputers is being governed by the
rapid development of new generations of powerful and inexpensive microcomputers.
MICROCOMPUTER DEVELOPMENT
The last decade has seen the turbulent growth of the fledgling microcomputer industry into
an industry which is impacting on everybody's lives. This growth has been led sometimes by
small innovative companies, sometimes by international giants. In 1977 the first true home
computers were released by Apple, Commodore, Radio Shack and other companies. Initial sales
growth were considered acceptable at the time with, for example, Apple taking 2 V2 years to sell
50 000 Apple n computers with 4 kbytes (4 kb) of memory. The following year the 5 W floppy
disk drive was announced, paving the way for future software development. This was followed
11
-------
in 1980 by the introduction of the 5V4" Winchester hard disk drive, the first large affordable mass
storage device.
In August 1981, IBM released the IBM PC and proceeded to sell 50,000 units in 7 months
exceeding all expectations of IBM. Within a year seven companies announced IBM "compatible"
computers. In 1983 the HP150 Touchscreen Computer and IBM PC!-XT were announced to be
followed in 1984 by the release of the Apple Macintosh and the IBM PC-AT. In January of 1984
Apple took 74 days to sell 50,000 Macintoshes and in April 1984, in a remarkable marketing
drive, sold 50,000 Macintoshes in only 7*/2 hours. This increase in the rate of sales clearly
shows the rapid acceptance of microcomputers at both the personal and corporate levels.
Nearing the end of 1985, networking and multi tasking were begining to redefine the use of
microcomputers, a trend which is expected to continue into the 1990's.
Prior to 1983 the microcomputer market was seen to be 75% recreation and 25% business.
By 1985 this ratio had been reversed to 80% business and 20% recreation. This trend can be
attributed to an intensive campaign originally by IBM and subsequentlyby IBM "compatible"
manufacturers to establish the IBM PC as the defacto industry standard for business personal
computers. What therefore will prevent IBM from continuing its domination in this field?
There appear to be two factors impeding the growth of the IBM PC market, its hardware
and its operating system limitations. The IBM PC computer is based on the Intel 8088 chip
which was originally developed from the 8 bit 8080 chip released in 1973. Since the Intel 8088
chip has only an 8 bit bus it must store a 16 bit value in memory in two parts causing it to operate
at half the speed of a chip with a 16 bit bus. The adoption by IBM of the 8088 chip in preference
to its predecessor the true 16 bit 8086 chip is a decision only IBM can explain. This decision has,
however, laid some ground rules for software development. A program cannot be contained
within data segments unless both the program and the data segments are in the same 64 kb of
memory. Nor can the stack be inserted in a data segment. Hence, at any one time the computer
can access only 64 kb of memory; a further limitation is that there can only be 1 Mbyte (1 Mb) of
memory on a chip. This memory is in turn further reduced by the 360 kb required when wiring
the board, leaving a maximum of 640 kb of user memory.
At first glance a 640 kb memory limitation does not appear to be a hinderance, however, it
is just this limitation which is preventing the IBM expansion into the graphics field. A 1200 x
1000 pixel bit-mapped display requires 150 kb of memory, and a grey-scale or colour display can
easily require 1.2 Mb or 2.4 Mb. This snares a large slice from an available 640 kb and makes
realistic CAD applications impractical. A large display also requires a more powerful processor to
refresh the screen, imposing a further limitation on real CAD for IBM PC family.
The IBM PC family is also linked integrally with the development of the MS DOS
operating system. In late 1980 Microsoft won the contract to supply an operating system for the
IBM PC. With the release of the IBM PC only approximately 6 months away and with
insufficient time available to develop a proprietry operating system Microsoft purchased SCP-
DOS from Seattle Computer Products. This low-powered operating system, which is seemingly
based on CP/M, the 8 bit industry standard and which showed no great advantages over Digital
Research's CP/M-86, was destined to become MS DOS Version 1.0. The release of MS DOS
Version 2.0 offered great improvements including new file access, memory management and a
hierachical directory structure. It seemed to draw heavily on UNIX ideas using the same file
structure; the development of "pipes" and "redirection" of both input and output and the power of
the "shell" iterative system level commands in the form of batch files. However, the original
rushed development of MS DOS Version 1.0 and the decision to maintain the compatibility of
12
-------
Version 2.0 with its predecessor has meant that the MS DOS system has never offered the
facilities required to exploit the full capabilities of the screen and serial and parallel ports.
It is now seen that computers based on the Intel 8086 chip are being superseded by
computers based on the Intel 80286 chip which may in turn be superseded before operating
systems are available to make use of this increased power. Alternatively an extremely powerful
and advanced computer based on the Intel 80386 chip may be the answer but the current non-
availibility of both application software and an operating system means that its power can not be
currently exploited.
In contrast, computers based on the Motorola 68000 chip family have a natural expansion
path. This path begins with the Motorola 68000 chip which runs at 0.6 MIPS (8 MHz) and
continues on to the to the Motorola 68020 and 68030 chips and the Motorola 68881 and 68882
floating point co-processor chips. By the end of 1987 the RISC (Reduced Instruction Set
Computer) technology Motorola 78000 chip, which will run at 20 MIPS (25+ MHz), will be
available to further enhance the power of this chip family. Currently there are more than 200
computers available which are based on the Motorola 68000 chip family including Sun, Apollo,
Domain, Hewlett Packard, Amiga and the Apple Macintosh.
The Apple Macintosh is an excellent example of the growth of computers based on the
Motorola 68000 chip. Its innovative WIMP interface (windows, icons, mouse and pull-down
menu), originally developed by Xerox but successfully implemented first by Apple, is now being
emulated by workstations costing $100 000 and more, and is even being emulated by IBM with
Microsoft Windows. To the scientific user it means that the former 128 kb home computer has
expanded into a microcomputer that is ideally suited to the implementation of mainframe software.
T)
§
c«
I
o
-a
u
CO
4_>
-------
The performance of the microcomputers can be also improved by installing third party
upgrades to the point where, for example, the Apple Macintosh can compete with mainframe
computers of the power of a DEC VAX 11/780. Ushijima and Foster (2) recently reported in
detail on four upgrades currently available for the Apple Macintosh. Of particular interest,
though, was the reported comparative performance of the four upgrades with the performance of
the VAX 11/780 and the VAX 11/725 computers.
A Whetstone program that employs double precision floating point calculations was one of
the tests employed to rate the performance of the various upgrades. The results of the Whetstone
benchmark are presented in Figure 1.
It is readily concluded from the results presented in Figure 1 that awesome power is
currently available to the Macintosh user and that the availibility of such power will increase as
new generations of microcomputers which are based on advanced Motorola chips are released.
STORMWATER MODELLING ON MICROCOMPUTERS
STORMWATER MODELS
The upsurge in interest and usage of mathematical models has been directly linked to the
development of mainframe computers and in recent years to the development of powerful
microcomputers. By 1985, former mainframe computer stormwater modes including HEC-1,
HEC-2, SWMM3, DAMBRK, DAMS2, ILLUDAS, RAFTS, RATHGL and CELLS were
available on the IBM PC family of microcomputers. In 1987 an increasing number of stormwater
models for the Apple Macintosh family, including HEC-2, SWMM3, DAMBRK, RAFTS,
RATHGL and CELLS are being released.
The development of the SWMM model is just one example of this process. The last major
revision of the EPA SWMM, SWMM3, was released in 1983 and coincided with the release of
the IBM PC. The burgeoning interest in microcomputers and their scientific applications was
reflected in the release in 1984 of an adaption of SWMM3 for the IBM PC and
compatibles.namely PC SWMM3 (CHI (3)). This version of SWMM represented both a
dramatic reduction of the complexity of SWMM3 data entry and a dramatic increase in the
availability of the SWMM3 model to users who previously did not have access to mainframe
computers. In 1987 PCSWMM3 has been joined by the recently released Apple Macintosh
version, namely MACSWMM (CHI (4)).
Software for computers based on the Motorola 68000 chip family is gaining in popularity
due to its ability to utilize the innovative features of this chip family to provide features including:
• windows / mouse control
• menu driven applications
• graphics capabilities
• simplicity of operation
14
-------
A typical example of the graphics capabilities available, for example, to Macintosh users is
the graphics output presented in Figure 2. This output presented in Figure 2 is output produced
by the Macintosh implementation of the RUNOFF module of SWMM3 (CHI (4)).
Results: Gutter -828
-,4.000
TIME (sec)
-.500.0
-.3000.
-.2000.
SUS.SOL
6000.
n2000.
TIME (sec)
COD
6000.
TIME (sec)
3.0000E+13
COLIFM
6000.
TIME (sec) n 600*0. ' TIME (sec) ' ' 6000. TIME (sec) 6000.
Figure 2 Sample Graphics Output from MACSWMM RUNOFF Module - INSTL3A Data Set
STORMWATER MODEL BENCHMARKS
The performance of three stormwater models, namely SWWM3, HEC-2 and RAFTS has
also been investigated on three families of microcomputer. The microcomputer configurations
and upgrades tested are listed in Table 1.
TABLE 1 MICROCOMPUTER CONFIGURATIONS AND UPGRADES
CODE
Al
A2
A3
Ol
O2
11
COMPUTER UPGRADE
Apple Mac Plus
Apple Mac Plus HD2000
Apple Mac Plus HD2000
Olivetti M24
Olivetti M24
IBM PC-XT
PROCESSOR
Motorola 68000 ( 8 MHz)
Motorola 68000 (12 MHz)
Motorola 68000 (12 MHz)
+ Motorola 68881*
Intel 8088
Intel 8088 + Intel 8087-2*
Intel 8088 + Intel 8087-3*
* The Motorola 68881, Intel 8087-2 and Intel 8087-3 chips are all co-processor chips
15
-------
A3 Ol
Configuration
Figure 3 SWMM Benchmark - INSTAL4A Data Set
~ o™ first performance test conducted was the execution of verification test data set
INSTL4A supplied with PCSWMM3 (CHI (3)). This data set simulates a single event (11/28/73)
for Lancaster, Pennsylvania drainage area. The only module executed is the RUNOFF module
The results of the INSTL4A benchmark test is presented in Figure 3.
~ 180°
I 1600
1400
1200
1000
800
600
400
200
0
I
-------
The second, third and fourth performance tests conducted were the execution of verification
test data set TEST1, TEST5 and TEST16 supplied with HEC-2 (U.S.Army Corps of Engineers
(5), (6)). These data set simulate a subcritical flow profile, special and normal bridge with
tributary flow and a split flow simulation respectively. The cumulative execution times of the
TEST1, TESTS and TEST16 benchmark tests are presented in Figure 4.
The fifth and sixth performance tesst conducted were the execution of the Wrights Basin
and Mogo test data sets supplied with RAFTS (Willing & Partners (7)). These single event data
sets simulate a 1000 Yr ARI, 1 hour duration storm for southern Canberra and a 5 Yr ARI, 12
hour duration storm for the southern New South Wales coast respectively. The cumulative
execution times of the Wrights Basin and Mogo benchmark tests are presented in Figure 5.
cfi
•a
CO
N»X
I
'fi
1800
1600
1400
1200
1000
800
600
400
200
0
Mac Plus
Olivetti M24
IBM PC-XT
Al
A2 A3 Ol
Configuration
O2
Figure 5 RAFTS Benchmarks - Wrights Basin and Mogo Test Data Sets
The results of the benchmark tests highlight both the practicality of conducting stormwater
model simulations on microcomputers and the processing power currently available. The
benchmark tests also highlighted the advisability of running stormwater models on a
microcomputer fitted with a floating point co-processor. The Apple Macintosh benchmark tests
also indicate the dramatic reduction in execution time to be expected as new generations of
microcomputers based on the the Motorola 68000 chip family are released in the future.
THE STORMWATER MODELLING FUTURE
The only certainty in the computer industry over the next five or more years is that the
breakneck pace of development of the last decade will not slow down. All else is speculation.
17
-------
It is speculated that desktop computers in the next few years will have a minimum 1 Mb of
random access memory (RAM) but more likely will support 8 Mb to 16 Mb of RAM. They will
operate at speeds in excess of 10 MIPS and will have much more mass storage. They will have
much better graphics capabilities with high resolution full page screens and will dump output onto
high resolution non impact quiet printers. When CD mass storage, voice driven input and optical
circuitry can be expected however is a matter for true conjecture.
With the ever increasing cost of human resources and the decreasing cost and increasing
power of computers we are already witnessing a shift in emphasis from hardware to software and
software support. It is likely that this trend will continue and it will therefore be found that a large
investment in software will control the upgrade path followed by computer users.
This upgrade path may be controlled by either maintaining continuously upgradeable chips,
which would require all computers to be based on the same chip family, or by establishing a
common operating system which could be overlayed on the hardware; such an operating system
would need to minimize the number of system dependant calls.
The former alternative is clearly impossible due to the wide range of proprietry hardware
currently in existence, however, a likely contender for the latter alternative is the impressive
UNIX operating system. Developed in 1969 by A T & T Bell Laboratories there are now at least
74 vendors including Sun, DEC, Hewlett Packard, Apollo, Microsoft (Xenix), Amdahl, Data
General, Burroughs/Sperry, NCR, IBM (AIX) and Apple with UNIX operating systems for
machines ranging from microcomputers to supercomputers. Of these, more than 50 vendors are
complying with die System V Interface Definition (SVID).
The independence of a particular hardware set or vendor means that the investment of a
software developer is protected since a multi-vendor standard minimizes risks and ensures the
continuity of software sales.
While the carefully nurtured growth of UNIX may not be evident when tracing the
chronological release of the software; Version 6 followed by Version 7, then 3.0BSD through
4.2BSD, System III, System V and now System V Release 2, it is well structured, with a
consistent and powerful philosophy of control and structure. Originally written for programmers,
UNIX utilities for debugging (eg. make, SCCS ) provide a productive environment for
developing software. For developers, applications written for a standard operating system (e.g.
SVID) and ported onto a compliant host only requires the software to be re-compiled and re-
linked.
The number of vendors already involved,the special interest groups, trade expositions,
journals and most importantly university curricula are encouraging the development of UNIX into
the standard operating system for technical computing.
Having established a standard operating system which will ensure a protected investment for
software developers what results can be expected?
The electronic paperless office obstinately refuses to arrive and the reams of output
accompanying computer programs is largely responsible. Graphics will undoubtedly play a key
role in output for the future in both presentation of information for the decision makers and as an
aid to the designer. An example of the first step in such a direction is the graphic output from
MACSWMM presented in Figure 2. A further step will be the integration of the results of
complex numerical models into CAD packages for plotting of construction details It suffices to
13
-------
say that the full impact of graphics on numerical modelling in the future is yet to be realized either
physically or conceptually.
The advent of the microcomputer has already seen the development of the "user friendly"
operating system with the Macintosh WIMP (windows, icons, mouse and pull-down menus)
interface often being declared to be an industry standard. But how far should this "user
friendliness" be transposed into the development of a numerical model? Should the model
become an "expert system" capable of being used by anybody ,ie. professional and non-
professional alike, or should the software retain some of the intricacies of parameter selection, for
example, and require the engineer or modeller to understand the model ?
Ideally, the expert system is the path to follow in the future. However, the cost of resources
necessary to develop true artificial intelligence in software is immense and anything less runs the
risk of still obeying the axiom - "garbage in / garbage out".
CONCLUSIONS
The advent of increasingly powerful microcomputers in the 1980's is inexorably changing
the method in which scientific calculations and mathematical modelling are undertaken. The
SWMM3, HEC-2, DAMBRK, RAFTS and CELLS models are just a few examples of the
downloading of former mainframe models into microcomputers in recent years. The
implementation of these and other models on microcomputers heralds the start of a new future for
stormwater modelling.
It is also evident that the impressive power of current microcomputers will be dwarfed by
the power of tomorrow's microcomputers. It is expected that furure developments in the
stormwater modelling field will utilize the increasing power of microcomputers to improve the
scientific basis of the models, improve the user friendliness for both input and output and will
integrate current models with sophisticated support packages.
ACKNOWLEDGEMENT
The assistance of staff in the Hydrology and Water Resources Unit of the Department of
Territories, Canberra, Australia in conducting benchmark tests is gratefully acknowledged.
The work described in this paper was not funded by the U.S. Environmental Protection
Agency and therefore the contents do not neccessarily reflect the views of the Agency and no
official endorsement should be inferred.
19
-------
REFERENCES
1. Huber, W.C., Heaney, J,P., Nix, S.j., Dickinson, R.E. and Polmann, D.j. Storm Water
Management Model Users Guide, Version III. EPA Project No. CR-805664, U.S.
Environment Protection Agency, Cincinnati, Ohio, 1984. 504pp.
2. Ushijima, D. and Foster, D.L. New ways to a faster Mac. Systems review, Macworld,
August, 1986. pp 88-94.
3. Computational Hydraulics Incorporated. PCSWMM3 Users Manual. 1-4: At the
University of Alabama, Tuscaloosa, Alabama, 1986.
4. Computational Hydraulics Incorporated. MACSWMM Users Guide. 1: Co-Published by
Willing & Partners Pty Ltd, At the University of Alabama, Tuscaloosa, Alabama, 1987.
22pp.
5. United States Army Corps of Engineers. HEC-2 Water Surface Profiles Users Manual.
The Hydrologic Engineering Center, Davis, California, January, 1981. 39pp.
6. United States Army Corps of Engineers. HEC-2 Water Surface Profiles Programmers
Manual. The Hydrologic Engineering Center, Davis, California, September, 1982. 30 pp.
7. Willing & Partners Pty Ltd. RAFTS, Runoff and How Training Simulation, Detailed
Documentation and Users Guide, Version 2.3. Willing & Partners Pty Ltd, Canberra,
Australia, 1986. 37pp.
20
-------
ENHANCING SWMM3 FOR COMBINED SANITARY SEWERS
by
William James T. Wayne Green
Cudworth Prof, of Hydrology Regional Mun. of Halton
University of Alabama 1151 Bronte Road.
Tuscaloosa, AL 35487-1468, USA Oakville, ON L6T 6E1,
Canada
ABSTRACT
Operating authorities responsible for the design and maintenance of sewer systems with
problems such as a large percentage of storm water, or frequent recurrence of basement flooding
during major storm events, or by-passes of excess flows into lakes and rivers, have identified a
need for a forecasting method. The method should compute the various components of flow
entering combined sanitary sewers.
Most computational models account for flow sources entering a sanitary sewer by
making coarse approximations of several flow sources in one aggregate flow calculation. This
paper examines procedures for forecasting the major components of inflow and infiltration from
surface and groundwater sources as well as sanitary flows from residential, commerical and
industrial sources.
The USEPA Stormwater Management Model (SWMM), in particular the version of
SWMM adapted for microcomputers (PCSWMM), was evaluated for forecasting combined
sewer flows. Two sewersheds in Oakville and Burlington in Ontario were used for this
assessment. New algorithms for the additional forecasting procedures are suggested.
21
-------
INTRODUCTION
The Water Pollution Control Federation 1970 (1) defines wastewater as:
"A combination of liquid and water-carried wastes from residences, commercial
buildings, industrial plants and institutions together with any groundwater, surface
water, and storm water that may be present."
Sewers carrying this type of flow are generally categorized as combined storm sewers.
The use of a single sewer to carry both storm and sanitary flows is no longer allowed, for
environmental and health reasons, although it is by far the least expensive of all the urban
servicing methods. Analysis of this type of sewer can be easily accommodated through the use
of hydrologic models such as the USEPA Stormwater Management Model (SWMM3) (Huber et
al. 1981 (2)).
It was a common method in many municipalities during the post war era of the 1950's
and 1960's to allow builders to service new subdivisions with sanitary sewers only. The
sewers were typically 200 mm to 300 mm diameter and designed to carry domestic wastes and a
small amount of groundwater infiltration.
Builders were allowed to connect foundation weeper pipes and, in some cases, roof
drainage pipes directly into the sanitary sewer. Storm drainage was provided by gutters,
roadside ditches, storm sewers and drainage swales. In the interests of minimizing construction
costs, the storm sewers were generally constructed at a minimum frost protection depth and too
shallow to intercept the basement weeper flows, which thus continued to outlet into the sanitary
sewers. Pipe deterioration further aggravated the problem by allowing groundwater to enter the
sanitary sewer system through pipe joints and broken or cracked pipes. Figure 1 illustrates a
sewer system with house connections typical of this servicing method.
Metcalf and Eddy (1979 (3) p.24) define the sources of stormwater which enter a
sanitary sewer as inflowAnfiltration:
"Infiltration:: Water entering a sewer system, including sewer service connections
from the ground, through such means as, but not limited to, defective pipes, pipe
joints, connections, or manhole walls. Infiltration does not include and is
distinguished from inflow."
"Inflow: Water discharged into a sanitary sewer system, including service
connections, from such sources as, but not limited to, roof leaders, cellars, yards, and
area drains, foundation drains, cooling water discharges, drains from springs and
swampy areas, manhole covers, cross connections from storm sewers and combined
sewers, catchbasins, storm waters, surface runoff, street wash waters, or drainage.
Inflow does not include, and is distinguished from, infiltration."
can
The type of sanitary sewer which is experiencing a high inflow/infiltration component
be defined as a combined sanitary sewer, as distinct from a combined storm sewer. Figure 1
22
-------
Figure 1
ROOF
DRAIN
MANHOLES
ROOF DRAIN
CATCH BASIN
STORM SEWER
SANITARY SEWER
TYPICAL CONNECTIONS TO STORM AND SANITARY SEWERS
SANITARY
VENT
nNGi
PERI^j
c=
t—.
TOILET, LAUNDRY ETC
Jfc
u
r-FLOOR
\ DRAIN
- ROOF DRAIN
, ROAD LEVEL
n
-SURCHARGE LEVEL
IN SANITARY SEWER
n
STORM SEWER
BASEMENT FLOODING FROM INFLOW/INFILTRATION
INTO SANITARY SEWERS
23
-------
illustrates a basement flooding condition which is one consequence of a combined sanitary
sewer.
In recent years, with the passing of the U.S. Water Pollution Control Act amendments
(1972) and the Fisheries Act and Environmental Contaminants Act in Canada, municipalities
are
required to separate the wastewater generated from urban development into two separate
systems:
one system to carry sanitary waste and a second system to carry storm runoff. The Water
Pollution Control Federation (1972 (4)) defines sanitary sewers and storm sewers as follows:
"Sanitary sewer - a sewer that carries liquid and water-carried wastes from residences,
commercial buildings, industrial plants and institutions together with minor quantities
of storm, surface and ground waters that are not admitted intentionally."
"Storm sewer - a sewer that carries storm water and surface water, street wash and
other wash waters or drainage, but excludes domestic wastewater and industrial
wastes."
Virtually all new sewer construction carried out in municipalities today requires sanitary
and storm flows to be collected in separate systems.
Thus we now distinguish between combined sewer systems designed primarily for
stormwater (combined storm sewers) and combined sanitary systems designed primarily for
sanitary flows (combined sanitary sewers).
Combined storm sewers can be readily analyzed using readily available storm sewer
design or analysis techniques. There is, however, a need for a method to forecast flows in
combined sanitary sewers with high inflow and infiltration. This paper reviews methods for
accounting for the various flow components which enter a combined sanitary sewer.
SCOPE OF THE PROBLEM
A recent court decision on a pump station discharge in British Columbia highlights the
concern over combined sewer discharges in lakes or rivers. The events surrounding the
discharge related to a pump station malfunction for a period of approximately 20 minutes during
which sewage was discharged into an adjacent creek. The District of North Vancouver was
charged and fined under The Fisheries Act for depositing a deleterious substance into an adjacent
water course through an emergency overflow. The overflow had been designed and approved
as part of the drainage system (Consulting Engineers of British Columbia, 1983 (5)).
Most municipalities have overflows designed into their systems at various low points to
prevent basement flooding during rainfall events. The Court decision to penalize the operating
authority, in this case the District of North Vancouver, may have wider implications for
approving authorities and the consulting engineering industry.
Environmental concerns relating to combined sewer overflows have been well
documented by various Government authorities and court decisions (Patinskas 1983 (6)). The
impact which excessive combined sewer flows have on homeowners and operating authorities
has not been as well recorded. In 1984 a review of treatment costs relating to combined sewer
flows was carried out for six treatment plants in southern Ontario serving approximately 2,000
hectares of urban development with a population of approximately 270,000 (Regional
24
-------
Municipality of Halton Annual Report, 1984 (7)). Using plant flow information, an
approximate annual volume of inflow and infiltration treated at 6 plants was determined.
Groundwater inflow and infiltration was estimated by comparing plant flow records for a low
groundwater season (such as mid January) to that of a high groundwater season (such as mid
March). To reduce the effect of contributions from homeowners and industries, the plant flow
record used in the comparison was the lowest point of recording, which occurs at approximately
2:30 a.m. This comparison provides an estimate of infiltration volumes in the sanitary sewer
system from groundwater conditions.
To estimate inflow from downspouts, window wells, and catchbasins connected
directly
to the sanitary sewer, a comparison was made between a dry summer daily flow rate and a storm
event flow rate for the same time of year. The comparison between the instantaneous flow rates
occurring immediately after the storm event gives an indication of the amount of direct inflow
from impervious areas to the sanitary system.
Estimating inflow and infiltration in this way for each of six treatment plants, it was
determined that an annual cost of approximately $400,000 per year is being expended on
treatment of groundwater and surface water inflow (approximately $1.50 per person per year).
In addition to the direct treatment cost, there are other capital cost losses which also affect the
true cost of groundwater and stormwater inflow and infiltration. These relate to the loss of
treatment plant capacity and sewer system capacity. The loss in capacity is more difficult to
assess and no doubt would exceed the operating loss estimated above.
As stated earlier the homeowner (or user of the system) also suffers through frequent
surcharging of the sanitary line which often results in flooded basement floor drains, also
illustrated in Figure 1. The resulting damage to household furnishings and, on occasion,
structural damage to floor slabs, results in numerous insurance claims. These costs are
reflected in litigation costs and higher premiums.
APPLICATION OF PCSWMM3
PCSWMM was applied to two sewersheds, A and B, which are considered typical of
most sewersheds with combined sanitary sewer systems. The computed results were then
compared to flows recorded by American Digital Systems (ADS, 1985 (8)). Eighteen
subdrainage areas were monitored in the City of Burlington Maple Drainage Area and 32
subdrainage areas were monitored in the Town of Oakville, South West Drainage Area. In
addition to flow data, American Digital Systems Inc. collected rainfall data. The purpose for the
data collection was to establish comparative flow results between the various drainage areas and
to pinpoint those areas which were displaying high inflow during rainfall events. The two
typical sites were selected from the data to represent the high goundwater infiltration condition
(Site A) and the high pipe inflow condition (Site B).
DESCRIPTION OF SEWERSHED A
Sewershed A comprises 126 acres of residential development including single family
homes and a small percentage of low density apartments as shown in Figure 2. The area was
serviced in the mid 1950's with vitrified clay pipe material. The condition of the sewer-pipes is
known from photographs taken on in-line camera. The photos indicate severe cracking around
25
-------
l-o
TOWN OF
OAKVILLE
a
op
I
SITE 'A'
LANDUSE AND SEWERSHED BOUNDRIES
-------
the crown of pipe and frequent displacement of joints between pipes. Groundwater infiltration
has been occurring for some time as evidenced by calcium build-up at the joints, and root
infiltration.
The as-constructed drawings contained in the Halton Region record system indicate that
most of the sewers throughout Sewershed A were excavated into a rock trench. At the time of
the initial servicing, only a sanitary sewer was installed; storm sewers were not constructed and
storm drainage was provided by roadside ditches in most areas. Limited storm sewers were
constructed as a means of draining the roadside ditches. Comments from the residents in this
area indicate a high groundwater table, particularly in the spring season. The soil conditions in
the area indicate a 3-6 foot layer of clay soil over limestone rock. A very early topographic map
of this area (1920) designates Sewershed A area a swamp or marsh. The high rock profile and
poor drainage of the subsoil material results in a high groundwater condition around most
basements and sewers.
DESCRIPTION OF SEWERSHED B
Site B comprises 97 acres of residential and semi-detached homes as shown in Figure 3.
The area was serviced between 1972 and 1975. Asbestos cement and clay pipe material were
used for the main sewer and house service connections. At the time of servicing, both storm
sewers and sanitary sewers were installed along with a curb and gutter urban class roadway.
The soil type typical of this area is a clay till classification. The water table elevation particularly
in the spring season rises to the elevation of the sanitary sewer and surrounding basements.
This high watertable condition is of short duration and remains below the sewer and basement
elevation for the remainder of the year.
Site B is subjected to a high sanitary inflow condition during rainfall events. A more
detailed investigation was undertaken by the Halton Region engineering staff to indentify the
source of the stormwater inflows. With the co-operation of adjacent residents, a program of
smoke testing and dye testing was undertaken. The testing program revealed 13 homes that had
at least one roof downspout connected directly to the sanitary sewer. In addition, one double
street catchbasin, a driveway catchbasin, and a hydro transformer vault were also found to be
directly connected.
DATA COLLECTION FOR SITE A AND SITE B
Pipe length, size and manhole data for Site A and Site B was collected from as-built
records provided by the Regional Municipality of Halton. Since Site A was constructed much
earlier than Site B, the accuracy of records and data is not as complete. Much of the data for Site
A was taken from a sewer network or operating map which has been kept up to date, and the
accuracy of these maps is adequate for the input requirements of PCS WMM.
The as-constructed plan and profile drawings for Site B were used to confirm pipe
lengths, pipe sizes and manhole information. In addition, the as-built drawings for Site B were
used for determining runoff areas for roadway and driveway catchbasin areas and transformer
vault drainage.
Information on the number of homes within each sub-sewer shed was obtained from
street index maps provided by the local area municipalities, Oakville and Burlington. Data on
house values, population information and statistics relating to the market value of homes,
percentage garbage grinders and average family income was obtained from the Planning
Department staff at the Regional Municipality of Halton. Information on water billing rates and
27
-------
Figure 3
CITY OF
BURLINGTON
SITE 'B'
LANDUSE AND SEWERSHED BOUNDIES
28
-------
water consumption rates was obtained from the Finance Department for the Region of Halton.
Data relating to sewer infiltration for the sewershed was also input to the PSWMM model.
The computed flows from PCSWMM were compared with flows recorded by on-site
monitoring equipment located at the outlet manhole locations for Sites A and B. The monitoring
equipment installed in the manholes comprised a pressure transducer located inside the first
length of pipe upstream from a pre-selected manhole site. The manholes selected for the
monitoring equipment were field-checked to confirm that the sensing equipment would be free
from flow turbulence due to a poorly benched manhole. The sensor detected the depth of water
flowing within the pipe and transmitted an analogue voltage signal to a small micro-computer
monitor hung under the manhole cover. The analogue signal was converted to a digital signal
and stored in a data collection system.
An ultra-sound velocity sensor was also provided at each monitoring site. The depth and
measured velocity were recorded at 15 minute time steps over the study period. Tipping bucket
rainfall information was also recorded at 15 minute intervals. The flow recording equipment
was closely monitored by field personnel. Each monitoring site was calibrated and velocity
checks were carried out to confirm the accuracy of the data being recorded.
COMPUTATION METHOD USING PCSWMM
Site A was chosen to represent a high groundwater infiltration condition. A three day
simulation period was selected, extending from 0:00 hour on September 18, to 0:00 hour on
September 21. This was a mid week time period and the observed flows display a uniform
diurnal pattern. Because the simulation involved infiltration due to groundwater, surface water
inflow from rainfall was not monitored. Submodels FILTH and INFIL within the Transport
Module were used in the simulation. The computed sewage flow was routed through the pipe
system to outlet manhole 67. An input hydrograph accounting for all upstream flows was input
at manhole 50. The dry weather sewage flow was computed in the FILTH submodel and input
at various manholes throughout the system.
A value for infiltration was selected for the INFIL submodel and a proportional amount
was entered at the upstream manholes of each subsewershed by the ENFIL routine. Based on an
"old sewer" criteria, the infiltration rate for the entire Site A sewershed was calculated to be 0.85
cfs.
For computing dry weather flows the FILTH model allows the use of water
consumption rates for computing dry weather flows or population statistics and land use
information. Simulations were carried out using both alternative methods and the results are
discussed below.
Site B was chosen as an example of a sanitary sewershed with surface inflow and sub-
surface infiltration. Input hydrographs were computed for each of the surface runoff areas
within the Runoff Module. These hydrographs were input at four manhole locations, 159, 153,
152 and 157. The storm hydrograph which was input at manhole 159 was combined with a
sewage hydrograph from the upstream sewage drainage areas. These hydrographs as well as
the dry weather flow computed for the various sub-drainage areas were routed through the pipe
system to the outlet point at manhole 158. A "new sewer" condition infiltration of 0.32 cfs
was input based on an average infiltration rate of 20 cubic metres per hectare per day, typical of
a new sewer infiltration rate.
29
-------
An average sewage flow rate of 90 gallons per capita per day was used for predicting the
dry weather flows for both Sites A and B. The computed and observed flows are presented and
discussed below.
SENSITIVITY ANALYSIS
To improve on the predictions, infiltration values were optimized, based on observed
flows during a 2:00 a.m. to 4:00 a.m. time period for each site. Flow rates in the early morning
hours have a minimum domestic sewage component and therefore provide a more accurate value
for groundwater infiltration.With the new infiltration values (Site A =0 .31 cfs and Site B = 0
.13 cfs), a significant improvement in the values computed by PCSWMM was obtained.
Additional computational runs were made for Site A using water meter records to
forecast the dry weather flow. Also, the population data for Site A was used as a third means
of predicting the sewage flows. As shown in Figure 4, the actual water meter record provides a
slightly more accurate prediction when compared to observed flows than do the other two
methods available to the user.
Additional predictions were made for Site A using varying input parameters . The
objective was to examine the sensitivity of each variable and thereby determine which
parameters must be selected with care and accuracy. The results are summarized in Table 1.
Table 1
SITE A - Sensitivity Comparison
Input variables INFIL and FILTH - PCSWMM
Variable
Infiltration
House Prices
Population Density
Persons Per
Dwelling
% Garbage
Grinders
Family Income
Diurnal Variation
Output Sensitivy to Variable Change
High
Medium
Low
From the computed results and sensitivity analysis , it can be concluded that the largest single
variable in the prediction of combined sanitary sewer flow is infiltration. In the present version
of PCSWMM, the user must select this value using his best judgment. There is no method to
predict the groundwater flow infiltration component for a combined sewer model. All other
variables contained in the FILTH algorithm are predictable and reasonable estimates can be
obtained through the various data sources within municipalities or operating authority records.
30
-------
Figure 4
a.o
3.
COMPflRISON-OBSERUED SEHftGE TO CftLCIlLflTEP PUF SHMM3
A-DHF FROM HATER HETER RECORDS
- MEASURED SEUAGE FLOWS
SITE A
31
-------
PROPOSED IMPROVEMENTS
GENERAL
The various sources of flow entering a combined sanitary sewer have been categorized
as follows:
a) sanitary sewage from domestic, commercial and industrial sources,
b) subsurface inflow and infiltration,
c) surface inflow.
Stormwater models such as PCSWMM3 are designed to compute surface runoff and can
be used to predict the storm water component of a combined sanitary sewer (referred to under
(c) above). Several changes and additions are now suggested for SWMM such that flows can
be determined for the sanitary sewer component described in (a) and inflow and infiltration flow
component described in (b). The sources of flow can be separated as shown in Figure 5.
For stormwater modelling, the design engineer usually first discretizes the drainage area
into various sub-catchment areas whose surface characteristics can be meaningfully estimated.
A similar approach should be taken to discretizing the sanitary flow catchment areas and
subsurface infiltration/inflow sewersheds. All significant sources of inflow and infiltration must
be accounted for. The three primary categories are:
a) Inflow from direct surface runoff sources,
b) Infiltration from existing groundwater sources,
c) Infiltration from surface water which percolates through the soil to the pipe system.
All significant sources of flow in a combined sanitary sewer should be disaggregated and
predicted separately. This will allow the user to predict each flow component and further to
examine alternatives for a sanitary sewer with high inflow and infiltration. The user will have a
greater ability to manage combined sanitary sewer flows and the varied complex solutions which
may be required.
The model must emphasize the need to account for all sources of flow entering the
sanitary sewer. In our approach here, no attempt is made to forecast these flows through the
use of the full St. Venant equations. Instead, simple mathematical expressions are used to
account individually for each source.
A continuous computation of groundwater storage is necessary to compute variations in
the depth of groundwater above or below the pipe system. The model should include site-
specific characteristics such as soil conditions, bedrock elevation, and surface infiltration. The
determination of surface inflow and subsurface inflow/infiltration should be based on different
discretized drainage areas. The sanitary flow calculation should be based on discretization of
residential, commercial and industrial land use areas. The modeler will, therefore, be required to
discretize three sets of areas:
1) sanitary sewershed areas,
2) surface runoff catchment areas directly entering the sanitary sewer (inflow),
3) subsurface catchment areas entering the sanitary sewer (infiltration).
32
-------
Figure 5
Flows Included in Prediction Modules SANF, DRAINF, Qinflow
33
-------
The user must therefore collect data on surface and subsurface characteristics as well as
land use information. All sources of flow in the upper sewershed should be determined in the
RUNOFF Module of any storm water model. The flows and solution alternatives in the larger
diameter sewer network can be analyzed in the TRANSPORT Module.
The prediction methods for sanitary flow and inflow/infiltration suggested here use
variables for which field data is available or readily obtained.
SANITARY COMPONENT (SANF)
A computational algorithm SANF is presented in Figure 6. The required data and steps
to be followed when using SANF are presented in Figure 7.
The land-use area (aO used as the basis for computing sewage flow is readily available
from aerial photography or land use maps. The land use areas are multiplied by a population
density (Pi) in the case of residential area, and a population density equivalent in the case of
commerical and industrial areas, to determine the net population (P) for the total study area. The
net population figure is multiplied by an average daily per capita flow value (q), to arrive at the
average daily sewage flow for the drainage area. Because of fluctuations in daily and weekly
flow patterns, a peaking factor (M) must be applied to the average daily flow to estimate a peak
sewage flow (Qs).
Typical factors for population densities for various land use types are given in Table 2
(Halton 1985 (8)).
TABLE 2
POPULATION DENSITIES FOR VARIOUS TYPES OF LAND USE
Land Use Type Persons Per ha
1 Single Family 55
2 Semi-Detached 100
3 Multi-Family (row housing) 135
4 Apartments (over 6 stories) 285
5 Light Commerical 90
6 Light Industrial 125
The average daily per capita sewage flow (q) exclusive of infiltration and inflow,
ranges from 225 to 450 litres per capita per day (M.O.E. 1984 (9)). A typical value used for a
southern Ontario community is 275 litres per capita per day (Moore 1985 (10)). This value,
determined from water use records, provides a best estimate for water entering the sewage
system, exclusive of water used for fire fighting, lawn watering and pipe loss.
If peak flow information from industrial and commerical sources is readily available,
the equivalent area calculation for these land uses may be omitted and the flows added directly to
the residential flow component to arrive at the peak flow for the drainage area.
The peaking factor (M) is the ratio of maximum to average daily sewage flow rates
(Babbitt and Baumann 1958 (11)). Sanitary sewage flow will typically follow a diurnal pattern.
Minimum flow occurs during the edttto morning hours when water consumption is lowest. The
first flow peak generally occurs in^Pcr morning when the peak morning water use reaches the
34
-------
Figure 6
Algorithm Schematic for Sanitary How (SANF)
ac, pr,
Determine Equivalent Population
P = aj.Pt
1000
Determine Peaking Factor
14
M = 1 +
4 + P0.5
Determine Sewage Flow
SANF = M * q * P
86.4
/
-SANF
Figure 7
Logic Schematic for Sanitary Flow
READ
ai
Pi
q
industrial, commercial and residential
areas IheO
residential population density and
industrial and commercial equivalent
population density (persons/ha)
unit flow rate
Determine Equivalent Population
Determine Peaking Factor
Determine Sanitary Sewage Flow SANF,
35
-------
treatment plant. A second peak flow generally occurs in the early evening hours between 7:00
p.m. and 9:00 p.m. This pattern is normally constant throughout the year, however, weekly or
seasonal variations may also occur. For example in areas which have a high industrial flow
component and definite shift times, or in areas which have a high percentage of seasonal users
e.g. in cottage or trailer park areas, such variations may occur. In all cases, for a sanitary sewer
with high inflow and infiltration, the diurnal variations are insignificant in comparison with the
seasonal variations experienced from high groundwater and rainwater inflow. For this reason,
diurnal, weekly or seasonal variations are not considered further here; the sanitary flow
component is determined on the basis of a daily peak flow.
To obtain the daily peak flow, a peaking factor (M) is applied to the average daily
flow. Various peaking factors have been developed from empirical relationships, for example
Harman (1918 (12)).
Alternatively, the user may prefer to develop a site-specific peaking factor, if sufficient
average daily flow and peak daily flow information is available.
INFLOW (QINFLOW) AND INFILTRATION (DRAINF)
To compute the inflow from direct surface runoff, existing storm runoff algorithms in
PCSWMM can be used. This flow determination will be referred to hereafter as QINFLOW.
The user must consider drainage area from roofs, pavement areas, grassed areas, etc. which are
directly connected to the sanitary sewer. A separate area determination must be made for each
sewershed. Surface water percolation and groundwater contributions are more difficult to
compute.
An algorithm for computing the groundwater elevation over a continuous time period is
presented in Figure 8. Also presented is a method for computing an infiltration flow (DRAINF)
into a sewer system given the groundwater elevation above the sewer. Figure 9 defines the
input variables and provides the user with a series of steps to follow when using DRAINF.
To calculate infiltration into the sanitary sewer a water balance must be carried out
for each integration interval - this gives the amount of groundwater present in the soil and its
head relative to the sanitary sewer. The terms which make up the groundwater accounting
model are as follows: The surface infiltration source (INF) as defined by El-Kadi and Heije
(1983 (13)) is the entry of water from the air side of the air/soil interface into the soil profile.
The amount of water which infiltrates may not directly contribute to the groundwater accounting
model. A percentage of the surface infiltration may be lost to evapotranspiration in the upper
soil zone (EVAPFR) or lost in deep percolation in the lower soil zone (DEEPFR). The
combined loss of surface infiltration is LOSSFR.
In the RUNOFF Module surface infiltration is computed by either the Horton or Green-
Ampt infiltration algorithms. Both of these algorithms subtract evaporation from rainfall depths
prior to calculating infiltration. In DRAINF, the computed infiltration from the Horton or
Green-Ampt algorithms is further reduced to reflect the loss of available groundwater due to
DEEPFR and EVAPFR. Both of these losses continue to occur after rainfall has stopped. The
losses of groundwater to deep percolation and evapotranspiration are treated as calibration
parameters.
Flow from external sources is difficult to compute without extensive knowledge of the
groundwater movement. For this reason, in this study, the external source flow is taken to be a
36
-------
Figure 8
Algorithm Subsurface Infiltration Flow (DRAINF)
'READ
S,de, Y0, DELT, Jc, f, A n
old tlie
INF, LOSSFR, EXTERNAL SOURCE
EQK
S2 * 2 * f
9 * A2 * k * DELT * (de + mo)
- de
EQK
DRAINF
(m0 - mi) * f* A,
tile
DELT
DRAINF
new old
- LOSSFR)-(ni0 - HIT ) + EXTERNAL
• SOURCE
old new
37
-------
Figure 9
Logic Schematic or Subsurface Infiltration to Sanitary Sewer (DRAINF)
READ
8
de
yo
DELT
K
f
:NF
LOSSFR
EXTERNAL
SOURCE
Atil6
sewer spacing
equivalent depth from impervious layer to
sewer
water table height above zero, datum
or impervious layer at start of each
time step DELT
time step
hydraulic conductivity
porosity
infiltration Into soil from Runoff Module
loss of infiltration to evapotranspiration
and deep perculation
inflow of groundwater
surface area of water table effected by
drain
Calculate water table height (mo) above
drain at start of time interval
no groundwater
flow to sewer
Using van Schilfgaarde's tile drain equation,
detrmine the new height of watertable at end
of time step (DELT)
EQK
- de
where
EQK >
1 + EQK
2 S2
DELT (de + roo)
Determine the groundwater flow (DRAINF)
entering the sanitary sewer
DRAINF - [m0- ra,] mt*tllel
DELT
Determine new water table-height Yj
for end of time step using a ground-
water accounting method
"YI - "Y0 4- 'INF 11 - LOSSFR) - (mo - nn)
External
f Source
Set new water table depth
and repeat calculations
for next time step
38
-------
calibration parameter. The external source parameter may also include lateral inflow from other
sewersheds. The external source is entered in units of inches of groundwater depth.
The final variable to be defined in groundwater accounting is DRAINF. This variable
represents pipe infiltration into a sanitary sewer through cracked pipes, joints and porous pipe
materials.
The hydraulic theory underlying the computation of DRAINF is considered to be the same
as a tile drain subjected to a groundwater head. An empirical relationship for tile drain spacing
with a falling water table condition was developed at North Carolina State College by van
Schilfgaarde (1963 (14)). It should be noted that the water table in van Schilfgaards's equation
refers to a water table condition in the immediate area of the tile drain or sewer being considered.
The equation may be rearranged so that the water table measured above the top of pipe (mi) can
be determined at the end of each time step (DELT).
The groundwater flow into the sanitary sewer (DRAINF) can then be determined from the
following relationship (Thompson 1979 (15)):
DRAINF =
DECT
where:
mo = watertable height above drain at start of DELT (m)
mi = watertable height above drain at end of DELT m)
f = soil porosity, (% of volume)
Atiie = surface area of water table affected by DRAIN (m2)
DELT = time step (sec)
The infiltration flow DRAINF is determined from the computed drawdown multiplied by
the drawdown area. The area is determined by the product of the length of sewer within the
drainage area and the width of drawdown on each side of the sewer. The width will depend
upon the amount of infiltration which is occurring into the sanitary sewer. This in turn will
depend upon the age or condition of the sewer and the amount of open joints, cracked pipes, and
porous materials.
It should be noted that the drawdown, mo-mi is multiplied by the soil porosity. The
drawdown relates to the soilwater depth within the ground profile. The soilwater depth may
comprise 80% soil material and 20% water for a soil having a porosity of 20% by volume. To
arrive at a cubic metre volume amount of water, the drawdown [momi] must be multiplied by
the soil porosity.
With all the variables defined, an accounting or water budget expression can be developed
as follows:
Ynew = Yold + Ay
where:
Ay = INFd - LOSSFR) - (m0 - mi) + External Source
f
The reader should recognize that Ay refers to a water depth in the soil profile. For this
reason, the infiltration value of the water budget equation is divided by the soil porosity value.
39
-------
Infiltration within the equation is measured in inches of rainwater and to convert this value to
inches of soil water, the infiltration value is divided by soil porosity (f). The amount of
drawdown which was computed in DRAINF will reflect the amount of soil water lost to pipe
infiltration. The final variable in the water budget equation reflects the increase in soil water due
to inflow from an external source, which should be input in units of soil water depth.
CONCLUSIONS
The loss of sanitary capacity to groundwater infiltration has become a problem for most
municipalities with sewer systems installed in the early 1900 era. For a variety of reasons, e.g.
construction methods, materials and quality control, many of the once separated sanitary sewer
systems must now be analyzed, maintained and managed as combined sanitary sewers.
The sources of flows in combined sanitary sewers are complex and varied. A broad range
of solution alternatives are available. Many of the solutions can be disruptive and expensive to
implement in a built-up urban environment. Sewer operating authorities and sewer design
engineers have identified a need for a method of analyzing and managing flows which are
occurring in combined sanitary sewers.
Most hydraulic models, SWMM being one of the most widely accepted and used, do not
provide a groundwater accounting procedure or a pipe infiltration prediction method. In the case
of SWMM, the user is required to input a "lumped" value for all pipe infiltration from the
various sources. The user's prediction is based on his personal experience or from suggested
textbook values or values used by other operating authorities. The combined sewer prediction
results for Site A and Site B demonstrate the sensitivity which this variable has on obtaining a
satisfactory output hydrograph. An error in judgment in selecting a proper infiltration parameter
will provide unsatisfactory results.
The groundwater accounting procedure and pipe infiltration procedure proposed in
DRAINF will allow the user to predict pipe infiltration and to compute groundwater fluctuations
on a continuous basis. The user will be able to disaggregate and compute the various flow
components which make up a combined sanitary sewer flow. The code, testing and verifying of
these new algorithms have not been carried out.
The review of SWMM has also indicated certain limitations in the present FILTH
algorithm contained in the SWMM Transport Block. The data required for this module is not
readily available from most municipal data bases. Further, the sensivivity of certain variables
contained within the FILTH algorithms do not provide any significant change to the output
results, e.g.diurnal variation parameter, % garbage grinders, house value, household income
value. An alternative sewer prediction method (SANF) which will generate only peak sewage
values using available data from a municipal data base source has been suggested as an
alternative computational algorithm for combined sanitary sewers. The diurnal or weekly
variation in sewage flow is insignificant in comparison to the total combined sewer flow. For
this reason, the new method predicts only peak sewage flow.
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.
40
-------
REFERENCES
1. Water Pollution Control and American Society of Civil Engineering 1970. Design and
Construction of Sanitary and Storm Sewers Manual of practice No. 9, pp. 1-331.
2. Huber, Wayne C., Heaney, James P., Stephan, Nix J., Dickinson, Robert E., and
Polmann, Donald J. 1981. Storm Water Management Model User's Manual, Version HI,
Project No. CR-805664, Municipal Environmental Research Center, USEPA.
3. Metcalf and Eddy Inc. 1979. Wastewater Engineering; Treatment Disposal Reuse, Second
Edition, McGraw Hill inc.
4. Field, R., and Struzeski, EJ. 1972. Management and Control of Combined Sewer
Overflows, Journal of Water Pollution Control Federation, Volume 14, No. 7, pp. 1393-1415.
5. Shore, A.G. 1983. Letter from Consulting Engineers of British Columbia to Member
Organizations.
6. Patinskas, J., Munno, T., Rehm, R., and Curtin, T. 1983. Combined Sewer Overflow
Loadings Inventory for Great Lakes Basin, USEPA Contract No. 68-01-6421, pp. 1-98.
7. Moore, R.W. J. 1985. Departmental Activity Report for Period of January 1, 1984 to
December 31,1984, Regional Municipality of Halton.
8. American Digital Systems Inc. 1985. Infiltration/Inflow Analysis of Burlington, Ontario
prepared for Regional Municipality of Halton, pp. 1-75.
9. Ministry of the Environment for Ontario 1984. Guidelines for the Design of Sanitary
Sewage Systems, pp. 1-44.
10. Moore, R.W.J. 1985. Design Criteria, Contract Specifications and Standard Drawings,
Regional Municipality of Halton, Design Criteria for Sanitary Sewers, pp. 1-18.
11. Babbitt, H.E., and Baumann, E.R. 1958. Sewage and Sewage Treatment, John Wiley &
Sons Inc., New York, 8th Edition, pp. 25-54.
12. Harmon, W.G. 1918. Forecasting Sewage at Toledo Under Dry-Weather Conditions,
Engineering News Record, No. 80, pp. 1233.
13. Ed-Kadi, A.I., and Vander Heijde, Paul K.M. 1982. A Review of Infiltration Models:
Identification and Evaluations, American Society of Agricultural Engineering, Paper No. 83-
2506, pp. 1-17.
14. Van Schilfgaarde, Jan. 1963. Design of Tile Drainage for Falling Water Tables, Journal of
Irrigation and Drainage Division, American Society of Civil Engineers, IR2, pp. 1-11.
15. Thompson, L.R., 1979. A Comprehensive Subcatchment Hydrologic Simulation Model
for Urban and Rural Applications MEng., Thesis presented to Waterloo University, Waterloo,
pp. 1-136.
41
-------
SEWERCADD
by: Michael H. Jackson and Jeff L. Lambert
Bovay Northwest Inc.
E. 808 Sprague ave.
Spokane, WA 99202
ABSTRACT
Sewercadd, developed by Bovay Northwest Inc., is a product of
the relatively recent availability of powerful and flexible
micro computer programs. Simply stated Sewercadd is a practical
application of a micro computer-based design model which has
been expanded to include drafting and database functions.
Sewercadd includes enhancements to the design model to assist
the design engineer both in design and construction of a sewer
system. These enhancements were developed around a database
management system. A high priority in the development of
Sewercadd was to provide consistency between the documents
required to design, bid and build a sewer system such as plans,
cost estimates, bid schedules and design calculations.
This paper describes the comprehensive system, Sewercadd, that
has been developed from these programs and used by the authors
to design and monitor the construction of several gravity sewer
systems in the Spokane area.
The purpose of this paper is twofold. The first is to encourage,
by example, computer design model users to take advantage of the
power and flexibility of the micro computer-based design models
available today. The second is to suggest that design model
developers provide flexibility in their models without
sacrificing the power and features now available.
42
-------
SEWERCADD
INTRODUCTION
Sewercadd, developed by Bovay Northwest Inc., is a product of
the relatively recent availability of powerful and flexible
micro computer programs. Simply stated Sewercadd is a practical
application of a micro computer-based design model which has
been expanded to include drafting and database functions.
Sewercadd includes enhancements to the design model to assist
the design engineer both in design and construction of a sewer
system. These enhancements were developed around a database
management system. A high priority in the development of
Sewercadd was to provide consistency between the documents
required to design, bid and build a sewer system such as plans,
cost estimates, bid schedules and design calculations.
This paper describes the comprehensive system, Sewercadd, that
has been developed from these programs and used by the authors
to design and monitor the construction of several gravity sewer
systems in the Spokane area.
The purpose of this paper is twofold. The first is to encourage,
by example, computer design model users to take advantage of the
power and flexibilty of the micro computer-based design models
available today. The second is to suggest that design model
developers provide flexibility in their models without
sacrificing the power and features now available.
THE SEWER DESIGN PROCESS
The design of a sewer system is an iterative process. The
preliminary system layout is done with limited field data and
approximate design flows. This information is then entered into
a computer model for calculation of the required size and slope
of the entire system. The model's results are then manually
drafted. As field information becomes available it is also
plotted, and conflicts with existing utilities are resolved by
the design engineer. The input to the model is then edited
based on this additional information about conflicts and any
revised design flows. The model is then rerun to refine the
design. The process is repeated until the design is finally
accepted. This is a time-consuming iterative process which
43
-------
requires a great deal of labor. The manual drafting results in a
significant time delay between the introduction of new
information and the ability of the engineer to analyze it in the
context of the design.
The Sewercadd design process (as shown in figure 1.) has the
same features and flow as traditional methods. Sewercadd,
however, simplifies the design process by integrating hydraulic
and hydrologic sewer design calculations with drafting and"data
management. Sewercadd manages all other project related data as
well, resulting in accurate and consistent production of
material takeoffs, estimates and bid schedules. The accuracy
and consistency of these documents helps create a bidding
environment that results in better and lower construction bids.
cici n r>A TA
rILLU Uf\ IA
DESIGN
CRITERIA
/-»/-! r* T r\ A TA
OC/o/ UAIA
*
E
W
R
A
D
D
FINAL
QUANTITIES
ESTIMATES
BID TABULATIONS
BID SCHEDULES
PAY REQUESTS
Figure 1. Sewercadd design process
SEWERCADD MODULES
The Sewercadd system is composed of three modules; the HYDRA
design module, the Database Manager, and the Drafter. The
modules accept input from a variety of sources and process then
pass along the data to the other modules of sewercadd to produce
all the required documents for a sewer design project as shown
in figure 2.
44
-------
DRAWINGS
1
QUANTITIES
DESIGN
DRAFTER
i
L
DATA
t
A/P/ / T<
^«\
HYDRA
I
Figure 2 . Module interaction
The building blocks for each of the Sewercadd modules are
popular micro computer programs;
HYDRA1,dBase III2 and AutoCAD3
The flexibility and programmability of both dBase III and
AutoCAD along with HYDRA's output data, made the complex and
sophisticated integration of the Sewercadd modules easier to
develop.
The Sewer Design Module, HYDRA
The sewer design module, HYDRA, is a flexible storm and sanitary
sewer system analysis and design program. Given design
criteria, a ground profile, and service area information, HYDRA
can provide a preliminary design of pipe size and vertical
alignment. HYDRA uses the traditional "peaking factor" concept
to generate sanitary flows but gives the user the option of
calculating storm flows by either the Rational Formula or by
using hydrological simulation techniques. HYDRA also has
functions to prepare cost estimates and financial analyses of
the sewer systems. An extensive set of data is output in a ASCII
text file.
?HYDRA is a registered trade mark of Pizer Assoc.
jdBase III is a registered trade mark of Ashton-Tate.
AutoCAD is a registered trade mark of Autodesk Inc.
45
-------
The basic unit in the HYDRA model is a link of pipe which is the
length of pipe between two manholes. HYDRA, after optimizing the
system design, outputs 75 pieces of information on each link.
This information includes all design information (pipe size,
flow, velocity, etc.) as well as cost information for that link.
Sewercadd moves this information from HYDRA into the Sewercadd
Data Manager were it serves as only part of the project
database.
Data Manager
The Data Manager manages all pertinent design information.
dBase III was used as the base to develop a collection of
routines to input, edit, create and manage information from a
variety of sources. The HYDRA model, as described above,
provides an extensive set of data on each link. In order for
Sewercadd to complete the design, other information is required
from the engineer. This information can be input through
standard dBase features and includes:
Existing utilities that cross the proposed design
Additional ground line information
Manhole numbering
Information used to combine the links output from
Hydra into drawings
Stationing
The Data Manager is the heart of Sewercadd. The Data Manager
stores all the project design data and can output this
information to the other modules of Sewercadd. The Data Manager
also can output reports for cost estimating, material takeoffs,
bid schedules and construction pay requests. This module helps
the engineer edit as well as enter data into the computer
quickly and efficiently and then combine it with the HYDRA
results.
While the Data Manager is key in the development of sewer
profiles, it also is used in the collection of additional
information to develop a complete cost estimate of the proposed
project. Although Hydra can develop quantities for pipe and
the associated excavation and paving, there are many more items
that will make up the entire estimate. These items such as
mobilization, tree removal, and drainage structures are entered
into the Data Manager to develop a complete cost estimate. The
list of items, when edited and refined, can then be used to
output a bid schedule that becomes part of the bid documents.
The Data Manager also has routines to prepare bid tabulations to
check each contractor's bid, as well as other routines to help
with construction management and calculate the payments due the
contractor through the construction phase.
The use of a consistent data base from design to project
closeout is a strong advantage of Sewercadd. For example, the
46
-------
database is automatically updated as the project design
proceeds, therefore any material takeoffs and cost estimates are
output from up to date information. Since the Sewercadd Drafter
uses the database as it's source of data, the drawings will be
consistent with the design.
The Drafter
1*10
1*10
1MO
1
I
!
MAOEUA
STREET
8
5.
l
1
1
1
1
1
1
1
1
1 J
.
CUB
I
S
i
j
»
I
§
s
?
•
W
L PAVEUE
1
V.
»
L
tt
PL
D
IT - 9EE
BOS
UMU
MCP. CLA!
' 222.44
- 2746
• AY
1
•LAN NO.
nw
H»^
t V
1
ft 9t i
UNION PARK ADDITION
u ( 41H TO 1 3RD - 3*3.42*
^ MH. 15
38.5'N C UBERTY /^-\
PARK PLACE (3)
t STA.37+5Z58 ,,;r
I
\
'/
»
L ' *
J-I-I
f
J
1C
ii
WM
3
I
C
1
1
1
J
i
*f
iK
||
uw
r*cp. a.
- 4101
L- 274S
1
I
<
iij
•or
». v. T~ — ._. _x*r A
""to S 3f>N3Av| |
-'--— K
i ' ^
1
1
»'SA — ^
SEE DWG. NO.
P-l-2
TO t3W>-149JO>
U
< « 4TH AVE.
STA. fe+75
1
.
*
\
^-
\
t 4TH TO t UBomwuy
3* V
1
i
1MO
1*10
iaw
I
f \
5 S
^-— 36*SA
SEE OWC. NO.
P-l-4
UKKTYPMK
\MH. 17
1« tt 41H AVE.
23*E « MAOEUA ST.
STA. 4O+1&01 BK.-
STA. 10+04.88 AHO.
B07
Nor
1
Figure 3. Sample plan and profile drawing
47
-------
The Drafter is the module that assists in the creation of the
drawings of the sewer system (see figure 3). It is a collection
of routines written in Autocad's Autolisp language.
The Drafter extracts information from the Data Manager to draw
the sewer profile including: groundline, pipe links, manholes
and existing utilities.
The Drafter accurately locates this information on the drawing's
grid and adds all required labels and notations. There are
routines to avoid text writing over other items already in the
drawing. The Drafter has a set of parametric drafting tools to
assist in the creation of the plan view of the proposed sewer
and the plan background showing all property lines, existing
utilities, and roadway curbs an sidewalks. The drawings (plan
and profile) can then be reviewed and the data edited in
Sewercadd's Data Manager and then the profiles redrawn until a
final solution is reached.
The Drafter also has a sophisticated method to automatically
organize the HYDRA output links into drawings. These routines
are based around a system diagram (see figure 4) drawn using
standard AutoCAD techniques. The sewer system diagram is
created with approximate manhole locations. Manhole numbers and
link identifiers are added during this process. The information
from the system diagram is passed to the Data Manager to assist
in determining drawing layout and connectivity of the links.
9NTO
SHARP
BOONE
Figure 4. System Diagram
-------
ADVANTAGES OF THE SEWERCADD SYSTEM
The Sewercadd system enhances the use of computer design
modeling techniques to ease the effort for the design engineer
not only in designing a sewer system but also in the other tasks
included in the engineer's list of responsibilities. As in any
computer program, a big advantage is the quick iterations which
can be run. However, in this case the iterations include not
only design calculations but also the production of graphical
output. This allows the engineer to review the proposed
alignment relative to utility interferences or other existing or
proposed improvements more quickly on either a hard copy plot or
on the computer monitor.
Using computer aided drafting methods allows all the
intermediate plan and profile plots to be color coded. These
plots make ideal check plots as often times there are complex
networks of existing utilities to be checked.
Because Sewercadd is an integrated system there is consistency
between the database functions, such as cost estimates and bid
schedules, and the graphical products (the plan and profile
drawings and details) produced by the system. Sewercadd's
comprehensive database insures consistent plans, cost estimates,
specifications, bid quantities, bid schedules and design
calculations. Retaining consistency is a persistent problem for
the practicing engineer as the design changes during the design
phase. Inconsistency between plans and other bid documents can
lead to expensive construction change orders.
In general, Sewercadd allows more time to be spent improving the
design and less on drafting, material takeoff and corrections.
The drafting quality is very high. Sewercadd is simple,
flexible and effective in improving the design and reducing the
construction cost. Additionally the computerized graphical
data is a valuable resource for the owner of the Project.
CONCLUSIONS
The advent of exceptionally powerful and fast micro computers
allows the design engineer to use powerful computer design
models, Cadd systems and database managers in a hands on
environment. These computer tools are tending to be flexible
with programming languages for customizing each application.
Because each engineer and design situation is unique no one
system can provide for all the needs. The answer for the
sophisticated user is to create his own system which meets the
specific needs that the engineer has identified.
There is a trend, by design model and computer programs in
general, toward allowing the user to customize as appropriate
for his own needs. To take full advantage of this power and
49
-------
flexibility will require two critical skills in the user. The
First required is a vision about what is possible and what will
work in their environment. The Second skill necessary the
complete understanding of the actual requirements of the task to
be automated. These skills enable users to turn powerful
programs into effective tools to serve their needs.
For the model writer, then, it may be more important not to try
to anticipate user needs but to allow the data to be manipulated
in a way to provide for them. The computer model HYDRA did not
provide the cost estimating functions suitable for our unique
situation, nor did it generate useful graphical output. The
model did, however, provide a data dump with every possible bit
of information included. Likewise, AutoCAD provided a
programming language, Autolisp, which could manipulate input
data into very complex output.
Finally, Sewercadd's Database Manager was developed around dBase
III and allowed input from several different sources and in
different formats which provided the nexus for design and
drafting functions.
It is anticipated the Sewercadd system will be further enhanced.
Currently Sewercadd requires a relatively sophisticated
computer user to interface between the design engineer and the
programs. The next step will be to provide a graphical method
of input which will interpret the user's requirements for sewer
location and capacity from a schematic basemap. This basemap
would be the basis for creating input data for the design
module. The engineer would be provided with a graphical method
to check this input data. The resulting improvement will allow
an engineer, who is a relative computer novice, to utilize the
entire system for design and drafting without the assistance of
a computer specialist. The engineer would then be left with a
powerful tool to deal with the drudgery of both design and
drafting while leaving all of the creative problems to be solved
with the imagination of the engineer.
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.
50
-------
CURRENT TRENDS IN AUSTRALIAN STORMWATER MANAGEMENT
by: A. G. Goyen, M.Eng., MIE.Aust.
Director, Willing & Partners Pty Ltd,
Canberra, Australia
ABSTRACT
Despite Australia's small population, it still suffers the same urban
sprawl as other densely populated countries. Over the last decade
significant advances in urban stormwater management have taken place.
Future directions in modelling techniques will revolve integrally with
the very rapid advances in computer technology; this in turn will open up
the ability for far more professionals to be involved in complex analysis.
The profession will need to address the problems associated with black box
analysis by persons with insufficient experience to question the results.
INTRODUCTION
Australia, with a land mass of 7,682,300 km^ experiences climatic
extremes and has a varied topography. There are rain forests and vast
plains in the north, snowfields in the south-east, desert in the centre and
fertile croplands in the east, south and south-west.
In hydrologic terms it is a land of contradictions. While the average
annual rainfall of only 465 mm represents the driest continent in the
world, mean peak annual floods, relative to mean annual runoff, are about
an order of magnitude larger than world figures, McMahon, 1982tl].
Additionally despite Australia's relatively low population density of
two persons per square kilometre representing a population of 16 million in
a continent 82% the size of North America, it is also one of the most
urbanised countries in the world with 70% of the population living in the
10 largest cities and 83% of the population classified urban. More than
six million people live in the country's two largest cities of Sydney and
Melbourne. See Figure 1.
51
-------
It is not surprising therefore that parts of Australia suffer the same
stormwater quality and quantity problems generally associated with far more
populated countries around the world.
DARWIN UOO
pop. 0.07m
IOOO.
BRISBANE
pop. 1.13m
U, CANBERRA //*. I*OO
pop. 0.2 5rn/7 SYONEY
pop. 3.3m
IOOO
HOBART
pop. 0.18m
ANNUAL RAINFALL (Median)(mm)
Figure 1. Australian representative rainfall and population statistics
COMPARATIVE STATISTICS
METEOROLOGY INPUTS
Stormwater management in Australia has to respond to the extreme
variations in hydrologic regime occurring across the country. Table 1
details typical meteorological inputs that directly affect urban stormwater
management.
As can be seen from Table 1, in a number of areas the combined affects of
relatively low design rainfall intensities, very intermittent rainfall
events and high evapotranspiration rates, make the estimation of design
loss rates and the consequential runoff extremely difficult.
WATER QUALITY
Water pollution emanating from non-point source urban stormwater
represents a growing problem in a number of Australian cities. Pollution
52
-------
TABLE 1. COMPARATIVE METEOROLOGICAL STATISTICS
1 hr Rainfall (mm)
Location
Adelaide
Alice Springs
Brisbane
Canberra
Darwin
Hobart
Melbourne
Perth
Sydney
Annual
Rainfall
(mm)
531
200
1157
639
1536
633
661
879
1215
Annual
Evaporation
(mm)
1700
3600
1900
1550
2800
1100
1550
1900
1600
1 in 1-yr AEP
12.0
13.6
36.0
17.0
50.0
11.0
15.0
15.0
31.0
1 in 100-yr AEP
40.0
46.4
90.0
50.0
95.0
30.0
45.0
35.0
85.0
loadings are in the same order as North American data for a range of
constituents, including total solids, suspended solids, nutrients and
bacteria. Table 2 indicates typical Australian annual constituent loads in
relation to some North American data.
Although nutrient washoff in urban stormwater at first sight would
appear to equate with North American data, total phosphorus for example
appears in Australia in far more particulate form. Australian soils are
generally much lower in organic content than northern hemisphere soils.
Based on local data, Lawrence, 1986t5'- available phosphorus to algae was
found to be only 30% urban runoff content plus 10% rural runoff content.
This has particular ramifications when examining the behaviour of local
lakes and problems of eutrophication.
RUNOFF
Recent research by McMahon, 1982 [1] and Finalyson et al, 1986[6] have
shown that Australian catchment runoff exhibit significant variations to
both North America and world data. Australian streams are more variable
than world rivers. Additionally, McMahon, 1982^l states that relative to
mean annual runoff, mean peak annual floods are about an order of magnitude
larger in Australian streams than world figures. However, when catchment
area is taken as the independent variable world streams produce larger mean
annual floods.
53
-------
TABLE 2. COMPARATIVE WATER QUALITY CONSTITUENT DATA
Australia (Canberra)*
kg/ha/yr
USA (Urban)
kg/ha/yr
Constituent
Rural
Urban
Durham,
Nth Carolina1'
Washington DC*
Sediment
Suspended Solids
Total Kjeldahl N
Total Phosphorus
293
15
.17
.12
2153
332
4.7
.61
5954
-
5.4
4.2
-
74
3.60
.66
* Willing & Partners Pty Ltd, 1986[21
t Colston, 1974W
* Randal, 19821*1
Analysis of 100-year flood data by McMahon suggests that Australian
catchments yield per unit area peak discharges that are about 60% more than
world values. Figures 2 and 3 indicate the typical findings of
McMahon, 1982 and Finlayson et al, 1986.
Cv
1.2
1.0
0.8
0.6
0.4
0.2
AUSTRALIA
WORLD,
"NORTH AMERICA
10* io» 10* io9
CATCHMENT AREA (km2)
WORLD
10' 10* 10s 10* 10s
CATCHMENT AREA (Km2)
Figure 2. Coefficient of Variation
of Annual Flows versus Catchment
Area
After Finlayson et al, 1986
Figure 3. 100-year Floods Expressed
as Ratios of Mean Annual Floods
and Area
After McMahon, 1982
54
-------
It is likely therefore that traditional flood frequency extrapolation
techniques based and tested on northern hemisphere data may be questionable
without detailed examination using local data.
URBAN STORMWATER MANAGEMENT CRITERIA
WATER QUANTITY
Over the last decade there has been a significant shift in Australia
to catchment wide stormwater management and in recent years to the
integration of this with wider urban planning inputs. Based on a number of
judicial findings the legal responsibilities for flood damages associated
with a "duty of care" has forced local councils and state government
authorities to think out the wider issues of stormwater planning and
advice.
In general a minor/major stormwater system has developed with a piped
urban system to take flow peaks between I and 10-year AEP and a surcharge
system via roads and formal floodways to take rarer events up to and
including the 1 in 100-year AEP flood. Where retention basins of
significant size have been included into a drainage system it has been
common to size emergency spillways and embankment protection to frequencies
of between 1 in 1000-year AEP and the MPF.
In some instances the setting of such prescriptive flood frequency
levels has led to over protective measures, excluding the development of
otherwise valuable land. Additionally the adherence to, say, the 1 in 100-
year AEP as a guide to flood protection in older areas under redevelopment
has led to adverse social reaction.
In the state of New South Wales a recent Floodplain Development Manual
(NSW Government, 1986)[7] has been issued to assist councils in developing
plans for the management of their floodplains.
The policy takes into account that "flood liable land is a valuable
resource and should not be sterilised by unncessarily precluding its
development". Central to the policy is the requirement that all
development proposals be treated on their merits.
This policy places considerable responsibility on individual councils
to carry out adequate catchment wide flood studies to base sound management
principles on flood hazard, economic factors, environmental planning and
development control. This departure from a standard flood frequency
requirement such as the 1 in 100-year AEP is likely to, in the shorter
term, involve difficult decisions. In the longer term it is expected that
the preparation and implementation of overall management plans will
incorporate the merit approach to its fullest extent.
At an individual development level a range of acceptance criteria is
usually applied to minimise both nuisance flooding and major hazard from
flooding of roadways and buildings. Table 3 indicates a typical set of
acceptance criteria being applied to urban areas within Canberra in the
Australian Capital Territory.
55
-------
TABLE 3. TYPICAL DESIGN ACCEPTANCE CRITERIA - CANBERRA*, A.C.T.
Surface Flow Regime
Subsurface Flow Regime
Situation
Limiting Criteria^
Situation
Return
Period
at Return
v.d Period
Roads - general
Access to emergency
facilities
Pedestrian trafficable
floodway
Other floodways
Other areas and§
Hospital and defence
facilities
< .2 < .4
< .2 < .4
< .75 <1.0
Neg Neg
< .1 < .1
100
100 +
100
100
5
100
Notation
v - velocity of flow (m/sec)
d - depth of flow (m)
Return Period - expressed in years
Roads - 1-lane clear^
Minor 2
Collector 5
Distributor 10
Ordinary arterial 20
Inter urban arterial 50
Access to emergency
facilities 100
Urban development*
Buildings and
trafficable areas to
be drained to prevent
damages to return
period specified
Residential -
- Low density 5
- Medium density 10
- High density 20
Shopping and
commercial -
- Local 10
- Regional 20
Industrial -
- Light 20
- Heavy 50
Hospitals and
emergency service
areas 100+
Notes:
* Limiting criteria set for Canberra region only. In other areas these
would need to be adjusted to local rainfall regime.
§ Limit set to restrict surface flows being routed through private
property.
* Criteria directly related to traffic density. Should be adjusted where
situation warrants.
* The surface flow regime should be sized to take into account partial
pipe failure through blockage wherever this could possibly occur.
t In all cases the affects from flows in excess of the proposed limiting
criteria should be minimised wherever possible.
56
-------
The stated criteria places as much importance on the control of
surface flows resulting from infrequent storm events as the removal of
frequent flows from on and about the urban pedestrian and vehicle transport
network.
The table sets three basic limits, being:
(a) the velocity-depth limit that has been found in association with
the depth of flow to govern the stability of vehicles and the
ability of pedestrians to "walk out" of flood flows,
(b) the depth limit, and
(c) the return period limit which is the economic criteria, for which
damages should not be occasioned.
Resulting from the approach described in Table 3 a constant pipe
design frequency is not followed. To maintain acceptable road surface
flows for example it may well be possible and sometimes necessary to either
decrease or increase the design frequency of the pipe system over
particular reaches of the network.
WATER QUALITY
Since the enactment and implementation of environmental protection
legislation in Australia, in the early 1970s, by the six states and by the
Federal Government there has been growing concern in Australia about the
actual and potential impacts upon receiving water bodies of polluted urban
runoff. Noteworthy studies have been carried out for the two principal
inland cities in Australia, ie, Canberra and Albury-Wodonga, by Cullen,
Rosich & Bek, 1978'8] and by Gutteridge, Haskins & Davey, 1974'91.
It has been found that phosphorus is the limiting nutrient for a range
of Australian freshwater lakes. Bliss, Brown & Perry, 1979 [10) reported on
investigations of the pollution potential of urban runoff in Sydney. They
reached the conclusion that the pollution potential of urban runoff in
Sydney was high and that both the more commonly determined pollutants such
as non-filterable residue, bio-chemical oxygen demand and nitrogen and
phosphorus forms, oils and polycyclic aromatic hydrocarbons may cause
severe degradation of certain Sydney receiving water bodies.
In recent years investigation, planning, design and implementation of
water quality control schemes have been carried out in Canberra,
Goyen, et al, 1985[11) and Lawrence & Goyen, 1987t12!, to combat future
water quality degradation due to continued urbanisation. Up until recently
however point source control and the treatment of wastewater has taken up
the majority of the country's resources in this field. Monitoring within
the A.C.T. has shown that a change from rural to urban land use has
entrained a seven to tenfold increase in the level of export of a range of
runoff constituents.
Receiving water quality objectives adopted for planning in Canberra,
Lawrence 1986[S1 have been based on the protection of designated uses of
the waters of lakes and streams and aquatic ecology. Ecological criteria
determined by bio-assay techniques have been found to have little relevance
57
-------
to local fauna and the quality of lakes and streams in the region as to
date they have generally been well within broad water quality objectives.
It has been general policy to closely monitor changes in lake and stream
ecology as the primary basis to reviewing pollution control strategies.
Rather than restricting absolute constituent concentrations from new
developments to downstream receiving waters, acceptance criterion in the
A.C.T. has relied more on maintaining rural water quality levels, or as
near as possible, after urban development has taken place.
CURRENT MODELLING TECHNIQUES
WATER QUANTITY
As with overseas practice there has been a progressive trend over the
last 10 years towards computerised analysis and design. This has occurred
at both the minor stormwater reticulation level as well as on the flood
mitigation level.
Minor Systems
For piped drainage systems the main analysis techniques still revolve
around the Rational Formula, Messner & Goyen, 1985113' although ILLUDAS,
O'Loughlin & Mein, 1983'14] and SWMM, Carleton, 1983115!, Attwater & Vale,
1986 [16^ have recently been applied to a limited number of Australian
catchments.
Table 4 indicates a range 'of urban models that have gained at least
rudimentary use in Australia.
ILLUDAS developed from the TRRL Method by Terstriep & Stall, 1974[17]
has been recently further developed in Australia by O'Loughlin, 1986[18].
O'Loughlin & Mein, 1983[14) stated that ILLUDAS even with recent Australian
improvements would not make it suitable for detailed pipe design, including
such considerations as pit energy losses and cover depths.
WASSP is a program suite developed by the UK National Water Council,
1981[191 which offers similar capabilities to ILLUDAS although to date has
not been widely used or tested on Australian systems.
SWMM is a comprehensive program suite supported by the US EPA,
Huber et al, 1981121] that concentrates on urban piped systems. It
provides full unsteady flow and backwater effects within the system through
the use of the EXTRAN block. •Overflow rerouting and limited inlet capacity
consideration is not presently covered. Carleton, 1983[151 when attempting
to model an existing catchment in Sydney for a range of severe storm events
was only partly successful, since the model could not take into account the
extensive blocking and restrictions associated with inlet pits which were a
major cause of the flooding.
PIPENET is a propriety drainage design model developed by
Bloomfield, 1981[21] based around the Rational Formula that is offered as
an interactive tool to design new piped systems. The model does not
directly address variable pressure change coefficients, surcharges or
surface flow rerouting.
58
-------
TABLE 4. CHARACTERISTICS OF AUSTRALIAN USED URBAN STORMWATER MODELS
Models.
Description
ILLUDAS WASSP SWMM PIPENET RAFTS/RSWM RATHGL
H H/H H/H H/H H H/H
Uses
Design of New Systems x x
Analysis of Existing System x x
Water Quality Analysis
Hydrology
Rational Method
Modified Rational Method x
Simple Hydrograph Routing x x
Complex Hydrograph Routing
Hydraulics
Simple x* x
Complex x
Empirical HGL Analysis
Solution of St Venant Eqns x
Energy Loss Estimates
Colebrook-White x
Mannings Equation x
Static Pit P.C. Coeff. x
Dynamic Pit P.C. Coeff.
x
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Other Features
Surcharge Allowed
Overflow Rerouting
Limited Inlet Capacity
X* X* X
X
X
X
X
X
P.C. - Pressure Change
H/H - Both hydrologic/hydraulic model
Notes: t total area only
* pipe flow based on bedslope as
the friction slope
H - Primarily hydrological model
* total and critical area
surcharge pools or is lost
from the system
RAFTS, Goyen[22] while providing detailed hydrologic input to complex
stormwater pipe and channel systems, covers only limited pipe hydraulics
roughly equating to the TRANSPORT block in SWMM.
RATHGL is a Rational Formula based hydrologic model with extensive
pipe hydraulic routines, Messner & Goyen, 1985'13^ . The main features of
the model include: network outlet (backwater) control, pit and channel
surcharge facilities, surface flow rerouting, limited inlet capacity to
pits, pipe and pit energy losses. Figures 5 and 6 indicate the general
arrangement of RATHGL.
59
-------
vv
\
COMPUTATION OF PIPE
CAPACITY AND/OR PIPE
FRICTION SLOPE
Figure 5. Typical network for RATHGL
CFN > CRIT. AREA
TOTAL FLOW TO
NODE £ SURFACE
PLUS PIPE
.'TAKEOFF' FOR
' NEXT REACH
£S.2l.l.,PITSHAPE,ETC\
Do Oo Do I
DIAMETER Du
K* • WATER SURFACE LEVEL
COEFFICIENT
Ku » PRESSURE CHANGE COEFFICIENT
Ku MAY NOT NECESSARILY ' K»
NODE OS
NODE N
Figure 6. Typical RATHGL single reach from pipe network showing two flow
limit states
6C
-------
To date the RATHGL type of model has gained acceptance over more
complex models such as ILLUDAS and SWMM in routine pipe design and
analysis. This is thought to be primarily due to their relationship with
the familiar rational formula approach and their flexibility in handling a
wide range of real like situations. Additionally, the hydraulic
algorithms, although numerically complex, follow similar techniques to that
previously carried out by hand, however with the advantage of accelerating
the process of examining alternatives many fold.
MAJOR SYSTEMS
Once urban systems leave the piped network to join the major
channel/floodway/river systems the use in Australia of rainfall/runoff
routing models has held sway.
The most widely used models in Australia are RORB, Laurenson & Mein,
1983'23) and RAFTS, Goyen and Aitken, 1976f24J, Goyen, 1983I22]. Both models
consider watershed wide analysis involving streams and reservoirs or
retention basins and allow the analysis and design of a wide range of flood
mitigation options.
In RORB the model considers the whole catchment as a unit and
describes internal concentrated storages related to a minimum of 5 to 20
internal subcatchments subdivided on watershed lines plus concentrated
special storages to represent retention basins and additional stream
routing effects.
All storage elements within the catchment are represented via the
equation
S = 3600k Qm
where k = represents a storage delay parameter and m represents a
measure of the catchment's non-linearity.
When m is set equal to unity the catchment's routing is linear.
The storage parameter "k" within the general storage equation is
modified to reflect not only the catchment storage but also the reach
storage by the form:
k = kc.kr
where m is a measure of the catchment's non-linearity, and
kc is an empirical coefficient applicable to the entire
catchment and stream network, and
kr is a dimensionless ratio called the relative delay time,
applicable to an individual reach storage. kr thereby is
modified to reflect the nature of the channel reach.
RORB has been used extensively throughout Australia on a range of
rural and urban catchments. Calibrated values for kc and m for a large
61
-------
number of regions have been developed throughout Australia which have been
used to estimate flows on relatively ungauged catchments.
RSWM was originally developed in the early-1970s, Goyen & Aitken,
197g [24] jointly by Willing & Partners Pty Ltd and the Snowy Mountains
Engineering Corporation. The Runoff Analysis and Flow Training Simulation
(RAFTS) is a proprietry model developed from RSWM by Willing & Partners Pty
Ltd, Goyen, 1983[225 to include separate routing of impervious and pervious
areas, continuous loss modelling, pipe/channel analysis, detailed retention
basin analysis including hydraulically interconnected schemes. The model
is flexible in that, as well as handling small urban catchments, is equally
comfortable with very large rural river basin analysis. RAFTS is further
described diagrammatically in Figure 7.
Each subcatchment model is represented by a series of ten non-linear
concentrated cascading storages based on the works of Laurenson, 1964'25! .
Within RAFTS each of the subareas in a subcatchment is treated as a
concentrated storage with a storage/discharge relation:
S = k(Q).Q
with
K(Q)
B.Qn
n and B represent the catchment non-linearity and subcatchment storage
delay coefficient respectively and roughly equate in relative terms to
RORB's m and k parameters.
BOUNDARY
SUB-CATCHMENT
NODE
MODEL STORAGE
SUB-CATCHMENT
INFLOW
NONLINEAR
CONCENTRATED
STORAGES
RUNOFF ROUTING MODEL
HYDROGRAPH MODULE
FOR EACH SUB-CATCHMENT
CHANNEL ROUTING
MODULE WITH
LATERAL INFLOW
ROR8
RAFTS
Figure 7. Diagrammatic Representation of RORB and RAFTS
-------
The number of subcatchments used in RAFTS is not important as each
individual subcatchment is represented by a complete model. RORB however
requires a minimum of 5 to 20 subcatchments to represent a valid catchment
model.
The RAFTS model incorporates more sophisticated loss routines than
other Australian models. In addition to an initial loss/continuing loss
rate option the model allows the use of the infiltration, wetting and
redistribution algorithms of the Australian Representative Basins Model,
Black & Aitken, 1977'26' and Goyen, 1983[22].
A further option that is provided with RAFTS is the SDLM Module which
is a stochastic/deterministic loss model that links the probabilities of
rainfall and soil moisture to estimate rainfall excess and runoff frequency
curves without the need to use traditional loss modelling techniques,
Goyen, 1983t22]. Figure 8 describes this option diagrammatically.
DNENS«ONLESS o
DESIGN 0
STORM
TEMPORAL H
PATTERN J
f\
h 0
OETERMINISTIC\
RAINR)LU^S-\
EXCESS OR PEAlOv
RMMMXI
UNDFF
\\\X
O
§
^
100 99
M
UPPER SOL
WETNESS »GEX
FREQUENCY,/
DISTRIBUTION
.
/
99
00
Re
r
s
2
<
1
RaWfftLL FREQ
DISTRIBUTION
^^^^"'~
S^
9999 0
' MW.DURATION RAtNBU. INTENSITY •>/. PROS OF EXCEED % PROB OF EXCEB
(a) ., (b)
yiic
SJ2
j?^
*
2s
ZK
(c)
A
RAINFALL
CONDITION
Al
PROBABILITES
OF Mk
X
(e)
I Pe
PL 1 .T\-
O-99mm IO-l99inn
£
f
1 n
r^S Ir^n I
EXCESS
RAINFALL
FR
-.Out
^CY
DISTRIBUTION
^
^
^-^
,(d)
1
Rc
CONDITIONAL
RAINFALL
PROBABILITY
ARRAY
(f )
_ Qo
i
RUNOFF
FREQUENCY
DISTRIBUTION
-^^^
20-299nm 3O-399mm
DIMENSIONLESS EXCESS RAIN
TEMPORAL PATTERN
(9)
9899 OOI
% PROS OF EXCEED
(h)
9999 OOI
% PROB OF EXCEED
(I)
Figure 8. Diagrammatic Representation of the RAFTS/SDLM Module
RAFTS, unlike RORB, is more closely related to SWMM being a true
network routing model with its basic element being the subcatchment
providing input into the channel network system.
Subcatchment outflows from RAFTS are further routed through the
channel network and retention basins by separate Muskingum-Cunge, Price,
1973t281 and level pool routing modules respectively. Separate routing of
pipe flows under channels and retention basins is also provided.
63
-------
In RAFTS n is set at a subcatchment level where m in RORB relates to
the total catchment.
Similarly B in RAFTS is defined separately for each subcatchment to
represent other than area considerations such as urbanisation, slope,
catchment roughness, etc. k in RORB is defined on a catchment wide basis.
In RORB m is varied together with k to calibrate the catchment
outflows to gauged data which means that the effects of both channel/stage
and catchment storage over the entire catchment are reflected in the
adopted k and m values.
RAFTS in contrast to RORB, which uses regionally derived k and m
values for ungauged catchments, usually sets n equal to a constant (0.715).
B is varied for each subcatchment based on measured characteristics and
developed regression relationships where B is a function of (total area,
slope, impervious area and surface roughness).
Over recent years research into catchment non-linearity, particularly
in relation to large runoff events, has indicated that response becomes
more linear with larger events, Bates & Pilgrim, 1983127! .
To allow for the rare event modelling, RAFTS allows a variable n
relationship relative to subcatchment discharge/stage.
Varying linearity problems is likely to be of less a problem with
RAFTS than RORB as only subcatchment routing is effected. In RORB the
effects of linearising channel/storage with increasing flows is a
significant factor in overall catchment routing as the combined effects
have to be absorbed in the m value.
Provided subcatchments and consequential storages are kept relatively
small compared to the effects of overall channel storage routing and
reservoir storage the n value selected with RAFTS should be relatively
insensitive to changes in flow regime.
WATER QUALITY
Modelling techniques to predict pollutant build up and wash off such
as STORM, and SWMM, have only been used to a limited degree in Australia.
The major modelling techniques have to date revolved around regression type
algorithms to relate pollutant exports to daily runoff. The relationships
shown in Table 5 have been based primarily on correlations with individual
storm analysis for a range of monitored urban and rural catchments in
Canberra.
In stream transfer models have, to date, been based around relatively
simple gradually varying conservations of mass flow type techniques
incorporating decay functions to account for loss in constituent mass flow
with flow downstream.
Lake Response Models have mainly revolved around an adaptation of the
Vollenweider Lake loading model, Lawrence & Goyen, 1987 to estimate the
effects of eutrophication abatement programs.
64
-------
TABLE 5. CORRELATION OF POLLUTANT EXPORTS WITH RUNOFF R (mm/day)
(AFTER LAWRENCE & GOYEN, 1987)
Catchment
Land Use
Urban
Rural
Coarse
Sediment
(kg/km2)
1 000 R1-4
400 R1-1
Suspended
Solids
(kg/km2)
200 R
20 R
Total
Phosphorous
(kg/ km2)
0.39 R0-8
0.115 R0-57
Total
Nitrogen
(kg/ km2)
3.0 R0-84
0.3 R1-6
4k
E-Coli
(Count/km2)
30 000 R0-9
500 R0-9
TRENDS IN URBAN STORMWATER MANAGEMENT
WATER QUANTITY
In the last ten years there has been a significant improvement in
catchment wide management techniques in Australia. New urban stormwater
systems are now generally designed for the minor piped system together with
a major surface flow floodway system. Additionally, future catchment
development is now taken into account when planning and sizing downstream
systems.
A number of regions now incorporate retention basins to maintain flow
peaks at or below predevelopment levels. Unlike North American practice,
however, retention basins are usually sized to optimise the attenuation of
major flow peaks in the order of 50 to 100-year return period events. In
general the size of basins are relatively large, typically 20 000 m3 plus
Mein, 1982t29^ and few in number per watershed.
Predominately retention basins have only been implemented to reduce
major flow peaks and water quality has not been a consideration.
The city of Canberra in the Australian Capital Territory representing
a model city for trends in urban stormwater management has recently begun
to incorporate wet basins to combine water quality and quantity aspects.
In recent years there has been an increasing trend to re-analyse older
areas on a catchment basis to retrofit these to new area standards or as
near as economically practical, Henkel & Goyen, 1980t301.
Management techniques have included the inclusion of retention basins
in existing parks, mid catchment diversions, upgrading of pipe and pit
systems and augmentation to channels and floodways.
In general the analysis techniques including models such as RATHGL,
RORB and RAFTS have been applied to isolate particular weaknesses-in the
stormwater system. The degree of augmentation and the management options
65
-------
selected have in recent times been based primarily on the acceptance
criteria described previously.
WATER QUALITY
Possibly the most significant recent trend in a number of parts of
Australia has been the inclusion of water quality elements into urban
stormwater systems.
In Canberra, Australia's largest inland city, an integrated water
quality/quantity approach is now in progress with seven water quality
control ponds and three larger lakes providing both water quality and
quantity control plus gross pollutant traps upstream of ponds and lakes
either constructed or planned to trap urban litter, debris and sediment.
Similar water quality strategies are presently expanding to other
areas in New South Wales in particular in areas showing specific stress
from stormwater pollution.
Urban stormwater quality in Canberra has been approached with a four
pronged strategy, Lawrence & Goyen, 1987 [12], namely:
(a) the establishment of urban lakes, primarily as biological
treatment systems,
(b) the utilisation of shallow ponds (water quality control ponds)
and wetlands, as physical and biological treatment systems,
upstream of urban lakes,
(c) the incorporation of gross pollutant traps on inlets to lakes or
water quality control ponds to intercept trash and debris and the
coarser fractions of sediment plus associated nutrient and other
toxic constituents, and
(d) the incorporation of "off-stream" and "on-stream" sediment
retention ponds into land development works to intercept and
chemically treat runoff prior to its discharge to the stormwater
system.
Additionally the Australian Capital Territory Water Pollution
Ordinance was enacted in 1984 to control discharges to lakes, streams or
stormwater systems. This Ordinance has provided an important enforcement
mechanism during the construction phase of land development in particular.
DISCUSSION ON STORMWATER MANAGEMENT
Stormwater management in Australia as in other developed countries, is
becoming extremely complex and now incorporates a wide range of constraints
and social objectives not previously included. In a significant sense this
has been made possible with the rapid advance in computing power available
to professional engineers. It is the author's opinion that in the
forthcoming depade one of the greatest challenges facing thre profession
will be the marriage of this ever accelerating computational power with the
knowledge and experience of those engineers having to make the complex
engineering/social decisions based on computer predictions.
66
-------
CONCLUSIONS
Urban stormwater management in Australia has progressed significantly
since the late 1970s with integrated strategies generally now being
enforced by state and local government authorities to ensure more
controlled development on a catchment wide basis.
Modelling techniques/ using modern computers, are now in widespread
use in both government and private bodies allowing stormwater master
planning to be carried out prior to development approvals plus detailed
analysis of various drainage options prior to construction. This has
allowed for more engineering and social issues to be investigated at the
planning and design stage providing input to complex decision making
processes which have significantly influenced stormwater management.
Detailed hydrologic and hydraulic simulation of existing urban pipe
and channel networks has allowed the accurate isolation of system
shortcomings and the formulation of appropriate management strategies for
upgrading.
Water quality control has now taken on serious proportions in
Australia with Canberra generally leading the way in control strategies and
management techniques. It is expected that other sensitive regions in
Australia will follow in Canberra's vein over the next few years.
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. McMahon, T.A. World Hydrology: Does Australia Fit? Symposium on
Hydrology and Water Resources, Melbourne, The Institution of
Engineers, Australia, 1982. pp 1-7.
2. Willing & Partners Pty Ltd. Lower Stranger Creek Water Quality
Control Pond. Final Design Report, Canberra, 1986. 10 pp.
(Unpublished).
3. Colston, N.V. Characterisation and Treatment of Urban Land Runoff.
Environmental Protection Technical Services. EPA-670/2-74-096.
December, 1974. •
4. Randal, C.W. Stormwater Detention Ponds for Water Quality Control.
Proceedings of Conference on Stormwater Detention Facilities,
Henniker, New Hampshire, 1982. pp 200-204.
5. Lawrence, A.I. Source and Fate of Urban Runoff Constituents and Their
Management. 12th Symposium on Stormwater Quality in Urban Areas,
Water Resources Foundation of Australia, 1986.
67
-------
6. Finlayson, B.L., McMahon, T.A., Srikanthan, R. and Haines, A. World
Hydrology: A New Data Base for Comparative Analysis. Hydrology and
Water Resources Symposium, Brisbane, The Institution of Engineers,
Australia, pp 288-291.
7. New South Wales Government. Floodplain Development Manual, NSW
Government Printer, 1986.
8. Cullen, P., Rosich, R. and Beck, P. A Phosphorus Budget for Lake
Burley Griffin and Management Implications for Urban Lakes, AGPS,
Canberra, 1978.
9. Gutteridge Haskins & Davey. River Murray in Relation to Albury-
Wodonga. A report to the Cities Commission, Australia, 19*74.
10. Bliss, P.J., Brown, J.D. and Perry, R. Impact of Storm Runoff from
Urban Areas on Surface Water Quality. Proceedings of Hydrology and
Water Resources Symposium, Perth, The Institution of Engineers,
Australia, 1979.
11. Goyen, A.G., Moodie, A.R. and Nuttal, P.M. Enhancement of Urban
Runoff Quality. Hydrology and Water Resources Symposium, The
Institution of Engineers, Australia. May, 1985. pp 202-208.
12. Lawrence, A.I. and Goyen, A.G. Improving Urban Stormwater Quality -
An Australian Strategy. Submitted to Fourth International Conference
on Urban Storm Drainage, Lausanne, 1987.
13. Messner, M.J. and Goyen, A.G. The Interaction of Hydrology and
Hydraulics in Urban Stormwater Modelling. Hydrology and Water
Resources Symposium, Sydney, The Institution of Engineers, Australia,
1985. pp 141-145.
14. O'Loughlin, G.G. and Mein, R.G. Use of Computer Models for Piped
Urban Drainage in Australia. Hydrology and Water Resources Symposium,
Hobart, The Institution of Engineers, Australia, 1983. pp 156-160.
15. Carleton, M.G. The Practical Application of the Computer Models
"ILLUDAS" and "SWMM-EXTRAN" to Urban Stormwater Runoff Problems.
Master of Local Government Engineering Project, New South Wales
Institute of Technology, Australia, 1983.
16. Vale, D.R., Attwater, K.B. and O'Loughlin, G.G. Application of SWMM
to Two Urban Catchments in Sydney. Hydrology and Water Resources
Symposium, Brisbane, The Institution of Engineers, Australia, 1986.
pp 268-272.
17. Terstriep, M.L. and Stall, J.B. The Illinois Urban Drainage Area
Simulator ILLUDAS. Bulletin 58, Illinois State Water Survey, Urbana,
1974.
18. O'Loughlin, G.G. The ILSAX Program for Urban Drainage Design and
Analysis. School of Civil Engineering, The NSW Institute of
Technology, Sydney, 1986.
68
-------
19. U.K. National Water Council. Design and Analysis of Urban Storm
Drainage. 5 Vols, London, 1981.
20. Huber, W.C., Heaney, J.P., Nix, S.J., Dickinson, R.E. and Polmaner,
D.J. Stormwater Management Model User's Manual Version III (SWMM).
Department of Environment, Engineering Services, University of
Florida, 1981.
21. Bloomfield, P. PIPENET - Computer Aided Urban Drainage Design - Users
Manual.
22. Goyen, A.G. A Model to Statistically Derive Design Rainfall Losses..
Hydrology and Water Resources Symposium, Hobart, The Institution of
Engineers, Australia, 1983. pp 220-225.
23. Laurenson, E.M. and Mein, R.G. RORB - Version 3, Runoff Routing
Program. Department of Civil Engineering, Monash University, 1983.
24. Goyen, A.G. and Aitken, A.P. A Regional Stormwater Drainage Model.
Hydrology Symposium, Sydney, The Institution of Engineers, Australia,
1976. pp 40-49.
25. Laurenson, E.M. A Catchment Storage Model for Runoff Routing.
Journal of Hydrology, Vol. 2, 1964. pp 141-163.
26. Black, D.C. and Aitken, A.P. Simulation of the Urban Runoff Process.
Water Resources Council, Technical Paper No. 26, 1977.
27. Bates, B.C. and Pilgrim, D.H. Simple Models for Nonlinear Runoff
Routing. Hydrology and Water Resources Symposium, Hobart, The
Institution of Engineers, Australia. November, 1983.
28. Price, R.K. Flood Routing for British Rivers. Hydraulic Research
Station, Wallingford, INT 111, March 1973.
29. Mein, G.M. Australian Detention Basins; Recent Developments.
Proceedings of Conference on Stormwater Detention Facilities,
Henniker, New Hampshire, 1982. pp 41-48.
30. Henkel, G. and Goyen, A.G. Retarding Basins, Best Upgrade Stormwater
Protection in Old Areas. International Symposium of Urban Storm
Runoff, University of Kentucky, Lexington. July, 1980.
69
-------
A NEW GROUNDWATER SUBROUTINE FOR THE STORM WATER MANAGEMENT MODEL
by: Brett A. Cunningham*, Wayne C. Huber* and Victor A. Gagliardo**
•Department of Environmental **Reynolds, Smith and Hill, Inc.
Engineering Sciences Box 22003
University of Florida Tampa, Florida 33622
Gainesville, Florida 32611
ABSTRACT
Due to the importance of groundwater in the prediction of runoff, SWMM
has been equipped with a groundwater subroutine to model the underlying water
table. The subroutine models two zones — an upper (unsaturated) zone and a
lower (saturated) zone. Outflow from the upper zone to the lower zone is con-
trolled by a percolation equation whose parameters can be calibrated or esti-
mated from soils data. Loss from the unsaturated zone occurs through upper
zone evapotranspiration; loss from the lower zone comes from both evapotran-
spiration and deep percolation. Groundwater flow from the lower zone is de-
termined by a user-defined power function of water table stage and tailwater
depth, and it can be routed to any previously defined inlet, trapezoidal chan-
nel, or pipe. (Pipes may be used to simulate under-drains.) Inflow to the
subroutine is the infiltration calculated in subroutine WSHED. In the cases
where the water table approaches the surface, infiltration that cannot be
accepted by the soil is added back to the surface water component by means of
a reduction in the variable RLOSS, the sum of infiltration and evaporation.
Both printed and graphical output can be obtained for groundwater flow, water
table stage, and moisture content in the unsaturated zone.
This paper has resulted from a project partially funded by the EPA, and
it has been reviewed in accordance with the U.S. Environmental Protection
Agency's peer and administrative review policies and approved for presentation
and publication.
70
-------
INTRODUCTION
Because the EPA Storm Water Management Model, SWMM (Huber et al., 1981)
was originally developed to simulate combined sewer overflows in urban catch-
ments, the fate of infiltrated water was considered insignificant. Since its
development, however, SWMM has been used on areas ranging from highly urban to
relatively undeveloped. Many of the undeveloped and even some of the devel-
oped areas, especially in areas like South Florida, are very flat with high
water tables, and their primary drainage pathway is through the surficial
groundwater aquifer and the unsaturated zone above it, rather than by overland
flow. In these areas a storm will cause a rise in the water table and subse-
quent slow release of groundwater back to the receiving water (Capece et al.,
1984). For this case, the fate of the infiltrated water is highly signifi-
cant. By assuming that the infiltration is lost from the system, an important
part of the high-water-table system is not being properly described (Gag-
liardo, 1986).
It is known that groundwater discharge accounts for the time-delayed
recession curve that is prevalent in certain watersheds (Fetter, 1980). This
process has not, however, been satisfactorily modeled by surface runoff me-
thods alone. By modifying infiltration parameters to account for subsurface
storage, attempts have been made to overcome the fact that SWMM assumes infil-
tration is lost from the system (Downs et al., 1986). Although the modeled
and measured peak flows matched well, the volumes did not match well, and the
values of the infiltration parameters were unrealistic. Some research on the
nature of the soil storage capacity has been done in South Florida (SFWMD,
1984). However, it was directed towards determining an initial storage capa-
city for the start of a storm. There remains no standard, widely-used method
for combining the groundwater discharge hydrograph with the surface runoff hy-
drograph and determining when the water table will rise to the surface. For
instance, HSPF (Johansen et al., 1980) performs extensive subsurface moisture
accounting and works well during average conditions. However, the model never
permits the soil to become saturated so that no more infiltration is permit-
ted, limiting its usefulness during times of surface saturation and flooding.
Another difficulty with HSPF occurs during drought conditions, since there is
no threshold saturated zone water storage (corresponding to the bottom of a
stream channel) below which no saturated zone outflow will occur. These dif-
ficulties have limited HSPF usefulness for application to extreme hydrologic
conditions in Florida (Heaney et al., 1986).
In order to incorporate subsurface processes into the simulation of a
watershed and overcome previously mentioned shortcomings, SWMM has been
equipped with a simple groundwater subroutine. The remainder of this paper
will describe the theory, use, and some limitations of the subroutine.
THEORY
INTRODUCTION
An effort was made to utilize existing theoretical formulations for as
many processes as possible. The purpose was to maintain semblance to the real
71
-------
world while enabling the user to determine parameter values that have meaning
to the soil scientist. Also, in the following discussion the term "flow" will
refer to water that is passed on to another part of the system, and the term
"loss" will refer to water that is passed out of the system. In addition,
in the groundwater subroutines, flows and losses have internal units of velo-
city (flow per unit area).
The groundwater subroutine, GROUND, simulates two zones — an upper (un-
saturated) zone and a lower (saturated) zone. This configuration is similar
to the work done by Dawdy and O'Donnell (1965) for the USGS. The flow from
the unsaturated to the saturated zone is controlled by a percolation equation
for which parameters may either be estimated or calibrated, depending on the
availability of the necessary soil data. Upper zone evapotranspiration is the
only loss from the unsaturated zone. The only inflow to subroutine GROUND is
the calculated infiltration from subroutine WSHED. Losses and outflow from
the lower zone can be via deep percolation, saturated zone evapotranspiration,
and groundwater flow. Groundwater flow is a user-defined power function of
water table stage and, if chosen, depth of water in the discharge channel.
The physical processes occurring within each zone are accounted for by
individual mass balances in order to determine end-of-time-step stage, ground-
water flow, deep percolation, and upper zone moisture. Parameters are shown
in Figure 1 and defined below. Mass balance in the upper (unsaturated) zone
is given by,
TH2 = {[(ENPIL-ETU)*PAREA-PERC]*DELT+(D1-D2)*TH2+TH*DWT1}/(DTOT-D2) (1)
In the lower (saturated) zone, for rising water tables,
D2 = {[PERC-ETD*PAREA-.5*(GWPLW+A1*(D2-BO)B1+A3*D2*TA-i-DEPPRC+DP*D2/DTOT) (2)
-TWFLW]*DELT+(D2-D1)*(TH-TH2)}/(PR-TH2) + D1
and for falling water tables,
D2 = {[PERC-ETD*PAREA-.5*(GWPLW+A1#(D2-BO)B1+A3*D2*TA+DEPPRC+DP*D2/DTOT) (3)
-TWFLW]*DELT}/(PR-TH2) + D1
where TH2 = end-of-time-step upper zone moisture content (fraction),
ENFIL = infiltration rate calculated in subroutine WSHED,
ETU = upper zone evapotranspiration rate,
PERC = percolation rate,
PAREA = pervious area divided by total area,
DELT = time step value,
D1 = beginning-of-time-step lower zone depth (elevation above a
datum),
D2 = end-of-time-step lower zone depth,
TH = beginning-of-time-step upper zone moisture content,
DWT1 = beginning-of-time-step upper zone depth,
DTOT = total depth of upper and lower zone = D1+DWT1,
ETD = lower zone evapotranspiration rate,
GWFLW = beginning-of-time-step groundwater flow rate,
72
-------
IMPERVIOUS
AREA
UPPER
ZONE
LOWER
ZONE
^
DET
Dl
rETD
EU
~i—r
DWTl
1 i v
'
GWFLW
V
PERC
\
'
DTOT
BO
DEPPRC
-7 7~
~> r
-BELEV
Figure 1. GROUND parameters and conceptualization.
-------
A1 = groundwater flow coefficient,
BO = bottom of channel depth (elevation above datum),
B1 = groundwater flow exponent,
DEPPRC = beginning-of-time-step deep percolation rate,
DP = a recession coefficient derived from interevent declines in
the water table,
PR = porosity, and
TWPLW = channel water influence rate,
A3 = groundwater flow coefficient, and
TA = depth of water in channel (elevation above datum).
Solving equation 1 for TH2 and using DWT1 = DTOT-D1, yields a much sim-
pler form which is not a function of the unknown D2,
TH2 = t(ENPIL-ETU)*PAREA-PERC]*DELT/DWT1 + TH (4)
Equation 4 is solved first, followed by a Newton-Raphson solution of equation
2 or 3. The sequencing will be described in more detail in a subsequent sec-
tion, following a description of the various simulated processes.
UPPER ZONE ET
Evapotranspiration from the upper zone (ETU) represents soil moisture
lost via cover vegetation and by direct evaporation from the pervious area of
the subcatchment. No effort was made to derive a complex formulation of this
process. The hierarchy of losses by evapotranspiration is as followsj 1)
surface evaporation, 2) upper zone evapotranspiration, and 3) lower zone tran-
spiration. Upper zone evapotranspiration is represented by the following
equations,
ETMAX = VAP(MONTH) (5)
ETAVLB = ETMAX-EVAPO (6)
ETU = CET*ETMAX (7)
IF(TH.LT.WL.OR.ENPIL.GT.O.) ETU = 0. (8)
IF(ETU.GT.ETAVLB) ETU = ETAVLB (9)
where ETMAX = maximum total evapotranspiration rate (input on card P1),
VAP(MONTH) = input maximum evapotranspiration rate for month MONTH,
ETAVLB = maximum upper zone evapotranspiration rate,
EVAPO = portion of ETMAX used by surface water evaporation,
GET = fraction of evapotranspiration apportioned to upper zone, and
WL = wilting point of soil.
The two conditions that make ETU equal to zero in equation 8 are believed to
simulate the processes actually occurring in the natural system. The first
condition (moisture content less than wilting point) relates to the soil sci-
ence interpretation of wilting point — the point at which plants can no
longer extract moisture from the soil. The second condition (infiltration
greater than zero) assumes that vapor pressure will be high enough to prevent
additional evapotranspiration from the unsaturated zone.
74
-------
INFILTRATION
Infiltration enters subroutine GROUND as the calculated infiltration from
subroutine WSHED. As before in SWMM, either the Horton or Green-Ampt equation
can be used to describe infiltration. For time steps where the water table
has risen to the surface, the amount of infiltration that cannot be accepted
is subtracted from RLOSS (infiltration plus surface evaporation) in subroutine
WSHED. In the event that the infiltrated water is greater than the amount of
storage available for that time step, the following equation is used to calcu-
late the amount of infiltration that is not able to be accepted by the soil.
XSINFL = ENFIL*DELT-AVLVOL/PAREA (10)
where XSINFL = excess infiltration over pervious area, and
AVLVOL = initial void volume in the upper zone plus total losses and
outflows from the system for the time step.
The second condition exists because of the algebra in equations 2, 3
i|. As the water table approaches the surface, the end-of-time-step moisture
value, TH2, approaches the value of porosity, which makes the denominator in
equations 2 and 3 go towards zero. Since a denominator close to zero could
result in an unrealistic value of D2, a different way of handling the calcula-
tions had to be implemented. When the initial available volume in the upper
zone plus the volume of total outflows and losses from the system minus the
infiltration volume is between zero and an arbitrary value of 0.0001 ft, sev-
eral assumptions are made. First, end-of-time-step groundwater flow and deep
percolation, which are normally found by iteration, are assumed to be equal to
their respective beginning-of -time-step values. This step is taken to ensure
that the final available volume remains in the previously mentioned range.
Second, TH2 is set equal to an arbitrary value of 90/E of porosity. It is
believed that this will allow the TH2 value in this special case to be reason-
ably consistent with the TH2 values juxtaposed to it in the time series.
Third, D2 is set close to the total depth — the actual value of D2 depends on
the value of porosity. Fourth, the amount of infiltration that causes the
final available volume to exceed 0.0001 ft is calculated in the following
equation and sent back to the surface in the form of a reduction in the term
RLOSS in subroutine WSHRD.
XSINFL = ENFIL*DELT+(.0001-AVLVOL)/PAREA (11)
Because of the way this special case is handled, it is possible for a falling
water table to have the calculated excess infiltration be greater than the
actual amount of infiltration. It is not desirable for the ground to pump wa-
ter back onto the surface! Hence, the difference between the calculated ex-
cess infiltration and the actual infiltration is added to the infiltration
value of the next time step. The number of occurrences of this situation in a
typical run is very small, as is the computed difference that is passed to the
next time step, so no problems should occur because of this solution.
75
-------
LOWER ZONE EVAPOTRANSPIRATTON
Lower zone evapotranspiration, ETD, represents evapotranspiration from
the saturated zone over the pervious area. ETD is the last evapotranspiration
removed, and is determined by the follovd-ng depth-dependent equation and
conditions.
ETD = (DET-DWT1)*ETMAX*(1-CET)/DET (12)
IPCETD.GT.(ETAVLB-ETU)) ETD = ETAVLB-ETU (13)
IPCETD.LT.O.) ETD = 0. (1i|)
where ETD = lower zone evapotranspiration rate, and
DET = depth over which evapotranspiration can occur.
Since ETD is typically very small compared to other terms and has to be
checked for certain conditions, it is assumed constant over the time step and
not solved for in the iterative process.
PERCOLATION
Percolation (PERC) represents the flow of water from the unsaturated zone
to the saturated zone, and is the only inflow for the saturated zone. The
percolation equation in the subroutine was formulated from Darcy's Law for
unsaturated flow, in which the hydraulic conductivity, K, is a function of the
moisture content, TH. For one-dimensional, vertical flow, Darcy's Law may be
written
v = -K(TH) dh/dz (15)
where v = velocity (specific discharge) in the direction of z,
z = vertical coordinate, positive upward,
K(TH) = hydraulic conductivity,
TH = moisture content, and
h = hydraulic potential.
The hydraulic potential is the sum of the elevation (gravity) and pressure
heads,
h = z + PSI (16)
where PSI = soil water tension (negative pressure head) in the unsaturated
zone.
Equating vertical velocity to percolation, and differentiating the
hydraulic potential, h, yields
Percolation = -K(TH)«(1+ dPSI/dz) (17)
A choice is customarily made between using the tension, PSI, or the moisture
content, TH, as parameters in equations for unsaturated zone water flow.
Since the quantity of water in the unsaturated zone is identified by TH in
76
-------
previous equations, it is the choice here. PSI can be related to TH if the
characteristics of the unsaturated soil are known. Thus, for use in equation
17, the derivative is
dPSI/dz = dPSI/dTH * dTH/dz (18)
The slope of the PSI versus TH curve should be obtained from data for the
particular soil under consideration. Relationships for a sand, sandy loam and
silty loam are shown in Figures 2, 3 and J| (Laliberte et al., 1966). The data
are based on laboratory tests of disturbed soil samples and illustrate only
the desaturation (draining) characteristics of the soil. The relationship
during the saturation (wetting) phase will ordinarily be different; when both
the wetting and draining relationships are shown the curves usually illustrate
a hysteresis effect. The figures also show the relationship between the hy-
draulic conductivity of the unsaturated soils and the moisture content. In
some cases (e.g., sand), K(TH) may range through several orders of magnitude.
Soils data of this type are becoming more readily available; for example, soil
science departments at universities often publish such information (e.g.,
Carlisle et al., 1981). The data illustrated in Figures 2, 3 and 4 are also
useful for extraction of parameters for the Green-Ampt infiltration equations.
Equation 17 may be approximated by finite differences as
Percolation = -K(TH)*[1 + (ATH/Az)*(APSI/ATH)] (19)
For calculation of percolation, it is assumed that the gradient, ATH/Az, is
the difference between moisture content TH in the upper zone and field
capacity at the boundary with the lower zone, divided by the average depth of
the upper zone, DWT1/2. Thus,
Percolation = -K(TH)#{1+[(TH-FD)*2/DWT1]*PCO} (20)
where FD = field capacity, and
PCO = APSI/ATH in the region between TH and FD.
PCO is obtained from data of the type of Figures 2, 3 and 4.
Finally, the hydraulic conductivity as a function of moisture content is
approximated functionally in the moisture zone of interest as
K(TH) = HKTH = HKSAT*EXP[(TH-PR)*HCO] (21)
where HKTH = hydraulic conductivity as a function of moisture content,
HKSAT = saturated hydraulic conductivity, and
HCO = calibration parameter.
HCO can be estimated by fitting the HKTH versus TH curve to the hydraulic
conductivity versus moisture content curve, if such data are available (e.g.,
Figures 2, 3, 4); three fits are shown in Figure 5. The fits are not optimal
over the entire data range because the fit is only performed for the high
moisture content region between field capacity and porosity. If soils data
77
-------
Touchet Silt Loam
Figure 2. Tension, PSI (squares, in.
of water) and hydraulic
conductivity, K (crosses,
in/hr, K multiplied by 200)
versus moisture content.
Derived from data of
Laliberte et al (1966),
Tables B-5 and C-3.
Porosity = 0.503, temp. =
26.5 °C, saturated hyd.
conductivity = 0.53 in/hr.
Columbia Sandy Loam
Figure 3« Tension, PSI (squares, in.
of water) and hydraulic
conductivity, K (crosses,
in/hr, K multiplied by 100)
versus moisture content.
Derived from data of
Laliberte et al. (1966),
Tables B-8 and C-5.
Porosity = 0.1(85, temp. =
25.1 °C, saturated hyd.
conductivity = 0.60 in/hr.
Unconsolidated Sand
Figure 4. Tension, PSI (squares, in.
of water) and log-10 of
hydraulic conductivity, K
(crosses, K in in/hr) versus
moisture content. Derived
from data of Laliberte et
al. (1966), Tables B-14 and
C-11. Porosity = 0.452,
temp. = 25.1 °C, saturated
hyd. conductivity = 91.5
in/hr.
78
-------
^•CONSOLIDATED SAND
rOUCHEJ SILT LOAM'
•wr O.4 ~
B
COLUMBIA SANDY LOAM
o O.3 -
Figure 5. Model representation of and measured hydraulic conductivity curves
for three types of soil.
79
-------
are not available, HCO can be estimated by model calibration.
Combining equations 20 and 21 gives the resulting percolation equation
for the model,
PERC = HKTH*[1+PCO*(TH-FD)/(DWT1/2)] (22)
where PERG = percolation rate (positive downward) and is only nonzero when
TH is greater than PD.
If data sources for parameters PCO and HCO are lacking, they may be
estimated through the calibration process. On the basis of preliminary runs,
the groundwater subroutine is relatively insensitive to changes in PCO and
HCO, so a lack of extensive soils data should not discourage one from using
the model.
If moisture content is less than or equal to field capacity, percolation
becomes zero. This limit is in accordance with the concept of field capacity
as the drainable soil water that cannot be removed by gravity, alone (Hillel,
1982, p. 2i»3). Once TH drops below field capacity, it can only be further
reduced by upper zone evapotranspiration.
The percolation rate calculated by equation 22 will be reduced by the
program if it is high enough to drain the upper zone below field capacity or
make the iterations for D2 converge to an unallowable value. Also, since
checks must be made on PERC, it is assumed to be constant over the time step
and therefore not determined through an iterative process.
DEEP PERCOLATION
Deep percolation represents a lumped sink term for unquantified losses
from the saturated zone. The two primary losses are assumed to be percolation
through the confining layer and lateral outflow to somewhere other than the
receiving water. The arbitrarily chosen equation for deep percolation is
DEPPRC = DP*D1/DTOT (23)
where DEPPRC = beginning-of-time-step deep percolation rate, and
DP = a recession coefficient derived from interevent water
table recession curves.
The ratio of D1 to DTOT allows DEPPRC to be a function of the static pressure
head above the confining layer. Although DEPPRC will be very small in most
cases, it is included in the iterative process so that an average over the
time step can be used. By doing this, large continuity errors will be avoided
should DEPPRC be set at a larger value.
80
-------
GROUNDWATER DISCHARGE
Functional Form
Groundwater discharge represents lateral flow from the saturated zone to
the receiving water. The flow equation takes on the following general form:
GWPLW = A1*(D1-BO)B1 - TWPLW + A3*D1*TA (24)
and TV
TWPLW = A2*(TA-BOr^ (25)
where GWPLW = beginning-of-time-step groundwater flow rate (per subcatchment
area,
TWPLW = channel water influence flow rate (per subcatchment area),
A1,A2 = groundwater and channel water influence flow coefficients,
A3 = coefficient for cross-product,
B1,B2 = groundwater and tailwater influence flow exponents,
BO = elevation of bottom of channel, and
TA = elevation of water in channel.
If D1 is less than BO or TA, GWPLW is set equal to zero. In addition, if TA =
BO and B2 = 0, then the indeterminant form of zero raised to the zero power in
equation 25 is set equal to 1.0 by the program. The functional form of equa-
tions 24 and 25 was selected in order to be able to approximate various hori-
zontal flow conditions, as will be illustrated below.
Since groundwater flow can be a significant volume, an average flow each
time step is found by iteration using equation 2 or 3. Groundwater flows can
be routed to any previously defined inlet, trapezoidal channel, or pipe, al-
lowing the user to isolate the various components of the total hydrograph, as
shown in Pigure 6. That is, the groundwater flow does not have to be routed
to the same destination as the overland flow from the subcatchment.
The effects of channel water on groundwater flow can be dealt with in two
different manners. The first option entails setting TA (elevation of water
surface in the channel) to a constant value greater than or equal to BO
(bottom-of-channel elevation) and A2, B2 and/or A3 to values greater than
zero. If this method is chosen, then the user is specifying an average tail-
water influence over the entire run to be used at each time step.
The second option makes the channel water elevation, TA, equal to the
elevation of water in an actual channel (trapezoidal channel or circular
pipe). Por this option, the groundwater must be routed to a trapezoidal chan-
nel or pipe — not an inlet. The depth of water in the channel (TA - BO) at
each time step is then determined as the depth in the channel or pipe from the
previous time step. (It is assumed that the bottom of the channel is at the
elevation BO.) The beginning-of-time-step depth must be used to avoid complex
and time-consuming iterations with the coupled channel discharge equations in
subroutine GUTTER. Unfortunately, because of this compromise the groundwater
flow may pulsate as D1 oscillates between just above and just below elevation
TA. This pulsing may introduce errors in continuity and is, of course, unrep-
81
-------
I TO
2O 4O
eo ao ioc
tlm«, mtn.
14O 16O 16O
Figure 6. Hydrograph of total flow and its two major components.
1
1 *
Flow "
f Impermeable
•\\ \ n f j \\ vv/V / \ \/
j, f
< \J
ha ^^~
i '
/ / / \ \ \///\\///\\\
Figure 1. Definition sketch for Dupuit-Forcheimer approximation for drainage
to adjacent channel.
^
///\ \ \
^V
Impermeable
/ // / \ ^ W
i h, -^
\ i
(
/ \ \ \ \///\\\
i. ->-
\
///\\///
Figure 8. Definition sketch for Hooghoudt's method for flow to circular
drains.
82
-------
resentative of the actual system. Shorter time steps and larger or less steep
channels (reducing the response of the channel) can be used to reduce the
pulses. Also, caution must be taken when selecting A1, B1, A2, B2 and A3 so
that GWPLW cannot be negative. Although this may occur in the actual system
and represent recharge from the channel, there is currently no means of repre-
senting this reverse flow and subtracting it from the channel. One way of
assuring that this cannot happen is to make A1 greater than or equal to A2 and
B1 greater than or equal to B2, and A3 equal to zero.
Because of the general nature of the equation, it can take on a variety
of functional forms. For example, a linear reservoir can be selected by set-
ting B1 equal to one and A2 and A3 equal to zero. Two drainage examples are
illustrated below.
Example; Infiltration and Drainage to Adjacent Channel
Under the assumption of uniform infiltration and horizontal flow by the
Dupuit-Forcheimer approximation, the relationship between water table eleva-
tion and infiltration for the configuration shown in Figure 7 is (Bouwer,
1978, p. 51)
K(h2 _ h2) = L2f (26)
where f = infiltration rate,
K = hydraulic conductivity, and other parameters are as shown on
Figure 7.
Before matching coefficients of equations 24 and 25 to equation 26, it should
be recognized that the water table elevation in SWMM, D1, represents an aver-
age over the catchment, not the maximum at the "upstream" end that is needed
for h1 in equation 26. Let D1 be the average head,
D1 = (h1 + h2)/2 (27)
Substituting h1 = 2 D1 - h^ into equation 26 gives, after algebra
(D12 - D1 h2) 4K/L2 = f (28)
from which a comparison with equations 24 and 25 yields A1 = A3 = 4K/L , A2 =
0, and B1 = 2. Note that GWFLW has units of flow per unit area, or length per
time, which are the units of infiltration, f, in equation 28.
Example; Hooghoudt's Equation for Tile Drainage
The geometry of a tile drainage installation is illustrated in Figure 8.
Hooghoudt's relationship (Bouwer, 1978, p. 295) among the indicated parameters
is
f = (2D + m) 4Km/L2 (29)
e
where De = effective depth of impermeable layer below drain center, and other
83
-------
parameters are defined in Figure 8. De is less than or equal to bQ in Figure
8 and is a function of bQ, drain diameter, and drain spacing, L; the compli-
cated relationship is given by Bear (1972, p. 412) and graphed by Bouwer
(1978, p. 296). The maximum rise of the water table, m = h1 - bQ. Once again
approximating the average water table depth above the impermeable layer by D1
= 2h.| - bQ, equation 29 can be manipulated to
f = [(h1 - b0)2 + 2De(h1-b0)] 4K/L2 =
= C(D1-b0)2 + DeD1 - Deb0] 16K/L2 (30)
Comparing equation 30 with equations 24 and 25 yields
A1 = 16K/L2,
B1 = 2
A2 = l6KDebQ/L2
B2 = 0
A3 = 16KIL / TA L2
C
and TA = BO = bQ = constant during the simulation. The equivalent depth, De,
must be obtained from the sources indicated above. The mathematics of drain-
age to ditches or circular drains is complex; several alternative formulations
are described by van Schilfgaarde (1974).
LIMITATIONS
Since the moisture content of the unsaturated zone is taken as an average
over the entire zone, the shape of the moisture profile is totally obscured.
Therefore, infiltrated water cannot be modeled as a diffusing slug moving down
the unsaturated zone, as is the case in the real system. Furthermore, water
from the capillary fringe of the saturated zone cannot move upward by diffu-
sion or "suction" into the unsaturated zone.
The simplistic representation of subsurface storage by one unsaturated
"tank" and one saturated "tank" limits the ability of the user to match non-
uniform soil columns. Another limitation is the assumption that the infil-
trated water is spread uniformly over the entire catchment area, not just over
the pervious area. In addition, just as for surface flow, groundwater may not
be routed from one subcatchment to another. The tendency of the tailwater
influence to cause pulses if TA-BO is equated to the dynamic water depth in
the adjacent channel is a limitation that will remain until the channel flow
and subsurface flow are solved simultaneously using a set of coupled equa-
tions. Such a solution would also permit reverse flow or recharge from the
channel to be simulated.
Finally, water quality is not simulated in any of the subsurface rou-
tines. If water quality is simulated in RUNOFF and the subsurface flow rou-
84
-------
tines activated, any loads entering the soil will "disappear," as if the soil
provides 100 percent treatment.
SUBROUTINE CONFIGURATION
A flowchart of the subroutine configuration is presented in Figure 9.
Initial values and constants used in subroutine GROUND come mostly from sub-
routine GRIN, designed specifically to read in these values. Subroutine GRIN
is called by RHYDRO. Other necessary values are transferred during the CALL
statement and from previously calculated values stored in COMMON.
Subroutine GROUND first initializes pertinent parameters, then calculates
fluxes that are constant over the time step. Beginning-of-time-step fluxes
are calculated next, and the value of percolation is checked to ensure that it
will not raise the water table above the ground surface.
After other constants are calculated and TH2 is determined from equation
^, the program branches to one of four areas, as indicated in Figure 9. The
first and second areas are for rising and falling water tables, equations 2
and 3> respectively. In both cases, Newton-Raphson iteration is used to solve
simultaneously for the final groundwater flow, depth of lower zone, and deep
percolation. Each iteration checks whether or not groundwater flow is possi-
ble (D1 greater than or equal to TA and BO). After the iterations converge,
final conditions are set as the next time step's initial conditions.
Tn the event of saturation (D1 = DTOT), the third area sets D2 equal to
DTOT, sets final groundwater flow equal to the maximum possible (D2 = DTOT),
and assumes DEPPRC remains constant over the time step. Any excess infiltra-
tion is then routed back to the surface for overland flow calculations, and
final conditions are set for the next initial conditions. However, if the
maximum groundwater flow and DEPPRC rates permit some infiltration into the
subsurface zone, the initial and final groundwater flow are averaged to be
used as the new initial groundwater flow, and the program branches back to
iterate for the solution. 'Phis pathway will rarely, if ever, be taken, but
must be included to minimize possible continuity errors.
In the event the available storage in the unsaturated zone is less than
0.0001 ft, the fourth area sets TH2 equal to 9056 of porosity and D2 close to
DTOT, and returns any infiltration to the surface that causes the final un-
filled upper zone volume to be greater than 0.0001 ft. This is to avoid os-
cillations as the water table hovers near the ground surface. Again, final
conditions are then set as the next time step's initial conditions.
EXAMPLE RUNS
CYPRESS CREEK CALIBRATION AND VERIFICATION
Two examples will illustrate the use of the new subroutine. The first
example is a year-long simulation of a J|7 mi portion of the 117 mi Cypress
Creek Watershed in Pasco County, Florida, about 30 miles north of Tampa (Fig-
ure 10). The region has been studied in relation to the interaction of sur-
85
-------
Input subsurface
parameters
QtfN
Calculate infiltration and
surface evaporation
WSHED
Ihitialias parameters. Calculate constant
rates and begirmng-of-tdne-stsp rates.
Cetermine TH2. If water table at surface,
than 1H2 = field capacity
GRUlsD
If end-of-time-step
groudwater flow is
allowable, it is max.
Use avg. groundwater"
flow as begiming-of-
tirne-step flow
Calculate excess
infiltration, set
and reset
rent variables
:or next tine step
Set 1H2 close to,, but
less than, porosity.
Sat end-of-time-step
flow eqjal to beginning-
of-time-step flow
Iterate to find
DZ Set E2=D1
if no convergenoe.
Calculate excess
infiltration. Sat
flag
Iterate to find D2 and end-
of-tinB-step groundwater
flow. Each iteration must
creek if grounci»ater flow
is allowable
Iterate to find D2 and end-
of-tine-step grouTdwater
flow. Each iteration oust
check if grotndwater flow
is allowable
If no oorwargence, use
initial conditions as
final conditions
If n9 conyergenoa, use
initial conditions as
final ccnditions
GFCC10
Recalculate
RU3SS, if necessary
WSHED
Figure 9.
Flowchart of subsurface and directly-connected surface
calculations.
86
-------
face water and ground water under the stress of heavy pumping and drainage
activities in the area (Heaney et al., 1986). The watershed is characterized
by sandy soils in which most water movement follows subsurface pathways. For
this example, only a single 47 mi area above State Road 52 (Figure 10) and
tributary to the USGS gage at San Antonio has been simulated.
Twenty-four parameters on three additional H-cards are required for each
subsurface subcatchment. (Many of these can be ignored or set to zero during
most runs; not all parameters are required for all runs.) Input parameters
are echoed on two new pages of output that immediately follow the surface
subcatchment information. Figure 11 is an example of these two new pages; the
values in Figure 11 are from the calibration run on Cypress Creek. In addi-
tion to the new output just mentioned, a subsurface continuity check is pro-
vided in addition to the existing surface continuity check. An example of
this amended page is shown in Figure 12.
The simulation is divided into two six-month runs: the first six months
for calibration, and the second six months for verification. Since Cypress
Creek is a very flat, pervious area with well-drained soils and very little
surface flow, it was modeled in a manner that would allow groundwater flow to
account for most of the flow in the channel. In other words, the groundwater
parameters represented by far the most critical part of the calibration. The
only complete rainfall data for the calibration period are for the gage at St.
Leo, out of the catchment to the east. Although these data are in daily in-
crements, the calibration process was relatively simple because of the exis-
tence of both flow and shallow-well stage data. In addition, only one sub-
catchment (surface and subsurface) was used, since the purpose of this example
was only to illustrate the use of subroutine GROUND, not to provide a thorough
simulation.
Figure 13 shows the predicted groundwater flow hydrograph and the mea-
sured total flow hydrograph for the calibration run, and Figure 14 shows a
comparison of the predicted total flow hydrograph to the measured total flow
hydrograph for the calibration run. Predicted and measured stages for the
calibration and verification can be seen in Figures 15 and 16. The calibra-
tion is not especially remarkable in light of the lack of detailed rainfall
data for the 47 mi area. The predicted stage hydrograph does not exhibit the
short-term variations that are measured, primarily because of the lack of
spatial detail in the rain. In addition, the measured stages are at one well
near the center of the modeled area and would be expected to show more varia-
tion than would the average water table over the 47 mi simulated by SWMM.
The existence of more than one gage in the 47 square miles of the catchment
and shorter increment rainfall data would have improved the fit seen in Fig-
ures 15 and 16. Figures 17 and 18 show flow results for the verification
runs. In general, the average recession of the water table is simulated
accurately, but not the fluctuations.
HYPOTHETICAL CATCHMENT WITH HIGH WATER TABLE
The second example is a 100 ac hypothetical subcatchment with the same
soil properties as Cypress Creek and a water table that is initially one foot
87
-------
82-30'
82* 231 !•«•». *. «' 8f««««He»
87* 20'
28-25')-
28-10 —
LEGEND
GAGES
Evapo-Transpirotion
Stream
---- Watershed
.............. Wellfields
01234
miles
P 0 J»«« !9S5
Figure 10. Map of Cypress Creek Watershed in Pasco County, Florida.
(Heaney et al., 1986)
88
-------
G R OUNOWA
SUBCAT.
NO.
21
GUTTER
OR INLET
22
GROUND
(FT)
20. 00
TER INPUT DATA
ELEVATIONS
INITIAL
BOTTOM STAGE BC
(FT) (FT)
-------
* * * CONTINUITY CHECK FOR QUANTITY -
TOTAL PRECIPITATION (RAIN PLUS SNOW)
TOTAL INFILTRATION
TOTAL EVAPORATION
TOTAL CUTTER/PIPE/SUBCAT FLOW AT INLETS
TOTAL WATER REMAINING IN CUTTER/PIPES
TOTAL WATER REMAINING IN SURFACE STORAGE
INFILTRATION OVER THE PERVIOUS AREA. . .
* * *
CUBIC FEET
3.434232E+-O9
2. 878862E+O9
5.298OOOE+O8
3. 559983E+07
O.OOOOOOE+OO
0.OOOOOOE+OO
2.B78862E+O9
INCHES OVER
TOTAL BASIN
30. 518
29. 583
4. 7O8
O. 227
O. OOO
0. OOO
25. 841
INFILTRATION + EVAPORATION +
SNOW REMOVAL + INLET FLOW •*•
WATER REMAINING IN GUTTER/PIPES +
WATER REMAINING IN SURFACE STORAGE +
WATER REMAINING IN SNOW COVER : . .
3. 344122E+09
29. 718
»*» CONTINUITY CHECK FOR SUBSURFACE WATER «**
TOTAL INFILTRATION
TOTAL UPPER ZONE ET
TOTAL LOWER ZONE ET
TOTAL GROUNOWATER FLOW
TOTAL DEEP PERCOLATION
INITIAL SUBSURFACE STORAGE
FINAL SUBSURFACE STORAGE
UPPER ZONE ET OVER PERVIOUS AREA
LOWER ZONE ET OVER PERVIOUS AREA
CUBIC FEET
2.878862E+09
1.149S78E+09
6.667578E+08
9.O13922E+O7
4.816257E+08
9.675055E+O9
1. 0164S9E+1O
1. 149578E+O9
6.667578E+08
INCHES OVER
TOTAL BASIN
25. 583
10. 216
5. 925
0. 801
4. 28O
85. 978
90. 33O
10. 319
5. 985
THE ERROR IN CONTINUITY IS CALCULATED AS
#»**»»**»************#**************«.**
* PRECIPITATION + INITIAL SNOW COVER *
* - INFILTRATION - *
*EVAPORATION - SNOW REMOVAL - *
*INLET FLOW - WATER IN GUTTER/PIPES - *
»WATER IN SURFACE STORAGE - *
*WATER_REMAINING IN SNOW COVER «
* PRECIPITATION + INITIAL SNOW COVER *
»»*******»*»»*»**********»*****«•*****»#
ERROR.
2.624 PERCENT
#**»*«»**#*********#******»»**************
* INFILTRATION + INITIAL STORAGE - FINAL »
* STORAGE - UPPER AND LOWER ZONE ET - *
* GROUNDWATER FLOW - DEEP PERCOLATION _ *
* INFILTRATION + INITIAL STORAGE - *
» FINAL STORAGE *
******************************************
ERROR
O.039 PERCENT
Figure 12. Continuity check for surface and subsurface for Cypress Creek
calibration. The relatively large surface continuity error does
not actually exist; it comes from a double accounting of the
groundwater flow — a problem that will be fixed.
90
-------
120. 000
FLOW
IN
CFS
SO. OOO
40 000
0 000
0
CYPRESS C
HYDRO8RAPH STATIS
CU8I
PREDICTED, 0. 14
TOTAL TIME
MEASURED, 0. 16
TOTAL TIME
PREDICTED, 0 14
OVERLAP? INO
TIME
+
+
*
##
4-* • +
* 4-
*
4- **
*»» *
+• » *
+ »#
+4-* 4.4- »»
+•4.4. 4- »»»
4-4. 4- 4- 4-4- ***
4 4 4-» 4- 4-4 **#+»
0 9OO. O 10OO. 0 1900.0 2OOO. 0 2300. O 3000.0 39OO. O 4OOO. 0 49OO. 0
TIME OF DAY, IN HOURS PREDICTED-*, MEASURED"*
REEK CALIBRATION RUN LOCATION 21
riCS FOR LOCATION 21
VOLUME PEAK FLOW DURATION NO.
: FEET INCHES TIME, HR FLOW, CFS START, HR END, HR LENOTH, HR POINTS
M7E*09 1.293 3109.000 73.842 0. OOO 4430.000 4430.000 194
399E+09 1.494 312O. 000 ISO. OOO 0. OOO 4392.000 4392. OOO 184
«63E*O9 1.289 3109. OOO 73.842 O. OOO 4393. OOO 4393. OOO 192
MEASURED, 0. 16399E+09 1.494
OVERLAPPING
TIME
DIFFERENCES,
X-'O^MEAS °-»»"E*°8 ,?;!$?
312O. OOO 18O. OOO
19. 000 IO6. 798
99.310
0. OOO 4392. OOO 4392. OOO
184
Figure 13. Predicted groundwater flow hydrograph and total measured flow
hydrograph for Cypress Creek calibration.
-------
for Cypress Creek calibration.
••«
H-
TO
C
CD
Total predicted flow hydrograph and total measured flow hydrograph
MEASURED, O 16:
OVERLAPPING
TIME
DIFFERENCES,
ABSOLUTE -0 63:
X OF MEAS
? «
m m
S £
M -0
1 1 ^
NIG* m
*»* A
ca
V> M
82 8
vjsj o
O»4 O
MM o
0
o
o
o
M
o
8
1
0
o
o
03
PREDICTED, 0. 17i
OVERLAPPING
TIME
s
I
u
u
o
K
CD
O
U
0
o
o
u
o
8
3
MEASURED, 0. 16:
TOTAL TIME
u
i
s
2
U
8
O
8
i
p
O
O
O
&
•O
N
I
o
o
o
8
PREDICTED. 0. 17
TOTAL TIME
fO
nt
4
O
B
o
o
o
w
CO
B
p
o
o
o
u
p
I
o
o
o
•0
*
r>
c
a
VOLUME PEAK FLOW DURATION NO.
C FEET INCHES TIME, HR FLOW, CFS START, HR END, HR LENGTH, HR POINTS
HYDROORAPH STATIS1
TICS FOR LOCATION 23
5
S
H
T)
s
n
o
r
a
I
o
H
§
KJ
U
H
m
?
2
g
CD
•o
m
o
o
o
•
*
2
to
i
s 3 I
o S § S S
88888
0 O O O 0
p. . „ „„ „„. ^^^^
4-
; 4-+
4-+
i ++
4-4
[ ++
[ +4
i +4
[ ++
[ +4-
[ +4-
[ ++
[ »++
[ »4-+
*4-4-
[ *+
[ *4-
[ •+
[ »*•
[ +*
t +»
[ +»
1 1
t 4-»
[
1 *
[ «*
[ * + *
I * »
I 4- *
I + *
4- **
[ +** *
I + • *
I + ••
I +~f+ ++ »»
I * + +4 + *** *
I» » ++ + + ++ »»« ««
I* 0* »++ +* 4 ++ ••• »*
I* ** * +* 4- +< + 4-f »»4>4.»
I# »**44 44++4-4- # +#» 4-#»»*»**44-4+ 4-4-4- 4-+**
0 300.0 10OO. 0 19OO. 0 2OOO. 0 25OO. O 3OOO. 0 33OO. 0 40OO. O 4SOO. O
N
g
C
c
c
t
t-
t-
1-
•-
1
-------
CYPRESS CREEK CALIBRATION
y
ac.
lu
I
72. -T-
71 -
r>8 -
PREDICTED
2O 40 GO
10O 12O 14O IfO ISO
TIME
Figure 15. Predicted and measured stages for Cypress Creek calibration.
CYPRLSS CREEK VERIFICATION
o
in
(t
2'»
MEASURED
I i --)•- f - 7- -| i - -f —(- -i 7 |— —p ~| i
220 240 260 28O 3OO 320 J4O 56O
TIMF
Figure 16. Predicted and measured stages for Cypress Creek verification.
93
-------
1000 000
8OO. 000
VD
FLOW
IN
CFS
40O 000
200. 000
0. OOO
•f
•f ++ +
•*•+ +• •*-
•*• ++• +-H-++++
•f
+•
+
+ +*#
++•*
•f-t-
*
***#
»
4-
•f
-f
+
* •*-
+*#
+***
•*•+**#**
OO 3OO. 0 10OO 0 130O. 0 2000.0
TIME OF DAY. IN HOURS
CYPRESS CREEK CALIBRATION RUN
30OO. O 3300. 0 40OO. 0 49OO. 0
PREDICTED-*. MEASURED-*
LOCATION 21
HYDROORAPH STATISTICS FOR LOCATION 31
VOLUME
CUBIC FEET
PREDICTED,
TOTAL TIME
MEASURED.
TOTAL TIME
PREDICTED.
OVERLAPPIN6
TIME
0
0.
0.
42333E+09
7I232E+09
42426E+0?
INCHES
3.
6.
3.
780
330
770
PEAK
TIME, HR
2112.
2112.
2112.
OOO
OOO
OOO
FLOW
FLOW. CFS
144.
9OO.
144.
983
OOO
383
MEASURED. 0 71232E+O9
OVERLAPPING
TIME
DIFFERENCES,
ABSOLUTE 0. 286O6E+0?
X OF MEAS
2. 960
40. 440
2112.000 3OO. OOO
0. 000 333. 417
71. O83
DURATION NO.
START, HR END, HR LENOTH, HR POINTS
O 000 4330. OOO 4330. OOO 199
O. OOO 4320. OOO 4320. 000 181
0. 000 4312. OOO 4312. OOO 197
0. OOO 4320. OOO 432O. 000 181
Figure 17. Predicted groundwater flow hydrograph and total measured flow
hydrograph for Cypress Creek verification.
-------
FLOW
IN
CFS
400 000
200. OOO
0. OOO H
+
•
4-
•*•
4-
4»
-f
+•
4-
4-
4- 4-
4. + +*** +
4-4- 4- 4-* * +•
4- •*•+•+• +4-» +•##
4-4- 4- 4- 4-+ +***
4- +4- •*-+-f+4-4-*"f» * 4-4-#*#»*
0.0 5OO. 0 10OO. O 1500. O 2OOO. O
TIME OF DAY. IN HOURS
PLOT OF TOTAL RUNOFF FOR CYPRESS CK CAHBRATON
25OO.~0 3000. O 330O. O 40OO. 0 4SOO. 0
PREDICTED-*, MEASURED"*
LOCATION 23
HYDROORAPH STATISTICS FOR LOCATION S3
VOLUME
CUBIC FEET
PREDICTED,
TOTAL TIME
MEASURED,
TOTAL TIME
PREDICTED,
OVERLAPPING
TIME
MEASURED,
OVERLAPPING
TIME
DIFFERENCES,
ABSOLUTE
X OF MEAS
0
0.
0
0.
0.
44397E+09
71232E+09
44289E»O9
71232E+O?
26943E+09
INCHES
3.
6.
3.
6
2.
37.
945
330
936
330
394
S24
PEAK
TIME.HR
2112.
2112.
2112.
2112.
0.
000
OOO
OOO
OOO
OOO
FLOW
FLOW
149.
5OO.
149.
SCO.
350.
70.
, CFS
90S
OOO
9OB
OOO
O92
018
DURATION NO.
START, HR END, HR LEN8TH, HR POINTS
0. 000 4350. 000 4350. 000 199
0. OOO 432O. OOO 4320. OOO 181
0. OOO 4312. OOO 4312. OOO 197
0. OOO 4320. 000 432O. OOO 181
Figure 18. Total predicted flow' hydrograph and total measured flow hydrograph
for Cypress Creek verification.
-------
from the surface. The 10-yr SOS Type II design storm for Tallahassee, Florida
is used for the rainfall input (Figure 19). This storm is characterized by
very high rainfall between hours 11 and 12.
In order to illustrate the influence of a high water table, runs were
made with and without the groundwater subroutine. Table 1 shows the disposi-
tion of the rainfall when a high water table is simulated as opposed to when
it is ignored. Note that evaporation is about the same, and the difference in
the amount of infiltrated water shows up as a direct difference in surface
runoff. (The runs were halted before all water had run off.) The two hydro-
graphs and the corresponding water table (for the run in which it is simu-
lated) are shown in Figure 20. A larger difference in peak flows would have
resulted if the flows had not been routed to a very large channel. Also, note
that the two hydrographs are identical until about hour eleven into the simu-
lation, when the simulated water table rises to the surface.
TABLE 1. FATE OF RUNOFF WITH AND WITHOUT HIGH WATER TABLE SIMULATION
Inches Over Total Basin
Water
Budget
Component
Precipitation
Infiltration
Evaporation
Channel flow at inlet
Water remaining in channel
Water remaining on surface
Continuity error
With Water Table
Simulation
8.399
6.637
0.103
1.495
0.015
0.150
0.001
Without Water Table
Simulation
8.399
1.731
0.104
2.407
0.038
4.124
0.005
Execution time on the IBM 3033 mainframe increased from 0.32 CPU seconds
without the groundwater simulation to 0.42 CPU seconds with the groundwater
simulation. Thus, some additional computational expense can be expected.
CONCLUSIONS
Although the subroutine is fairly simple in design and has several limi-
tations, the new groundwater subroutine should increase the applicability of
SWMM. Preliminary test runs have determined it to be accurate in the simula-
tion of water table stage and groundwater flow. Further calibration and veri-
fication tests need to be done on other areas to confirm these preliminary
results. Also, estimation of parameters, although fairly numerous, appears to
be relatively uncomplicated. In addition, parameters are physically-based and
should be able to be estimated from soils data. The flexible structure of the
algorithm should permit a more realistic simulation of catchments in which a
major hydrograph component is via subsurface pathways.
96
-------
02)
tuoaj
9~[qi3q. aeq.'BM
eoisjjns
SMil
PUB MO"[J
(OOOl/lX
seqcnsea
J63.BM
jd
*02
aaivnnwis
ION aovjunsans
uStsep
aoj
gog
^uaiuqoq.'eoqns -[-BOTq.9qq.od/q aoj
•6L
O
NI awn
O 02
0 "91
0 'SI
O '8
O'O
— »»— 1 J 1 1 —
*****************
******
******
*
*** 4
* <
*
*
*
*
****
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
#
*
*
*
*
*
»*4
*************<
********
****
***
t***
*
>
T T
r wu v/
t
I
:
t
i
'.
OOB 'O
':
[
[
- OO9 I
[
[
[
[
[
[
:
:
[
- OO* 'Z
OOZ 'E
OOO 'V
UH / NI
NI
-------
ACKNOWLEDGMENTS
This work was supported by EPA Cooperative Agreement CR-811607. The
authors thank Mr. Robert Dickinson for technical support.
REFERENCES
1. Bear, J. Dynamics of Fluids in Porous Media. Elsevier, New York, 1972.
2. Bouwer, H. Groundwater Hydrology. McGraw-Hill, New York, 1978.
3. Capece, J.C., Campbell, J.C. and Baldwin, L.B. Estimating peak rates and
volumes from flat, high-water-table watersheds. Paper No. 84-2020, Amer-
ican Society of Agricultural Engineers, St. Joseph, Michigan, June 1984.
4. Carlisle, V.W., Hallmark, C.T., Sodek, P.,Ill, Caldwell, R.E., Hammond,
L.C. and V.E. Berkheiser. Characterization data for selected Florida
soils. Soil Science Research Report No. 81-1, Soil Science Department,
University of Florida, Gainesville, Florida, June 1981. 305 pp.
5. Bawdy, D.R. and T. O'Donnell. Mathematical models of catchment behavior.
J. Hydraulics Division, Proc. ASCE. 91(HY4)i123-137, July 1965.
6. Downs, W.C., Dobson J.P. and Wiles, R.E. The use of SWMM to predict run-
off from natural watersheds in Florida. In; Proceedings of Stormwater and
Water Quality Model Users Group Meeting, Orlando, Florida, EPA/600/9-
86/023, Environmental Protection Agency, Athens, Georgia, March 1986, pp.
109-120.
7. Fetter, C.W., Jr. Applied Hydrogeology. Charles E. Merrill, Columbus
Ohio, 1980.
8. Gagliardo, V. A subsurface drainage model for Florida conditions. ME
Project Report (unpublished), Dept. of Environmental Engineering Sci-
ences, University of Florida, Gainesville, Florida, 1986. 46 pp.
9. Heaney, J.P., Huber, W.C., Downs, W.C., Hancock, M.C. and C.N. Hicks.
Impacts of development on the water resources of Cypress Creek, north of
Tampa. Publication No. 89, Water Resources Research Center, University of
Florida, Gainesville, Florida, January 1986. 355 pp.
10. Hillel, D. Introduction to Soil Physics. Academic Press, Orlando, Flor-
ida, 1982.
11. 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. 531 PP-
98
-------
12. 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, Environ-
mental Protection Agency, Athens, Georgia, 1980. 684 pp.
13. Laliberte, G.E., Corey, A.T. and R.H. Brooks. Properties of unsaturated
porous media. Hydrology Paper No. 17, Colorado State University, Port
Collins, Colorado, November 1966. 40 pp.
14. South Florida Water Management District. Permit information manual. Vol-
ume IV. Management and Storage of Surface Waters. South Florida Water
Management District, West Palm Beach, Florida, January 1984.
15. van Schilfgaarde, J., ed. Drainage for Agriculture. Agronomy Series No.
17, American Society of Agronomy, Madison, Wisconsin, 1974.
99
-------
SWMM APPLICATIONS FOR MUNICIPAL STORMWATER MANAGEMENT;
THE EXPERIENCE OF VIRGINIA BEACH
by: John A. Aldrich, P.E.
Camp Dresser & McKee
Annandale, Virginia
John E. Fowler, P.E.
Department of Public Works
City of Virginia Beach, Virginia
ABSTRACT
The Stormwater Management Model (SWMM) plays a significant role in the
stormwater management program of the City of Virginia Beach, a rapidly
growing municipality covering 250 square miles in southeastern Virginia.
Numerous subdivision designs are based upon SWMM simulations performed by
consultants to land developers. SWMM use has increased due to the need to
evaluate complex hydraulic phenomena present in flat coastal areas, and
because of easier access to sophisticated computer models and microcomputers.
As a rule, RUNOFF is used for hydrologic predictions while EXTRAN is used to
size storm sewers, culverts, and detention basins. Guidelines have been
issued for SWMM studies of development projects that must be approved by the
City, and submittal and review of SWMM input datasets is now required.
Problems encountered by municipal public works departments when reviewing
land development designs based on computer simulations will be discussed.
SWMM is also the principal planning tool being used to develop a storm-
water master plan for the City of Virginia Beach. The master plan will
recommend structural and nonstructural control measures for peak flow control
and nonpoint pollution management under ultimate development conditions.
SWMM is ideal for this study because of the importance of backwater, flow
reversals, interconnected canal and lake systems, and tidal boundary con-
ditions. A key use of the EXTRAN Block is the evaluation of the primary
drainage system (40 mi), which consists of several large interconnected
canals controlled by three major tidal outlets with different boundary
conditions. SWMM has been enhanced for this study to accept multiple
boundary conditions and simulate channels of irregular cross-section input in
a HEC-2 format. Upon completion of the master plan, the enhanced SWMM model
and master plan data sets will be used by the City to evaluate changing land
use patterns, to establish tailwater design conditions for drainage projects,
and to evaluate individual development proposals.
100
-------
INTRODUCTION
The City of Virginia Beach in the southeastern, or Tidewater, portion of
Virginia, is the fastest growing city on the east coast (Figure 1). A wide
diversity of land uses is found within the. borders of this 250-square mile
City, including dense suburban development in the north and northwest,
commercial and retail space along the toll road corridor, high rise hotels
along the beachfront, several major military bases, a rapidly developing
region in the center of the City, and vast farm land and unspoiled wetlands
in the southern half of the City. Development is encroaching rapidly on the
Back Bay, a large estuary whose water quality has been declining in recent
years. Environmental concerns have prompted the establishment of a planned
limit to development—the Green Line—which currently precludes most
development in the Back Bay Watershed.
Hydraulically, the stormwater conveyance system in Virginia Beach is
characterized by an interconnected canal system which provides primary
drainage for over half of the City. This system has three major boundaries:
Chesapeake Bay; the Elizabeth River, a tributary of the Chesapeake Bay; and
Currituck Sound in North Carolina. Many of these streams flow through major
freshwater and saltwater wetlands, constraining potential channelization
projects. The primary stormwater management controls currently used in
Virginia Beach are on-site detention ponds serving large subdivisions.
The purpose of this paper is two-fold. First, the experience of the City
in stormwater management is presented. A key highlight is the effort
required by the City to review stormwater facility designs based upon SWMM
simulations submitted for individual subdivisions. In several cases, the
review process has been hindered because models have been misapplied or
designs have been based upon erroneous results. The City has issued
guidelines on the use of SWMM for subdivision design and is considering the
submission and analysis of SWMM run streams as a component of subdivision
review.
The need for a stormwater model to aid in subdivision review, coupled
with the rapid growth of the City, prompted the development of a stormwater
master plan for the City. The second purpose of this paper is to present the
key issues involved in the development of the master plan, the techniques
required for modeling the stormwater conveyance system, modifications to SWMM
required for the model, and future uses of the master plan model by the City.
VIRGINIA BEACH STORMWATER MANAGEMENT PRACTICES
CITY SUBDIVISION REVIEW PROCEDURES
Virginia Beach regulates stormwater management of new development through a
three stage subdivision review process. Requests for rezoning are evaluated
based on general impact to the City drainage system, impacts to wetlands, and
location of the project with respect to known floodplains. As development
proceeds, a hydrology/hydraulic study for an entire subdivision site are
prepared by consultants for the developer and reviewed by the City. The
City's review ensures that overall site drainage is designed according to
City standards and criteria. Finally, detailed subdivision site plans are
101
-------
figure 1. Virginia Beach Watersheds and Primary Channels
102
-------
reviewed, with attention given to storm sewer design, inlet/outlet sizing and
location, and adequate site grading. Improvements to existing stormwater
facilities required for existing development are capital improvement projects
(CIP) evaluated and designed by the City's Special Projects section and
financed through the General Fund.
USE OF SWMM FOR SUBDIVISION DESIGN
Each year the City Plan Review Bureau receives a limited number of
subdivision hydrology studies performed for developers by engineering
consultants. Approximately six were received in 1986; however, the number is
increasing each year.
These subdivision studies typically rely upon SWMM to model systems of
interconnecting detention lakes for large (greater than 100 acres) proposed
subdivisions. The Virginia Beach Department of Public Works (VAB-DPW)
reports that few SWMM studies are approved on the first submittal, largely
because a "final" version of the SWMM modeling study, with a voluminous
printout and limited documentation, is submitted without prior consultation
with the City engineering review staff. To ease this problem, the City
issued guidelines for hydrology studies based upon SWMM. These guidelines
can be summarized as follows:
• A clear schematic of the drainage system and tables of parameters
for each subcatchment, channel, and lake must be provided,
• Either 12-hour design storms with wet antcedent moisture conditions
or 24-hour design storms with dry antecedent moisture conditions may be
simulated, and
• Equivalent conduit calculations must be documented and EXTRAN
simulations must be shown to be numerically stable.
VAB-DPW has found that problems with subdivision hydrology studies using
SWMM result from four main factors:
1. Failure to coordinate the SWMM study activities with City
engineering review staff prior to submittal; i.e., the guidelines are not
followed.
2. Lack of understanding of the theoretical basis of the program,
specifically that stability criteria must be satisfied when approximating
solutions to the governing St. - Venant equations with the explicit finite
difference methods used by EXTRAN.
3. Failure to check the SWMM results for either numerical errors,
printed error messages, or accuracy versus simple hand calculations.
4. Use of SWMM for subdivision design in lieu of easier to apply models
which are more appropriate for certain drainage analyses simply because SWMM
is considered to be a "trendy" and sophisticated tool.
103
-------
The following example illustrates some of the problems encountered by
VAB-DPW with subdivision hydrology studies based on SWMM. The SWMM modeled
hydrology study for a proposed subdivision (430 acres, 9 detention lakes
outfalling to an existing ditch) was first submitted to the City without
prior coordination on input data. The study was disapproved six weeks later
(the long review time due to insufficient documentation of the study) for 12
reasons, primarily involving RUNOFF block data and large continuity errors.
The study was modified and resubmitted in six weeks. Large continuity
errors, for some runs on the order of 35%, still existed. The consultant
submitted hand calculations attempting to show the continuity error was due
entirely to EXTRAN's failure to compute the initial volume. However, after
accounting for this volume of water in the wet detention system continuity
errors were still between 13% and 16%. After many discussions and meetings,
the study was approved—approximately five months after the first submittal.
VIRGINIA BEACH STORMWATER MASTER PLAN
PRODUCTS OF THE STORMWATER MASTER PLAN
A stormwater master plan analyzes the watershedwide impacts of stormwater
runoff and proposes an appropriate mix of controls to alleviate stormwater
impacts. A master plan focuses on an overall framework of management
alternatives which may include the following:
• Regional detention systems,
• improvements to the primary stream or sewer within a subbasin (about
200-300 acres),
• Improvements to major stream crossings on this stream, and
• Nonstructural measures within the subbasin, such as flood plain
zoning, land acquisition, land use controls, etc.
The objective of master planning is to locate facilities and propose
management schemes which provide the greatest benefits, minimize capital and
O&M costs, and provide the greatest environmental sensitivity throughout the
entire watershed. Master planning usually does not address local subdivision
and highway drainage systems since these are typically evaluated for lesser
design events and seldom influence watershedwide drainage individually. The
primary concerns of the master planning study are the interactions between
large stormwater facilities.
SWMM was selected for the master plan study because EXTRAN allows dynamic
simulation of interactions between the major facilities proposed for storm-
water management. On-site stormwater controls typically limit peak flows to
predevelopment levels. Analysis seldom proceeds outside the subdivision,
where the timing and duration of peak runoff from on-site controls may
interact with runoff flows from other parts of-the watershed to increase
downstream flows and water surface elevations. In many cases, these dynamic
interactions cause flooding of the downstream conveyance system, causing
backwaters which may be severe enough to impact the performance of the
on-site facility.
104
-------
For regional facilities, design is typically based on more severe
rainfall conditions and larger drainage areas, thus the potential for adverse
impacts due to regional interactions is increased. By dynamically predicting
the time history of flow and water surface level throughout the entire
watershed, EXTRAN is able to identify and indicate solutions to adverse
conditions caused by hydrograph interaction, backwater, varying tidal
boundary conditions, and flow diversions.
KEY ISSUES OF THE VIRGINIA BEACH STORMWATER MASTER PLAN
Virginia Beach is currently in a period of rapid growth. Figure 2 shows
the recent and planned growth within the past five to ten years in a typical
watershed in the City. This growth is stressing both the hydraulic
conveyance and the environmental quality of drainage systems throughout the
City, a phenomenon which is drawing increased public awareness to the costs
of rapid growth. Master planning for the entire City is desired to determine
the overall capacity of the City drainage system, identify capacity problems
under existing and ultimate land use conditions, and develop controls which
solve these capacity problems. The City has found that on-site drainage
controls and a strict program of subdivision review cannot ensure that the
impacts of rapid growth are controlled.
Water quality protection is a serious concern in the City. The major
estuaries within the City are severely stressed, with water quality
perceived to be worsening under development pressure. Much new development
in Virginia Beach includes lakes with permanent pools which provide
recreation and aethstetics as well as drainage control. Thus many existing
and proposed lakes, ponds and detention facilities may provide pollutant
removal if operated and maintained properly. The master plan will propose
detention facilities suitable for pollutant removal as well as drainage
control.
Virtually all major streams within Virginia Beach flow through freshwater
and estuarine wetlands. Many drainage projects and a few subdivision
proposals have been denied under Federal wetlands regulations. Federal
opposition has arisen from channelization projects through wetlands, drainage
diversions out of wetlands, and drainage bypasses of lakes providing wetlands
protection to the stressed estuaries. Because of the increased emphasis on
wetlands protection, the cost to study, design, and construct major drainage
improvements has increased dramatically. Thus alternatives based upon
regional detention storage and various non-structural measures (flood plain
zoning, land acquisition, land use controls, down zoning) are proving to be
attractive. These measures correspond well with the current mood of many
residents demanding controls and limits to growth, but may be resisted by
major developers and large land owners unless compensation is provided.
Currently, drainage projects for existing development are financed from
the City's general fund. These capital improvements are budgeted annually.
Developers must design and construct stormwater facilities on new development
sites. These stormwater facilities are reviewed by the City and
traditionally have been deeded to the City following construction. Recently,
the City has advocated continued private ownership of stormwater facilities
since insufficient public funding is available for O&M of these facilities.
These O&M requirements are increasing as regulations and environmental
105
-------
LEGEND
DEVELOPED PRE-1980
DEVELOPED 1980-1986
(INCLUDES COMMITTED
DEVELOPMENT)
DEVELOPMENT UNDER
ULTIMATE CONDITION
UNDEVELOPED UNDER
ULTIMATE CONDITION
Figure 2. Development Trends - Redwing Lake Study Area
106
-------
constraints promote detention/retention ponds as the bulk of stormwater
facilities within the City. Regular programs for routine maintenance of
ditches and channels are performed regularly by the City's Highway
Department. Maintenance requirements for detention ponds which provide
effective water quality control have not been established by the City, nor is
this maintenance typically funded either publiclly or privately. Therefore,
feasible institutional and financial methods for funding stormwater
management are another focus of the stormwater master plan. Existing and
potential City ordinances and regulatory proceedures will be explored, as
will innovative financing methods for the construction and maintenance of
stormwater facilities. Innovative financing methods include stormwater
utilities, which impose a fee on 'users' of stormwater facilities, and pro
rata share assessments, where new development pays a portion of the costs for
regional stormwater management controls.
SWMM MODELING APPROACH
Watershedwide Models
Twenty-five major watersheds were identified in Virginia Beach. Regions
where maximum flood elevations are caused by coastal storm surges were
excluded from the master plan since stormwater management controls for
rainfall-runoff flooding would be ineffective. Watershed sizes ranged
from 20 square miles to less than one square mile, with average subbasin size
ranging between 200 and 300 acres. Only those channels and storm sewers
which provide the primary drainage for one or more subbasins are modeled for
the master plan. A separate SWMM model of each watershed was established to
study the performance of the existing regional stormwater drainage system, to
identify flooding locations under existing and ultimate land use conditions,
and to study alternative plans for relief, control, or management of this
flooding. Runoff from Design storms with return periods of 10, 25, and 50
years was predicted to evaluated drainage system performance according to
Virginia Beach design standards. All stormwater management alternatives were
screened to insure that their interactions did not worsen flooding elsewhere
in the City. Water surface profiles under the master plan recommendations
for the 10-year storm will serve as tailwater elevations for subdivision
drainage design. Minimum floor elevations will be set at one foot above the
100-year storm water surface elevation under the master plan recommendations.
City-wide Model
The interconnected canals which comprise the bulk of the primary
stormwater drainage system for the City are a unique concern in coastal areas
such as Virginia Beach. The major north-south drainage divide moves
depending on interactions between tidal conditions, rainfall intensity,
drainage improvements and growth patterns throughout the City. Therefore a
City-wide SWMM model of the entire interconnected canal system was created to
help understand these interactions and permit delineation of detailed master
planning models of smaller areas. The 200-300 acre subbasins from the
watershedwide models were also used in the City-wide model. EXTRAN channels
in the City-wide model were generally greater than 3000 feet long, permiting
a three minute computational time step and simulations of 30-40 hours. For
initial screening of overall hydraulic performance, only culverts shown to be
significant hydraulic constraints were included in the City-wide model. This
decreased computation time and permitted estimation of maximum conveyance of
107
-------
the canals. Spreadsheets were used to compute culvert capacity under various
headwater and tailwater conditions. Comparisons between culvert capacity and
peak channel flows allowed identification of additional stream crossings as
hydraulic constraints.
Tidal Boundary Conditions
Tidal boundaries are an important consideration for stormwater management
in Virginia Beach. Even though the stormwater master plan does not study or
recommend controls for tidal flooding, the interaction between fluvial and
tidal flooding must ensure that flooding in tidal zones is not increased by
stormwater controls for fluvial flooding. For this study, stormwater
management controls were designed based on average annual astronomical or
wind-driven high tides (i.e., the 50-year rainfall event coinciding with a
one-year tide is a 50 year event). The performance of stormwater controls
during tidal flooding events was checked by predicting flood elevations
caused by a design storm surge coinciding with the average rainfall observed
during historical surge events (typically rainfall on the order of a two-year
storm). From this analysis, three zones were identified within each
watershed. The zone of fluvial flooding is bounded downstream at the point
where rainfall-runoff does not raise the water surface above high tide
elevation. The zone of tidal flooding is bounded upstream by the limit of
tidal backwater during a design storm surge event. The third zone lies
between the first two and defines where tidal flooding sets the design flood
elevation but where stormwater controls may be effective during more frequent
events. The stormwater master plan focuses on control in the zone of fluvial
flooding, but investigates benefits of controls in the intermediate zone
where tidal and fluvial flooding is significant.
SWMM ENHANCEMENTS FOR MASTER PLANNING
Several modifications of SWMM were required for watershedwide master
planning:
• Irregular channel cross-sections can now be entered in the same
format used for HEC-2, permitting data transport between hydraulic models,
• Multiple boundary conditions allow specification of different
constant or time-varying tidal stages at different locations within the
model, and
• Variable stage/storage/discharge relationships in RUNOFF and
variable stage/surface area relationships in EXTRAN permit simulation of
lakes, ponds, and detention/retention facilities.
FUTURE USES OF THE MODEL BY THE CITY
A key goal of the stormwater master plan for Virginia Beach was to
establish appropriate algorithms and modeled representations of City drainage
for future planning, evaluation, and design. Extensive documentation of
sources for model parameters is being compiled to aid understanding of the
model and modeling concepts. Spreadsheets and CAD drawing files compiled
during the study will be used by the City to help maintain the model and
perform modifications as growth occurs. The City is considering the
108
-------
submission of SWMM data sets by land developers to assist with review of
subdivision drainage.
DISCLAIMER
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.
109
-------
THE EFFECT OF SUBWATERSHED BASIN CHARACTERISTICS ON DOWNSTREAM
DIFFERENCES IN STORM-RUNOFF QUALITY AND QUANTITY
by: E.G. Brown
U.S. Geological Survey
St. Paul, Minnesota, USA
ABSTRACT
Runoff quantity and quality were calculated in four subwatersheds of
Lamberts Creek, located near St. Paul, Minnesota, during 12 storms in 1985.
Downstream differences in storm-runoff quantity and quality in Lamberts Creek
are affected by four basin characteristics of the subwatersheds; urban land
use (impervious areas), presence of wetlands (surface-water storage), basin
slope, and channel slope. Storm-runoff quantity is smallest in subwatersheds
that have (1) small amounts of urban land use (impervious area), minimizing
surface runoff, (2) gentle basin slopes, impeding subsurface flow, and (3)
large amounts of surface-water storage (wetlands), temporarily retaining storm
runoff. Storm-runoff loading of total suspended solids, total phosphorus, and
total nitrogen is smallest in subwatersheds that have (1) gentle channel
slopes, minimizing channel erosion and (2) large wetland areas, allowing for
retention of loads through sedimentation. Channelized wetlands are not as
effective as unchannelized wetlands in storing storm runoff or in retaining
loads.
INTRODUCTION
BACKGROUND
Lamberts Creek, located near St. Paul, Minnesota, USA, flows into a lake
from which St. Paul obtains its municipal-water supply. During the summer,
the water supply commonly has an undesirable taste and odor that has been
linked to algal species associated with eutrophication of the lake (1). The
eutrophication is the result of sediment and nutrient input from several point
and nonpoint sources, including nonpoint sources in storm runoff from Lamberts
Creek. Therefore the quantity and quality of storm runoff from Lamberts Creek
required assessment. Lamberts Creek channel is a drainage ditch that
initially was constructed to drain wetlands for vegetable farming and
subsequently was used to drain urban areas. Additional urban development is
proposed in the watershed which may cause additional sediment and nutrient
input to the lake from Lamberts Creek.
no
-------
PURPOSE AND SCOPE
Storm-runoff quantity and quality in Lamberts Creek is largely affected
by basin characteristics of the four subwatersheds. This paper presents the
results of a study evaluating the effects of basin characteristics on storm-
runoff quantity and quality during 12 storms in 1985 within each subwatershed.
Storm runoff is defined as the runoff that occurs during the hydrograph
resulting from a storm. The paper addresses the downstream differences in
storm runoff by evaluating the differences in basin characteristics between
subwatersheds. The differences .in selected basin characteristics are analyzed
to determine what effect a particular basin characteristic has on storm-runoff
quantity and quality in Lamberts Creek. The selected basin characteristics
include land use, basin slope, channel slope, soil type, and surface-water
storage. The downstream differences in storm runoff will be determined by
comparing the storm-runoff quantity and quality exported from each
subwatershed. The interpretation and discussion is limited to storm-related
differences in runoff as annual-related differences (which include storm and
nonstorm runoff) are likely different.
METHODS
STUDY AREA
Figure 1 shows the subwatershed basins, data-collection sites, major
storm-sewer outlets, and land use. The data-collection sites were placed at
the downstream and upstream ends of each subwatershed basin: (a) subwatershed
1 is the basin upstream from data-collection site 1, (b) subwatershed 2 is the
basin between data-col lection sites 1 and 2, (c) subwatershed 3 is the basin
between data-collection sites 2 and 3, and (d) subwatershed 4 is the basin
between data-collection sites 3 and 4.
Basin characteristics of each subwatershed are given in table 1. Urban
land use areas are those used for residential and commercial purposes.
Nonurban land use areas are those presently undeveloped, excluding wetlands or
lakes. Vegetated wetlands are the dominant type of waterbody in the study
area except for Goose Lake (located southeast of data-collection site 1).
Lamberts Creek originates as a large storm-sewer system in subwatershed 1 that
discharges into an open ditch at data-collection site 1. Lamberts Creek is
channelized through all wetlands except through the wetland immediately
downstream from data-collection site 1. The storm-sewered urban area
southeast of Goose Lake drains directly into the lake, which has an outlet
into the wetland immediately downstream from data-col lection site 1. Soils in
each subwatershed are classified as either sand, loam, or clay. Surface-water
storage is the maximum surface-water storage capacity within each subwatershed
expressed in centimeters (cm) derived from cubic meters (nP) of water storage
per square kilometer (km ) of drainage area.
Ill
-------
MINNESOTA
ST. PAUL (Study area)
Base from Minnesota Department of Transportation
General Highway Map of Ramsey County, 1977
1 MILE
1 KILOMETER
EXPLANATION
Urban land use Watershed boundary
Wetland and lakes Subwatershed boundary
. . A1 Data-collection site and
Nonurban land use subwatershed number
stream D Major storm-sewer outlet
Figure 1. Location of Lambert Creek watershed and subwatersheds showing
land use and data-collection sites.
112
-------
TABLE 1. BASIN CHARACTERISTICS OF THE LAMBERTS CREEK SUBWATERSHEDS
Basin Characteristics 123
Drainage area (knr)
Urban land use (km )
Wetlands and lakes (km )
Nonurban land use (km )
Basin slope (meters per kilometer)
Channel slope (meters per kilometer)
Channelized wetlands (km )
Un channelized wetlands (km )
Soil type (loam, sand, or clay)
Surface-water storage (cm)
2.77
2.27(82)
0.03 (1)
0.47(17)
1
0.19
0
0.03
loam
2
9.04
4.10(45)
2.01(22)
2.93(33)
3
0.56
0.77
0.63
loam
10
4.80
1.78(37)
1.69(35)
1.33(28)
10
0.64
1.51
0.18
sand
6
2.92
1.32(45)
0.51(17)
1.09(38)
4
1 .2
0.46
0.05
sand
5
number in parentheses is the percent of basin
DATA COLLECTION
Storm-runoff quantity and quality data at each of the four data-
collection sites were obtained during 12 storms in 1985: one in March, three
in April, one in May, one in July, three in August, two in September, and one
in October. Storm-runoff quantity was derived from a continuous record of
streamflow during each storm hydrograph, and storm-runoff quality was
determined by chemical analysis of v/ater samples. Streamflow was calculated
from stream-stage data collected every 15 minutes by use of a relation between
stage and measured flow (2).
Discrete samples for water-quality analysis were collected throughout the
storm hydrograph at the data-collection sites by automatic samplers. Runoff
quality at each site was measured from either the discrete samples collected
during each storm or a flow-weighted composite sample of the discrete samples.
Flow-weighted composite samples represent the flow-weighted-mean concentration
of the discrete samples collected during the event and were calculated by
using an equation based on the theory of "mid-interval determination of
suspended-sediment discharge" (3).
Discrete and composite samples were analyzed for concentrations of total
suspended solids, total phosphorus, and total nitrogen (ammonia, organic,
nitrate, and nitrite nitrogen) according to methods described by Fishman and
Friedman (4). Storm-runoff loads of each constituent, in kilograms (kg), were
calculated by multiplying the volume of streamflow associated with the sample
by the constituent concentration (5).
Storm runoff (in cm derived from m' per km ) and storm-load yields (in kg
per km ) from each subwatershed were calculated by subtracting the streamflow
(in m^) and load (in kg) determined at the site upstream from the subwatershed
from those determined at the site downstream from the subwatershed and
dividing the difference by the subwatershed drainage areas (in km^). Storm
runoff (or runoff quantity) and storm-load yields (or runoff quality) of each
subwatershed are the data used for evaluating the downstream differences in
storm runoff as affected by subwatershed basin characteristics.
113
-------
RESULTS AND DISCUSSION
DIFFERENCES IN STREAM DISCHARGE BETWEEN SUBWATERSHEDS
The hyetograph and hydrograph during the first 24 hours of storm 4 are
shown in figure 2. The figure illustrates the downstream differences in the
shape of the Lamberts Creek hydrograph between data-collection sites. The
shape of the hydrograph at data-collection site 1 is affected by storm runoff
from the impervious area within the highly-urbanized subwatershed 1. The
shape of the hydrograph includes a large sharp peak in discharge occurring
over a short time period which is typical of urban runoff hydrographs (6).
The long and flat shape of the hydrograph at data-collection site 2
(compared to data-collection site 1) is a result of the surface-water storage
capacity in both the unchannelized wetland and Goose Lake. The storm-sewer
discharge from subwatershed 1 enters the wetland, disperses throughout it, and
is temporarily stored. Storm runoff from the urban area in the southeast
section of subwatershed 2 enters Goose Lake, disperses throughout, and also is
temporarily stored.
17 18 19 20 21 2223
Figure 2. Total-hourly rainfall and average-hourly discharge for data-
collection sites during the first 24 hours of storm 4.
114
-------
The shape of the hydrographs at data-collection sites 3 and H have
greater ascending slopes as compared to data-collection site 2 because of
storm runoff that enters the channel, primarily from subwatershed 3. The
increase in stream discharge between data-collection sites 2 and 3 differs
substantially from that between data-collection sites 3 and 4. This large
increase in stream discharge is primarily attributable to storm runoff from
the steeply-sioped, sandy-soil basin in subwatershed 3-
DIFFERENCES IN STORM RUNOFF BETWEEN SUBWATERSHEDS
Storm runoff and storm-load yields from subwatersheds during each of the
12 storms are shown in figure 3. The storm runoff from subwatershed 3 is
generally greater than that from the other subwatersheds during all 12 storms.
The greatest difference in basin characteristics between subwatershed 3 and
the other subwatersheds is basin slope (table 1). Basin slope within
subwatershed 3 is more than twice that of any of the other subwatersheds. The
rate at which subsurface runoff moves within the basin is directly
proportional to basin slope, all else being equal (7). Soils generally are
more permeable on steep slopes than on gentle slopes because fine particles
have been removed from steep slopes allowing subsurface-flow velocity that is
proportional to basin slope. Soils on gentle slopes contain fine particles
which impede lateral subsurface flow because the soil is less permeable (8).
The steeply-sloped basin and sandy soils in subwatershed 3 allow for
substantially greater storm runoff during the hydrograph than that observed in
the other subwatersheds.
The initial hypothesis was that storm runoff would be greatest from the
predominantly urbanized subwatershed 1 because of the amount of impervious
area associated with urbanized basins. However, the relatively flat basin
slope in subwatershed 1 allows the few pervious areas to become temporary
surface-water storage areas of runoff during storm periods. The retention of
runoff in the pervious areas allows for substantial infiltration of runoff
into the soil and, therefore, reducing the amount of runoff leaving the basin.
During the hydrograph period following the rainfall event (36
millimeters) of storm 6, the amount of storm runoff leaving subwatershed 2 (at
data-collection site 2) was less than the amount of storm runoff entering the
subwatershed (at data-col lection site 1). This can be attributed to two
factors related to surface-water storage capacity for the storm: (1) storm
runoff from within the subwatershed was small, as a result of low antecedent-
soil moisture conditions, and a portion of the storm runoff was stored within
the wetland and Goose Lake and (2) a portion of the storm runoff entering the
subwatershed from subwatershed 1 also was stored in the wetland. Surface-
water storage capacity in the subwatershed was substantially greater for storm
6 compared to other storms because of a 21*-day dry period prior to storm 6.
115
-------
TOTAL NITROGEN,
IN KILOGRAMS PER
SQUARE KILOMETER
TOTAL PHOSPHORUS,
IN KILOGRAMS PER
SQUARE KILOMETER
(D
CO
09
d-
1
I
I
o P
b
H.O.
OQ CO
•'J
VO
A
ca
«T
SO
d-
(D
B-
(D
P.
TOTAL SUSPENDED
SOLIDS, IN
« « « « « KILOGRAMS PER
°P?°??°° SQUARE KILOMETER
o o _ _ _
o p p p
o Li to u A
RUNOFF, IN
CENTIMETERS
-i )0
(H
I __
I I I
-------
DIFFERENCES IN STORM-LOAD YIELDS BETWEEN SUBWATERSHEDS
Storm-load yields of all three constituents were generally highest from
subwatershed 4 during the 12 storms. The high yields are probably the result
of channel erosion within the subwatershed because (a) the channel is steeply
sloped, nearly twice that of the other subwatersheds, (b) high flow is common
in the channel as a result of the large amount of runoff entering the channel
from subwatershed 3» and (c) the organic soils in the channel are highly
erosive. Erosion of organic soils is typically associated with high
concentrations of suspended solids, phosphorus, and nitrogen (8). Although
organic soils are present in the channel reaches of subwatersheds 2, 3, and 4
the steeply-sioped channel and high flow in the channel reach of subwatershed
4 results in substantial greater channel erosion compared to the other
channels reaches.
Storm-load yields of all three constituents leaving subwatershed 2 (at
data-collection site 2) during the 12 storms were generally less than the
storm-load yields entering the subwatershed (at data-collection site 1). The
storm-load yields leaving the subwatershed are lower because (a) loads from
subwatershed 1 are retained in the wetland and (b) loads from the urbanized
area within the subwatershed are retained in Goose Lake. Retention of
suspended material in the wetland (and lake) is directly related to a decrease
in water velocity as the water enters the wetland (and lake). As flow
velocity decreases, sedimentation increases (9). Vegetation in a wetland
tends to decrease water velocity beyond that of pooling alone (such as in the
lake) and promotes fallout of suspended material (10). Nitrogen and phosphorus
from the suspended material is deposited within the wetland (and lake) and
removed from the water column (11). The wetlands in subwatersheds 3 and 4 are
not as effective in retaining suspended material because of channelization.
CONCLUSIONS
Downstream differences in storm-runoff quantity and quality in Lamberts
Creek are affected by four basin characteristics of the subwatersheds; urban
land use (impervious areas), presence of wetlands (surface-water storage),
basin slope, and channel slope. Storm-runoff quantity is smallest in
subwatersheds that have (1) small amounts of urban land use (impervious area),
minimizing surface runoff, (2) gentle basin slopes, impeding subsurface flow,
and (3) large amounts of surface-water storage (wetlands), temporarily
retaining storm runoff. Storm-runoff loading of total suspended solids, total
phosphorus, and total nitrogen are smallest in subwatersheds that have (1)
gentle channel slopes, minimizing channel erosion and (2) large wetland areas,
allowing for retention of loads through sedimentation. Channelized wetlands
are not as effective as unchannelized wetlands in storing storm runoff or in
retaining loads.
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.
117
-------
REFERENCES
1. Walker, W.W. Analysis of 1984 monitoring data from the Vadnais Lakes
diagnostic study. Progress Report to the Board of Water Commissioners,
City of St. Paul, Minnesota, 149 p., 1985.
2. Kennedy, E. J. Discharge rating at gaging stations. U.S. Geological
Survey Techniques of Water- Re sources Investigations 3(A10): 59 p, 1984.
3. Porterfield, G. Computation of fluvial sediments. U.S. Geological
Survey Water- Re sources Investigations Book 3 Chapter C3 , 66 p., 1972.
4. Fishman, M. J. and Friedman, L. C. Methods for determination of inorganic
substances in water and fluvial sediments. U.S. Geological Survey
Open-File Report 85-495, 709 p., 1985.
5. Nelson, L. , and Brown, R.G. Streamflow and water-quality data for lake
and wetland inflows and outflows in the Twin Cities Metropolitan area,
Minnesota, 1981-82. U.S. Geological Survey Open-File Report 83-5434,
182 p., 1983.
6. Novotny, V. and Chesters, G. Handbook of Nonpoint Pollution,
Van Nostrand Reinhold, New York, New York, 555 p., 1981.
7. Whipkey, R. Z. and Kirkby, M.J. Hillslope Hydrology, John Wiley and Sons,
New York, New York, p. 121-141., 1978.
8. Bay, R. R. Runoff from small peatland watersheds. Journal of Hydrology
91-112, 1969.
9. Boto, K. G. and Patrick, W. H. Jr. Role of wetlands in the removal of
suspended sediments. In; Proceedings of the symposium on Wetland
Functions and Values: The State of Our Understanding. American Water
Resources Association, Minneapolis, Minnesota, 1978. pp. 479-489.
10. Fetter C.W. Jr., Soley, W. E. , and Spangler, F.L. Use of a natural marsh
for wastewater polishing. Journal of the Water Pollution Control
Federation 50:290-307, 1978.
11. van der Valk, A. G. , Davis, C.B. , Baker, J.L. , and Beer, C.E. Natural
freshwater wetlands as nitrogen and phosphorus traps from land runoff.
In: Proceedings of the symposium on Wetland Functions and Values: The
State of Our Understanding. American Water Resources Association,
Minneapolis, Minnesota, 1978. pp. 457-467.
118
-------
SOME THOUGHTS ON THE SELECTION OF DESIGN RAINFALL INPUTS
FOR URBAN DRAINAGE SYSTEMS
by: Ivan Muzik
Department of Civil Engineering
The University of Calgary
Calgary, Alberta, Canada, T2N 1N4
ABSTRACT
The art of runoff simulation has made significant advances during the
past decade, owing to improvements in computer hardware and software.
Further development would be however impaired if the information content of
rainfall input, used in runoff simulation, is not improved. Dynamics of
rainfall fields is far too complex to be adequately represented by a design
storm. The design storm concept needs to be replaced by continuous simulat-
ion using distributed rainfalls in time and space for input.
INTRODUCTION
One of the key objectives of urban hydrology is the capability of
predicting hydrographs or peak discharges of storm runoff corresponding to a
certain frequency of occurrence. Statistical analysis of long series of
observed runoff events could provide an approach to deal with the problem.
However, measured data are seldom available. It would seem reasonable to
assume that a plausible alternative approach would be to derive the runoff
series from observed rainfall. All that is necessary to accomplish this
task is to simulate the rainfall-runoff process reasonably well. This
approach, formalized perhaps for the first time by Mulvaney(l) in 1847 when
he introduced the rational formula, has become today the main tool of modern
hydrology. Throughout the years it has also become accepted that the
rainfall input, which drives the equations of our mathematical models, is
known, that it can be sufficiently well described. Consequently, the major
effort has been expanded on building and refinement of runoff models while
the study of rainfall characteristics has been relatively neglected, or
perhaps put more fairly, whatever new information about rainfall became
available has been mostly ignored by runoff modellers.
One can reason that urban runoff, in contrast with rural runoff, will
be relatively sensitive to the variability of rainfall input due to the fact
119
-------
that urban catchments consist of many hydraulic components with fast re-
sponse times, such as impervious surfaces and storm sewers. However, it was
not until recently that researchers(2,3,4,5,6) began to study the effect of
storm dynamics on urban runoff. The present paper attempts to discuss the
problem of rainfall input selection for the use in runoff simulation in view
of the presently available information on rainfall properties. A short
overview of the historical development of rainfall and runoff concepts is
given, followed by a discussion of problems related to the determination of
rainfall over a catchment in terms of its temporal and spatial distribution,
and of its frequency. The non-correspondence of rainfall and runoff fre-
quencies is also addressed.
RAINFALL AND RUNOFF CONCEPTS
It may be useful to start the discussion of rainfall-runoff modelling
by recounting first a few facts from the history of hydrology. The reader,
equipped with a historical perspective, may judge for himself the state of
our present day knowledge and practice.
HISTORICAL BACKGROUND
According to Biswas(7) the earliest reference to a rain gage was made
by Kautilya in his book about the science of politics and administration (in
India) written towards the end of the fourth century B.C. However, the
causality of rain and runoff was not recognized for many centuries to come.
A typical view upon the origin of water in streams during the Middle Ages is
offered, for example, by Leonardo da Vinci(7): "...if the body of the earth
were not like that of man, it would be impossible that the waters of the
sea, being so much lower than the mountains, could by their nature rise up
to the summits of these mountains. Hence it is to be believed that the same
cause which keeps the water at the top of the head in man keeps the water at
the summits of the mountains."
It was not until the seventeenth century that Perrault and Mariotte(7)
established a quantitative link between rainfall and runoff. By measuring
both rainfall and river discharge they showed that rainfall was adequate to
supply the water flowing in streams and rivers. A first proposal to predict
discharge on the basis of rainfall appeared in 1850 when Mulvaney published
a paper(7) describing the use of a well known rational formula. Then, it
was not again until 1932 that a new concept of the rainfall-runoff relation-
ship was introduced, the unit hydrography concept. Since the nineteen
sixties the development of rainfall-runoff modelling took on a new face.
This was made possible chiefly by advances made in computer technology and
the proliferation of numerical techniques for solution of unsteady flow
equations. Distributed models became a possibility.
PRESENT-DAY PRACTICE
Today a catchment is viewed upon as a complex- system with multiple
inputs and outputs. Runoff simulation models are in essence logically
arranged mathematical relations between major variables controlling runoff.
These relations can be express in form of state and output equations. The
120
-------
state equations are usually continuity equations and the output equations
are often given in form of a momentum equation or a rating curve. The input
which drives these equations is a hyetograph.
The hyetograph used in most simulation models usually belongs to one of
the following groups: (1) a block of rainfall of constant intensity, based
on IDF curves and the time of concentration concepts, (2) a synthetic design
storm hyetograph, such as the Chicago design storm, and (3) observed point-
rainfall hyetograph. In all three cases the rainfall is usually assumed to
be uniformly distributed over the catchment area, and movement of the storm
is not considered. The frequency of the generated runoff is considered to
be the same as that of the used rainfall input.
It would appear that a gap has developed between our best distributed
models, capable of simulating unsteady flow throughout a catchment, and the
quality of input information we are able to supply to these models. This
view follows from the results of recent studies (3,5) showing that the
dynamic properties of rainfall are significant for the runoff generating
processes in urban setting.
STRUCTURE OF RAINFALL FIELDS
Far from being stationary and isotropic, as depicted by idealized
design storms, the actual rainfall fields are complex dynamic systems
composed of rainfall cells undergoing periods of grows and decay. Three
aspects of rainfall dynamics will be discussed here. The movement of the
storm over the catchment, the areal distribution of rainfall, and the
temporal distribution of rainfall at single points on the ground.
STORM MOVEMENT
The velocity of raincells producing heavy rainfall over urban catch-
ments was observed to vary between 7 m/s and 25 m/s (5,8). The size of the
raincells was calculated by Niemczynowicz (5) to vary between 2.0 km2 and
7.6 km2. The storm movement effects the runoff in such a way that the
maximum discharge is produced when the direction of storm movement coincides
with the main direction of sewers in the catchment, and the velocity of
storm movement is about the same as the sewer flow velocity. According to
Niemczynowicz (5) the peak discharge from a storm moving downstream may be
about 80% higher than discharge which derives from a storm moving upstream,
or about 30% higher than the peak discharge caused by a stationary storm.
Assessment of the probability of occurrence of storm movement in a
certain direction with a certain speed is difficult at the moment. It is
obvious, however, that this probability influences the probability of
occurrence of peak discharge.
AREAL DISTRIBUTION
It is generally recognized that areally averaged intensities are lower
than point intensities. It is not an uncommon practice to base design
storms on areal IDF curves or area reduction factors applied to point
-------
intensities. This approach, however, still cannot account for the actual
spatial distribution of rainfall intensity during a storm event. This
spatial variability is often significant in terms of runoff production (5).
TEMPORAL DISTRIBUTION
As a result of internal dynamics and the storm's movement both areal
and temporal distributions of rainfall intensity are very seldom close to
uniform. Smaller catchments are believed to be generally more sensitive to
temporal (and spatial) variability of rainfall than very large basins. Yen
and Chow (9) defined a "small" basin as such that "...its sensitivities to
high intensity rainfalls of short durations and to land use are not suppress-
ed by the channel-storage characteristics". Most urban catchments will
display a high degree of sensitivity to temporal variation of rainfall.
Only a few researchers have attempted to model the temporal distribution of
rainfall (10,11). Such models can be subdivided into: (a) models assuming
that the rainfall series consists of internally independent random values,
(b) the Markov chain type models, which allow for sequential dependence of
data and (c) the time series models, which try to preserve also trends,
periodicities and persistence observed in rainfall series. Niemczynowicz
(5) gives a review of existing models.
THE DESIGN STORM CONCEPT
Experience shows that for a given rainfall total depth over a part-
icular duration, the hyetograph may differ considerably for different
storms. Yen and Chow (9) stated that "one has to yet find two natural
rainstorms that are identical. Therefore, the design hyetograph for
drainage facilities, which are expected to serve future needs, can only be
estimated through statistical analysis of past records".
The design hyetograph, or the socalled design storm, is a synthesized
rainfall sequence of a duration varying from several minutes to several
hours, supposedly preserving some statistical characteristics of observed
(usually point data) rainfalls. The design storm is assigned a return
interval, usually on the basis of elementary statistical analysis of a
simple parameter, such as the rainfall total or average intensity over a
given duration (IDF curves). The return interval of the computed design
discharge is deemed to be the same.
This procedure, while expedient, has a number of serious flaws. First,
the design storm concept does not account for storm movement or spatial
variability. Second, temporal variation of rainfall intensity is neglected
or oversimplified. Third, the return interval of the design storm is a
fictitious value which is not based on any probability considerations of
occurrence of real sequences of rainfall, storm movement (direction and
speed) and spatial variability. Lastly, the assumed equality of return
intervals of a design storm and the generated runoff is only valid for
blocks of rainfall of constant intensity uniformly distributed in space, and
for hydrographs having exactly the same initial antecedent moisture con-
122
-------
ditions. Since such conditions are unlikely to occur naturally, the design
storm concept does neither allow computation of the correct return interval
of the "design" discharge, nor does it provide for the estimation of its
tolerance limits.
CONTINUOUS SIMULATION
The foregoing discussion shows that conventional frequency analysis,
such as applied to maximum discharge series, cannot be used to analyze
statistical character of rainfall fields. This is because unlike the peak
discharge, which can be considered as a point variable, rainfall is multi-
dimensional, having three space and one time coordinates. Because of this
complexity there is no technique or sufficient data available as yet to
assign a probability value to an observed rainfall event. A number of
researchers (3,5,12) suggested, as a possible way of overcomming the short-
comings of the design storm procedure, to use long rainfall series in
continuous simulation of runoff. This approach allows making statistical
analysis of the output - the discharge series, without actually assigning
probability values to the input rainfall series. With the advent of micro-
computer hardware and software the continuous simulation is likely to
replace the conventional design storm, a one event type of analysis.
CONCLUSIONS
Rain and runoff have been part of man's everyday experience since the
origin of man. Yet, it took over 2200 years, according to preserved written
information, for man to progress from making first measurements of rainfall
to the stage when he realised the connection between the two phenomena, and
was able to make an estimate of runoff from rainfall by means of rational
formula. The rational formula requires only a block of uniform rainfall as
an input. Today, the rational formula is still extensively used along a
number of relatively sophisticated computerized models. While the degree of
sophistication of these models has been steadily increasing over the past
decade, the information content of rainfall input, which drives these
models, has stagnated at the design storm level. The design storm concept
cannot adequately represent the complexity of rainfall fields dynamics. The
return interval of design discharge cannot be inferred correctly from the
design storm. The return interval assigned to a design storm is a rather
fictitious value itself. Future research should concentrate on continuous
simulation, and the use of upgraded information on rainfall input in terms
of its spatial and temporal variations. Continuous simulation, which uses
long series of realistic rainfall events, does not require assignment of a
probability of occurrence to individual events. Furthermore, the pro-
bability of a given design discharge is estimated by statistical analysis of
the simulated discharge series, rather than being considered identical with
the rainfall event.
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 the Agency and no official endorsement should be
inferred.
123
-------
REFERENCES
1. Biswas, A.K. History of Hydrology. North-Holland Publishing Comp.,
Amsterda, 1970. p.336
2. Foroud, N., Broughten, R.S., and Austin, G.L. The effects of a moving
rainstorm on direct runoff properties. Water Resources Bulletin. 20:
87, 1984.
3. James, W. and Scheckenberger, R. Storm dynamics for urban runoff. In:
Proceedings of the Internat. Symp. on Urban Hydrology, Hydraulics and
Sediment Control. University of Kentucky, Lexington, Kentucky, 1983.
4. Ngirane - Katashaya, G.G. and Wheater, H.S. Hydrograph sensitivity to
storm kinematics. Water Resources Research. 21: 337, 1985.
5. Niemczynowicz, J. An investigation of the areal and dynamic properties
of rainfall and its influence on runoff generating processes. Report
No. 1005, Department of Water Resources Engineering, University of Lund,
Lund, Sweden, 1984. 215 pp.
6. Sargent, D.M. An investigation into the effects of storm movement on
the design of urban drainage systems. Public Health Engineering.
9:201, 1981.
7. Mulvaney, T.J. On the use of self-registering rain and flood gauges in
making observations of the relations of rainfall and flood discharges in
a given catchment. In; Proceedings of the Institution of Civil
Engineers of Ireland, 4: 18, 1850.
8. Nimmrichter, P. and James, W. Dynamic of storms on the western shore of
Lake Ontario. In; Proceedings of the Conference on Stormwater and
Water Quality Management Modeling. Toronto, Ontario, 1985. p.29.
9. Yen, B.Ch. and Chow, V.T. Design hyetographs for small drainage
structures. Fournal of the Hydraulics Division, ASCE. 106:1055, 1980.
10. Nguyen, V.T.V. and Rouselle, J. A stochastic model for the time
distribution of hourly rainfall depth. Water Resources Research. 17:
399, 1981.
11. Amorocho, J. Stochastic modeling of precipitation in space and time.
In; Proceedings of Int. Symp. on Rainfall-Runoff Modelling. Mississ-
ippi State Univ. 1981. p.l.
12. Harremoes, P., Jensen, N.B., and Johansen, N.B. A staged approach to
application of rainfall data to urban runoff calculations. In;
Proceedings of the Specialized Seminar on Rainfall as the Basis for
Design and Urban Runoff Analysis. IAHR, Copenhagen, Denmark, 1983.
p.221.
124
-------
FIELD MEASUREMENT AND MATHEMATICAL MODELING
OF COMBINED SEWER OVERFLOWS TO FLUSHING BAY
by: Guy Apicella, Donald Distante, Michael J. Skelly
Lawler, Matusky & Skelly Engineers
Pearl River, New York 10965
and Les Kloman
New York City Department of Environmental Protection
New York, New York 10013
ABSTRACT
Virtually all of the direct loading of conventional pollutants to
Flushing Bay, a tidal embayment connected to the East River (Figure 1),
comes from combined sewer overflows (CSOs). As part of the New York City
Department of Environmental Protection's effort to improve the water quality
of Flushing Bay and Creek, a comprehensive study that included modeling of
CSO discharges and water quality was performed. This paper discusses three
aspects of the overall study: (1) measurement and evaluation of CSO dis-
charge and pollutant loadings in tidally affected outfalls, (2) application
of the Stormwater Management Model (SWMM) to simulate the major CSO, and (3)
development and application of the Storage Pumping Model (SPM) to evaluate
CSO retention for the major CSO discharge.
The method for measuring the net (nontidal) CSO discharge and pollu-
tant loading consisted of profiling velocity over the depth of flow in the
outfall and compositing a sample based on the flow in the depth intervals.
The largest outfall, CS4, which has three conduits or barrels with dimen-
sions of 18.5 ft (width) by 10 ft (height), accounted for more than half the
total CSO discharge and load to Flushing Bay and Creek.
The CS4 system, which has a drainage area of 7409 acres, was modeled
in two stages using SWMM: (1) the upper half of the system, which is not
tidally affected, was calibrated to field survey data collected at sampling
locations approximately 12,000 ft upstream of the outfall bulkhead; (2) both
the upper and lower parts were then verified to data collected at the out-
fall during three surveys. SWMM was applied to evaluate alternative loca-
tions for CSO control; Extran was used to hydraulically analyze in-line
storage within the tidally affected outfall referred to as the Kissena
Corridor storm sewer.
SPM was developed to determine the appropriate capacity for a storage
facility that would pump CSO retained during storms to one of New York
125
-------
SSQHffi^^Hfc
COILLEGE POINJ
QLZDQr=^^^H^
Sc^FHRR°°
BD
r3;—'U=r//
(
LEFEND
OUTFALL
NO.
SIZE
CS-I ll'-O' X 7'-6'
CS-? OBL 13'-9' X B'-O'
CS-3 4 6L 10--6' X 9'-3'
CS-4 3 BL 18--6- X 10'-0l
CS-5 7'-0" X B'-6'
CS-7 DBL B'-O' X e'-O'
CS-B 10' D1A.
CS-9 IB' OH..
CS-IO 12' 01*.
CS-11 22' X 15'
CS-le 60" DJA,
CS-lj !?• DJA.
CS-i« IB' X 14'
OS-15 !?' DJA.
Figure 1. Location of combined sewer overflow (CSO) discharges to Flushing Bay and Creek.
-------
City's water pollution control plants. The SPM results projected that a 40
million gallon (MG) offline storage facility would yield a 58% reduction in
CSO discharge and 73% and 76% reductions in BOD and total suspended solids
loading, respectively. Water quality models of Flushing Bay and Creek pro-
jected a substantial improvement in dissolved oxygen and coliform bacteria
concentrations from such a facility, including disinfection of overflows at
CS4.
INTRODUCTION
The work described herein was performed by Lawler, Matusky & Skelly
Engineers (LMS) under subcontract to URS Company, Inc. The relationship
of the work to the Flushing Bay Water Quality Facility Plan is shown
schematically in Figure 2. Since the CSO discharges to Flushing Bay are
affected by the tide, sampling and flow measurement techniques were devel-
FIELD MEASUREMENTS
SAMPLING 1 ANALYSIS
FLUSHING BAY * CREEK
"• FIELD MEASUREMENT
'"!.» SAMPLING OF CSO«
J EVALUATION OF NET
(NON-TIDAL) DISCHARGE '
.! i POLLUTANT LOADING ..'.
SIMPLE LOAD
GENERATOR
ALL CSO DISCHARGES
(RRMP)
STORAGE PUMPING
MODEL
WATER QUALITY
MODELS
FLUSHING BAY 1 CREEK
PLAN SELECTION
SEWER NETWORK .-.
SIMULATOR
MAJOR CSO* ;'
(SWMM) N
EVALUATION OF
ALTERNATIVES
<«.B.. IN-LINE STORAGE,
REGULATOR MODIFICATION)
Figure 2. Linkage of this paper to Flushing Bay Water
Quality Facility Plan.
127
-------
oped to evaluate the net (nontidal) discharge and pollutant loadings. The
sewer networks of the largest CSO, which accounted for over half the total
CSO discharge, was simulated by using the Storrawater Management Model
(SWMM). The purpose of SWMM was to assist in evaluating CSO abatement
alternatives (e.g., in-line storage) as well as to provide a check on the
simpler Rainfall Runoff Modeling Program (RRMP) used to generate loads for
all CSO discharges as input to the water quality model (1). The Storage
Pumping Model (SPM) was developed to evaluate the relationship between CSO
storage capacity and the reduction in loadings of biological oxygen demand
(BOD) and total suspended solids (TSS). Calibrated mathematical models of
water quality were applied to project the water quality improvement from an
array of alternatives, and a facility plan for CSO abatement was selected.
CSO MEASUREMENTS
For water quality modeling purposes and the planning and design of CSO
abatement facilities, the discharge and pollutant loading from CSOs must be
measured accurately. The typical Flushing Bay CSO outfall consists of one
or more rectangular sewers from 10 to 18 ft wide and approximately 8 to 10
ft. high with the crown above high water elevation and the invert below low
water. Salinity, tidal flow, and tidal stage (which has a mean range of 6.5
ft) extend into the outfall up to the regulators. Because of the many
complex interconnections of sewers and regulators, the combined sewage flow
was reregulated and the outfall was the only location where the CSO
discharge could be sampled and measured.
The method for CSO sampling performed at the shoreline bulkheads was
devised to measure the net discharge and loading into the receiving water
body. Preparation of the sites was required. Platforms were constructed
outboard of the bulkhead for access by the sampling crews (Figure 3).
Probes for velocity and conductivity meters and sampling hoses were attached
to the bottom of stadia rods that were raised and lowered through braces
mounted to the platform.
The measurement and sampling procedures were as follows:
1. Velocity, conductivity, and temperature were measured at depth
intervals generally 1 to 2 ft (or less at the surface).
2. A water sample was pumped from each depth position where the
velocity was positive (into the receiving water).
3. Flow-weighting factors equal to the ratio of the positive flow in a
depth interval to the total positive flow were calculated.
4. Aliquot volumes of sample were composited according to the
flow-weighting factors.
The frequency of sampling during a storm event was geared to the CSO
discharge rate, which depended on the rainfall pattern, ambient tide, and
other conditions. High velocity, turbidity, and negligible salinity that
128
-------
SHORELINE
STADIA ROD
PLATFORM
MOUNTING BRACE
'SAMPLING HOSE*"
PROBES TO VELOCITY
1 CONDUCTIVITY METERS '
BULKHEAD
Figure 3. Sketch of typical CSO sampling station.
occurred over the entire depth just after intense rainfall marked the onset
of a "first flush," when measurements and sampling were performed as rapidly
as possible. Sampling frequency was approximately every 15 to 30 min during
a first flush and 30 to 60 min at other times.
An electromagnetic meter was used to measure velocity profiles manually
at all tidally affected CSO locations. At the largest outfall (CS4), acous-
tic flow-measuring equipment and manual measurements were used for the
second and third surveys; manual measurements alone were used for the first
survey and acoustic equipment alone was used for the fourth through seventh
surveys. The equipment was installed in two sewers with ultrasonic signal
transmission paths at four depth positions in each (Figure 4). Depth and
velocity data were output to a console printer and a data logger.
Comparison of the velocity measurements between the electromagnetic
meter and the acoustic equipment generally showed differences of less than
20% over the range of velocity up to approximately 2 fps. The electromag-
netic meter produced lower readings than the acoustic equipment when veloci-
ties were greater than 2 to 3 fps. The manually deployed electromagnetic
meter was probably in error at these high velocities because of bending of
the stadia rod, which yielded a component of the normal velocity, and/or
debris getting caught on the probe, which interfered with the meter func-
tioning. Measures were taken to minimize these interferences; however,
extremely high outfall velocities necessitated the removal of the stadia
rods to prevent damage.
129
-------
OJ
o
&
If
•m.
$$&&
PLAN VIEW
;^ v _^f wsg^ferk >> -
-------
Acoustic flow-measuring equipment, although expensive to use, has the
following advantages: it measures laterally averaged velocity; it is accu-
rate at velocities (up to 10 fps) that would bend manually deployed rods;
and measurements are automatically recorded at short intervals (e.g., 5
min).
Combined sewers were sampled during one survey at locations upstream of
the tidal effects. The dual purpose of this survey, which was conducted in
the Kissena Corridor system upstream of CS4 (sampled routinely during this
CSO survey), was (1) to compare upstream data with the data collected at the
CSO outfall, and (2) to obtain data from locations that would facilitate the
calibration of SWMM.
Six CSO outfalls were surveyed during three to seven storm events.
Flow-weighted composite CSO samples were analyzed in the laboratory for BOD,
suspended solids, coliforms, and nutrients.
The tidal phase affected the timing of peak flow because flood tide
generally held back the CSO discharge and slack or ebb tide allowed it to
pass. Interaction of the tide and rainfall can cause various sequences of
flushing. For example, an initial flush can be cut short by an incoming
tide and then a flush will occur later during the outgoing tide. Although
the surging turbid discharge did not always occur at the beginning of a
storm, sampling crews usually noticed peak discharge and solids loading that
generally fits the term "first flush."
EVALUATION OF NET DISCHARGE AND POLLUTANT LOADING
The results of the CSO surveys are summarized in this paper for the
largest outfall, CS4, which discharges at the upstream end of Flushing
Creek. Data for other outfalls are presented in LMS 1986 (2). The CS4 out-
fall has a total drainage area of 7409 acres, or 44% of the entire drainage
area to the bay and creek. Sewers convey stormwater runoff from approxi-
mately 20% of this primarily residential area; combined sewers are used for
the remainder.
Two of the three CS4 outfall barrels (width 18.5 ft, height 10.0 ft)
were sampled; the remaining barrel, identified as CS4A, was assumed to be
identical to CS4B. Water quality data for these two sampling stations are
summarized in Table 1. The mean concentrations reflect tidal dilution since
the outfall - in essence the Kissena Corridor storm sewer - is tidal for
approximately 12,000 ft from the bulkhead, and the total mean tidal volume
is approximately 22 million gallons (MG). The fourth and fifth surveys con-
ducted during light rainfall have B0b$ concentrations approaching ambient
Flushing Creek levels. The BOD concentrations for periods of greater rain-
fall, which are relatively low compared to the normal range for CSO (3), are
attributable to the predominance of storm water as opposed to sanitary
wastewater. TSS concentrations show a relationship with total rainfall,
suggesting that stormwater runoff, which flushes solids from the streets as
well as scours deposited solids from the combined sewers, appears to control
the solids loading.
131
-------
TABLE 1. SUMHARy OF WATER QUALITY DATA FOR SAMPLING OF LARGEST CSO OUTFALL
Station
CS4B
CS4C
Survey
1
2
3
4
5
6
7
1
2
3
4
5
6
7
Total
rain
(in.)
1.02
0.80
1.24
0.09
0.31
1.35
2.66
1.02
0.80
1.24
0.09
0.31
1.35
2.66
Total
BOD5
(HK/1)
9.5
23.8
19.5
8.3
10.0
13.8
14.5
8.8
15.2
16.8
7.0
9.4
13.0
16.4
TSS
(mg/1)
67.2
69.0
52.1
12.5
17.3
75.5
55.8
47.2
38.5
70.0
22.5
15.8
47.6
61.4
Mean
Coliforms
(106 counts/
100 ml)
0.85
0.84
1.48
2.35
2.02
1.49
1.52
0.54
0.74
2.18
2.55
1.56
2.07
0.98
concentration*
Ammonia
ta/i)
1.34
2.64
3.66
3.96
1.49
2.75
2.70
3.73
2.96
1.78
3.07
TKN
8.66
7.39
6.81
6.76
4.00
3.97
8.79
4.63
6.29
4.80
4.43
fa/1)
0.65
0.90
0.28
0.53
0.21
0.38
1.04
0.45
0.30
0.50
0.20
0.37
TP
0.86
0.75
1.11
0.52
0.60
0.67
1.02
0.65
Mean ratios
Filtered:
total BQDe;
0.58
0.40
0.56
0.82
0.65
0.55
0.45
0.61
0.68
0.71
0.56
0.62
TVSS:
TSS
0.25
0.54
0.44
0.76
0.59
0.49
0.40
0.22
0.45
0.40
0.67
0.57
0.48
0.43
Fecal:
total
coliforms
0.10
0.43
0.39
0.37
0.19
0.46
0.19
0.12
0.28
0.10
0.20
0.30
0.18
0.22
*Arithnetic average of all samples except for coliforms, which are geometric averages.
-------
The net discharge of CSO to the receiving water body, evaluated from
the velocity measurements at each depth interval, is computed as the summa
tion of flow in each depth interval:
Qnet - Evi di wi
i=l
where
Qnet = net discharge (cfs)
v^ = velocity of depth interval (fps)
dj = depth of interval (ft)
W£ = width of interval (ft)
i = interval number from bottom to water surface
n = total number of depth intervals
For the rectangular outfall sewers, the width is constant:
n
Qnet = * 2>i di
The depth of an interval was normally constant except for the top
interval, which may be less.
A schematic of typical vertical profiles of velocity in CSO outfalls
illustrates two flow regimes (Figure 5). Regime 1 is defined as having uni-
directional velocity at all depth intervals. Regime 2 is characterized as
having an upper layer of seaward flow and a lower layer of landward flow.
The timing of the tide and rainfall dictated the resulting flow regime. In
general, unidirectional (Regime 1) flow was more common, particularly during
flood and ebb and during high rainfall runoff that produced high positive
velocity. Regime 2 was sometimes encountered when rainfall accumulated just
before low water slack, causing CSO bypasses of low density water that
flowed seaward as the higher density ambient (bay or creek) water began to
flood. Heavy rainfall during a flood phase triggered high CSO outflow that
essentially forced the tidal volume out of CS4 with Regime 1 flow just as
the tide approached high water.
The total CSO discharge for a storm survey is computed as the integra-
tion of the net discharge over the time of measurements. To account for the
total CSO discharge, velocity measurements should extend to the low water
following the end of rainfall. The latter surveys covered full tidal
periods that encompass all rainfall during the surveys and provide reliable
data for evaluation of total CSO discharge.
The net pollutant loading is analyzed separately for Regime 1 and
Regime 2 flow. The equations for net pollutant loading are presented below
with reference to Figure 5, the schematic illustrating the two regimes.
133
-------
VELOCITY (Ips)
EBB -2 -1 01 2
REGIME 1 —I ' ' ' t-
Land
FLOOD
REQIME 1
REGIME 2
SALINITY (0(00) DEPTH IN POLLUTANT LOAD Ob/day)
'P 2,0 COMPOSITE 9
Bay
~554«
Low0r Layer*
iblenl Salinity
Upper Layer
Figure 5. Typical vertical profiles of flow in CSO outfalls.
For Regime 1 flow, the calculation of pollutant mass loading is the
product of the net discharge and the concentration of the flow-weighted com-
posite sample.
M = Qnet Ccomp (3)
where
M = mass loading of pollutant (Ib/day)
Ccomp = pollutant concentration of flow-weighted
composite (mg/1)
(The units are different for coliforms and conversion factors are
necessary.)
Regime 2 is analogous to the two-layered transport in partially strati-
fied estuaries. For Regime 2 flow, the pollutant mass loading is computed
separately for the upper and lower layers of stratification.
UL
(4)
134
-------
MLL = QLL CLL
where the subscripts UL and LL refer to upper layer and lower layer, respec-
tively.
The upper layer loading is similar to Regime 1 in that the concentra-
tion is a flow-weighted composite of the depth intervals that have positive
flow. The lower layer concentration is not measured by the CSO surveys but
is computed based on the degree of mixing between the upper layer and the
ambient water. The equation for the concentration of the lower layer
(CLL) is:
CLL = f ca + U-f) CUL <6>
where
f = mixing factor, decimal fraction of
ambient water in the lower layer
Ca = concentration of the ambient water
outside the outfall
Conductivity and temperature measurements yield salinity that is used
as a tracer of ambient Flushing Bay or Creek water. The salinities of the
upper and lower layers are known from the measurements at each depth inter-
val. Water quality data showed that the ambient salinity was generally
found at the deepest sampling point (depth interval 1) within the outfall.
Solving Equation 6 for salinity yields a mixing factor, f, that is then -used
to compute the lower layer concentration for pollutants (BOD, TSS, and coli-
forms). The total mass loading for Regime 2 is the sum of the upper and
lower layers.
The variability of ambient pollutant concentrations was evaluated by
sensitivity analysis: concentrations were reduced by 50% and increased by
100% and the net pollutant loading for several outfalls was computed. The
resulting variation in total pollutant loading for the first two surveys was
less than 5%, primarily because lower layer flow into the outfall was low in
comparison to the total positive flow.
The total discharge, BOD5, TSS, and total coliform loadings for seven
surveys performed at the two CS4 sampling stations are summarized in Table
2.
The results for net discharge, BOD, and TSS loading for sampling sta-
tion CS4B are illustrated for one survey in Figure 6; the rainfall and tidal
stage are plotted vs time in Figure 7. The accumulation of rainfall through
approximately 0200 hr on 5 November caused a substantial CSO outflow just
before the end of flood tide, with a peak discharge of 370 cfs in CS4B
occurring at high water slack. Although the 6005 concentration varied mini-
mally over time, the maximum TSS concentration increased to over four times
its average and resulted in a peak solids loading of almost 300 tons/day.
135
-------
.COMPOSITE CONCENTRATION. MGT.
TIME. HOURS
BO
•0
100
to
0
jraf
JOOOtKB
Joootwe
.nooetoe
NET MASS LOADING. IB/DAY
5
•00 TO • '•
.•"•' ".
Figure 6. Net discharge, composite concentration,
and net mass loading at Station CS4B for
CSO Survey No. 6, 4-5 November 1985.
RAINFALL INCHES
T1ME.HO(5BS
TME.HOUW
Figure 7. Rain and tidal stage vs time for CSO Survey No. 6,
4-5 November 1985,
136
-------
CSO discharge from the upper half of the CS4 drainage area was measured
for the same survey in the outfall (labeled Kissena Corridor storm sewer)
and at the Regulator 40 bypass. The upstream field measurement locations
are approximately 2.5 miles upstream of the outfall. The travel time to the
outfall was measured as approximately 2.5 hr. Beginning at about 0100 hr on
5 November the upstream flow shows a response to rainfall attributable to
the approximately 1200 acres with separate storm sewers (Figure 8). The
bypass from R40 occurs after the Kissena Corridor flow has peaked. The flow
at the CS4 outfall shows an attenuated peak flow with some fluctuation dur-
ing the ebb tide.
The total flow and loadings from the upstream drainage area (referred
to as the upper Kissena Corridor) are compared with the total CS4 outfall
results. Flow from the upstream area is approximately 34% of the flow at
the CS4 outfall; the BOD, suspended solids, and coliform loadings are about
15 to 20% of the pollutant loadings at the outfall. The reasons for these
differences, which are shown in the schematics of the SWMM model network,
are:
• Approximately 40% of the sanitary flow from the upstream
area is conveyed by separate sewers to combined sewers in
the lower Kissena Corridor that are then regulated.
• The portion of the combined sewer flow that is not by-
passed at Regulator 40 and passes along the interceptor
sewer is reregulated and can be discharged to the CS4 out-
fall.
• The lower Kissena Corridor has a greater percentage of im-
perviousness that yields higher runoff and CSO discharge.
STORMWATER MANAGEMENT MODEL (SWMM)
The version of the U.S. Environmental Protection Agency's Stormwater
Management Model (known as PCSWMM3.2) adapted for the personal computer and
distributed by Computational Hydraulics, Inc. (4) was applied to the CS4
system in two parts. First, the upper half of the CS4 or Kissena Corridor
system, which is unaffected by tide, was modeled and calibrated to field
survey data; second, SWMM output from the upper part was used as input to
the model of the lower Kissena Corridor, which was verified to data col-
lected at the CS4 outfall during three surveys.
The order in which the SWMM modules were applied to the CS4 system is
shown in Figure 9; pertinent characteristics are listed in Table 3. Schema-
tics of the SWMM networks for the upper and lower Kissena Corridor CSO sys-
tems are shown in Figure 10. Our approach was to use the Transport module
for simulations of the field survey periods; however, a hydraulic analysis
of regulators and interceptors was performed using Extran. As most of the
regulators in the CS4 system are diversion chambers having transverse or
side-flow weirs, the hydraulic capacity of key regulators was analyzed'to
evaluate any variations in flow to the interceptor as a function of stage
137
-------
CMB
CMC
8U4 CMMHCMC
ROW CM
1ME
KIMCNA OUHHUOH
UP5THEAMRXW8
(MO
now en
aoo
"vW^" oioo
CSO Survey No.6, 4-5 November 1985, comparison of flows
at CS4 and upstream locations.
Figure 8.
138
-------
TABLE 2. SttMAFY OF WATER QUALITY DATA FOR SAMPLING OF LARGEST CSO OUTFALL
Discharge
Station
CS4B
CS4C
Survey
1
2
3
4
5
6
7
1
2
3
4
5
6
7
Flow
(cfs)
40.8
91.2
137.5
0.6
11.2
101.8
202.5
69.7
43.5
48.5
-9.8
-1.3
12.3
44.0
Volume
(MS)
9.7
21.5
25.9
0.2
3.4
35.1
82.7
16.0
9.0
6.9
-4.0
-0.4
4.2
18.1
BOX
Loading
Ib/day)
1544.0
14755.0
12617.8
171.2
434.8
14966.5
15176.6
4880.7
3077.9
6588.8
-226.4
-0.6
3137.6
4478.6
Load
(lb)
570.6
5379.4
3680.2
109.5
202.9
7982.1
9592.9
1738.7
988.8
1460.5
-144.6
-0.3
1673.4
2845.8
TSS
Loading
(Ib/day)
14634.9
67159.0
56511.9
-72.3
993.1
85051.8
87472.4
32769.7
8324.1
27779.7
-3235.5
-40.7
3664.5
25314.9
Load
(lb)
5408.8
24485.1
16482.6
-46.2
463.4
45361.0
55289.8
11674.2
2674.1
6157.8
-2066.7
-20.1
1954.4
16085.5
Total Coliform
Loading
(counts/day)
8.5E+14
1.4E+15
6.8E+15
-8.9E+14
6.7E+14
9.0E+15
7.8E+15
2.3E+15
1.9E+15
4.1E+15
-6.8E+14
-5.0E+13
2.3E+15
2.3E+15
Load
(counts)
3.1E+14
5.1E+14
2.0E+15
-5.7E+14
3.1E+14
4.8E+15
4.9E+15
8.2E+14
6.1E+14
9.1E+14
-4.3E+14
-2.5E+13
1.2E+15
1.5E+15-
Note: Negative discharge indicates landward flow due to tidal effect during period of nfiasurement.
-------
over the weir. Diversion chambers within the study area can be classified
according to three categories, as illustrated in Figure 11. The dependency
of interceptor flow on stage above the weir is summarized for each category
of diversion chamber: Category I, low; Category II, high; Category III,
inversely proportional to weir length.
Figure 9. CSO drainage areas and SWMM order of module application:
I upper Kissena Corridor, II lower Kissena Corridor, III
Bowery Bay. (See Reference 5 for III Bowery Bay.)
Table 3. PERTINENT CHARACTERISTICS OF UPPER/LOWER KISSENA CORRIDOR
SWMM model characteristics Upper Kissena Corridor Lower Kissena Corridor
Drainage area (acres)
Area with separate sewers
(acres)
Percent impervious
Number of regulators
Number of catchment areas
3700
1200
3709
366
43.2
8
13
50.3
12
15
140
-------
Upper Kissena Corridor CSO System
«. fM
'" *m. '" - U "'. _!
o —
Lower Kissena Corridor CSO System
1O •OWMV •*'
Figure 10. Schematics of SWMM model network, upper and lower
Kissena Corridor CSO system.
141
-------
•CATEGORY 1
Dlvnilon Chtmbvr wllh Tr«n»*«r«« W»lr
- Crown «f Interceptor lt>w«r then wolr er«t
Combined ft*w*r
Storm Conduit
Diversion Chamber with Treni«*ree W«k
• Crown of Interceptor Higher then weir creel
Combine S.wer
Owereton Chftir.boi with Sldi Fipw W«k
Combined S*w«f
item ConKuil-^^__x/
Figure 11. Simplified schematic of diversion chambers.
Regulators with a significant flow-stage dependency were defined as
stage-dependent flow dividers (referred to as Type 20) in the Transport
module, which was used to simulate discharge and pollutant concentration.
Water quality constituents (BOD5 and total coliform) were modeled by setting
concentrations for sanitary wastewater and storm water based on previous New
York City data (Table 4):
TABLE 4. POLLUTANT CONCENTRATIONS USED IN SWMM
Sanitary wastewater Storm water
BOD5 (mg/1)
Coliform (counts/100 ml)
130
1.1 x 1Q7
15
1.5 x IO6
The model calibration of flow in the upper Kissena Corridor is illus-
trated in Figure 12. The total CSO volume computed by SWMM is approximately
142
-------
10% greater than the field measurements. The observed peak flow precedes
the model peak by nearly 1 hr. Hourly rainfall data from National Weather
Service gages indicate that the storm tracked over the model study area
prior to reaching the La Guardia rain gage, which was the basis for model
input. The observed BOI>5 concentration of the CSO is compared to SWMM re-
sults with stormwater concentrations of 10 and 15 mg/1 (Figure 13). Ini-
tially, the SWMM simulation with 15 mg/1 fits the observed data; for the
latter part of the survey the 10 mg/1 simulation is closer. This suggests a
diminishing pollutant concentration in storm water for high rainfall accumu-
lations due to a finite pollutant source in the drainage area.
4-5 NOVEMBER STORM
3
u_
ac.
O
Q
K
K
O
O
LU
tfl
500
400 -
300 -
OBSERVED VOLUME: SU MO
MODEL VOLUME: 2WMQ
18 20 22 24 26 28 30 32 34 36
200 -
100 -
D FIELD DATA
TIME (HOURS)
RUNOFF/TRANSPORT
Figure 12. Upper Kissena Corridor flow calibration.
The SWMM model of the entire CS4 system was verified by simulating the
CSO discharge and pollutant loads during three surveys with cumulative rain-
fall of 0.31, 1.35, and 2.67 in. The SWMM results are compared graphically
with the observed net CSO discharge volume, BOD, and coliforra loadings in
Figure 14. It should be noted that a BOD5 concentration of 10 mg/1 in storm
water was used in SWMM for the highest rainfall survey, based on the hy-
pothesis of diminishing concentration with high rainfall accumulations. The
agreement between the computed and observed results demonstrates the accura-
cy of SWMM in modeling the CS4 system over a wide range of rainfall condi-
tions .
143
-------
4-5 NOVEMBER STORM
in
a
o
CD
o
o
a:
-------
Corridor system so that control of one or two specific regulators would not
be sufficient. Thus, the most effective abatement strategy would be to
locate a CSO storage facility near the outfall so that it would encompass
bypasses from all regulators. Second, a hydraulic analysis of the Kissena
Corridor storm sewer as an in-line storage facility was performed with
Extran, by generating backwater profiles for two scenarios involving storage
dams (simulated as transverse weirs). When the in-line storage capacity of
approximately 11 to 17 MG is exceeded, overflows through the systems would
cause CSO to back up over the weir at Regulator 30 (Figure 15). Potential
backups in a sewer system already prone to flooding made this option un-
attractive. Furthermore, velocities of 1.0 fps would occur during typical
overflows so that in-line storage would not be effective in settling solids.
STORAGE-PUMPING MODEL
SPM was developed to evaluate the effectiveness of the CSO retention
facility in reducing the CSO discharge, BOD, and TSS loadings. The theo-
retical development of the SPM computer program entails flow and mass
balance equations for the operational sequence of storage and pumping, as
shown schematically in Figure 16. Hourly CSO inflow is routed to a storage
volume and the settling of suspended solids and associated BOD is computed.
After each rainfall ends, CSO is pumped from the storage facility to a
treatment plant according to the available capacity based on the plant's
diurnal inflow.
A relationship between percent removal of TSS by settling and overflow
rate was developed using settling column results for CSO samples taken at
the upper Kissena Corridor. The following equation relates the fraction of
TSS settled to the overflow rate:
Fsed - 0.80e-°-00042(OR) (7)
where
Fse
-------
40
SCENARIO; Two 8 ft. Storoge Dams at Nodes 18 and 25
I
5
2
en
in
20 -
10 -
0 -
-10
40
30 -
20 -
10 -
0 -
i 1 1 r
9 11 •• 15
DISTANCE FROM CS4 OUTFALL (1000 ft)
SCENARIO: Two 6 ft. Storage Darns at Nodes 17 and 25
I
-10
17
19
T 1 1 1 1 T
9 11 13 15
DISTANCE FROM CS4 OUTFALL (1000
Figure 15. In-line storage backwater profiles for Kissena Corridor
storm sewer.
146
-------
capacity was found to maximize load reduction effectiveness as additional
load reductions diminish for further increase in storage volume. Water
quality models were applied to evaluate the response in dissolved oxygen and
coliforms to various alternatives including an array of CSO storage
capacities for the two largest discharges. Based on those results a 40 MG
storage facility is recommended for CS4 to improve water quality
particularly in Flushing Creek.
VOLUME
M.
to
90.
10.
TluttKini
M.
M.
f .
1 -
Qvaftow
TUeOouil
..
TMCton)
TMCCmn).
Figure 16. Schematic of CSO storage and pumping to WPCP for SPM.
100
M
u
70 •
M -
M -
40 -
10 •
0
CS4 3TCMWX FAC1U1Y
«4 nauec FACILITY
D MTHUrTUm
otntcnrtiu)
t wrmour scmwo
U •flMKnUM
CM>>arr(uc)
wrTHour serriMa
Figure 17. Load reduction vs storage capacity for CS4 storage facility.
147
-------
The work described in this paper was not funded
by the U.S. Environmental Protection Agency and
therefore the contents do not necessarily re-
flect the views of the Agency and no official
endorsement should be inferred.
REFERENCES
1. HydroQual, Inc. Steady-state water quality analysis, development of
Flushing Bay rainfall-runoff model and evaluation of alternatives.
2. Lawler, Matusky & Skelly Engineers (LMS). Task 3.2 report. Measure-
ments and evaluation of discharge and pollutant loadings from combined
sewer overflows to Flushing Bay and Creek. 1986.
3. Field, R. and Struzeski, E.J. Management and control of combined sewer
overflows, J. of Water Eollut. Control Fed. Vol. 44, No. 7, July 1972.
4. Computational Hydraulics, Inc. (CHI). User manual, PCSWMM 3.2. 1985.
5. Lawler, Matusky & Skelly Engineers (LMS). Task 5.2 report. Stormwater
management model and storage pumping model of major combined sewer over-
flows to Flushing Bay and Creek. 1987.
148
-------
ACCOUNTING FOR TIDAL FLOODING IN DEVELOPING URBAN
STORMWATER MANAGEMENT MASTER PLANS
by: Stergios Dendrou, Ph.D. and Kelly A. Cave
Camp Dresser & McKee Inc.
Annandale, Virginia 22003
ABSTRACT
Typical receiving waters in urban storm drainage systems are rivers
(conveyance systems) and lakes (storage). Depending on their size and
hydrologic behavior, they can be treated as true boundary conditions, i.e.,
as conditions not affected by the storm-drainage system. Coastal receiving
waters present a different challenge. First, there is the diurnal tidal
fluctuation, which poses the problem of phasing with the runoff hydrograph.
Second, there is the question of surge-induced coastal flooding and its
coincidence with riverine flooding.
Flood insurance studies, which are concerned with the delineation of
flood zones under existing conditions, use the joint probability method.
Stormwater management studies, with the goal to improve inland flooding
conditions, require a different approach. Such an approach was developed for
the stormwater master plan for Virginia Beach, Virginia. It encompasses the
following steps: delineation of tidal zones affected mostly by the astronomic
tide; delineation of fluvial zones affected exclusively by inland runoff; and
delineation of transition zones usually affected by runoff and base flow but
occasionally affected by inland propagating surges. Stormwater management
alternatives are thus designed to improve conditions primarily in the fluvial
zone. They are also designed to improve conditions in the transition zone
under runoff conditions, with the constraint that they must perform
adequately during extreme surge events.
Procedures are developed for the delineation of the above three zones,
including addressing the question of astronomic tide phase lag, and joint
probability of occurrence of surge and runoff. A separate procedure is also
developed for non-tidal, wind set-up prone embayments. Application of the
methodology to the Virginia Beach stormwater master plan is presented as a
case study.
-------
INTRODUCTION
The effect of urbanization on the rainfall-runoff portion of the
hydrologic cycle is manifested by an increase in peak runoff-rate and total
runoff volume. This phenomenon is primarily due to the increase in connected
impervious areas. Flooding may result from these changed conditions when the
"receiving waters" (i.e., the area where the urbanized watershed ties into
the larger scale natural drainage basin) do not have the conveyance to drain
away the increased peak runoff rate nor the volume to store or temporarily
accommodate the excess runoff volume. Stormwater management plans are
developed to alleviate such conditions.
Coastal receiving waters present a different challenge: they are large
masses of water that are at once more accommodating of large runoff volumes
but which also exhibit their own flooding conditions that require interaction
and coordination with the inland flooding conditions.
Receiving waters are often considered as boundary conditions in storm-
water models such as SWMM. However, accounting for coastal receiving waters
in such models clearly requires more than a simple incorporation of an
"elevation-versus-time" relationship for these boundary conditions. The
purpose of this paper is to present a general methodology to account for
tidal flooding in stormwater studies and its implementation in the context of
the SWMM model. This methodology was developed as part of the Virginia
Beach, Virginia stormwater master plan but is generally applicable to most
coastal communities.
PROBLEM STATEMENT
Coastal communities in the mid-Atlantic states, from New Jersey to
Florida, and along the shores of the Gulf of Mexico, are built on the coastal
plain which is characterized by a flat, almost relief-less topography. These
communities are notorious for their flooding problems. For example, several
communities in the Florida panhandle were flooded three consecutive times in
the same season, the hurricane season of 1985. Yet, all flooding is not
always caused by tidal waves (surges). Inland runoff in the absence of
coastal surges also causes flooding, and the poor natural drainage of the
flat lands ususally exacerbates the problem.
One way to deal with the coastal flooding problem is to avoid development
in these areas. The government provides incentives to this end by offering
affordable insurance to communities participating in the National Flood
Insurance Program. To participate in this program, the communities have to
adopt floodplain management measures to reduce future losses. Flood-risk
zones are delineated to establish the insurance rates. The critical level of
risk is the 1% risk of flooding per year. Equivalently, protection is sought
against the 100-year event. Where many causes of flooding are present, a
composite risk is estimated by means of the joint probability of occurrence.
For example, if a certain flood level is exceeded on the average once every
100 years from riverine flooding (0.01 probability of exceedance) and once
every 100 years from coastal flooding (0.01 probability of exceedance), and
if these are independent, then that level will be exceeded on the average
150
-------
twice in 100 years. That is, the average return period is 50 years. For
events that are not independent, the procedure is amended by use of
conditional probabilities.
The joint probability method (Myers, 1970) is adequate for describing
existing conditions. Stormwater management plans, however, deal with
measures to improve drainage conditions, not just to describe them as they
exist. In Stormwater management plans, therefore, it is necessary to
discriminate between the various modes of flooding, to address each one
separately, to encourage synergisms and to coordinate conflicts.
The questions that need to be addressed are:
(1) What are the areas that are predominantly affected by coastal
conditions?
(2) What are the areas that are predominantly affected by inland
flooding?
(3) What is the extent of the areas that are affected by both coastal
and riverine flooding?
(4) Are there any measures that can improve either coastal or riverine
flooding conditions?
(5) Are there any measures that can improve both coastal and riverine
flooding conditions?
(6) Are there any measures that could exacerbate the situation under
adverse conditions? E.g., a weir or flood wall to protect against coastal
surges that would also impede inland runoff.
METHODOLOGY - APPROACH
The first step in addressing the above questions is the delineation of
the three zones, respectively, of exclusive astronomic tidal influence,
inland runoff influence, and transition zone of occasional coastal surge
flooding. This is accomplished by performing the following tasks:
(1) Determine astronomic tide range, amplitude, and their variation
through the lunar cycle at the closest tidal gage.
(2) Simulate a 25-year storm of duration comparable to the watershed
time of concentration for various phases and amplitudes of the astronomic
tide. From these runs, establish the uppermost extent of propagation of the
astronomic tide. Note whether this section is sensitive to tidal phase
and/or amplitude.
(3) Repeat the same above simulations with a fixed tidal boundary
condition of high tide. Establish whether these simulations produce similar
or identical envelopes of high water marks along the tidal reaches of the
watershed. If they do, then the analysis can be simplified by neglecting the
tidal phasing parameter.
151
-------
(4) Analyze all hurricanes and winter storm surges of record at the
nearest tidal gage. In particular, establish correlation between surges and
associated rainfall.
(5) Simulate several hurricanes and/or winter storms with their
attendant rainfall. Observe maximum extent of propagation of surge.
Establish sensitivity of that location to surge amplitude.
(6) Repeat above simulations with constant downstream boundary condition
at peak surge level. Establish whether these simulations produce similar or
identical envelopes of high water marks along the tidal reaches of the
watershed. If they do, then the analysis can be simplified by not having to
account for the time variation of the surge.
The above determined river cross-sections delineate three zones as
illustrated in Figure 1; namely, the zone of exclusive coastal flooding, the
zone of riverine flooding, and the buffer zone which is occasionally affected
by surges. Design storms or other conventional methods can be used for
stormwater management in the riverine zone, with the additional constraint
that these practices should perform adequately during surge events as well.
This would actually be monitored in the transition zone.
Where representative tidal gages are not available, the above procedure
should be augmented to include simulation of representative surges. A
special case is that of landlocked embayments subject primarily to local wind
set-up. Furthermore, when multiple tidal boundaries exist, the above
procedure should be expanded to incorporate analysis of the interaction of
the various embayments. The above procedure was implemented using the model
SWMM.
APPLICATION - IMPLEMENTATION
Implementation of the above methodology is shown for the Virginia Beach,
Virginia, stormwater master plan. The City of Virginia Beach is located
along the Atlantic coast at the mouth of the Chesapeake Bay, in the south-
eastern corner of Virginia (see Figure 2). This 250-square mile city is one
of the most rapidly growing areas of the country. It is drained primarily by
a complex system of interconnected channels, canals, and lakes, and has few
large storm sewer systems.
The City was divided into 25 major watersheds for master planning
analysis. The majority of watersheds in Virginia Beach drain into coastal
receiving waters; these receiving waters vary significantly in tidal
influence. For example, as shown in Figure 3, the watersheds in the northern
section of the City are bounded by a large bay system which drains into the
Chesapeake Bay; thus, these watersheds are directly influenced by the tides
and the storm surges of the Atlantic Ocean. Several watersheds in the
eastern section of the City are bounded by the Elizabeth River estuary, which
is also influenced by the astronomic tide fluctuations of the Chesapeake Bay.
In the southern section of the City, several of the eastern watersheds are
influenced by wind-driven tides in the landlocked Back Bay. The watersheds
in the southwestern section of the City drain to the North Landing River,
which ties into the Atlantic Ocean.
152
-------
I igure I
Interaction Of Coastal flooding
With Urban Runoff
-------
MARYLAND^
WEST
VIRGINIA
Chesapeake
VIRGINIA
VIRGINIA
BEACH
NORTH CAROLINA
SOUTH
CAROLINA
Figure 2- Virginia Beach Location Map.
154
-------
Chesapeake
Elizabeth
River
LEGEND
if TIDAL INFLUENCE
Figure 3. Location of Tidal Boundaries for Virginia Beach,
155
-------
The variety of tidal boundary conditions in the Virginia Beach area
present an interesting application of SWMM to study city-wide interactions.
Because the primary drainage system for a large portion of Virginia Beach is
interconnected, it is important to understand the operation and sensitivity
of the entire interconnected drainage system under a variety of scenarios in
order to provide good master plans.
An analysis of tidal conditions and typical surge events with corre-
sponding rainfall was performed as the first step for delination of tidal
influence zones. The highest tides at four long-term control tidal stations
near Virginia Beach are shown in Table 1; summary information on Norfolk,
Virginia (adjacent to Virginia Beach) rainfall occurring at the time of the
highest tides is shown in Table 2. Comparison of these tables shows that all
high tide events were accompanied by rainfall. Table 3 lists the temporal
distributions of each of the rainfall events given in Table 2 for which
hourly data are available. As can be seen from the tables, no consistent
rainfall pattern can be generalized because of the random nature of each
event. For example, some storms are fully advanced while others are delayed;
in addition, most events are multi-peaked. From Table 2, the average
duration of rainfall events associated with the highest tides near Virginia
Beach is 16.5 hours and the mean total rainfall is 2.3 inches. Thus, the
average rainfall intensity is approximately 0.14 inches per hour; this
intensity relates to a 16-hour storm having a return period of just under
2 years when compared to the Virginia Department of Highways and Trans-
portation (VDH&T) published intensity-duration-frequency curves for Norfolk,
Virginia. Based on this analysis, the joint probability method would be used
with conditional probabilities to account for a surge/rain coincident event.
For example, the conditional probability of occurrence of a 50-year return
period tide (0.02 probability) with a 50-year return period_rainfall event
that would actually be even smaller than 0.02 x 0.02 - 4.10~4; that is,
extremely small. Rather, surges should be combined with rainfall amounts
that usually occur during such storms. Using the joint probability method
for stormwater planning analysis could lead to over-design of stormwater
control facilities.
An example of the tidal reach delineation methodology previously
presented is shown for Watershed 7 (see Figure 3), which drains into the
Chesapeake Bay. Watershed 7 has been almost completely developed into single
family residential land use. The most upstream subwatersheds drain into a
small lake controlled by a weir; just downstream of this lake is a series of
box culverts. A schematic of the model setup is shown in Figure 4.
To delineate the (channel zone) area of Watershed 7 subject to tidal
fluctuations, a typical astronomic tide was applied as the downstream
boundary condition for SWMM analysis. For one scenario, this tide was
determined from the Norfolk, Virginia tide gage for September 27, 1956 (see
Figure 5). A second scenario applied a constant 1.83 foot-tide (i.e., high
tide for 9/27/56) as the downstream boundary condition. The timing of the
astronomic tide was set such that the peak runoff coincided with high tide
in the first scenario and with low tide in another, in all scenarios, the
initial water level throughout the drainage system was set at 1.83 feet.
A 25-year, 24-hour Soil Conservation Service (SCS) type II design storm was
applied to the watershed in the RUNOFF block of SWMM. Figures 6 and 7 show
that the high water mark at most locations was virtually the same for the
156
-------
TABLE 1. HIGHEST SURGES AROUND VIRGINIA BEACH, VIRGINIA
(ELEVATIONS IN FEET NGVD)
STORH
NAME
ST 14
SI 18
ST 113
ST 113
FLOSSY
DONNA
DORA
6LORIA
CHARLEY
NORFOLK
DATE
09/19/28
08/23/33
09/16/33
09/18/36
10/05/48
04/11/56
09/27/56
10/06/57
10/21/58
09/12/60
10/21/61
03/07/62
09/13/64
10/14/77
04/27/78
10/25/82
09/27/85
08/17/86
TIHE
N/A
N/A
N/A
10:18
11:42
22:06
02:12
08:30
17:36
06:24
19:54
10:24
06:06
10:12
00:30
04:36
04:48
N/A
ELEV
W N/A
N/A
N/A
7.03
4.93
6.03
5.43
5.23
4.73
5.33
4.63
6.83
4.93
4.86
5.87
5.38
5.03
N/A
HAMPTON
TIHE
00:00
09:00
18:00
09:48
12:00
22:00
02:00
08:00
17:18
06:12
19:00
10:00
15:48
09:42
00:00
04:00
04:54
22:00
ROADS
ELEV
4.52
7.22
5.32
5.92
4.62
5.52
5.12
4.82
4.52
5.12
4.42
6.42
4.82
4?62
5.61
5.10
4.25
4.52
CHESAPEAKE BAYtttt VA BEACH
TIHE
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
08:54
23:00 II
03:18
05:00
02:24
ELEV
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
5.87
6.24
5.81
6.10
5.37
TIHE
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
18:00 t
13:06
N/A
N/A
N/A
N/A
N/A
N/A
ELEV
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
4.12 t
4.22
N/A
N/A
N/A
N/A
N/A
N/A
t High tide of 4.6 feet NGVD recorded 06:00 10/22/61
t* High tide occurred on 4/26/78
HI N/A indicates data not available
till Elevations in feet-HLH (gage not tied into NGVD)
157
-------
TABLE 2. RAINFALL ACCOMPANYING HIGHEST SURGES
VIRGINIA BEACH, VIRGINIA
STORM
NAME
ST
ST
ST
ST
#4
#8
#13
#13
FLOSSY
DONNA
GLORIA
CHARLEY
START
DATE
09/18/28
OS/ 22/33
09/15/33
09/17/36
10/04/48
04/11/56
09/26/56
10/05/57
10/21/58
O9/ 11/60
10/21/61
03 / O6/62
09/13/64
10/14/77
04/27/78
10/24/82
09/27/85
08/17/86
TIME
'?'?
04
14
21
13
23
18
18
01
07
06
18
13
10
N/A
N/A
N/A
N/A
: 00
: OO
: 00
: 00
: 00
: 00
: 00
: 00
: 00
: 00
: 00
: 00
: 00
: 00
END
DATE
09/19/28
08 / 2O/33
09/16/33
09/18/36
10/05/48
04/12/56
09/27/56
10/06/57
1O/21/58
09/12/60
10/22/61
03/07/62
09/13/64
1O/ 14/77
04/27/78
10/25/82
09/28/85
OS/ 18/86
TI
12
06
08
10
00
08
09
07
23
22
18
10
06
04
STORM
DURATION
ME (MRS)
N/A
N/A
N/A
N/A
: OO
: 00
: OO
: OO
: OO
: 00
: 00
: OO
: OO
: 00
: 00
: OO
: OO
: OO
N/A
N/A
N/A
N/A
15
27
19
14
12
10
16
14
23
15
13
17
18
18
TOTAL
RAIN
( IN. )
•-'
1
1
4
1
1
2
>—i
~)
3
0
0
4
1
0
o
5
1
.57
.31
.59
.06
.90
.85
.57
.10
.25
.81
.53
. 79
. 73
.10
.24
.28
.65
.08
MEAN
NOTE: ONLY DAILY RECORDS AVAILABLE PRE-1948
16.5
158
-------
TABLE 3. HISTORICAL STORMS ACCOMPANYING HIGHEST SURGES
(RAINFALL IN HUNDREDTHS OF INCHES)
TINE li
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
50
60
70
80
90
110
120
TOTAL
)/48 0
1
1
3
5
8
8
8
18
19
19
12
31
33
19
5
190
4/56 0
5
8
6
8
4
6
5
7
5
6
2
3
11
3
0
22
21
14
11
3
8
13
5
5
2
1
1
0
0
0
0
0
0
0
0
185
19/56 1
1
3
1
1
12
25
9
26
27
32
17
17
24
25
18
1
12
5
1
257
STORH DATE
0/57 10/58 09/60 1'
4 1 9
3 17 13
1 24 25
7 21 87
4 5 111
32 5 55
27 37 43
16 49 13
13 13 18
17 10 7
7 41
22 2
43
14
210 225 381
o/6i o;
1
20
6
7
5
0
1
1
1
0
0
4
3
2
1
1
53
3/62 0
4
4
8
8
8
8
8
11
10
3
2
T
4.
2
1
79
9/64 1
21
12
29
17
28
7
14
17
35
34
12
49
52
44
27
19
13
14
15
6
2
3
3
473
0/77 0'
2
14
45
9
0
0
6
1
1
0
4
8
11
5
4
110
1/78 1
2
3
4
1
0
1
1
4
4
2
1
0
1
24
0/82 0
5
3
0
10
21
9
12
12
14
14
24
43
9
16
11
14
11
228
9/85 C
34
16
33
11
72
23
5
7
14
26
33
5
36
60
95
78
16
1
565
18/86
0.1
0.1
1
15
14
7
10
13
0
4
6
2
1
7
10
10
7
1
0.1
108
159
-------
0
120
9325
LEGEND
STORAGE NODE
NODE
ROAD CULVERT
SUB-BASIN
CHANNEL
LYNNHAVEN
BAY
7205 ^ 7225 X 7245
220^240
Figure 4. Model Representation for Watershed 7
-------
-1
NORFOLK TIDES
SEPTEMBER 27, 1956
D ASTRONOMIC TIDE
TIME (HOURS)
TOTAL WAVE
O
UJ
4 -
-2
NORFOLK TIDES
SEPTEMBER 27, 1985
40
TIME (HOURS)
TOTAL WAVE
60
Figure 5. Tides Used In Case Study.
80
161
-------
o
20
15 -
10-
III
ui
u.
~ 5H
z
o
UJ
_l
111
0-
-5-
-10-
Figure 6
WATERSHED 7 - MAXIMUM WATER SURFACE ELEVATIONS
25-Year,24-Hour SCS Design Storm
LEGEND
(l60) NODE NUMBER
PEAK RUNOFF WITH ASTRONOMIC HIGH TIDE;
INITIAL WATER SURFACE ELEVATION = 1.83'
CONSTANT 1.83' TIDE
CHANNEL INVERT
2000 4000 6000 8000 10000 12000
DISTANCE UPSTREAM FROM SYSTEM OUTFALL (FEET)
14000
-------
20 -\
15 -
10 -
111
III
u.
*•* 5 -\
s OH
u
-5 -
-10-
Figure 7
WATERSHED 7 - MAXIMUM WATER SURFACE ELEVATIONS
25-Year,24-Hour SCS Design Storm
LEGEND
NODE NUMBER
PEAK RUNOFF WITH ASTRONOMIC LOW TIDE;
INITIAL WSE = 1.83'
i•••CONSTANT - 0.97' TIDE
CHANNEL INVERT
I
0
2000 4000 6000 8000 10000 12000
DISTANCE UPSTREAM FROM SYSTEM OUTFALL (FEET)
14000
-------
astronomic tide and the constant tide scenarios. In addition, a plot of the
high water surface elevations over time for Junction 220 (see Figure 8) shows
the high water mark to be the same for a constant tide and for a time-varying
tide. That is, for purposes of determining the envelope of high water marks
along the tidal reaches of the watershed, it is sufficient to set the
downstream boundary condition at high tide.
The transition zone influenced by occasional storm surges was delineated
by applying the September 27, 1985 storm (Figure 5) as the downstream
boundary condition for SWMM analysis. This storm, also known as Hurricane
Gloria, was chosen because it produced a high storm surge and was also
accompanied by 5.65 inches of rainfall, making it one of the most severe
storms to hit the Virginia Beach area. Both the RUNOFF and EXTRAN blocks of
SWMM were used with actual rainfall and tide data; the peak rainfall occurs
almost simultaneously with the storm surge for this hurricane. A second
scenario applied a 2.34 foot constant boundary condition to the system; this
elevation was chosen because it was shown to be the typical high astronomic
tide for several days preceding the storm. Figure 9 illustrates that
Junction 250 (i.e., the downstream end of the box culvert system) was the
limit of the zone affected by the storm surge. In addition, a plot of the
water surface elevations over time for Junction 250 (see Figure 10) shows the
high water mark to be the same for both a constant tide and for a time-
varying tide.
Thus, it can be concluded that Watershed 7 is relatively insensitive to
the timing of the tide and a constant tidal boundary may be used for analysis
of stormwater management alternatives. This represents a significant
simplification. However, for larger systems such as the interconnected
citywide drainage system in Virginia Beach, such a simplification cannot be
made.
Analysis of other watersheds subject to open water boundary conditions
such as the Chesapeake Bay and its estuaries was completed following this
methodology. The watersheds which are bounded by the landlocked Back Bay in
the southeast corner of the City deviate from this methodology, however,
because the water levels in Back Bay are influenced primarily by wind set-up.
In the absence of astronomic tides and/or surges, that boundary condition is
established by estimating the local wind set-up. Factors affecting wind
set-up are: wind speed, fetch length, water depth, and duration. For a
large, deep body of water, set-up increases in proportion to the square of
wind speed (i.e., wind energy), and to the fetch length. Set-up also
increases with duration up to the steady-state, fully arisen state. The time
to steady-state set-up is longer for larger fetches. For closed bodies of
water, the wind set-up is limited by the size of the basins (limited fetch),
and by depth. The maximum sustainable set-up is reached relatively rapidly.
Careful examination of the Back Bay system reveals that there are
principally two bodies of water: namely, North Bay/Shipps Bay in the north;
and Redhead Bay/Back Bay in the south. These water bodies are separated by
Long Island and linked through stable and deep channels at Great Narrows
(Figure 11). The northern system has a north-south fetch length of about
6 miles and an average depth of 3.5 feet. The southern system has an
approximate fetch of 12 miles and on average depth of 4.4 feet. For
164
-------
a) CONSTANT TIDE - 1.83 FEET (NGVD)
?
2
b
£
o
l.SO -
1.89 -
1.88 -
1.87 -
1.86 -
1.85 -
1.84 -
1.82 -
1.81 -
1.SO -
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
TIME (HOURS)
b) ASTRONOMIC TIDE
-0.8
TIME (HOURS)
Figure 8. Effect of Constant vs. Time-Varying Astronomic Tidal
Boundary Condition at Node 220 in Watershed 7.
165
-------
Figure 9
WATERSHED 7 - MAXIMUM WATER SURFACE ELEVATIONS
20
15 -
5 io-
o
z
Ul
Ul
u.
** 5 -
" 0 -J
ui
-5 -
-10 -
LEGEND
NODE NUMBER
TOTAL WAVE FOR HURRICANE GLORIA
INITIAL WSE = 2.34'
CONSTANT 2.34' TIDE
CHANNEL INVERT
WEIR
LAKE
i i i i i t i
2000 4000 6000 8000 10000 12000
DISTANCE UPSTREAM FROM SYSTEM OUTFALL (FEET)
14000
-------
ELEVATION (FEET. NGVD)
ELEVATION (FEET. NGVD)
CO
o :
Q.
01
O
0>
'
O
Q.
I
O
c
en
O
-s
to
(B
O
>
m
IT
c
o
>
z
m
o
o
O
O
z
CO
>!
2
O
m
ho
m
m
o
<
D
-------
20 Miles
SCALE-MILES
210 2
6 Miles
12 Miles
Figure 11. Major Wind Set-Up Zones of Back Bay, Virginia.
(City of Virginia Beach, 1984)
168
-------
comparative purposes, Table 4 shows estimates of the wind set-up for each
subsystem separately.
TABLE 4. SIGNIFICANT WAVE SET-UP HEIGHT
(Bretschneider, 1966)
Body of Water
North/Shipps
Redhead/Back
Bay,
Bay
Fetch
Miles
6
12
Depth, Significant height
ft Wind 20 30 40 60
3
4
.5
.4
<1 1
1.2 1
.2
.6
1
1
.5
.8
1.9
2.2
, ft
80 m/hr
2.
2.
2
6
Wave set-up in both above subsystems is depth-limited rather than
fetch-limited (i.e., basins with the same size but deeper would sustain a
higher set-up). The most important finding is that the deeper Back Bay can
sustain a higher set-up such that a 0.5-ft gradient can exist between North
Bay and Back Bay. Sustained flows through Great Narrows can therefore cause
higher elevations in North Bay if sustained winds prevail over many hours.
Thus, the following combination of boundary and initial conditions was used:
a flat 2-ft level representing a frequent (at least once a year) highwater
level in North Bay; and a 3-ft level representing more extreme conditions.
Because the wind set-up is relatively insensitive to wind speed, and because
storms can last long enough (2-3 days) to fill the entire system, tidal
boundary conditions should be set at 2 ft msl for North Bay/Shipps Bay and 3
ft msl for Redhead Bay/Back Bay.
Watershed 9 (Figure 3) is an example of a watershed with Back Bay as the
downstream boundary condition. Back Bay was modeled as a node with a
constant water surface elevation of 2 ft msl; this elevation is likely to
occur about once a year and was thus selected for the master plan analysis
for this basin.
CONCLUSIONS
Stormwater management for coastal communities requires consideration of
the interaction of inland and coastal causes of flooding. Dealing with the
probability of their joint occurrence is only one part of the problem, as is
the incorporation of "level-as-a-function-of-time" boundary conditions in
SWMM. Rather, the questions to be addressed are:
(1) delineation of zones exclusively influenced by inland runoff;
(2) delineation of zones influenced by the astronomic tide;
(3) determination of zones occasionally influenced by surges; and
(4) development of measures to alleviate inland flooding which do not
exacerbate coastal flooding conditions.
We have presented a methodology that is practical, reliable, implement-
able, and generally applicable to most coastal communities along the Atlantic
169
-------
and Gulf Coasts. An example of application was shown for two typical water-
sheds of Virginia Beach, Virginia. A city-wide SWMM network was also
developed to study drainage system interactions. A modified version of SWMM
which includes multiple boundary conditions was used to accurately model the
diverse boundary conditions for this system. Work with the citywide model is
ongoing; results will be presented in further publications.
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
view of the Agency and no official endorsement should be inferred.
REFERENCES
1. Bretschneider, C.L. "Wave Generation by Wind; Deep and Shallow Water,"
Estuary and Coastline Hydrodynamics. A.T. Ippen, ed., McGraw-Hill Book
Co. Inc., New York, New York, 1966. pp. 192.
2. Chow, V.T. Open Channel Hydraulics. McGraw-Hill Book Co. Inc., New
York, New York, 1959.
3. City of Virginia Beach, Virginia. "A Management Plan for the Back Bay
Watershed." Prepared by Roy Mann Associates, Inc., May 1984.
4. Federal Emergency Management Agency. Flood Insurance Study. Prepared
for the City of Virginia Beach, Virginia, July 17, 1984.
5. Myers, V.A. "Joint Probability Method of Tide Frequency Analysis." ESSA
Technical Memorandum WBTM, Hydro 11, April 1970.
6. Soil Conservation Service. "Urban Hydrology for Small Watersheds."
Technical Release No. 55. U.S. Department of Agriculture, Washington,
D.C., January 1975.
7. Virginia Department of Highways and Transportation. "Drainage Manual."
1985.
8. Weather Bureau. "Rainfall Frequency Atlas of the United States."
Technical Paper No. 40. January 1963.
170
-------
WASTELOAD ALLOCATION FOR CONSERVATIVE SUBSTANCES
by: Main R. Hutcheson
Oklahoma Water Resources Board
Oklahoma City, OK 73152
ABSTRACT
The Environmental Protection Agency is implementing its third round
strategy for National Pollution Discharge Elimination System permits. A
primary goal is to develop permits which protect toxics criteria in
water quality standards. This requires a wasteload allocation (W. A.)
that mathematically predicts the amount of substance which may be
allowed in an effluent without violating in-stream numerical criteria.
While wasteload allocation methods for conventional pollutants
(oxygen demanding substances, for example) are well established and
widely used, W. A.'s for conservative substances (toxic metals, for
example) are only now being considered. The most common W. A. for
conservative substances uses the assumption that the pollution is mixed
uniformly across a stream. Since most states standards require that a
zone of passage be maintained across a stream, the use of the uniform
mixing assumption may result in hundreds of miles of the nation's waters
being in violation of water quality standards.
A W. A. has been developed which protects the zone of passage while
retaining the desirable features of the mass balance assumption: a
minimum of input data required and ease of computation. The development
of the W. A. is discussed, the assumptions upon which it is based are
examined, and the analytical nature of the W. A. explained.
171
-------
WASTELOAD ALLOCATION FOR CONSERVATIVE SUBSTANCES
INTRODUCTION
The EPA is now issuing their "third round" of NPDES permits to
confirm that aquatic life is being adequately protected on a
site-specific receiving stream basis, and the need for a viable
wasteload allocation (W.A.) for conservative substances is obvious. A
conservative substance'remains in the water column and does not undergo
chemical alteration. A viable wasteload allocation will yield permit
limits which protect instream water quality standards for many toxic
substances.
REQUIREMENTS WHICH SHOULD BE MET BY A VIABLE WASTELOAD ALLOCATION
An economical wasteload allocation should use input parameters
which may be obtained without the necessity of data collection
specifically for allocation purposes. State and federal permitters do
not have the resources to perform special measurements every time a
permit is drafted. The main advantage of the mass balance allocation,
which incorporates an assumption of complete mixing of effluent in the
receiving stream, is that it requires only the background concentration,
C • the stream flow, Qu; the effluent flow, Q£; and the water quality
standard, C. If Cg is unknown it may be assumed zero. Qu is either the
low flow value obtained from USGS analyses or the minimum flow at which
numerical water quality standards apply. C is the concentration of the
conservative substance allowed in the receiving stream. Most permitters
are experienced in obtaining these input parameters. Furthermore, there
is no reason to believe that a more resource intensive wasteload
allocation will yield more accurate permit limits on a routine basis.
Because it is undesirable to require data collection for an
allocation method which will be routinely used, the dispersion equation
(upon which an allocation for conservative substances is based) must be
analytical, rather than empirical. An empirical model requires field
measurements for calibration each time it is used. Wasteload
allocations for conventional parameters (such as dissolved oxygen) are
empirical and, therefore, very resource intensive.
The W. A. should not be based on a premise which leads to water
quality standards violations. This is not the case for the mass
172
-------
balance allocation in states which prescribe a zone of passage. A
simplistic depiction of the zone of passage is presented in Figure 1.
Usually it consists of a fraction of the flow volume or cross sectional
area in the receiving stream. Since the mass balance allocation is
based on the assumption of complete mixing, the standards will not be
protected in those portions of the zone of passage where complete mixing
has not occurred. Therefore, the wide use of mass balance allocations
creates the potential for standards violations in hundreds of miles of
the nation's waters.
//////// //////A ////////////////
Figure 1. Mixing zone, zone of passage and plume dispersion in a
receiving stream.
Since numerical standards are enforced only in the zone of passage,
a regulatory mixing zone, where numerical criteria are not applicable,
is created (Figure 1). Any resemblance between the regulatory mixing
zone and the mixing zone created by dispersion of a conservative
substance in a receiving stream is purely coincidental. The dispersion
is represented in Figure 1 by isopleths of concentration, with C( being
the maximum and C4 the minimum concentration depicted. The maximum
concentration on the boundary between the regulatory mixing zone and the
zone of passage is C2; and this point is labeled Cmax.
DERIVATION OF THE WASTELOAD ALLOCATION EQUATION
The law of conservation of mass may be used to develop a wasteload
allocation which will protect the zone of passage. In Cartesian
coordinates, conservation of mass may be expressed as (1)
99 9(eWx)
at ax
8 (9Wz)
(A)
where & (x, y, z, t) is the instantaneous concentration of the
conservative substance in the receiving stream, and W , W and W are
instantaneous stream velocities in the x (downstream along the bank), y
(vertical) and z (transverse) directions.
If assumptions of questionable validity are used in the derivation
173
-------
of a W. A. procedure, then a verification program which would tax the
resources of state and federal agencies would be required before the W.
A. could be used for NPDES permitting activities. For this reason, (A)
was chosen as the premise upon which the allocation is based. A rather
complicated set of assumptions is necessary to obtain a useful W. A..
The desired solution to (A) requires the following assumptions:
1. Mass is conserved
The pollutant does not change form chemically or volatilize
during travel from S to CThe pollutant is neutrally
buoyant during this travel time, so it does not settle out of
the water column.
2. Steady state conditions exist.
The effluent flow, effluent concentration and mean ambient
flow must remain constant for a longer period than the travel
time from the source S to the point of maximum concentration
on the mixing zone boundary (C in Figure 1).
3. Mo persistent transverse currents.
While random transverse currents (in the form of turbulence
for example) are necessary for dispersion at the rate
observed, no large whirlpools which create a persistent
transverse current can be tolerated.
4. Complete vertical mixing
In shallow streams, vertical mixing occurs within a few
hundred feet of the discharge (1). If the fraction of the
flow allocated to the zone of passage is sufficiently small,
C will be far downstream of the point of complete vertical
mixing.
5. Concentration is half-normally distributed in the transverse
direction.
Yotsukura and Sayre noted that in the Natural coordinate
system concentration distributions are normal in the
transverse direction (1).
6. Negligible reflection from the far bank
If the fraction of flow allocated to the zone of passage is
sufficiently large, then only a small fraction of the mass of
the conservative substance will have even reached the far bank
at Cmax'
174
-------
Stream flow remains constant between S and C
lu&X
There can be no flowing tributaries or significant water
withdrawal between S and C . Only larger tributaries flow
during critical conditions.
The discharge is a point source located at the bank.
Most discharges are via pipes which project a negligible
distance into the stream and do not have enough velocity to
produce a transverse flow at C
9. Stream depth and velocity change gradually.
This assumption is more valid for tranquil valley streams than
for turbulent mountain streams. Fortunately, there are
relatively few discharges to turbulent streams.
10. The background concentration is constant
No sources or sinks of pollution exist between S and C
11. The dispersion coefficient is constant in the vicinity of
max*
While the dispersion coefficient is not constant close to the
source, if the fraction of the flow allocated to the zone of
passage is small enough, then C will occur far enough
downstream so that the assumption is valid.
Using these assumptions, an analytical solution to (A) may be
obtained (2) :
where c is the steady state concentration, W is the wasteload (the
product of the effluent flow and concentration, i.e. W= C Q where C
is the effluent concentration) °" (x) is the concentration standard
deviation in the transverse direction, q is the stream flow between the
injection bank and the point in the stream where the concentration is c
(Figure 1 shows that at the far bank q=Q, the total flow in the stream
and at the mixing zone boundary q = CL). Equation (B) is underspecified
because both c and ff are unknown.
The solution (B) yields the concentration of the conservative
substance at any point in the receiving stream. However, for wasteload
175
-------
allocation purposes, the only point at which the concentration must be
known is at C. Figure 1 shows that if the standard is not exceeded
at Cmax ^ will n?t be exceeded anywhere in the zone of passage, because
at every other point the concentration is less than C2.
The concentration distributions along the mixing zone boundary ana
along the transverse cross section A, A1 are displayed in Figures 2a
and 2b, respectively. Figure 2a shows that the gradient of the
concentration distribution is zero at the point of maximum concentration
(3). Since the only other variable in (B) which is dependent on x is
a, (B) may be differentiated w/r to x to obtain
-------
ANALYSIS OF THE MIXING ZONE WASTELOAD ALLOCATION
If C is large in relation to the water quality standard, Cg may be
very small or even negative. This is caused by substituting the water
quality standards for c in (C) . For water quality management purposes,
C should never be required to be smaller than the standard.
ree quart
_ ^u + ^
Many states reserve three quarters of the stream flow for a zone of
passage. In this case
(C) may be rearranged by substituting for Qm and setting Cg = 0 to obtain
max
= .5165
+Qe*
(D)
where
Q * — Q_/n
e — — e/ \i
max
is tne water quality standard.
r /r
Figure 3 shows ^e'^max plotted against Q *. When Q * is large
(effluent flow is large in comparison with the stream flow), the
effluent concentration (C ) allowed by the mixing zone allocation (solid
line ) is small. When the effluent and stream flows are the same size
(Q * = 1), the effluent is required to nearly meet water quality
standards (C /C —*>!.). When the effluent is much less than the stream
flow (Q *«1), then Co is allowed to be much greater than the standard
6 18.00-1
17.00-
18.00-
16.00-
14.00- -\
13.00-
12.00-
11.00-
10.00-
I
o
9.00-
8.00-
7.00-
6.OO-
6.00-
4.00-
3.00-
2.00-
1.00-
0.00
0.00
1
0.10
0.20
—I
0.30
0.40
—I
0.60
—I
0.60
0.70
0.80
0.80
1.00
Figure 3.
Mixing zone waste allocation (solid line) and technology
based permit limit (dashed line). Q * is the ratio of the
effluent discharge flow to the upstream dilution flow.
C /C is the the ratio of the effluent concentration to
the water quality standard.
177
-------
e' max»l). The mixing zone allocation accounts for the dilution
capacity of the stream by allowing a higher effluent concentration when
the dilution capacity is greater.
If the dilution capacity is extremely large (Q *^»1), then the
mixing zone wasteload allocation allows the effluent concentration to
be almost unlimited. In this case, a technology based limit is
required. In Figure 3, a technology based limit is represented by a
dashed line. Because it does not depend upon dilution capacity, C /C-^x
is constant. For water quality management purposes, the more stringent
of the technology based or water quality based permit criteria should be
used to limit effluent concentration. In Figure 3, if, Q ^.025, a
technology based permit is appropriate. Otherwise, the permit limit
obtained from the mixing zone wasteload allocation should be used.
It may be shown that the mixing zone allocation always yields a
more stringent permit limit than the mass balance allocation does. If
CB = 0, the mixing zone allocation may be written as (2)
Ce
C Qe* (E)
A comparison of (D) and (E) reveals that the mixing zone allocation is
nearly twice as stringent as the mass balance allocation.
CONCLUSIONS
The mixing zone allocation is as simple to use as mass balance, but
does not allow standards violations. It is easily incorporated into a
microcomputer program, which, depending on state standards and
regulations, may incorporate technology based permit limits and other
wasteload allocations. In Oklahoma, a microcomputer accepts input data
which does not require field collection, determines the type of permit
required, develops permit limits, and writes a portion of the
rationale for the discharge permit.
Field validation of the mixing zone wasteload allocation method is
not necessary, because the assumptions upon which it is based are valid
for a natural stream. This is fortuitous since the location of ^max is
unknown and the concentration is changing rapidly in the transverse
direction in its vicinity (Figure 2b).
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.
178
-------
REFERENCES
1. Yotsukura, N. and Sayre W. Transverse mixing in natural channels.
12:695,1976;
2. Hutcheson, M. R. Wasteload allocation for conservative substances.
OWRB 97-2, Oklahoma Water Resources Board, Oklahoma City, Oklahoma,
1987. 33 pp.
3. Gowda, T. Critical point method for mixing zones in rivers. Journal
of Environmental Engineering 110:244, 1984.
179
-------
THE USE OF DETAILED COST ESTIMATION FOR DRAINAGE DESIGN
PARAMETER ANALYSIS ON SPREADSHEETS
by: S. Wayne Miles, Thomas G. Potter, and James P. Heaney
Florida Water Resources Research Center
University of Florida
Gainesville, Florida 32611
ABSTRACT
Extensive research in the field of sewer system design has produced many
excellent computer programs based on optimization and simulation techniques.
The use of these programs, however, has been limited because they require
specialized skills in programming and applied mathematics. This paper
describes the computerization of the current Florida Department of Trans-
portation (FDOT) drainage design procedures with the use of spreadsheets. By
replicating the current manual procedure, the spreadsheet permits the engineer
to perform calculations in a familiar format. The spreadsheet design proce-
dure allows the user to vary pipe sizes and slopes manually in a trial and
error method while monitoring the system constraints. The spreadsheet also
automatically provides a detailed cost estimate of the current system based on
the FDOT itemized drainage cost database. The template will update all system
calculations of pipe flow capacities, velocities, pipe elevations, and system
cost estimates with a change in a pipe size or slope.
The advantages of the spreadsheet design procedure over currently used
techniques are discussed. The advantages of using a detailed cost estimate
over a functionalized cost estimate, which may depend on one or two parameter
values, are also discussed. Savings induced by a redesign and cost estimate
using the spreadsheet template on the FDOT Thomasville Highway project are
presented to illustrate the method.
INTRODUCTION
Research in the field of sewer system design has brought steady improve-
ments in design procedures through the use of computerized simulation and
optimization models. The use of these new procedures for actual designs,
however, often lags far behind. Because of precedent and time and money
constraints, many designers continue to use established design procedures.
Often the only considerations of system cost in the design process is through
180
-------
use of rules of thumb and post-design cost reviews. Even while remaining
within the guidelines of a design procedure, many combinations of pipe sizes,
pipe depths, and structure types may be feasible for a working system. The
cost of these feasible systems, however, may vary greatly. If the design
calculations are performed manually, the number of alternatives evaluated may
be severely limited. The intent of this paper is to present a methodology
whereby design procedures are improved and computerized in a manner that will
be accepted by practicing designers. This process will necessarily involve
incremental changes in the current design procedures.
The spreadsheet provides a computerized environment where these incre-
mental changes may take place. In a spreadsheet, a design calculation proce-
dure may be organized into a format that is familiar to engineers who use the
analogous hand calculation procedures. Also the input and output of data is
done in a natural manner. These features enable the computerization of hand
calculation procedures to be done relatively easily on spreadsheets.
For this paper a spreadsheet replication of the Florida Department of
Transportation (FDOT) highway storm drainage design procedures was created.
The Lotus 1-2-3 spreadsheet package was used to computerize the FDOT design
procedures so as to duplicate their design calculations as closely as
possible. The spreadsheet procedure, however, offers many advantages over the
hand calculations. Most importantly, computerization allows the calculations
to be performed many more times and encourages improvement of the design by
trial and error. Also, an automatic cost estimation scheme has been included
in the spreadsheet. A change in the system design will produce a change in
the cost estimate of the design. With this ability, the user may develop a
heuristic algorithm to proceed through the system design. A spreadsheet
design procedure also allows the user to transfer any personal style or tech-
niques which may have been used in performing hand calculations to the
computerized procedure.
LITERATURE REVIEW
The history of computerized optimization of sewer systems dates back two
decades to papers by Liebman (1) and Holland (2). Since this beginning, opti-
mization algorithms have been used which made simplifying assumptions in order
to solve the sewer design problem. Liebman's linear programming algorithm
only dealt with the network layout, and Holland's nonlinear algorithm could
not handle discrete pipe sizes. The most common simplification in these tech-
niques is the use of a cost estimation function. Many of the dynamic program-
ming models use functions to determine cost as a function of pipe size and
invert depth (e.g. Zepp and Leary (3), Merritt and Bogan (!})). In each of
these optimization techniques much time is spent on defining the assumptions
so that errors are minimized. System constraints must also be carefully
defined. The definition of these constraints is a difficult job and Merritt
and Bogan (4) admit that, "It is unlikely that any optimization method
achieves a true optimum when the full scope of a real world setting is con-
sidered."
181
-------
A survey of water resources personnel performed by Austin (5) concludes
that a lack of models that represent "real world" situations was a common
complaint of model users. This survey also showed that simulation models are
much more widely used than optimization models. Simulation models are help-
ful, but are awkward for system design because of their batch run format. A
reliable simulation model used for verification of a simple optimization tech-
nique however, would be very valuable for system design. Spreadsheets have
been used successfully in a wide range of water resources applications from
stormwater permitting and groundwater modeling (Hancock (6)) to model prepro-
cessing (Miles and Heaney (7)). Hancock developed a decision support system
for following stormwater discharge permitting procedures on a spreadsheet.
Spreadsheets have been very useful for database needs and have extensive
calculation and programming capabilities.
BACKGROUND OP STUDY
This study began with the intention of designing an experiment which
would determine whether the use of computerized models in designing stormwater
drainage systems could be economically justified. In order to conduct such an
experiment, past projects must be analyzed and redesigned using new proce-
dures. The search for past projects led to the Florida DOT.
The Florida DOT has drainage design procedures which are very well docu-
mented in their Drainage Manual (8). They also have a large number of past
project designs available in blueprint form as well as planning calculation
form. The most extensive database of their past projects, however, is found
in their cost estimating department. An itemized unit cost which is based on
average bid prices from past projects is available for each highway construc-
tion item. This database is updated every six months and the item numbering
system allows drainage related items to be determined easily.
The FDOT Drainage Manual lists a mainframe Fortran program called
"Draino" (PEGDRG32) as available to assist in drainage design. The program
uses a heuristic algorithm that minimizes pipe costs of a drainage system.
This program is seldom used, however, because of its difficulty in handling
real world problems and its tedious data input procedure. Users find it dif-
ficult to define problem constraints and the program makes the assumption that
all pipes are designed to flow full.
CURRENT DRAINAGE SYSTEM DESIGN PROCEDURE
Presently, Florida Department of Transportation (FDOT) personnel perform
most of their drainage calculations using the worksheet shown in Figure 1.
Explanations of a few of the entries needed on this tabulation form are given
in an excerpt from the FDOT Drainage Manual shown in Figure 2. These calcu-
lations are most commonly performed by hand or with a nomograph. Repetitive
hand calculations are subject to error and do not encourage the engineer to
"push the limits" of the design criteria to find an optimal design. The tabu-
lation form procedure also does not explicitly include system cost as a design
criterion. The tabulation form has become the accepted practice for highway
182
-------
CO
STATE OF FLORIDA DEPARTMENT OF TRANSPORTATION
STORM DRAIN TABULATION FORM SIIEKT NO.
I»ATF I'HlllKCT NO. BOAn COUNTY RV tif
LOCATIOD
OF
llri'KH (Nil
STATION jnlSTJS
/7\
^ IMMg
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
i.
I
K
V)
(§)
IY?I C-
sisrcrrss
Cz)
0
&
e
1
. DRAINACI
AREA (ACM
C- s-^
O
80
O
-M-
(ii)
XJi
S)
r
fpl
i4$i-
ICG 07 CO.XCEM-
ntiiios ocra.)
(0)
i!
®
DTIEISm
©
3
H
©
j h -*1
©
ocfc
(fs>
^S'
ELEV. OF H.G.
CROWN ELEV.
FLOUL1NE ELEV.
h
i^yi
riim *:
ZONK -Hll.UUt NOT VR
n-.0)2 CnMCHKTF. rim * CUI.VF.R1S
d)
H I'M ARKS
T
i(iy
Figure 1. Recommended Storm Drain Tabulation Form
-------
14. Time of Flow in Section (min)
This is the time it takes the runoff to pass
through the section of pipe in question; it
depends on the velocity as well as the condition
of flow (i.e., gradient or physical flow time
based on proper condition and velocity).
15. Intensity
Intensity values are determined from one of the 11
intensity-duration-frequency (IDF) curves
developed by the Department and presented in
Chapter 5 of this volume. Intensity depends on
the design frequency and the time of
concentration.
16. Total (CA)
The total CA is the sum of the subtotal CAs.
17. Total Runoff (cfs)
Total runoff is the product of the intensity and
the total CA, less inlet bypass and exfiltration.
Figure 2. Excerpt from FDOT Drainage Manual (8) showing descriptions
of tabulation form entries.
drainage design. Designs obtained with the use of a computer model or alter-
native method must be compared to the tabulation form design. With the
present time and budget constraints, the prospect of extra work is a deterrent
to the use of modeling. Therefore, any improvement in design procedures must
also include an improvement in the efficiency of the time spent on the design.
The use of a spreadsheet in performing these design calculations may help to
increase this efficiency.
PROPOSED SPREADSHEET DESIGN PROCEDURE
A spreadsheet template has been created that replicates the design proce-
dures used by the FDOT on their tabulation forms. The spreadsheet template is
divided into four areas: input area, initialization -area, interactive design
area, and database. Each of these areas will be described in the following
sections.
184
-------
DATA INPUT AREA
Information is input into the spreadsheet much like it would be written
onto the PDOT tabulation forms (see Figure 3). This information includes the
pipe identifications (to and from nodes), pipe lengths, ground elevation at
nodes, and peak flow values. If the Rational Method is used to calculate peak
flow values, then drainage areas, runoff coefficients, times of concentration,
and a design frequency must also be entered. The spreadsheet can determine
the storm intensity from a time of concentration and design frequency by using
the PDOT intensity-duration-frequency curve regression equations (PDOT Drain-
age Manual 1987). The user must also include the type of manhole or inlet at
each node if this cost is to be contained in the cost estimate used for system
optimization. These costs are important in the cost estimation because struc-
tural costs are given as a function of depth in the PDOT database. A survey
station number and the type of line (main, stub, etc.) may also be included in
the input area for user convenience.
D9: CMS]
MENU
Start [initialize! Recalculate Optisiize Base cost Change
Calculates initial values of pipe systen paraieters.
A B C D E F
1
2
3
4
5
6
7
9
10
11
12
13
14
IS
16
17
18
19
20
DOT Program Sinulator
Input area
Pipe
ID Struc.
Fr To Station Struc. Spec.
62 61 1131+00 Inlet E
61 60 1127+58 Inlet 6
40 58 1126+48 Inlet J-3
59 58 1124+58 Inlet P-2
6 H 1 J K
Enter values only in
areas
Type ten. Incre.
L M
unprotected
Mm
of Len. to area Runoff t.c.
Line (ft) outlt (acr) Coef
M 365 3275 3.80 Cl=
0.00 C2=
0.00 C3=
M 150 2910 22.20
0.00
0.00
M 190 2760 0.00
2.10
2.00
S 76 2646 0.00
0.80
CUD CflLC
(sun)
0.45 10
0.8
0.25
12
0
10
Figure 3. Spreadsheet data input area with menu.
185
-------
INITIALIZATION AREA
Once the input data has been entered into the storm drainage design tem-
plate, the user may choose to use the pipe size and slope initialization algo-
rithm. The initialization algorithm is analogous to the one used by the DOT
"Draino" program (8) and described here. It is not required, however, to
perform the initialization algorithm in the spreadsheet design procedure. The
algorithm first determines the lowest allowable hydraulic grade line elevation
in the system. The minimum slope from this point to the outlet is then calcu-
lated and all pipes downstream from this point are assigned this slope. All
pipes upstream from this point are assigned the ground slope. Initial pipe
sizes are then calculated with respect to these slopes and the design flow. A
macro program may now be executed to assist the user in moving the initial
pipe sizes to the design area.
The flow process in this template is controlled by a Lotus 1-2-3 macro
program. Menus have been created which are much like the menus that execute
the Lotus 1-2-3 commands. These menus, also shown in Figure 3» allow the user
to move easily throughout the spreadsheet and also begin the execution of
macro programs. The Lotus menus also provide a brief description of each
macro choice to which the cursor is moved. The initialization macro is one of
the algorithms which may begin by calling the menu and choosing the desired
macro.
INTERACTIVE DESIGN AREA
The design area in the template is constructed as shown in Figure ij.
This area is designed to keep the most important system design parameters
showing on the screen. Lotus 1-2-3's ability to create column and row titles
is used here. Titles can remain at the top of the screen while the cursor is
moved down to view pipes lower in the network (below row 20). Similarly, the
three columns to the far left of this area (A, B, and C) are also titles and
will always remain on the screen. This feature enables the user to always
know the current pipe identification.
The optimization of the drainage system is based on varying a pipe size
and slope while monitoring the resulting calculations such that they remain
within designated criteria. For example, a pipe may be reduced by one size
and then checked to make sure that its flow capacity exceeds the required
design flow. If it does not, the user may then try to increase the slope of
the pipe to increase the flow capacity while verifying that the velocity does
not exceed the maximum and that the minimum cover is kept above the pipe
crown. This feature allows the engineer to employ personal strategies in a
system design. Excavation can be minimized by following the ground slope as
closely as possible, or steeper grades and smaller pipes can be used in the
upstream regions of the network.
The user may also easily define system constraints in this area. For
example, if an existing utility constrained a pipe invert to be at a 100 foot
elevation, then the cell formula which had previously calculated the invert
186
-------
BN9: [W6] +INVERT+D1AHETER/12
READY
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
A B CBH BI BJ 8K BL BH BN BO BP BQ BR BS BT
Old systei cost= * 108112
New systei cost= * 108112
Pipe laterial = rep III
T Guttr/Crown Cronn D
Pipe Y Tot. Full Grate Invrt Invrt Brnd Pipe R Pipe
ID P Flo* Flow Vel Elev. Upper Lower Slope Slope 0 Dia. Pipe Struc
Fr To E cfs cfs Ifps) (ft) (ft) (ft) S P (in) Cost Cost
62 61
61 60
60 58
59 58
144.9 133.5
M 8.87 12. 10.0 148 143.6 132.3 0.032 0.031 1.5 15 6872. 4360
133.3 130.9
H 56.6 63. 12.9 136 130.8 128.4 0.003 0.016 0 30 7807. 1734
130.9 127.5
H 66.3 67. 13.8 135.5 128.4 125.0 0.008 0.018 0 30 9889. 2785
130.8 127.7
S 4.35 11. 9.56 133.8 129.5 126.5 0 0.04 1.5 15 1431. 1691
CALC
Figure 4. Spreadsheet interactive design area.
elevation could be replaced with the constant 100. If the constraint limited
the invert elevation to be greater than 100 feet, then the previous cell
formula could be replaced with a conditional formula which gives the constant
value of 100 if the calculated value is less than 100. In this manner, the
spreadsheet provides a simple method for defining individual system
constraints.
Cost Estimation
The optimization of the drainage system is performed primarily to save
on costs while not decreasing the reliability. Since a reduction in size of
almost any pipe will reduce the reliability in some way, the optimization
should be an effort to conform more precisely to the defined criteria (e.g. to
withstand runoff from the 25 year storm). The procedure would, in effect,
minimize the costs of meeting the prespecified standards.
187
-------
The first step in reducing the cost of a system is to acquire a feel for
the distribution of cost within the system. If a detailed cost estimate of a
system is known during the planning stage, then the search may concentrate on
the areas of most probable savings. For example, more effort should be spent
on reducing a 200 foot 72 inch diameter pipe by one size than on reducing a 50
foot 18 inch diameter pipe.
A detailed cost estimate of a drainage system includes quantities and
unit costs of installed pipes, inlets, and manholes. The spreadsheet template
can update the system cost to reflect changes in the system design by using
DOT itemized average bid data. With this ability, the user may quickly find
the design areas in the system with the largest potential savings and may
easily define tradeoffs between parameter refinement and system cost.
The spreadsheet provides an escape from functional!zed cost estimates
which may depend on one or two parameters. The most common sewer cost esti-
mation functions give cost as a function of pipe diameter (Grigg and O'Hearn
(9), Arnell (10)). Others include parameters such as invert depth and flow
(Tyteca (11), Han et al. (12)). If used as design criteria, these functions
may deceive the user into believing that a parameter value is of more or less
importance than it actually is. The itemized cost estimation allows the user
to see the true relationships between parameter values and cost and to opti-
mize the system accordingly.
Hydraulic Design
In addition to cost information, the design area of the spreadsheet .
template provides automated calculations for the pipe flow capacity at the
given slope, the velocity at the design flow, and the crown and inlet ele-
vations at each structure. For each refinement in a pipe slope or size, the
resulting calculations are performed throughout the remainder of the system.
This feature encourages a trial and error approach to the system refinement
since the user is not required to perform extra calculations if the refinement
proves faulty. The template also provides a series of automated criteria
checks which will alert the user to a criterion violation in the system which
was the result of parameter change.
The automation provided in many of the steps in this design template is
not required in order to perform the calculations. It is often the case that
the design may be performed more easily without automatic criteria checks.
This is especially true when a large (greater than 20 pipes) drainage system
is being designed. The recalculation time of the template increases propor-
tionately to the number of pipes in the system. A long recalculation time
hinders the desired trial and error approach to the system design. To solve
this problem, a small macro program has been written that will recalculate
only the parameters for the pipe that is presently being optimized. This
procedure has the disadvantage that the automated criteria checks are dis-
abled, but the recalculation time is drastically reduced. This procedure
actually simulates the hand calculation procedure very closely by moving up
the system pipe by pipe, but encourages a fine-tuning of the system
parameters.
188
-------
The template does not yet include a procedure for calculating the
hydraulic grade line, but work on this addition is currently underway. The
template does, however, correct velocities in pipes that are not flowing full
at design flow. This correction is done using a method developed by
Christensen (13). The method was derived by a Fourier analysis of experi-
mental data relating the depth in a partially full pipe to the water velocity
and flow in that pipe. Since the design flow and the full flow of the pipe
will be known in the design, the partial flow depth and velocity may be calcu-
lated using the iterative method developed by Christensen.
Database Area
The database area of the spreadsheet template contains the FDOT itemized
unit costs for all drainage items. These unit costs are extracted from the
database and used elsewhere in the spreadsheet with the Lotus lookup table
function. The database area also contains the regression coefficients for the
FDOT intensity-duration-frequency curves. These curves give storm intensity
to be used in the Rational Method given the duration, frequency, and zone
number of the location. The spreadsheet template automatically updates the
intensities obtained from the regression curves given a design change in the
storm frequency or in the time of concentration of a subcatchment.
THOMASVILLE HIGHWAY CASE STUDY
Thomasville Highway was a reconstruction project which consisted of
widening a two lane rural highway into a four lane urban highway in Talla-
hassee, Florida. The project was carried out by the FDOT in three phases
over a span of six years.
Several characteristics of the Thomasville Highway project made it a
desirable case study. The topography of the area showed significant relief
for Florida (maximum of 3% grade) and so there existed the potential for a
pipe size and slope tradeoff in the system design. The availability of the
project blueprints and drainage planning calculations allowed an analysis of
the design procedures. Since the drainage design calculations were given on
the tabulation forms, a direct comparison to the spreadsheet design procedure
could be made. Each project phase consisted of an independent system that
emptied to a single outfall. Each of the phases also involved similar general
geographic, topographic and urban development characteristics. These charac-
teristics were deemed desirable for the evaluation of the effectiveness of
using detailed cost estimation as opposed to cost functions.
The evaluation of cost estimation requirements for design involved the
analysis of pipe length and pipe cost distributions. A plot of actual design
pipe lengths for each pipe diameter, expressed as a percentage of the total
pipe length, for each project phase reveals dissimilar distributions. Anal-
ogous plots for pipe cost distributions also are dissimilar. Figures 5a&b
show a large percentage of pipe length and cost centered in smaller diameter
pipes (18-36 in. dia.). For phase 3516, Figures 5c&d reveal a bimodal distri-
bution of pipe length with no intermediate pipe sizes (42-56 in. dia.)
required. However, the cost distribution for this phase is unimodal with 15%
189
-------
A PIPE LENGTH ''. TOTAL R PIPE iSST v. TOTAL
PROJ 35S6 PROJ 3506
X
o 25V
5
g 15-V
1 18V
S 5V
UJ
_J
1
1
1
77
I
f,
& 25V
8
_, 20* •
p 15V
1QX •
8 5V
O
ft'/ •
1
mm
i
|
1
II
15 18 24 30 36 42 43 54 60 66 72
PIPE DlfiMETER (in.)
15 18 24 30 36 42 48 54 60 66 72
PIPE DIAMETER (in.)
c
oo^.
X
0 23V
" 28X-
-t
? 1ffi-
u
_j
ex-
I
1
15
PIPE LENGTH -: TOTAL
PROJ 3516
E^
I i
IP ^
t i
\
1 1
18 24 38 36 42
f
f [
^ ^
1 1
48 54 60 66
I
|
72
PIPE DIAMETER (in.)
EPIPE LENGTH '/, TOTAL
I
ti 255: •
UENCTH/TOTAL LEMI
& tf § S 8 I
KvvVv^vv^vv.vki
1
PROJ 3517
V.
.1
^
ill
5 P
z
PIPE COST \ TOTAL
PROJ 3516
15 18 24 38 36 42 43 54 68 66 72
PIPE DIflMETER (in.)
50* • -
40:-. •
30V
1GV
15 18 24 30 36 42 48 54 60 66 72
PIPE DIAMETER (in.)
PIPE COST V. TOTAL
PROJ 3517
25:-. •
20* •
15'/. •
10V
5V
rfc-
CT
J^R^I^.,.^
\
TO
I
i
15 18 24 30 36 42 48 54 60 66 72
PIPE DIAMETER (in.)
_
-X
fe
X
PIPE LEHGTH \ TpTflL
COMBINED 3506,3516,3517
ii' "ir'"irl '
15 18 24 30 36 42 43 54 60 66 72
PIPE DIAMETER (in.)
H
PIFt CT557 H TOTRL
COMBIMED P3506,3516,3517
25V
2XR-
15V
10'. -
5V
o::
15 18 24 30 36 42 48 54 60 66 72
PIPE DIAMETER (in.)
Figure 5. Pipe length and cost distributions for Thomasville
Highway Project.
190
-------
of the total cost associated with the large diameter pipe. This suggests the
possibility of large savings through reduction of pipe size for large diameter
sections. A third pattern of length and cost distribution is shown in Figure
5e&f for phase 3517 with length and cost being centered in intermediate dia-
meter pipe. When the distributions for the overall project are generated
(Figures 5g&h) the pipe length distribution appears lognormal with larger
percentages in the smaller diameter pipes and a decreasing percentage as pipe
diameter increases. The cost distribution for the overall project reveals the
inverse relationship between cost and length distributions. These combined
distributions smooth out the dissimilarities between each of the individual
phase distributions.
The well behaved distributions for the overall project, if taken alone,
would not reveal the significance of cost estimation and cost feedback in the
design process. The dissimilarities of the distributions for separate project
phases can possibly be related to distance and slope from the outfall to
inlets where large areas contribute inflows to the system. Cost functions
based on and requiring such data appear more difficult to use at the design
level than straightforward detailed system costing. Each independent system
within the overall project requires a different design emphasis in order to
incur savings recognized by a detailed cost estimation. Without the use of
computer assistance, a detailed cost estimation as a part of the design deci-
sion process would be a very labor intensive task. The spreadsheet design
template, however, easily incorporates the detailed costing into the design
procedure.
Cost estimates for each of the three sections of the Thomasville Highway
were performed using the current FDOT itemized unit costs. A redesign of
phase 3517 using the spreadsheet design procedure produced a savings of over
10? in pipe costs over the original design. The pipe costs are estimated to
be about 75% of the total drainage system costs.
CONCLUSIONS
Optimization and simulation methods are being used very little in the
design of storm drainage systems because of a lack of understanding of
their techniques and a difficulty in defining real world problem
constraints. Also the time and effort needed to run computer programs
are often difficult to justify when working with limited time and money.
A spreadsheet design technique which is simply a modified version of
presently used design procedures will be more readily accepted by
drainage engineers.
The detailed cost estimates provided by the spreadsheet allow the engi-
neer to directly examine the components of the system which will incur
the most savings. Itemized cost databases have the advantage over cost
functions of being updated as a response to current item costs while cost
function update requires a reanalysis of cost data.
191
-------
3. Spreadsheet drainage design allows for easy definition of system con-
straints and allows for engineering judgment which is not easily pro-
grammed into a computerized algorithm.
4. Users feel comfortable with a solution obtained by fine-tuning the system
by trial and error because they have performed the calculations. Users
quickly get a feel for relationships between system parameters and cost
while directing the procedure toward a solution which meets both engi-
neering and design budget criteria.
ACKNOWLEDGMENTS
The authors would like to thank the Florida Department of Transportation
for their cooperation in this research effort. The work presented in this
paper was supported by Florida Water Center annual allotment funds from the
U.S. Geological Survey.
REFERENCES
1. Liebman, J.C., 1967, A Heuristic Aid for the Design of Sewer Networks.
J. of the Sanitary Engineering Division, ASCE, Vol. 93, No. SA4.
2. Holland, M.E., 1966, Computer Models of Wastewater Collection Systems.
Thesis, The Division of Engineering and Applied Physics, Harvard
University, Cambridge, Massachusetts.
3. Zepp, P.L. and Leary, A., 1969, A Computer Program for Sewer Design and
Cost Estimation. Regional Planning Council, Baltimore, Maryland,
available as PB 185 592, Clearinghouse, U.S. Dept. of Commerce,
Springfield, Virginia.
4. Merritt, L.B. and Bogan, R.R., 1973, Computer-Based Optimal Design
of Sewer Systems. J. of the Environmental Engineering Division, ASCE,
Vol. 99, No. EE1.
5. Austin, T.A., 1986, Utilization of Models in Water Resources. Water
Resources Bulletin. Vol. 22, No. 1.
6. Hancock, M.C., 1986, Analysis of Water Resources Problems Using
Electronic Spreadsheets. University of Florida Water Resources Research
Center Publication No. 92, Gainesville, Florida.
7. Miles, S.W. and Heaney, J.P., 1986, Application of a Lotus Spreadsheet
for a SWMM Preprocessor. In: Proceedings of Stormwater and Water
Quality Model Users Group Meeting. EPA/600/9-86/023, U.S. Environmental
Protection Agency, Athens, Georgia.
192
-------
8. Florida Dept. of Transportation, 1987, Drainage Manual. Tallahassee,
Florida.
9. Grigg, N.S. and O'Hearn, J.P., 1976, Development of Storm Drainage Cost
Functions. J. of the Hydraulics Division, ASCE, Vol. 102, No. HY4.
10. Arnell, V., 1982, Rainfall Data for the Design of Sewer Pipe Systems.
Report Series A:8, Dept. of Hydraulics, Chalmers University of
Technology, Goteborg, Sweden.
11. Tyteca, D., 1976, Cost Functions for Wastewater Conveyance Systems. J.
Water Pollution Control Federation, Vol. 48, No. 9.
12. Han, J., Rao, A. R., and Houck, M.H., 1980, Least Cost Design of
Urban Drainage Systems. Purdue University Water Resources Research
Center Technical Report No. 138, West Lafayette, Indiana.
13. Christensen, B.A., 1984, Analysis of Partially Filled Circular Storm
Sewers, In: Proceedings of Water for Resource Development. Hydraulics
Division, ASCE, New York.
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.
193
-------
CORRECTIVE PHOSPHORUS REMOVAL FOR URBAN STORM RUNOFF AT
A RESIDENTIAL DEVELOPMENT IN THE TOWN UP PARKER, COLORADO
by: William C. Taggart, Mary S. Wu
McLaughlin Water Engineers, Ltd.
2420 Alcott Street
Denver, CO 80211
ABSTRACT
This paper discusses runoff water quality management for a residential develop-
ment in the Town of Parker, Colorado. The project is located in Cherry Creek
drainage basin, southeast of Denver, where reduction of non-point pollutants, particu-
larly phosphorus, is required for the development. The primary intent of this program
is to reduce phosphorus loading and eutrophication of Cherry Creek Reservoir.
Facilities include two detention ponds, storm sewers, grass-lined channels, a combined
detention and sedimentation basin, pump station and an irrigation system. The system
would comply with "Criteria for the Control of Erosion and Non-Point Source
Pollution", a runoff water quality enhancement guideline for the Town of Parker.
The irrigation system, which was already needed for the development, is felt
to be a preferable phosphorus removal system over present guidelines which suggest
constructing a filtration system.
The soils are conducive to infiltration. An underdrain system was proposed to
augment treatment during wet periods and to provide monitoring opportunities. It
is perceived that the filtration system will have significant maintenance problems
because of sediment accumulation on the filtration bed and reliability/performance
problems with actual phosphorus removal because filtration cannot remove dissolved
phosphorous. Other than processes involving chemical treatment, it is more reliable
to involve treatment where a soil column and plant uptake of phosphorus is involved.
The system has other benefits such as reducing the need for groundwater for irrigation
needs.
BACKGROUND
The Town of Parker is in the Cherry Creek basin, which drains to Cherry
Creek Reservoir. Cherry Creek Reservoir is located southeast of Denver. The
reservoir was determined to be slightly eutrophic in the National Eutrophication
Survey (Ref. 1) (NES) conducted by U. S. Environmental Protection Agency between
194
-------
1972 and 1975. In 1984, the Cherry Creek Reservoir Clean Lake Study was conducted
by Denver Regional Council of Governments (DRCOG) to establish water quality
goals and standards related to eutrophication as well as recommend treatment levels
to achieve those goals and standards.
A total phosphorus standard of 0.035 mg/1 for Cherry Creek Reservoir was
adopted by the Colorado Water Quality Control Commission (CWQCC) in September,
1985 (Ref. 2). In order to maintain this phosphorus standard in Cherry Creek
Reservoir, the annual load of total phosphorus has to be reduced. Non-point
stormwater runoff was estimated as the major contributor (77%) of phosphorus to
the reservoir. Therefore, water quality control measures were called for which
would be capable of removing 50% of the total phosphorus load for non-point storm
water runoff for all the developments in Cherry Creek basin.
The Town of Parker had the firm of HydroDynamics prepare a manual entitled
"Criteria for the Control of Erosion and Non-point Source Pollution" (Ref. 3). In
order to achieve the desired goal, performance standards are cited (Table 1) for
various types of development. These are based on the regional studies (Ref. 3).
TABLE 1: PERFORMANCE STANDARDS FOR PHOSPHORUS REMOVAL FROM
DEVELOPED LANDS
Land Use Phosphorous Removal
Residential 45%
Commercial 70%
Public Areas 50%
Control measures such as retention, filtration, infiltration, and wetland applica-
tion are discussed. General efficiency factors are given in Table 2.
TABLE 2: MITIGATION MEASURE EFFECTIVENESS FOR REDUCING TOTAL
PHOSPHOROUS (Ref. 3)
Measure Reduction Efficiency
Retention 25%
Infiltration 90%
Filtration 50%
Wetland Application 75%
Retention is basically storage, sedimentation and very slow release. Infiltration
generally includes methods to utilize the native soil capability. This method is
preferred, but concerns are expressed (Ref. 3) as to the inherent loss of area where
this can take place because of development. Filtration refers to a sand bed filter,
which is typically downstream of a retention basin. Figure 1 illustrates a schematic
of the pond and filtration basin. Figure 2 depicts the pond outlet and Figure 3 the
filter drain layout. (Figures copied from Ref. 3). Wetlands application refers to
flow through wetlands where both sedimentation and phosphorous removal take place
as the plants trap and utilize phosphorous.
195
-------
DETENTION POND WITH WATER
QUALITY ENHANCEMENT FEATURES
J Standard IO 9 fOOrr
[\OutM Slruetwt
Filtration Piping
MiliraHof or
'illrttiwi Oat/a
NOTES- I. LENGTH/WIDTH > f.O
Tables 3, 4, and 5 are the Reduction factors provided by retention, filtration,
and infiltration methods for various amounts of runoff treated (Ref. 3). As the
amount treated increases the reduction factor approaches the efficiencies cited in
Table 2. A relative removal efficiency is used to adjust for higher or lower removal
efficiencies, such as might be documented by pilot tests.
Relative Removal Efficiency = Corrected Removal Efficiency (pilot test or other data)
General Removal Efficiency (Table 2)
Thus the effectiveness can be computed for each type of development receiving
a given treatment by multiplying the area treated (in terms of percentage of the
study area) by the reduction factor and the relative removal efficiency. the
cumulative total of all treatments is then the overall effectiveness.
PROJECT DESCRIPTION
Total drainage area of the proposed project was 358.5 acres. An initial drainage
plan was prepared by another consulting firm. The primary facilities included three
detention ponds, storm sewers, grass-lined channels and a sand/gravel filtration bed
below the lower pond (Pond No. 3) in compliance with standard recommendations.
The volume for water quality treatment was provided in Pond No. 3. Figure 4
illustrates the basin boundaries and a sketch map of the. storm runoff control facilities
for the development proposed in the drainage plan. The key facilities are described
below:
196
-------
(1) A storm sewer and grass-lined channel conveyance system which directs
the majority of runoff from more common events to Pond 3. Pond 2 is bypassed by
this conveyance system except during larger storm events.
(2) Provision of 6.2 acre-feet of water quality storage at Pond 3 which is
one-half inch of runoff from the impervious surfaces.
(3) An outlet structure for Pond 3 which incorporates coarse removal of
debris and sediment.
(4) A filtration basin, which is to be constructed according to Parker and
Douglas County Guidelines with imported sand and gravel filtering medium and
perforated PVC pipe drains.
The primary intent of the treatment process is to remove phosphorous. However
there is little performance history of such non-point phosphorus removal facilities.
It is perceived that there will be significant problems with maintenance because of
STORM DRAINAGE DESIGN AND TECHNICAL CRITERIA FIGURE 2
WATER QUALITY OUTLET
.FOR DRY POND
IQOyr Dtttntion ro/umi tlcvation
10/r Dtttntton ro/umt tltrotion
Maximum Wattr
Ouolitr Voter Surf a
tfte to prevent r° InfUtration/
hytlrottotic uplift Filtration Bosin_
Nc.lt: 1. Rixr pipe to be pilvenized tteej
2- Dtlipi ault provide for permanent (ccess to the outlet
• itructure et ill times
DATE: NOV 19B4 I REFERENCE-
REV: '
197
-------
STORM DRAINAGE DESIGN AND TECHNICAL CRITERIA) FIGURE 3
*
K;
&&.
w
x
^f
•^
_J j V V
_i_ ]
Cdfi of
filtration
basin
-/
From
detention
pond
^^-Emtrgency Spillway
To Outfall Sfirfr or
Major Drainagtway
REFERENCE.
sediment accumulation on the infiltration bed and the effectiveness for phosphorous
removal. Specifically the removal of dissolved phosphorous in a sand media filter
was questioned. Therefore, the water quality management alternative studied here
was oriented toward retention, sedimentation and subsequent treatment utilizing irri-
gation. The effort specifically addressed on-site soils characteristics and open spaces
plans conducive to such a treatment scheme. A key reason for utilization of this
system is phosphorous removal efficiency. Basically irrigation such as this would be
a land treatment system capable of 97% to 99% removal efficiency (Ref. 4).
ON-SITE SOIL INVESTIGATION
Most of the soils in- the area of Pond No. 2 are classified as Newlin gravelly
sandy loam or the Newlin-Santanta Complex by the Soil Conservation Service. The
Newlin gravelly sandy loam has reasonably high, hydraulic conductivity character-
istics with permeability of 0.63 to 2.0 inches per hour at the upper layer and 6.3 to
20 inches per hour at the lower layer.
The permeability of Santanta loam varies from 0.63 to 2.0 inches per hour for
both layers.
198
-------
Table 3: Residential Development: Mitigation Measure Effectiveness (Ref. 3)
Total Phosphorus Reduction Factor
Inches
Treated
0.00
0.02
0.04
0.06
0.08
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.60 or more
Retention
0.000
0.012
0.026
0.040
0.053
0.065
0.089
0.112
0.132
0.153
0.169
0.182
0.194
0.207
0.213
0.219
0.224
0.230
0.239
0.248
Filtration
0.000
0.024
0.052
0.080
0.108
0.130
0.178
0.223
0.263
0.307
0.338
0.364
0.389
0.415
0.426
0.438
0.449
0.411
0.478
0.496
Infiltration
0.000
0.044
0.094
0.145
0.194
0.234
0.253
0.401
0.473
0.551
0.607
0.654
0.700
0.746
0.766
0.787
0.807
0.828
0.859
0.893
Note: Intermediate values may be determined by linear interpolation
A subsurface drainage system under Pond 2 could be expected to accept from 6
to 12 inches of water depth for the first day. For the second day, the intake rate
may decrease by half and again by half the third day.
Soils in the area of Pond 3 and downstream of Pond 2 consist of Sampson loam.
The permeability of the upper layer varies from 0.2 to 0.63 inches per hour according
to the Soil Conservation Service (Ref. 5). Obviously, these soils are much slower
draining than the Newlin, gravelly, sandy loams or even the Santanta loam complex.
Because of the excellent subsurface conditions for Pond 2, it was first considered
that water quality volume and treatment by infiltrating be provided in Pond 2.
However, it was subsequently learned that a soccer field was planned, which would
dictate dry conditions soon after a rain storm. Also, it was desirable to keep
sediment accumulations to a minimum in order to maintain infiltration rates.
Therefore, it appears more practical to direct the majority of site runoff to Pond
3. Thus, only surplus runoff during major events will be spilled into Pond 2 directly.
The Pond 2 area characteristics for irrigation and infiltration treatment are
good, and thus the study indicated its use as a primary treatment area.
The area that requires irrigation in the development was estimated at 17.4
acres. The 6.2 acre-feet of water quality volume stored can provide a 2 days supply
for normal irrigation in this area. This irrigation water in area of Pond 2 could
be collected by the underdrains that would drain directly into Cherry Creek. Most
199
-------
Table 4: Grassland/Open Space: Mitigation Measure Effectivenesc (Ref. 3)
Inches
Treated
0.00
0.02
0.04
0.06 or more
Total Phosphorus Reduction Factor
Retention Filtration Infiltration
0.000
0.037
0.155
0.227
0.000
0.078
0.316
0.457
0.000
0.133
0.555
0.806
Note: Intermediate values may be determined by linear interpolation
Table 5: Commercial Development: Mitigation Measure Effectiveness (Ref. 3)
Total Phosphorus Reduction Factor
Inches
Treated
0.00
0.02
0.04
0.06
0.08
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.90
0.95
1.00
1.20
1.40
1.60
1.80
2.00 or more
Retention
0.000
0.039
0.065
0.076
0.087
0.098
0.114
0.127
0.139
0.149
0.158
0.164
0.171
0.177
0.181
0.185
0.191
0.196
0.199
0.202
0.209
0.212
0.218
0.225
0.237
0.238
0.241
0.244
Filtration
0.000
0.076
0.130
0.152
0.174
0.196
0.227
0.254
0.277
0.297
0.315
0.329
0.342
0.354
0.363
0.371
0.381
0.390
0.397
0.403
0.410
0.417
0.431
0.450
0.465
0.476
0.482
0.489
Infiltration
0.000
0.138
0.234
0.274
0.313
0.352
0.410
0.458
0.499
0.535
0.568
0.592
0.615
0.638
0.653
0.667
0.685
0.702
0.713
0.725
0.750
0.763
0.776
0.809
0.836
0.856
0.868
0.880
Mote: Intermediate values may be determined by linear interpolation
200
-------
-r IN
SUBBASIN DRAINAGE AREA SCHEMATIC
Mclaughlin Wat* EnginMra, ltd.
STUDY AHA
SUMASIN IOUNOAIV
STOIM MAIN Pitt
GIASSLINED CHANNfl
FIGURE 4
of the phosphorous and much of the nitrogen will be removed by the soil profile.
Plant uptake should be able to remove a high percentage of each element.
A subsurface drainage system under Pond No. 3 would drain water from the
pond slowly, probably too slowly to add much to the volume of irrigation water.
However, it would be desirable to be able to dry up the pond and the soil profile
under the pond in between rainfalls. This would allow machinery to work in the pond
to scrape off the sediment and also to cut grass or other vegetation.
CONCEPTUAL ALTERNATIVE SYSTEM FOR RUNOFF WATER TREATMENT
Based on the soils analyses and review of the proposed drainage facilities in
Master Drainage Planning, an alternative runoff water quality management system
was developed. As depicted in Figure 4, all frequent runoff events will find their
way to Pond 3, except sub-basins A5A,. A5B, and BIA. Sub-basins A5A and A5B
(23.9 acres, or 7%), would usually discharge to the overbank flood plain meadow
along Cherry Creek. This is similar to wetland application because this area is to
be dedicated as an open space park, runoff will filter through the vegetation and
soils there and thus be treated. Sub-basin BIA (11.1 acres, or 3.0% of the developable
land) is basically Jordan Road and drains to a drainageway outside the development
boundary.
Runoff from the remaining 323.5 acres (90%) of the development will find its
way to Pond 3. Based on the imperviousness of 40% for the development 6.2 acre-
feet of water quality volume is required to capture 0.5 inches of runoff, which is
201
-------
often regarded as an event that flushes approximately 90% - 95% of the available
pollutants to the water course. In the concept proposed in Figure 5, some of this
water would slowly infiltrate through underdrains to Cherry Creek, although this
water could potentially be pumped to irrigation. This also helps to address water
quality during the winter.
During the normal irrigation season, the majority of the water quality volume
in Pond 3 will be pumped to the irrigation for the site, especially the high infiltration
area around Pond 2. Auxiliary water supply would be needed during drier periods.
The irrigation system and pump station would necessitate a level of initial
filtration/screening/operation at Pond 3 that would be more involved than the water
quality outlet originally designated, but within reason. Pond 3 can be managed in
several different modes. The pond will be drawn down by the. irrigation pump station,
and dewatered by the underdrains. Thus, the pond bottom could be cleaned of
debris, particularly at the intake facility, and periodically disked to enhance plant
growth and infiltration. Once every 2 to 10 years, depending on development and
the success level of erosion-control practices, sediment accumulated in Pond 3 will
have to be removed.
With major sediment control taking place in Pond 3 and the pump intake system,
sediment application to the irrigated area should be minimized. It is anticipated
that frequent soil aeration will be necessary, and at worst, portions of the sod in
Pond 3 might have to be replaced (once every 10 to 30 years) if high infiltration
rates are to be sustained.
AUXILIARY WATER SUPPLY
IIIIGATION LINE
SCHOOL FIELDS -
USE FILL MATERIAL
WITH HIGH INFIL-
TRATION DATES
POND 2 - OPEN SPACE &
HIGH RATE APPLICA- RECREATION
TION ARE A, IMPROVE .AREAS
SOILS IN SELECT AREAS I
POND 2
OUTLET
MONITOR
POND 3 -
SEDIMENTATION, WATEI QUALITY
MANAGEMENT STORAGE
FLOOD CONTROL
POND 3
OUTLET
MONITOR
UMTEI QUALITY STORAGE AND
INITIATION GALL IK/
DfUIS REMOVAL AND
SfCONOARV INTAKi
UNDERDRAINS FOR
TRICKLE FLOWS AND
MAINTENANCE DEWATERING
OTHER COMPONENTS NOT SHOWN •
I SOURCE MANAGEMENT INCLUDING CONSTRUCTION
SEDIMENT CONTROL ON SITE AND ONGOING
HOUSEKEEPING ORDINANCES AND ACTIVITIES.
2. VEGETATION SHOULD IE PERIODICALLY MOWED
AND CUTTINGS REMOVED FROM SITE.
ALTERNATIVE RUNOFF WATER QUALITY MANAGEMENT PROGRAM
MCLAUGHLIN WATER ENGINEERS.LTD. — FIGURE s
202
-------
ESTIMATED SEDIMENT LOADS
Potential sediment loads during construction periods were estimated according
to sampling results for Cherry Creek (Ref. 2). A potential erosion rate of 156
tons/Ac-yr. was assumed. This converts to settleable solids of 1,750 ft.^/Ac-yr.
Erosion control measures would reduce this rate to 430 ft.-'/Ac-yr. For this sediment
loading rate, roughly 5,000 cubic yards per year may accumulate.
Sediment loads for developed conditions were estimated according to the study
for the Cherry Creek Dam tributary area (Ref. 3) and total suspended solids estimates
from DRCOG (Ref. 3). An annual sediment loading volume of 160 cubic yards/year
was indicated. Assuming that all this sediment arrives at Pond 3, and that it would
be practical to remove sediment when 3 inches of sediment had accumulated; a
sediment storage zone of 1,500 cubic yards (approximately 1 acre foot) was called
for. This data would indicate frequent sediment removal during early years of
construction.
EFFECTIVENESS OF RUNOFF WATER QUALITY MANAGEMENT PROGRAM
Using the evaluation criteria of the Town of Parker, the effectiveness of this
proposal is estimated at 61% which satisfies 50% (46% for this mix of development)
phosphorus removal standard adopted by CWQCC in Cherry Creek basin.
Monitoring of the system during normal events should be fairly straightforward
as the discharge from the underdrain system provides an absolute control point for
flow and quality. The pump station provides a logical point for determining volumes
and quality of water to be applied by irrigation. The overflow outlets out of each
pond provide points for determining quantities of water receiving lesser treatment.
ADDITIONAL SYSTEM NEEDS
The proposed system has many components and operational needs that would be
required with other on-site facilities. The discussion below highlights the additional
needs or special coordination items that should be recognized and refined.
(1) Pump Station and Intake. This is a key facility to the-success of the
system. Although some delay can occur before application, irrigation should take
place within a few days after the event. The controls would have to be tied to
the irrigation controllers so that during key situations application could take place.
(2) Irrigation System. A conventional irrigation system may be utilized, except
the hydraulics of the auxiliary source will have to be coordinated to make up the
difference when the pump station is only providing a portion of the needs. This is
not especially difficult, but needs recognition.
(3) Underdrain System. The underdrain system would use conventional agricul-
tural practices, which are fairly economical as they are installed with trenchers and
automatic pipe laying systems. The collection pipe to be routed to Cherry Creek
will also be desirable to allow monitoring and prevent salt buildup.
203
-------
CONCLUSIONS
This project investigated an alternative phosphorus control approach which
utilized detention storage, sedimentation and irrigation of open space areas. Although
three detention storages are called for only one provides water quality volume to
store the sediment and the initial flush of storm runoff.
Phosphorous removal by irrigation and land treatment can be 97 to 99% effective
(Ref. 4), and thus can have a higher removal efficiency than Infiltration cited in
the Parker Manual (Ref. 3).
Irrigation will take place in most development and thus an opportunity exists
to make efficient use of on-site soil infiltration as a preferred method and take
advantage of facilities that are usually planned. It is percieved that the effectiveness
of sand filtration beds as a phosphorous removal system is over-estimated in the
Parker Manual (3) because dissolved phosphorous will be carried directly through,
and because there will be excessive operational and maintenance costs.
ACKNOWLEDGMENTS
Credits: This study was conducted under a contract with MDC Construction
Company, Denver, Colorado.
Arthors: William C. Taggart is a principal of McLaughlin Water Engineers, Ltd.,
Denver, Colorado. Mary S. Wu is an engineer of McLaughlin Water Engineers, Ltd.,
Denver, Colorado. Correspondence should be addressed to William C. Taggart,
McLaughlin Water Engineers, Ltd. 2420 Alcott Street, Denver, Colorado 80211.
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. Denver Regional Council of Governments. Cherry Creek Reservoir Clean Lake
Study. Denver, Colorado, April, 1984. 165 pp.
2. Denver Regional Council of Governments. Cherry Creek Basin Water Quality
Management Master Plan. Denver, Colorado, September, 1988. 47 pp.
3. Hydrodynamics Incorporated. Criteria for the Control of Erosion and Non-point
Source Pollution. Prepared for the Town of Parker and the Parker Water and
Sanitation District. Parker, Colorado, July, 1985. 85 pp.
4. U. S. Environmental Protection Agency. Process Design Manual for Land
Treatment of Municipal Wastewater. Cincinnati, Ohio, October, 1981.
5. U. S. Department of Agriculture, Soil Conservation Service. Soil Survey of
Castle Rock Area, Colorado. Washington, D.C., November, 1974. 124 pp.
204
-------
EVALUATION OF SEDIMENT EROSION AND POLLUTANT
ASSOCIATIONS FOR URBAN AREAS
Kim Irvine1, William James2, John Drake1, Ian Droppo1 and Stephen Vermette1
ABSTRACT
The algorithms for erosion of pervious land in the USEPA stormwater management model
(SWMM) are examined. Field equipment for measurement of simulated rain erosion is
described. Field experiments and results are summarized. The relationship between eroded
solids and rainfall energy is evaluated for different land types. Associations between fractionated
eroded solids and metal concentrations are examined. Pollutographs show a relationship
between eroded solids and selected metals concentrations.
INTRODUCTION
The general term 'pervious land' includes all erodible areas such as gravelled parking lots,
bare industrial land, railway land, cemetaries, golf courses, parks, ravines and lawns. Clearly,
these different pervious land types will have different erosion response rates to rainfall energy
and different potential for pollutant (organics and trace metal) input to stormwater runoff. Urban
stormwater runoff quality models typically do not consider the dynamics of erosional processes,
if erosional processes are considered at all. The USEPA Stormwater Management Model
(SWMM), for example, uses the Universal Soil Loss Equation (USLE) to simulate erosion for
small time intervals (1-5 minute steps). In this study the PC version of SWMM (James, 1985
(1)) is being used. The USLE was originally developed from data obtained from rural
experimental soil plots in 21 states in the U.S.A. with the intention of simulating average annual
soil loss in agricultrual areas (Smith and Wischmeier, 1962 (2)). The time frame to which the
USLE is applied in SWMM, and the agricultural origin of some of the parameters, now
transposed to an urban environment, probably constitutes a misapplication of the USLE.
Several researchers (eg. Ammon, 1979 (3)); Malmquist, 1983 (4)) have suggested that
pollutant and paniculate output from pervious land may be significant factors contributing to poor
quality stormwater runoff. However, a literature review to date has revealed only one report
(Pitt, 1985 (5)) in which pollutant outputs are quantified for pervious land other than
construction sites. The purpose of this paper is to present some preliminary results from our
ongoing study on the role of pervious land in stormwater runoff quality in the city of Hamilton,
Ontario, Canada.
1 Dept. of Geography, McMaster University, Hamilton, Ontario
2 Cudworth Professor of Computational Hydrology, Univ. of Alabama, Tuscaloosa, AL
35487
205
-------
SEDIMENT EROSION
It has been found that fine particulates (less than 4 mm in diameter, and often referred to as
dust and dirt (DD) can significantly pollute stormwater runoff because: a) water turbidity is
affected; and b) trace metals are often adsorbed at the particulate surface (Ammon, 1979 (3);
Manning et al., 1977 (6); Ongley et al., 1981 (7); Pitt, 1985 (5); Sartor and Boyd, 1972 (8)).
Although the toxicity of a given trace metal is species dependent (Benes et al., 1985 (9);
Crecelius et al., 1982 (10); Florence, 1977 (11); Simkiss and Mason, 1984 (12)) it is time
consuming and difficult to identify labile species in water/sediment samples (Chau et al., 1983
(13); 1984 (14) ; Ricci et al., 1981 (15)). Furthermore, the physical, chemical and biological
processes that interact to produce the species distribution within a sample are very complex and
difficult to model in any detail (Chapman, 1982 (16)). It is possible to model particulate (DD)
movement within the urban system and to derive relationships between DD and selected
pollutants such as phosphorus and various trace metals. Toxicity cannot be directly evaluated,
but by monitoring and modelling DD buildup, redistribution and washoff processes, pollutant
sources may be identified and abatement programs can be instituted and evaluated.
A mass balance approach is used to assess particulate input, storage and output for
pervious/bare land, Components of the mass balance include inputs from rainfall, dry dustfall,
human activities, vegetation, wind redistribution from impervious land, and outputs through
water and wind erosion and biological activity. The major processes include: a) stormwater
erosion output; and b) rainfall, dry dustfall and human activity input. Sampling and modelling
procedures therefore concentrate on these processes.
There are four essential processes to be considered when evaluating water erosion from
pervious land: a) particle detatchment by rainsplash; b) particle detatchment by overland flow; c)
rill erosion/development; d) transport capacity of particles through rill and interrill areas. Each
component of the erosion process should be evaluated in simple quantitative and descriptive
terms to provide a fundamental understanding of particulate output from pervious land. Different
types of pervious/bare land, such as lawns, golf courses, cemetaries, railway yards, gravelled
parking lots and industrial yards will have different responses to similar rainfall inputs. A
classification system based on process response should therefore be developed to facilitate
system modelling.
Soil erosion by water in agricultural areas has basically been modelled by one of four
methods: a) Universal Soil Loss Equation (USLE) or modified Universal Soil Loss Equation
(MUSLE) (Smith and Wischmeier, 1962 (1); Wischmeier and Smith, 1978 (17); Williams, 1975
(18)); b) conceptual consideration for the physical processes involved (Donigian and Crawford,
1976 (19); Li et al., 1977 (20); Simons et al., 1977 (21)); c) combination of the USLE, interrill
and rill erosion and routing processes (Foster et al., 1980 (22); Khanbilvardi and Rogowski,
1984 (23)); d) combination of deterministic and probabilistic techniques (Moore, 1984 (24);
Rojiani et al., 1984 (25)). There are numerous problems with applying the USLE or MUSLE
directly to an urban catchment on an event basis, including: a) the inability to model detachment
and transport separately; b) no consideration of transported grain sizes; c) variables such as the
cropping factor were developed for agricultural areas and are not meaningful in an urban
environment. The (M)USLE is easy to use however, and some of the parameters in other, more
complex models may be difficult to obtain for urban land. The Storm Water Management Model
(SWMM) is used extensively in design and water quality studies, but the water quality
subroutines clearly need improvement (Boregowda, 1984 (26); Cermola et al., 1979 (27);
MacRae, 1979 (28)). The USLE is currently employed in SWMM to simulate erosion.
The USLE in SWMM does not provide any estimate of the grain size distribution of the
eroded sediment. In our study, eroded particle sizes are being analysed to determine which sizes
206
-------
are most important in pollutant transport and how eroded material from pervious land might
affect total DD transport in the urban system. Such information would be particularly useful in
the the development of nonstructural pollution abatement programs (eg. street sweeping, sewer
flushing) and in the optimization of settling/holding tank hydraulic conditions (Klemetson, 1985
(29); Randall, 1982 (30)).
There are at least three methods by which eroded grain sizes can be predicted: a) empirical
and quasi-conceptual deterministic models (Li et al., 1977 (20); Williams, 1980 (31)); b)
regression models that consider soil characteristics (Frere et al., 1975 (32); Young and Onstad,
1976 (33)); c) laboratory disaggregation techniques (Meyer et al., 1983 (34)). Approaches (a)
and (b) are most appropriate for the USLE type of erosion model. Using the predicted grain size
distribution, the transport capacity for each size class could be calculated by Yalin's equation, for
example.
PREVIOUS RESEARCH
It has long been known that construction and general urbanization can have a significant
impact on the sediment regime of a catchment (Armstrong, 1978 (35); Burton et al., 1976 (36);
Diseker and Richardson, 1962 (37); Keller, 1962 (38); Walling and Gregory, 1970 (39)).
Sediment yields from construction sites may be many times greater than from nearby agricultural
land and in general, progressive urbanization causes an initial rise in total sediment, which
ultimately is followed by a decline as pervious land is paved (Wolman and Schick, 1967 (40)).
During the 1960's and early 70's, research concerns in urban sedimentology were directed
primarily at environmental effects of the eroded sediment load itself, (for example, changes in
channel hydraulic geometry) rather than with respect to the various pollutants transported by the
particulates (eg. Guy, 1967 (41)). However, with increasing awareness that urban runoff was a
significant contributor to receiving body degradation, research programs such as the Nationwide
Urban Runoff Program (NURP) in the U.S., were initiated in the late 1970's (Cole et al., 1984
(42)).
In the review of research needs on urban stormwater pollution, Heaney (1986) (43)
suggested that the influence of soils, land use and season should be investigated. As mentioned
above, many researchers have suggested that pollutant output from pervious land may be
important, but our literature review to date revealed only one paper (Pitt, 1985 (5)) in which
outputs are quantified. Pitt found that for two residential areas in Bellevue, Washington, front
and back yards supplied approximately 83% of the total solids, 25% of COD, 42% of
phosphates, 39% of TKN, 2% of Pb and 4% of Zn loads for 2.5-65 mm rain events. Although
the total solids in this case may be relatively 'clean' of trace metals at source, they can provide a
transport medium for metals absorbed during movement over impervious areas.
Many of the field and modelling techniques, as well as a general description of water
erosion processes can be borrowed from the extensive work done in agricultural/rural areas.
Empirical equations to predict soil loss, such as those of Musgrave or Browning and his
coworkers, began to appear in the 1940's (Smith and Wischmeier, 1962 (1)). Wischmeier and
Smith (1958) (44) and Wischmeier (1959) (45) devised a rainfall erosion index for general use in
the United States based on the research of Laws and Parsons (1943) (46), and by 1960 the
USLE had been developed. The USLE was orginally based on 8000 plot years of basic
hydrometeorologic and soil loss data from experimental soil plots in 21 states. Use of the USLE
and modifications to individual parameters have been discussed by Mitchell and Bubenzer (1980)
(47); Smith and Wischmeier (1962) (1); and Wischmeier and Smith (1978) (17).
A great deal of research has been done on characterizing and quantifying rainsplash
detachment and transport, overland flow detachment and transport and rill development (eg. Al-
Durrah and Bradford, 1982 (48); Bryan, 1976 (49); 1979 (50); Emmett, 1970 (51); Evans, 1980
(52); Luk, 1979 (53); Luk and Hamilton, 1986 (54); Morris, 1986 (55); Poesen and Savat, 1981
207
-------
(56)). However, Morris (1986 (55)) has pointed out several difficulties in isolating and
rigorously quantifying the individual components of the erosion process, and Kirkby (1980) (57)
suggests that: a) sediment yields from rainsplash are low; and b) interrill mechanisms are so
complex that at present largely empirical models for these processes are sufficient. Thus, while it
is not profitable in the present context of research to model the individual erosional processes in
much detail, it is useful to consider detachment and transport separately, on a storm basis. Such
an approach overcomes several of the criticisms made of the USLE (Foster et al., 1980 (22);
Khanbilvardi and Rogowski, 1984 (23); Kirkby, 1980 (57)) and facilitates eroded grain size
distribution modelling (Foster et al., 1980 (22); Williams, 1980 (31)).
A mass balance approach has recently been used to model DD processes on impervious land
surfaces with some success (Boregowda, 1984 (26); James and Boregowda, 1985 a. (58) b.
(59); Novotony et al., 1985 (60)) and there would be some common processes for both pervious
and impervious land. Hamilton and Chatt (1982) (61) and Tanaka et al. (1981) (62) have
successfully determined paniculate metal concentrations directly from filters, and have indicated
that precipitation particulates may form an important part of pollutant load input to the surface.
Slinn (1977) (63) has developed relationships to predict particulate scavaging during rainfall
events. Dry dustfall has been examined by numerous researchers (eg. Jeffries and Snyder, 1981
(64); Malmquist, 1983 (4): Ontario Research Foundation et al., 1982 (65)) and it is now
generally accepted that dustfall can be a significant component in urban pollution. However, the
relationships between pervious land and dry dustfall have yet to be determined. Various
empirical relationships have been developed for population and vegetation inputs to impervious
land (Boregowda, 1984 (26); Prasad et al., 1980 (66)) and these are potentially applicable to
pervious land, although this will have to be investigated. Erosion and transport by wind has
been given similar theoretical and practical treatment as erosion and transport by water. Wind
velocity profiles in fully turbulent conditions have been described by the Prandtl and von Karman
equation, while Bagnold (1941 (67)) defined threshold shear velocity in terms of grain density,
air density, the gravitational constant and grain diameter. He also related the rate of sand flow per
unit width to wind shear velocity, grain diameter and air density. Numerous empirical equations
have been developed relating erosion rates directly to wind velocity, and in 1965 Woodruff and
Siddoway developed a wind erosion equation with a form similar to the USLE. Wilson and
Cooke (1980) (68) note that wind erosion can be highly localized and de Ploey and Gabriels
(1980) (69) examined the difficulties of measuring wind erosion. At the present time it may be
worthwhile to consider wind erosion in the simplest terms for an urban area.
STUDY CATCHMENT AND DATA
The Chedoke study catchment in Hamilton is 26.8 km2 in area of which 82.5% is pervious
(Boregowda, 1984 (26)). Land use is predominantly low to medium density, single family
residential, although institutional (eg. McMaster University and Medical Centre), commercial and
light industrial uses are also present.
Runoff was sampled at 7 sites between May and November, 1986: 1. light industrial
gravelled receiving area; 2. sewer inlet draining the paved road and lawns adjacent to site 1; 3.
light industrial bare gravelled storage lot; 4. side of a railway track embankment; 5. small grassed
plot at sewer overflow; 6. playing field; 7. two sites in a ravine receiving combined sewer
overflow: site (a) was at the sewer outfall and site (b) was downstream in the ravine. Runoff at
the sites was generated either by natural rainfall or by the rainfall simulator described below.
Simulated rain closely resembled natural rainfall characteristics.
Rainfall intensity data for 1 minute intervals for the natural events was obtained using a
Drop Counter Precipitation Sensor (DCPS) system installed on the roof of the Engineering
building at McMaster. The DCPS system, developed at McMaster ( James and Stirrup, 1986
(70)), provided reliable intensity data during the study period, the average error between
208
-------
observed and calculated (from the intensity at each time step) total storm volumes being 10%.
The greatest errors were recorded for rainfalls of less than 2 mm.
Total trace metal concentrations were determined by Instrumental Neutron Activation
Analysis (INAA) at the McMaster Nuclear Reactor under the direction of Dr. S. Landsberger.
Sample collection, storage and handling procedures were designed to limit sample contamination.
RAINFALL SIMULATOR
Rainfall simulators are often used for studying erosional processes in agricultural soils (eg.
Bryan, 1970 (71); Bryan and de Ploey, 1983 (72); Bubenzer and Jones, 1971 (73); Imeson,
1977 (74); Luk, 1979 (53); Luk and Hamilton, 1986 (54)), but they have also been used for
such diversified research as Karst landform development (Glew, 1976(75)), pollutant washoff
on city streets (Sartor and Boyd, 1972 (8)) and assessment of linear and initial storage theories to
describe the relationship between rainfall and runoff characteristics (Johanson, 1967 (76)).
Many of the simulators described in the aforementioned studies are strictly for use in a
laboratory setting. Reproduction of some pervious land types encountered in an urban area,
such as gravelled parking lots, bare industrial land, and railway land, would be difficult.
Therefore, a rainfall simulator was needed that could be easily transported for in situ simulations,
or for laboratory studies, if desired. Rainfall simulation is being used in this study to augment
data obtained from natural rainfall, since a large amount of data can be collected through
simulation at a time and place of the researcher's choosing, under carefully controlled conditions.
The four factors that were of greatest concern in the development and uses of the simulator
were: a) ease of installation and simplicity of design, due to time and financial constraints; b)
ability to simulate a range of rainfall intensities; c) even distribution of water over the sample plot;
and d) reasonable imitation of the size distribution and fall velocity of naturally occurring
raindrops. The simulator was developed and modified from a design proposed by Dr. S. Luk
for the Geography Department at the University of Toronto. The total costs of the 2-stand
simulators is $355 (Canadian dollars in 1986).
The 2-stand version can be assembled by 2 people in about 30 minutes, depending on the
slope of the land. More adjustment is generally required to ensure that both nozzles are at equal
elevation and that the upright galvanized steel pipe is vertical when the slope is greater than about
15%. The test plot can be expanded by adding more simulator pairs, although it may be
necessary to connect a second water pressure regulator (one for each side of the test plot) if more
simulator pairs are added.
The maximum test plot size to ensure an evenly distributed rain is 2 m x 2 m (4m2 or 10.8
ft2) and plots should be defined by garden edging or a wood frame, depending on the surface
type. Test plots in this study were typically smaller than 4 m2 (3.5-3.9 m2) because the
downslope plot edging is angled into a collector trough.
One stand is placed on either side of the test plot and the 1.9 cm (3/4") upright pipe should
be 1 m (3'3") from the top and bottom of the plot and 0.5 m (1'8") back from the plot edge.
This positioning ensures that the conical spray pattern of the individual spray nozzles overlaps on
the plot to produce an evenly distributed rain.
Gerlach-type overland flow troughs were used in the simulations to collect runoff and
eroded sediment at the downslope end of the plot. These troughs are constructed of PVC tubing
to limit sample contamination. The troughs are 0.66 m (2'2") in length (the dimension of a
typical sewer grate) and have one lip (3.5 cm or 9" wide) which projects into the soil, and an
optional second lip which helps to: a) limit rainfall input to the trough; and b) catch splash-
eroded particulates. In the absence of the second lip, the open top of the trough should be
covered with plastic to limit rainfall input. Overland flow from the plot is routed down the
trough, through a funnel, into a Nalgene collection bottle. Overland flow may be collected at a
sewer inlet if the surface does not permit installation of a flow trough.
209
-------
Rainfall intensities range from 52 to 87 mm hr1. The average variability of intensity
between individual rain gages during a single event (as measured by the standard deviation of
intensity for the gages) is about 5 mm hr1. The range of standard deviations of intensity for
individual runs is 0.97 to 10.5 mm hr1. Higher standard deviations were recorded when wind
velocity was high. The standard deviation of intensity is typically 2-3 mm hr1 provided that calm
winds prevail.
Simulated rainfall intensity is related to water pressure, which can be adjusted by turning a
screw on top of the pressure regulator. Tests indicate that water pressure must be changed by
3.45-5.17 Kpa (1/2-3/4 psi) for a measureable, consistent change in rainfall intensity. Rainfall
intensity decreases with increasing water pressure, being approximately 71 mm hr4 at 62.06 Kpa
(9psi), 62 mm hr1 at 65.5-68.95 Kpa (9.5-10 psi) and 56 mm hr1 at 75.84 Kpa (11 psi). Using a
similar rainfall simulation system, S. Luk (pers. comm.) found that a pressure of 67.23 Kpa (9.75
psi) results in a rainfall intensity of 65 mm hr1, which suggests that results for the two systems are
reproducible.
Water consumption averages approximately 1240 1 hr4 (272 gal. hr1) and the difference in
flow rate between the two uprights varies, but is normally around 6%.
The median drop diameter produced by the Spraco nozzle decreases in size from 810 um at
62.06 Kpa (9 psi) to 72- um at 103.42 Kpa (15 psi). Similarly, the sauter mean drop size
(diameter of a droplet whose ratio of volume to surface area is equal to that of the entire spray
sample) decreases from 648 um at 62.06 Kpa (9 psi) to 576 um at 103.42 Kpa (15 psi) (Spraco,
1985 (77)). These values appear to be slightly less than that produced by natural rainfall.
RESULTS
A total of 21 simulated and natural events were sampled. The simulated events were used
strictly in the collection of erosion rate information. The greatest number of events (7) were
sampled at site 4, a 3.26 m2plot with an 18% slope. Six of the seven events have a runoff and
sediment record for the entire event and average sediment yield (gm m2) for these six events was
regressed against a rainfall erosivity factor (R):
R = Ely) [1]
Iso is the maximum 30 minute rainfall intensity (in hr1) for the event; E is the total kinetic energy of
rainfall (ft tons ac4) for the event (from Wischmeier and Smith, 1958 (44)):
E = S[(916 + 3311oglj) Ij tj] [2]
where L is the rainfall intensity (in hr1) for the given time interval, t; (hours). All measurements in
this study were done using the metric system and although [2] was derived from imperial units, for
simplicity sediment yield, Y, (gm nr2) was regressed against [1] without conversion. The
regression equation for the 6 events is:
Y = 15.2 + 0.382(EI30) [3]
Equation [3] explains 70% of the variance in the data (63% when adjusted for degrees of freedom)
and the bi coefficient (0.382) is significantly different from zero (P=0.03). Other researchers (eg.
Foster et al., 1982 (78)) have also had reasonable success in relating EI^ to sediment yield.
The effect of pervious land characteristics on erosion rates becomes obvious when different
test site responses are compared for the same or similar rainfall input. For example, a
210
-------
thunderstorm on July 17 (El3o=175) resulted in an average sediment yield rate of 95.34 gm m-2
from site 4. The average sediment yeld from site 5, a 3.84 m2 grassed plot with a 24% slope was
0.018 gm nr2 for the same event. In four rainfall simulations at site 5 (Rl3o= 1800-7000) average
sediment yeld ranged between 0.01-0.054 gm m-2. At site 3, a 3.52 m2'gravelled plot with 1-2%
slope, the average yield for 2 rainfall simulations (Rl3o=3000) was 20-25 gm m-2. Rainfall on
October 27 (EIso=83) resulted in an average sediment yeild of 0.216 gm m-2 from site 1 (area=390
m2, slope=3.5%) and 0.009 gm m-2 at site 2 (area=5688 m2, slope-0.6%). The effect of lawn
cover and paving are obvious in this last example. Clearly, the different response rates for the
different land types would result in different bi coefficients in equation [3] and these different
response rates should be considered when modelling pollutant input from pervious land.
Various researchers (eg. Ackermann et al., 1983 (79); Forstner and Wittman, 1981 (80))
have found that metals and organic compounds are preferentially adsorbed and transported by finer
particles. The sediment grain size distribution also affects overland and in-sewer transport
dynamics. Sediment in the samples from most events was therefore separated into sand (>62 um),
silt (5-62 um) and clay (<5 um) fractions by wet filtering. Sediment in the samples to be analysed
for metals was not dispersed prior to filtration.
The average observed eroded sediment grain size distributions calculated from all samples for
selected events at the various sites were calculated . As expected, there are obvious differences in
the eroded grain size distributions from the various sites. Some of the sites (eg. 4 and 5) also
exhibited considerable variability in grain size distribution with time. This variability is in part
related to clay enrichment at the end of an event as transport capacity decreases. Clay content is
also less than average at the beginning of most events, probably due to its cohesive nature and
resistance to erosion. Other factors affecting grain size distribution variability are being
investigated.
Concentration of most metals analysed increase progressively from the sand to clay fraction
although some exceptions do occur, as Mn concentrations, for example, can occasionally be
greater in sand than silt. Average observed concentrations of V associated with the different size
fractions from selected events at site 4 were also calculated Ackermann et al. (1983) (79)
suggested that only the <20 um fraction of sediment need be analysed. However, given the high
proportion of coarse material in our samples, a great deal of the metals load can be carried by this
fraction even though the concentrations are lower. Our field observations show that 65% fo the
total V load is transported by sand, 30% by silt and 5% by clay. We suggest that all size fractions
by analysed to maximise information about pollutant movement.
Higher metal concentrations are associated with sediment eroded from sites that are industrial
in nature or near major transportation routes. In particular, the average Mn (1778 ppm) and V (142
ppm) concentrations on clay for selected events at the gravelled industrial receiving area (site 1) are
greater than from the nearby lawns and roadway (Site 2) and also greater than typical background
(ie. natural) levels (Lisk, 1972 (81)). Vermette et al (82) also found progressively higher Mn
concentrations towards the steel mills in Hamilton when grab samples from different pervious land
types were analysed. Average Mn and V concentrations at Site 6 (samples not fractionated) were
909 and 96 ppm respectively, which are closer to natural levels.
Finally, Mn and V concentrations assocated with the clay fraction were plotted with total
solids (TS) against time for one event from site 1 and one event from site 4. It is typically
assumed that metals concentrations, being conservative, can be related to TS concentrations and TS
data can be more easily and cheaply obtained (James, 1985 (1). The available data showed that in
general the metals concentration time series obtained from the clay fraction most resembled the
solids time series. The Mn and V concentrations appear to have some relationship with the TS
concentration (and also with clay concentration which is not plotted but which exhibited a pattern
similar to the TS concentration) although there may be a lag in response and some deviations from
this general relationship due to variable source areas and to variable atmospheric input. It should
also be noted that not all metals analysed (eg. Ca, Cl) had a similar close relationship with TS.
211
-------
CONCLUSIONS
1. The rainfall simulator is simple, portable and easy to erect and may be used for
simulation in both a laboratory setting and in-situ.
2. Test runs with the simulator indicate that the variation of simulated rainfall intensity
within the test plot is small, typically 2-3 mm hr1 under calm conditions.
3. Variation in simulated rainfall intensity during a single run, and between successive
runs using the same water pressure is due to: a) effect of winds; b) fluctuations in water supply
that are not totally controlled by the pressure regulator, c) unequal elevation of the spray nozzles;
and d) catch errors in the rain gauges.
4. Generally, higher water pressures produce lower simulated intensity rainfalls and the
different intensities are reproducible at adjacent sites.
5. The rainfall simulator can be used on plots with slopes of at least 24.5%.
6. In many areas of a city, erosion from pervious land may provide significant inputs of
particulates, metals and organic compounds (see also Pitt, 1985 (5)). The current practice of
assuming an exponential decay in pollutant load washoff through an event may be one reason that
results are poor for urban stormwater quality models. The results show that metal and total solid
concentrations from pervious land have multiple peaks throughout an event.
7. Although the USLE may not be suited for short time step simulations in an urban
environment, it appears that a rainfall erosivity factor such as the EIso of the USLE may be useful
in determining the amount of sediment detatched from pervious land. We suggest that the erosivity
factor be linked to some type of dynamic transport capacity model as is done in CREAMS. Work
in this area is being carried out.
8. The importance of evaluating eroded grain size distribution is apparent when
examining metal concentrations and potential pollutant transport. The highest metal concentrations
were assocated with the clay fraction of the sediment. However, clay typically made up a small
proportion of the eroded grain size distribution from most sites and a greater proportion of the total
metal load could be transported by the coarse fraction. There appears to be a good relationship
between concentrations for certain metals (eg. Mn, V) and the total solids and clay concentrations,
and further investigation into these relationships will be helpful in pollutant transport modelling.
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. James, W. 1985. PCSWMM User's Manual Runoff Module, CHI Publications, Markham,
Ont.
2. Smith, D.D. and Wischmeier, W.H. 1962. Rainfall erosion. Advances in Agronomy,
14:109-148.
3. Ammon, D.C. 1979. Urban Stormwater Pollutant Buildup and Washoff Relationships.
unpub. M.Eng. thesis, University of Florida, Gainesville, FL.
4. Malmquist, P.A. 1983. Urban Stormwater Pollutant Sources, An Analysis of Inflows and
Outflows of Nitrogen, Phosphorus, Lead, Zinc and Copper in Urban Areas. Chalmers University
of Technology, Goteborg, Sweden, 372 p.
5. Pitt, R. 1985. Characterizing and Controlling Urban Runoff Through Street and Sewerage
Cleaning. USEPA report, EPA/600/S2-85/038, Water Engineering Research Laboratory,
Cincinnati, OH.
212
-------
6. Manning, M.J., Sullivan, R.A. and Kipp, T.M. 1977. Nationwide Evaluation of Combined
Sewer Overflows and Urban Stormwater Discharges - Vol. HI: Characterization of Discharges.
USEPA report 600/2-77-064 c (NTS DB 272 107, Cincinnati, OH.
7. Ongley, E.D., Bynoe, M.C. and Percival, J.B. 1981. Physical and geochemical
characteristics of suspended solids, Wilton Creek, Ontario. Canadian Journal of Earth Sciences,
18:1365-1379.
8. Sartor, J.D. and Boyd, G.B. 1972. Water Pollution Aspects of Street Surface
Contaminants. USEPA report EPA-R2-72-08 (NTIS PB-214 408), Washington, D.C.
9. Benes, P., Cejchanova, M. and Havlick, B. 1985. Migration and speciation of lead in a
river system heavily polluted from a smelter. Water Research, 19:1-6.
10. Crecelius, E.A., Hardy, J.T., Gibson, C.I., Schmidt, R.L., Apts, C.W., Gurtisen, J.M.
and Joyce, S.P. 1982. Copper bioavailability to marine bivalves and shrimps: relationship to
cupric ion activity. Marine Environmental Research, 6:13-26.
11. Florence, T.M. 1977. Trace metal species in fresh waters. Water Research, 11:681-687.
12. Simkiss, K. and Mason, A.Z. 1984. Cellular responses of Molluscan tissues to
environmental metals. Marine Environmental Research, 14:103.118.
13. Chau, Y.K., Wong, P.T.S. and Kramar, O. 1983. The determination of dialkyl lead,
trialkyl lead, tetraalkyl lead and lead (II) ion in water by chelation/extraction and gas
chromatography/atomic adsorption spectrometry. Analytica Chimica Acta, 146:211-217.
14. Chau, Y.K., Wong, P.T.S., Bengert, G.A. and Dunn, J.L. 1984. Determination of dialkyl
lead, trialkyl lead, tetraalkyl lead and lead (II) compounds in sediment and biological samples.
Analytical Chemistry, 56(2):271-274.
15. Ricci, G.R., Colovos, G. and Hester, N.E. 1981. Ion chromatography with atomic
absorption spectrometric detector for determination of organic and inorganic arsenic species.
Analytical Chemistry, 53(4):610-613.
16. Chapman, B.M. 1982. Numerical simulation of the transport and speciation of
nonconservative chemical reactants in rivers. Water Resources Research, 18(1):155-167.
17. Wischmeier, W.H. and Smith, D.D. 1978. Predicting Rainfall Erosion Losses: A Guide to
Conservation Planning, USDA, Agriculture handbook No. 537, 58 p.
18. Williams, J.R. 1975. Sediment Yield Prediction with Universal Equation Using Runoff
Energy Factor. USDA ARS-S-40:244-252.
19. Donigian, A.S. and Crawford, N.H. 1976. Modelling Pesticides and Nutrients on
Agricultural Lands. USEPA Environmental Protection Technology Series, EPA-600/2-76-043,
Washington, D.C. 317 p.
20. Li, R.M., Simons, R.K. and Shiao, L.Y. 1977. Mathematical modeling of on-site soil
erosion. Proceedings, International Symposium on Urban Hydrology, Hydraulics, and Sediment
Control, University of Kentucky, Lexington, KY. 87-94.
21. Simons, D.G., Li, R.M. and Ward, T.J. 1977. A simple procedure for estimating on-site
soil erosion. Proceedings, International Symposium on Urban Hydrology, Hydraulics and
Sediment Control, University of Kentucky, Lexington, KY. 95-102.
22. Foster, G.R., Lane, L.J., Nowlin, J.D., Laflen, J.M. and Young, R.A. 1980. A model to
estimate sediment yield from field-sized areas: development of model, in CREAMS: A Field
Scale Model for Chemicals, Runoff, and Erosion from Agricultural Management Systems, W.G.
Knisel, ed., USDA Conservation Research Report No. 26, Ch. 3.
23. Khanbilvardi, R.M. and Rogowski, A.S. 1984. Mathematical model of erosion and
deposition on a watershed. Transactions, ASAE, 27(l):73-83.
24. Moore, R.J. 1984. A dynamic model of basin sediment yield. Water Resources Research,
20(1):89-103.
25. Rojiani, K.B., Tarbell, K.A., Shanholtz, V.O. and Woeste, F.E. 1984. Probabilistic
modeling of soil loss from surface mined regions. Transactions, ASAE, 27(6): 1798-1804.
26. Boregowda, S. 1984. Modelling Stormwater Pollutants in Hamilton, Canada, unpub.
Ph.D. thesis, Dept. of Civil Engineering, McMaster University, Hamilton, Ontario. 252 p.
213
-------
27. Cermola, J.A., Decarli, S., Sachdev, D.R. and El-Baroudi, H.M. 1979. SWMM
Application to Combined Sewage in Newhaven. J. Env. Eng.Div. 105 (EE6) 1035-1048
28. MacRae, C.R. 1979. Hydrologic and Water Quality Modelling in a Small urban Catchment
in London, Ontario unpub. M.Sc. Thesis, Dept. of Geography, university of Western Ontario,
268 p.
29. Klemetson, S.L. 1985. Factors affecting stream transport of combined sewer overflow
sediment. Journal Water Pollution Control Federation, 57 (5):390-397.
30.. Randall, C.W. 1982. Stormwater detention ponds for water quality control. Proceedings,
Stormwater Detention Facilities Conference, New England College, NH., W. DeGroot, ed., 200-
204.
31. Williams, J.R. 1980. SPNM, a model for predicting sediment, phosphorus, and nitrogen
yields from agricultrual basins. Water Resources Bulletin, 16(5):843-848.
32. Frere, M.H., Onstad, C.A. and Holtan, H.N. 1975. ACTMO, An Agricultural Chemical
Transport Model. USDA, Agricultural Research Service, ARS-H-3, 43 p.
33. Young, R.A. and Onstad, C.A. 1976. Predicting particle-size composition of eroded soil.
Transactions, ASAE, 19:1071-1076.
34. Meyer, L.D., Willoughby, W.E., Whisler, F.D. and Rhoton, F.E. 1983. Predicting size
distributions of sediment eroded from aggregated soils. Transactions, ASAE, 26(2):486-489.
35. Armstrong, J.J. 1978. Erosion control methods during construction. Conference
Proceedings No. 5, Modern Concepts in Urban Drainage, Canada-Ontario Agreement on Great
Lakes Water Quality, 335-374.
36. Burton, T.M., Turner, R.R. and Harriss, R.C. 1976. The impact of highway construction
on a North Florida watershed. Water Resources Bulletin, 12(3):529-538.
37. Diseker, E.G. and Richardson, E.G. 1962. Erosion rates and control methods on highway
cuts. Transactions, ASAE, 5(2):153-155.
38. Keller, FJ. 1962. Effect of urban growth on sediment discharge, northwest branch
Anacostia river basin, Maryland. U.S. Geologic Survey Professional Paper 450-C, 129-131.
39. Walling, D.E. and Gregory, K.J. 1970. The measurement of the effects of building
construction on drainage basin dynamics. Journal of Hydrology, 11:129-144.
40. Wolman, M.G. and Schick, A.P. 1967. Effects of construction on fluvial sediment, urban
and suburban areas of Mary land, Water Resources Research, 3(2):451-464.
41. Guy, H.P. 1967. Research needs regarding sediment and urbanization. Journal of the
Hydraulics Division, 93(HY6):247-254.
42. Cole, R.H., Frederick, R.E., Healy, R.P. and Rolan, R.G. 1984. Preliminary findings of
the Priority Pollutant Monitoring Project of the Nationwide urban Runoff Program. Journal Water
Pollution Control Federation, 56(7)-.898-908.
43. Heaney, J.P. 1986. Research needs in urban storm-water pollution. Journal of Water
Resources Planning and Management, 112(l):36-47.
44. Wischmeier, W.H. and Smith, D.LD. 1958. Rainfall energy and its relationship to soil loss.
Transactions, American Geophysical Union, 39(2):285-291.
45. Wischmeier, W.H. 1959. A rainfall erosion index for a Universal Soil Loss Equation. Soil
Science Society of America, Proceedings, 23:246-249.
46. Laws, J.O. and Parsons, D.A. 1943. Relation of raindrop size to intensity. Transactions,
American Geophysical Union, 24:452-460.
47. Mitchell, J.K. and Bubenzer, G.D. 1980. Soil loss estimation, in Soil Erosion, MJ. Kirkby
and R.P.C. Morgan, eds., John Wiley and Sons, Ltd., Toronto, ch. 2.
48. Al-Durrah, M.M. and Bradford, J.M. 1982. Parameters for describing soil detachment due
to a single waterdrop impact. Soil Science Society of America Journal, 46:836-840.
49. Bryan, R.B. 1976. Considerations on Soil Erodibility Indices and Sheetwash. Catena 3:99-
111.
50. Bryan, R.B. 1979. The influence of slope angle on soil entrainment by sheetwash and
rainsplash. Earth Surface Processes, 4:43-58.
214
-------
51. Emmett, W.W. 1970. The Hydraulics of Overland Flow on hillslopes. U.S. Geological
Survey Professional Paper 662-A, Washington, D.C. 68 p.
52. Evans, R. 1980, Mechanics of water erosion and their spatial and temporal controls: an
empirical viewpoint, in Soil Erosion, M.J. Kirkby and R.P.C. Morgan, eds., John Wiley and
Sons, Ltd., Toronto. Ch. 4.
53. Luk, S.H. 1979. Effect of soil properties on erosion by wash and splash. Earth Surface
processes, 4:241-255.
54. Luk, S.H. and Hamilton, H. 1986. Experimental effects of antecedent moisture and soil
strength on rainwash erosion of two luvisols, Ontario. Geoderma, 37:29-43.
55. Morris, S.E. 1986. The significance of rainsplash in the surficial debris cascade of the
Colorado Front Range Foothills. Earth Surface Processes and Landforms, 11:11-22.
56. Poesen, J. and Savat, J. 1981. Detachment and transportation of loose sediments by
raindrop splash. P art II Detachability measurements. Catena, 8:19-41.
57. Kirkby, M.J. 1980. Modelling water erosion processes, in Soil Erosion, J.J. Kirkby and
R.P.C. Morgan, eds., John Wiley and Sons, Ltd., Toronto, ch. 6.
58. James, W. and Boregowda, S. 1985a. Continuous mass-balance of pollutant build-up
processes in Urban Runoff Pollution, Springer-Verlag, Berlin 243-276.
59. James, W. and Boregowda, S. 1985b. Stormwater pollution modelling: build up of dust
and dirt on surfaces subject to runoff. Canadian Journal of Civil Engineering, 12(4):906-915.
60. Novotony, V., Sung, H.M., Bannerman, R. and Baum, K. 1985. Estimating nonpoint
pollution from small urban watersheds. Journal Water Pollution Control Federation, 57(4):339-
348.
61. Hamilton, E.P. and Chatt, A. 1982. Determination of trace elements in atmospheric wet
precipitation by instrumental neutron activation analysis. Journal of Radioanalytical Chemistry,
71:29-45.
62. Tanaka, S., Darzi, M. and Winchester, J.W. 1981. Elemental analysis of soluble and
insoluble fractions of rain and surface waters by particle-induced x-ray emmision. Environmental
Science and Technology, 15(3):354-357.
63. Slinn, W.G.N. 1977. Some approximations for the wet and dry removal of particles and
gases from the atmosphere. Water, Air and soil Pollution, ll(7):513-543.
64. Jeffries, D.S. and Synder, W.R. 1981. Atmospheric deposition of heavy metals in Central
Ontario. Water, Air and Soil Pollution, 15:127-152.
65. Ontario Research Foundation, United Technology & Science Inc., Concord Scientific Corp,
and the MEP Company, 1982. An Assessment of Street Dust and Other Sources of Airborne
Paniculate Matter in Hamilton, Ontario. Report ISBN 0-7743-7728-0, prepared for Environment
Canada and MOE, 325 p.
66. Prasad, D., Henry, J.G. and Kovacks, R. 1980. Pollution potential of autumn leaves in
urban runoff. International Symposium on Urban Storm Runoff, Lexington, KY. 197-201.
67. Bagnold, R.A. 1941. The Physics of Blown Sand and Desert Dunes. Chapman and Hall,
London, 265 p.
68. Wilson, SJ. and Cooke, R.U. 1980. Wind erosion, in Soil Erosion, MJ. Kirkby and
R.P.C. Morgan, eds., John Wiley and Sons, Ltd., Toronto, ch. 7.
69. de Ploey, J. and Gabriels, D. 1980. Measuring Soil Loss and Experimental Studies in Soil
Erosion M.J. Kirkby and R.P.C. Morgan, Eds., John Wiley and Sons Ltd. Toronto Ch. 3.
70. James, W. and Stirrup, D.M. 1986. Microcomputer-based precipitation instrumentation in
Urban Drainage Modelling Program on Press, 471-482.
71. Bryan, R.B. 1970. An improved rainfall simulator for use in erosion research. Canadian
Journal of Earth Sciences, 7:1552-1561.
72. Bryan, R.B. and de Ploey, J. 1983. Comparability of soil erosion measurements with
different laboratory rainfall simulators, in J. de Ploey, ed. Rainfall Simulation Runoff and Soil
Erosion, Cantena Supplement 4, 33-56.
215
-------
73. Bubenzer, G.D. and Jones, B.A. 1971. Drop size and impact velocity effects on the
detachment of soils under simulated rainfall. Transactions of the ASAE, 15:625-627.
74. Imeson, A.C. 1977. A simple field-portable rainfall simulator for difficult terrain. Earth
Surface Processes, 2:431-436.
75. Glew, J.R. 1976. The Simulation of Rillenkarren. unpub. M.Sc. thesis, Dept. of
Geography, McMaster University, Hamilton, Ont. 116 p.
76. Johanson, R.C. 1967. System Analysis of the Rainfall-Runoff Process, unpub. M.Eng.
thesis, University Natal, Durban, South Africa. 95 p.
77. Spraco, 1985. Spray Nozzle User's Manual. Spraco, Inc., Nashua NH. 125 p.
78. Foster, G.R., Lombardi, F. and Moldenhauer, W.C. 1982. Evaluation of rainfall-runoff
erosivity factors for individual storms. Transactions, ASAE, 25:124-129.
79. Ackerman, F., Bergmann, H. and Schleichert, U. 1983. Monitoring of heavy metals in
coastal and estuarine sediments - A question of grain-size: <20 um versus <60 um.
Environmental Technolgy letters, 4:317-328.
80. Forstner, U. and Wittmann, S. 1981. Metal Pollution in the Aquatic Environment, 2nd ed.,
Springer-Verlag, New York, 486 p.
81. Lisk, S. 1972. Trace Metals in Soils, Plants and Animals. Advances in Agronomy.
24:2676-325.
82. Vermette, S., Irvine, K. and Drake, J. (in preparation) Elemental and Size Distribution
Characteristics of Urban Sediments: Hamilton Ontario.
216
-------
UNCERTAINTY IN HYDROLOGIC MODELS:
A REVIEW OF THE LITERATURE
by: T. V. Hromadka II
Director of Water Resources,
Williamson and Schmid, Irvine, CA 92714,
and Research Associate, Princeton
University, Princeton, NJ 08544
ABSTRACT
Advances in hydrologic modeling techniques typically involves the incor-
poration of higher complexity into the hydrology model by use of improved
hydraulic submodels or a more refined approximation of the several subpro-
cesses integrated in the hydrologic cycle. With over 100 models reported in
the open literature, it is appropriate to review the progress achieved by the
complexification of hydrologic models. That is, it is time to evaluate
whether the general level of success afforded by the many types of complex
models provide a marked improvement over that achieved by the more commonly
used and simpler models such as the unit hydrograph method. Such a review
indicates that it is still not clear, in general, whether as modeling com-
plexity increases, modeling accuracy increases. It appears that a major
limitation to the successful development, calibration, and application, of
any hydrologic model is the uncertainty of the effective rainfall distribution
over the catchment.
INTRODUCTION
A review of the literature indicates that a substantial evolution in modeling
complexity has occurred over the last two decades. The majority of changes have
occurred in the incorporation of soil moisture accounting techniques and intricate link-
node model discretization using approximations for hydraulics. However in spite of the
advances made in the modeling complexity, the accuracy of models (in general) has not
been significantly improved in the correlation of rain gage data to stream gage runoff
data. Only a handful of papers and reports are available in the open literature which
compare modeling performance, and each of these reports note that simpler models do as
good as or better than complex models. Additionally, many of the papers indicate that
the uncertainty in the effective rainfall distribution over the catchment may be a key
factor in the lack of major gains in the development, calibration, and application, of
hydrologic models. As a result of this lack in demonstrated success in the use of any
particular advanced modeling technique or approach, there is continued reliance by the
217
-------
engineering community to use the more simpler modeling approaches such as the rational
method for peak flow rate estimates, or the classic unit hydrograph method when a runoff
hydrograph is needed.
In this paper, the literature is reviewed to identify trends which support using
simpler models such as the unit hydrograph method (more specifically, the design
storm/unit hydrograph approach). From these trends, suggestions as to how the more
complex modeling techniques may "prove themselves to practitioners and, therefore, find
more use in the engineering community, may be formulated.
MODEL SELECTION
In the selection of the hydrologic model, the need for both runoff peak flow rates
and runoff volumes (for the testing of detention basins) require the selection of a model
that produces a runoff hydrograph. The U.S. Army Corps of Engineers (COE) Hydrologic
Engineering Center (HEC) Training Document (TD) No. 11, (1980) (1) categorizes all
hydrologic models into eight groupings of which three develop a runoff hydrograph;
namely, single event (design storm), multiple discrete events, and continuous records
(continuous simulation). These models can be further classified according to the
submodels employed. For example, a unit hydrograph or a kinematic wave model may be
used to represent the catchment hydraulics.
In a survey of hydrologic model usage by Federal and State governmentaal agencies
and private engineering firms (U.S. Department of Transportation, Federal Highway
Administration, Hydraulic Engineering Circular No. 19, October 1984 (2), it was found
that "practically no use is made of watershed models for discrete event and continuous
hydrograph simulation." In comparison, however, design storm methods were used from 24
to 34 times more frequently than the complex models by Federal agencies and the private
sector, respectively. The frequent use of design storm methods appear to be due to
several reasons: (1) design storm.methods are considerably simpler to use than discrete
event and continuous simulation models; (2) it has not been established in general that the
more complex models provide an improvement in computational accuracy over design
storm models; and (3) the level of complexity typically embodied in the continuous
simulation class of models does not appear to be appropriate for the catchment rainfall-
runoff data which is typically available. Consequently, the design storm approach is most
often selected for flood control and drainage policies (considerations in the choice of
modeling approach are contained in the latter sections).
The next decision is whether to use the standard unit hydrograph method or the
more recently advanced kinematic wave method to model catchment hydraulics. Again, it
has not been clearly established that the kinematic wave approach (e.g., the overload flow
place concept) provides an improvement in modeling accuracy over the unit hydrograph
approach that has been calibrated to local rainfall-runoff data.
For the choice of design storm to be used, the work of Beard and Chang (1979) (3)
and HEC ("Hypothetical Floods", 1975) (4) provide a logical motivation for developing a
design storm using rainfalls of identical return frequency, adjusted for watershed area
effects.
Finally, specific components of the modeling approach must be selected and
specified. Inherent in the choice of submodels is the ability to calibrate the model at two
levels: (1) calibration of model parameters to represent local or regional catchment
rainfall-runoff characteristics, and (2) calibration of the model parameters (or design
218
-------
storm) to represent local rainfall intensity-duration-frequency characteristics. Beard and
Chang (1979) (3) note that in a hydrologic model, the number of calibration parameters
should be as small as possible in order to correlate model parameters with basin
characteristics. They also write that a regional study should be prepared to establish the
loss rate and unit hydrograph characteristics, "and to compute from balanced storms of
selected frequencies (storms having the same rainfall frequency for all durations) the
resulting floods."
LITERATURE REVIEW
CHOICE OF WATERSHED MODEL
In developing a flood control and drainage policy, the first, and possibly the most
important question to answer is: what type of model should be used to form the basis for
design calculations? To answer this question, the literature was reviewed extensively.
Based on the research findings summarized in the following paragraphs, the design
storm/unit hydrograph (UH) method appears to have continued support among
practitioners. The question naturally arises as to why the simple UH method continues to
be the dominant hydrologic tool when considerably more complex models (e.g., the
continuous simulation class of models which has a mathematical approximation for each
component of the hydrologic cycle, and typically utilizes physically based hydraulic flow
routing approximations. The Stanford Watershed Model is an excellent example of this
class of approach.) are available for public use. As explanation frequently cited in the
literature appears to be that the uncertainty in the effective rainfall over the catchment
overshadows the improved accuracy that may be possibly achieved by more complex
models.
A criterion for complex and simple models is given by Beard and Chang (1979) (3) as
as the "difficulty or reliability of model calibration—Perhaps the simplest type of model
that produces a flood hydrograph is the unit hydrograph model"...and..." can be derived to
some extent from physical drainage features but fairly easily and farily reliably calibrated
through successive approximations by relating the time distribution of average basin
rainfall excess to the time distribution of runoff." In comparison, the "most complicated
type of model is one that represents each significant element of the hydrologic process by
a mathematical algorithm. This is represented by the Stanford Watershed Model and
requires extensive data and effort to calibrate."
The literature contains several reports of problems in using complex models,
especially in parameter optimizations. Additionally, it has not been clearly established
whether complex models, such as in the continuous simulation or discrete event classes of
models, provide and increase in accuracy over a standard design storm unit hydrograph
model.
There are only a few papers and reports in the literature that provide a comparison
in hydrologic model performance. From these references, it appears that a simple unit
hydrograph model provides as good as or better results than quasi-physically based (or
QPB, see the work of League and Freeze (1985)) (5) or complex models.
In their paper, Beard and Chang (1979) (3) write that in the case of the unit
hydrograph model, "the function of runoff versus rainfall excess is considered to be linear,
whereas it usually is not in nature. Also, the variations in shapes of unit hydrographs are
not derivable directly from physical factors. However, models of this general nature are
usually as representative of physical conditions as can reasonably be validated by
219
-------
available data, and there is little advantage in extending the degree of model
sophistication beyond validation capability." It is suggested that "if 50 yr-100 yr of
streamflow were available for a specified condition of watershed development, a
frequency curve flows for that condition can be constructed from a properly selected set
of flows."
Schilling and Fuchs (1986) (6) write "that the spatial resolution of rain data input is
of paramount importance to the accuracy of the simulated hydrograph" due to "the high
spatial variability of storms" and "the amplification of rainfall sampling errors by the
nonlinear transformation" of rainfall into runoff. Their recommendations are the model
should employ a simplified surface flow model if there are many subbasins; a simple
runoff coefficient loss rate; and a diffusion (zero inertia) or storage channel routing
technique. Hornberger, et al. (1985) (7) writes that "Even the most physically based
models...cannot reflect the true complexity and heterogeneity of the processes occurring
in the field. Catchment hydrology is still very much an empirical science."
In attempting to define the modeling processes by the available field data forms
Hornberger, et.al. find that "Hydrological quantities measured in the field tend to be
either integral variables (e.g. stream discharge, which reflects an integrated catchment
response) or point estimates of variables that are likely to exhibit marked spatial and/or
temporal variation (e.g., soil hydraulic conductivity)." Hence, the precise definition of
the physics in modeling sense becomes a problem that is "poorly posed in the
mathematical sense." Typically, the submodel parameters cannot be estimated precisely
due to the large associated estimation error. "Such difficulties often indicate that the
structural complexity of the model is greater than is warranted on the basis of the
calibration data set."
Schilling and Fuchs (1986) (6) note that errors in simulation occur for several reasons
including:
"1. The input data, consisting of rainfall and antecedent conditions, vary
throughout the watershed and cannot be precisely measured.
2. The physical laws of fluid motion are simplified.
3. Model parameter estimates may be in error."
By reducing the rainfall data set resolution from a grid of 81 gages to a single
catchment-centered gage in and 1,800 acre catchment, variations in runoff volumes and
peak flows "is well above 100 percent over the entire range of storms implying that the
spatial resolution of rainfall has a dominant influence on the reliability of computed
runoff." It is also noted that "errors in the rainfall input are amplified by the rainfall-
runoff transformation" so that "a rainfall depth error of 30 percent results in a volume
error of 60 percent and peak flow error of 80 percent."
Schilling and Fuchs (1986) (6) also write that "it is inappropriate to use a
sophisticated runoff model to achieve a desired level of modeling accuracy if the spatial
resolution of rain input is low" (in their study, the rainage densities considered for the
1,800-acre catchment are 81-, 9-, and a single centered gage).
In a similar vein, Beard and Chang (1979) (3) write that in their study of 14 urban
catchments, complex models such as continuous simulation typically have 20 to 40
parameters and functions that must be derived from recorded rainfall-runoff data.
"Inasmuch as rainfall data are for scattered point locations and storm rainfall is highly
variable in time and space, available data are generally inadequate in this region for
220
-------
reliably calibrating the various interrelated functions of these complex models."
Additionally, "changes in the model that would result from urbanization could not be
reliably determined." They write that the application "of these complex models to
evaluating changes in flood frequencies usually requires simulation of about 50 years of
streamflow at each location under each alternative watershed condition."
Garen and Burges (1981) (8) noted the difficulties in rainfall measurement for use in
the Stanford Watershed Model, because the Kl parameter (rainfall adjustment factor) and
UZSN parameter (upper level storage) had the dominant impact on the model sensitivity.
This is especially noteworthy because Dawdy and O'Donnell (1965) (9) concluded that
insensitive model coefficients could not be caibrated accurately. Hence, they could not
be reliably used to measure physical effects of watershed changes.
Using another complex model, Mein and Brown (1978) (10) write that on "the basis of
several tests with the Boughton model it is concluded that for this model at least,
relationships derived between any given parameter value and measureable watershed
characteristics would be imprecise; i.e., they would have wide confidence limits. One
could not be confident therefore in changing a particular parameter value of this model
and then claiming that this alteration represented the effect of some proposed land use
change. On the other hand, the model performed quite well in predicting flows with these
insensitive parametrs, showing that individual parameter precision is not a prerequisite to
satisfying output performance."
According to Gburek (1971),"...a model system is merely a researcher's idea of how a
physical system interacts and behaves, and in the case of watershed research, watershed
models are usually extremely simplified mathematical descriptions of a complex physical
situation...until each internal submodel of the overall model can be independently
verified, the model remains strictly a hypothesis with respect to its internal locations and
transformations..." (also quoted in McPherson and Schneider, (1974) (12)).
The introduction of a paper by Sorooshian and Gupta (1983) provides a brief review
of some of the problems reported by other researchers in attempting to find a "true
optimum" parametr set for complex models, including the unsuccessful two man-year
effort by Johnston and Pilgrim (1973) (14) to optimize parameters for a version of the
Boughton model cited above.
In the extensive study by League and Freeze (1985) (5), three event-based rainfall-
runoff models (a regression model, a unit hydrograph model, and a kinematic wave quasi-
physically based model) were used on three data sets of 269 events from three small
upland catchments. In that paper, the term "quasi-physically based" or QPB is used for
the kinematic wave model. The three catchments were 25 acres, 2.8 mi2, and 35 acres in
size, and were extensively monitored with rain gage, stream gage, neutron probe, and soil
site testing.
For example, the 25 acre site contained 35 neutron probe access sites, 26 soil
parameter sites (all equally spaced), an on-site rain gage, and a stream gage. The QPB
model utilized 22 overland flow planes and four channel segments. In comparative tests
between the three modeling approaches to measured rainfall-runoff data it was concluded
that all models performed poorly and the QPB performance was only slightly improved by
calibration of its most sensitive parameter, hydraulic conductivity. They write that the
"conclusion one is forced to draw...is that the QPB model does not represent reality very
well; in other words, there is considerable model error present. We suspect this is the
case with most, if not all conceptual models currently in use." Additionally, "the fact
221
-------
that simpler, less data intensive models provided as good or better predictions than a QPB
is food for thought."
Based on the above selected sample of literature, the main difficulty in the use,
calibration, and development, of complex models appears to be the lack of precise rainfall
data and the high model sensitivity to (and magnification of) rainfall measurement errors.
Nash and Sutcliff (1970) (15) write that "As there is little point in applying exact laws to
approximate boundary conditions, this, and the limited ranges of the variables
encountered, suggest the use of simplified empirical relations."
It is noteworthy to consider HEC Research Note No. 6 (1979) (16) where the
Hydrocomp HSP continuous simulation model was applied to the West Branch DuPage
River in Illinois. Personnel from Hydrocomp, HEC, and COE participated in this study
which started with a nearly complete hydrologic/meteorologic data base. "It took one
person six months to assemble and analyze additional data, and to learn how to use the
model. Another six months were spent in calibration and long-record simulation." This
time allocation applies to only a 28.5 mi^ basin. The quality of the final model is
indicated by the average absolute monthly volume error of 32.1 and 28.1 percent for
calibration and verification periods, respectively. Peak flow rate average absolute errors
were 26 and 36 percent for calibration and verification periods, respectively. It was
concluded that "Discharge frequency under changing urban conditions is a problem that
could be handled by simpler, quicker, less costly approaches requiring much less data; e.g.,
design storms or several historial events used as input into a single-event model, or a
continuous model with a less complex soil-moisture accounting algorithm."
The complex model parameter optimization problem has not been resolved. For
example, Gupta and Sarooshian (1983) (17) write that "even when calibrated under ideal
conditions (simulation studies), it is often impossible to obtain unique estimates for the
parameters." Troutman (1982) (18) also discusses the often cited difficulties with the
error in precipitation measurements "due to the spatial variability of precipitation". This
source of error can result in "serious errors in runoff prediction and large biases in
parameter estimates by calibration of the model."
Because it still has not been well established in the open literature whether there is
a significant advantage in using a watershed model more complex or physically based than
a design storm unit hydrograph approach, the design storm unit hydrograph method will
probably have continued widespread use among practitioners for flood control design and
planning studies.
NONLINEARITY: USE OF NONLINEAR KINEMATIC WAVE METHOD
OR A LINEAR UNIT HYDROGRAPH METHOD
The dominant method used in runoff hydrograph development for presenting
catchment runoff response is the unit hydrograph (UH). The next most frequently used
method is the kinematic wave overland flowplane concept (KW). HEC TD#15 (1982) (19)
provides a description and comparison of these two alternatives. The relative usage of
KW by 1983 is indicated in Cermak and Feldman (1983) (20) who write that "actual
applications by Corps field offices have been few to nonexistent. Even at HEC the KW
approach has not been utilized in any special assistance projects." The relatively small
usage of KW were then explained as being due to the slack in hydrologic studies and due to
unfamiliarity with the technique.
222
-------
Watt and Kidd (1975) (21) write that in the comparison of so-called 'physically-based'
or "black-box' modeling types (e.g., UH or n-linear reservoirs) the differences are not
clear. For example, "except for certain 'ideal' laboratory catchments, the flow does not
conform to the sheet-flow model but instead occurs in many small rivulets...The choice is
then between a 'black-box' model and a 'physically-based' model which is based on a
physical situation quite different than the actual field situation, i.e., a *black box' model."
However, use of KW implies a non-linear response whereas the UH implies a linear
response. Nash and Sutcliffe (1970) (15) write tht "the UH assumption of a linear time .
invariant relationship cannot be tested because neither the input (effective rainfall) nor
output (storm runoff) are unequivocally defined." Although watershed response is often
considered to be mathemtically nonlinear, the nonlinearity of the total watershed response
has not been shown to be exactly described as a KW. Indeed, a diffusion hydrodynamic
model, DHM (Hromadka and Yen, 1986) (22), provides another nonlinear watershead
response that includes an additional term in the governing St. Vanant flow equations and
that may differ significantly in response from a KW model (e.g. overland flow planes with
KW channel routing). There are an infinity of nonlinear mathematical representations
possible as a combination of surface runoff and channel routing analogs, therefore, merely
claiming that the response of a watershed model can be classified as 'nonlinear' is not
proof that the model represents the true response of the catchment.
Given that the KW analog is only used to obtain an approximation to catchment
response, the KW approach does not appear to provide significantly better computational
results (for floods of interest in flood control design and planning) than the commonly used
UH method. Dickenson et al. (1967) (23) noted that "in the range of discharges normally
considered as flood hydrographs, the time (of concentration) remained virtually constant.
•In other words, in the range of flood interest, the nonlinear effect approached linearity."
An explanation was advanced that "at low discharges, the mean velocity may vary
considerably with discharge. However, for higher discharges contained within banks, the
mean velocity in the channel remains approximately constant."
In actual travel time measurements of flows in a 96-acre catchment using a
radioactive tracing technique, Pilgrim (1976) (24) noted that although the flood runoff
process "is grossly nonlinear at low flows, linearity is approximated at high flows."
Pilgrim also writes that "simple nonlinear models fitted by data from events covering the
whole range of flow may give gross errors when used to estimate large events." It is
noted that overbank flow was one of the factors for linearity in this study.
Seven (1979) (25) proposed to place limits on the nonlinearity associated to KW by
the specification of a constant flow velocity for catchment runoff for large floods. He
proposes "a nonlinear channel system at low flows and a linear system at high flows into a
single model." Hence for flood flows of interest in flood control planning and design,
Seven's model would reduce to a linear representation of the catchment hydraulics.
A physical test of the KW concept was provided by Hjelmfelt and Burwell (1984)
(26), who studied a set of 40 similar erosion plots and the net response to storm events.
Due to the large variability in measured runoff quantities from the plots, however, it was
concluded that a criterion for a valid rainfall-runoff model "is that it predicts the mean
runoff for each event." However, it is noted that this test may be more of a test of
effective rainfall variability over the catchment than a test of KW response.
In HEC Technical Paper No. 59 (1978) (27), six models, plus two variants of one of
these models and a variant of another, were calibrated and tested on a 5.5 mi2 urban
223
-------
catchment in Castro Valley near Oakland, California. Both single event and continuous
simulation models based on both UH and KW techniques were used in the test. The study
concluded that for this watershed "the more complex models did not produce better
results than the simple models..." An examination of the test results between the KW and
HEC-1 UH models did not show a clear difference between the methods.
It is of interest with Singh (1977) (28) concluded that "if one is not very confident in
estimates of watershed infiltration then in some circumstances linear models may have an
advantage over nonlinear models in runoff peak predictions because they do not amplify
the input errors." That is, the uncertainity in effective rainfall quantities may be
magnified by a nonlinear model; consequently, there is an advantage in using a linear
model when there are errors in loss rate and precipitation estimates.
Because it has not been well established whether the nonlinear KW method for
modeling surface runoff provides an improvement in accuracy over the linear UH based
hydrologic models, the UH model will probably continue to be the most often used runoff
model among practitioners.
DESIGN STORMS
HEC (Beard, 1975) (29) provides an in-depth study of the use of design runoff
hydrographs for flood control studies. "Hypothetical floods consists of hydrographs of
artifical flood flows...that can be used as a basis for flood-control planning, design and
operation decisions or evaluations. These floods represent classes of floods of a specified
or impied range of severity." Such "floods are ordinarily derived from rainfall or
snowmelt or both, with ground conditions' that are appropriate to the objectives of the
study, but they can be derived from runoff data alone, usually on the basis of runoff
volume and peak-flow frequency, studies and representative time sequences of runoff."
In complex watershed systems that include catchment subareas, and channel and
basin routing components, Beard (1975) (29) writes that "it is usually necessary to simulate
the effects of each reservoir on downstream flows for all relevant magnitudes of peaks
and volumes of inflows. Here it is particularly important that each hypothetical flood has
a peak flow and volumes for all pertinent durations that are commensurate in severity, so
that each computed regulated flow will have a probability or frequency that is comparable
to that of the corresponding unregulated flow...In the planning of a flood control project
involving storage or in the development of reservoir operation rules, it is not ordinarily
known what the critical duration will be, because this depends on the amounts of reservoir
space and release in relation to flood magnitude. When alternate types of projects are
considered, critical durations will be different, and a design flood should reflect a degree
of protection that is comparable for the various types of projects."
Beard (1975) (29) notes that the balanced storm concept is an important argument
for not using a historic storm pattern or sequence of storm patterns (e.g., continuous
simulation or discrete event modeling) as "No one historical flood would ordinarily be
representative of the same severity of peak flow and runoff volumes for all durations of
interest." Indeed, should a continuous simulation study be proposed such that the "project
is designed to regulate all floods of record, it is likely that one flood will dictate the type
of project and its general features, because the largest flood for peak flows is also usually
the largest-volume flood." Hence a continuous simulation model of say 40 years of data
can be thought of as a 40 year duration design storm with its own probability of re-
occurrence, which typically reduces for modeling purposes to simply a single or double day
storm pattern.
224
-------
Beard and Chang (1979) (30) write that for design storm construction, "it is generally
considered that a satisfactory procedure is to construct an aproximately symmetrical
pattern of rainfall with uniform areal distribution having intensities for all durations
corresponding to the same recurrence interval and for that location and size of area" (i.e.,
depth-area effects).
The nested design storm concept is developed in detail in HEC TD#15 (1982) (19),
including the use of depth-area adjustments. Again, because the current alternatives to
the design storm approach (i.e., continuous simulation or discrete event modeling) have
not been well established in the literature to provide more accurate estimates of flood
frequency values, the design storm approach probably will have continued widespread
usage among practitioners.
MODEL SELECTION
Of the over 100 models available, a design storm/unit hydrograph model (i.e.,
"model") is currently the most widely used modeling technique among practitioners. Some
of the reasons are as follows: (1) the design storm approach—the multiple discrete event
and continuous simulation categories of models have not been clearly established to
provide better predictions of flood flow frequency estimates for evaluating the impact of
urbanization and for design flood control systems than a calibrated design storm model;
(2) the unit hydrograph method—it has not been shown that the kinematic wave modeling
technique provides a significantly better representation of watershed hydrologic response
than a model based on unit hydrographs (locally calibrated or regionally calibrated) that
represent free-draining catchments; (3) model usage—the "model" has been used
extensively nationwide and has proved generally acceptable and reliable; (4) parameter
calibration—the "model" usually is based on a minimal number of parameters, generally
giving higher accuracy in calibration of model parameters to rainfall-runoff data, and the
design storm to local flood flow frequency tendencies; (5) calibration effort—the "model"
does not require large data or time requirements for calibration; (6) application effort—
the "model" does not require a large computational effort for application; (7)
acceptability—the "model" uses algorithms (e.g., convolution, etc.) that have gained
acceptance in engineering practice; (8) model flexibility for planning—data handling and
computational submodels can be coupled to the "model" (e.g., channel and basin routing)
resulting in a highly flexible modeling capability; (9) model certainty evaluation—the
certainty of modeling results can be readily evaluated as a distribution of possible
outcomes over the probabilistic distribution of parameter values.
FURTHER RESEARCH NEEDS
Even though there are many hydrologic and "physically based" models reported in
the literature which contain algorithms to model each of the specific hydrologic cycle
processes, the basic design storm/unit hydrograph method continues to be the most widely
used approach among practitioners. It appears that for another class of model to become
the engineering standard tool, the model must clearly demonstrate the benefits in its use
for the corresponding increase in computational effort. Such demonstrations include
exhaustive comparisons of modeling performance in accuracy and reliability; not only in
the reproduction of known storm events, but in the estimate of flood frequency peak flow
rate estimates. Finally, storm runoff hydrograph analyses should be conducted as
verification runs, where the storms tested are not elements of the calibration data set.
225
-------
CONCLUSIONS
Although modern hydrologic models have reached a high level of elegance in the
mathematical approximation of the rainfall-runoff process, the simpler models such as the
classic unit hydrograph approach continue to be the most widespread used study tool
among practitioners. In this paper, a review of the literature indicates that the more
complex models have not sufficiently proven themselves to be significantly better
computational tools. Indeed, many key reports indicate that the simpler UH modeling
approach provides computational results which are as good as or better than those
achieved by the more complex models.
Based on the selected literature, the uncertainty in the effective rainfall
distribution over the catchment appears to be a major limitation to the successful
development, calibration, and usage, of any hydrologic model. And this uncertainty in
effective rainfall appears to be more of a problem to complex model performance than
for simpler model's performance.
REFERENCES
1. U.S. Army Corps of Engineers, The Hydrologic Engineering Center, Adoption of
Flood Flow Frequency Estimates at Uhgaged Locations, Training Document 11,
February, 1980.
2. U.S. Department of Transportation, Federal Highway Administration, Hydrology,
Hydraulic Engineering Circular No. 19, October, 1984.
3. Beard, L., Chang, S., Journal of the Hydraulics Division, Urbanization Impact of
Streamflow, June, 1979.
4. U.S. Army Corps of Engineers, The Hydrologic Engineering Center, Hydrologic
Methods for Water Resources Development: Volume 5 - Hypothetical Floods,"
March, 1975.
5. Loague, K., Freeze, R., A Comparison fo Rainfall-Runoff Modeling Techniques on
Small Upland Catchments, Water Resources Research, Vol. 21, No. 2, February,
1985.
6. Schilling, W. Fuchs, L., Errors in Storm water Modeling—A Quantitative Assessment,
Journal of Hydraulic Engineering, Vol. 112, No. 2, February, 1986.
7. Hornberger, et al, Schenandoah Water Shed Study: Calibration of a Topography-
Based, Variable Contributing Area Hydrological Model to a Small Forested
Catchment, Water Resources Research, Vol. 21, No. 12, Dec. 1985.
8. Garen, D., Burges, S., Approximate Error Bounds for Simulated Hydrographs, Journal
of The Hydraulics Division, Proceedings of the American Society of Civil Engineers,
ASCE, Vol. 107, No. HY11, November, 1981.
9. Dawdy, D., O'Donnell, T., Mathematical Models of Catchment Behavior, Journal of
the Hydraulics Division, Vol. 91, No. HY4, July, 1965.
10. Mein, R. G. and Brown, B. M., Sensitivity of Optimized Parameters in Watershed
Models, Water Resources Research, Vol. 14, 1978.
11. Gburek, W. J., Discussion of Hydrologic consequences of rainfall augmentation,
Journal of The Hydraulic Division, ASCE, Vol. 97, HY12, 2114-2115, 1971.
12. McPherson, M., Schneider, W., Problems in Modeling Urban Watersheds, Water
Resources Research, Vol. 10, No. 3, June 1974.
13. Sorooshian, S., Gupta, V., Automatic Calibration of Conceptual Rainfall-Runoff
Models: The Question of Parameter Observability and Uniqueness, Water Resources
Research, Vol. 19, No. 1, February, 1983.
14. Johnston, P., Pilgrim, D., Parameter Optimization for Watershed Models, Water
Resources Research, Vol. 12, No. 3, June, 1976.
226
-------
15. Nash, J., Sutcliffe, J., River Flow Forecasting Through Conceptual Models Part 1 -A
Discussion of Principles, Journal of Hydrology, 10, 282-290, 1970.
16. U.S. Army Corps of Engineers, the Hydrologic Engineering Center, Continuous
Hydrologic Simulation of the West Branch Dupage River Above West Chicago: An
Application of Hydrocomp's HSP, Research Note No. 6, 1979.
17. Gupta, V., Sorooshian, S., Uniqueness and Observability of Conceptual Rainfall-
Runoff Model Parameters: The Percolation Process Examined, Water Resources
Research, Vol. 19, No. 1, pp 269-276, February, 1983.
18. Troutman, B., An Analysis of input in Preception-Runoff Models Using Regression
with Errors in the Independent Variables, Water Resources Research, Vol. 18, No. 4,
pp 947-964, August, 1982.
19. U.S. Army Corps of Engineers, The Hydrologic Engineering Center, Hydrologic
Analysis of Ungaged Watersheds Using HEC-1, Training Document No. 15, April,
iy 82.
20. Cermak, R., Feldman, A., Urban Hydrologic Modeling Using HEC-1/Kinematic
Wave, Presented at The 19th Annual AWRA Conference, October 9-13, 1983, San
Antonio, Texas.
21. Watt, W., Kidd, C., Quurm-A Realistic Urban Runoff Model, Journal of Hydrology,
27 (1975) 225-235, Elsevier Scientific Publishing Company, Amsterdam-Printed in
The Netherlands.
22. Hromadka II, T.V. and Yen, C. C., A Diffusion Hydrodynamic Model, Advances in
Water Resources, Vol. 9, No. 3, pp 118-170, 1986.
23. Dickinson, W., et al, An experimental Rainfall-Runoff Facility, No. 25, Hydrology
Papers, Colorado State University, Fort Collins, Colorado, September, 1967.
24. Pilgrim, D., Travel Times and Nonlinearity of Flood Runoff from Tracer
Measurements on a Small Watershed, Water Resources Research, Vol. 12, No. 3,
June, 1976.
25. Beven, K., On the Generalized Kinematic Routing Method, Water Resources
Research, Vol. 15, No. 5, October, 1979.
26. Hjelmfelt, A., Burwell, R., Spatial Variability of Runoff, Journal of Irrigation and
Drainage Engineering, Vol. 110, No. 1, March, 1984.
27. U.S. Army Corps of Engineers, The Hydrologic Engineers Center, Testing of Several
Runoff Models on an Urban Watershed, Technical Paper No. 59, 1978.
28. Singh, V. P., Sensitivity of Some Runoff Models to Errors in Rainfall Excess, Journal
of Hydrology, 33 pgs, 1977.
29. Beard, L., Impact of Hydrologic Uncertainties on Flood Insurance, Journal of the
Hydraulics Division, American Society of Civil Engineers, Vol. 104, No. HY11,
November 1978.
227
-------
UNCERTAINTY IN FLOOD CONTROL DESIGN
by: T. V. Hromadka II
Director of Water Resources,
Williamson and Schmid, Irvine, CA 92714,
and Research Associate, Princeton
University, Princeton, NJ 08544
ABSTRACT
The classic single area unit hydrograph (UH) approach to modeling runoff
response from a free draining catchment is shown to represent several impor-
tant modeling considerations including, (i) subarea runoff response (in a
discretized model), (11) the subarea effective rainfall distribution including
variations in magnitude, timing, and storm pattern shape, (iii) channel flow
routing translation and storage effects, using the linear routing technique,
(iv) subarea runoff hydrograph addition, among other factors. Because the UH
method correlates the effective rainfall distribution to the runoff hydrograph
distribution, the resulting catchment UH should be considered a correlation
distribution in a probabilistic sense. Should the uncertainty in rainfall
over the catchment be a major concern in modeling reliability, then the UH
output in the predictive setting must be considered to be a random variable.
228
-------
INTRODUCTION
The current trend in hydrologic surface runoff model development is
to discretize the catchment (assumed to be free draining) into several small
subareas, each linked by a channel hydraulic flow routing algorithm. The
resulting model is then formulated as a link node model which responds
hydraulically according to a specified effective rainfall in each subarea.
While over 100 such models have been developed in the open literature
(Hromadka, 1987), none have been shown to provide consistently "bptter"
results than the classic single area unit hydrograph (UH) methods in the
estimation of severe storm runoff of interest ir. flood control. It is
shown in this paper that the classic UH technique provides, (i) a rational
modeling structure which properly represents several hydrologic effects
which a highly discretized model misrepresents; (ii) a correlation distri-
' bution (distribution frequency of UH's) which correlates the effective
rainfall to be measured runoff hydrograph; and (iii) a probabilistic model
which represents the model output as a random variable, whose variance
represents the natural variance between effective rainfall and runoff.
CATCHMENT AND DATA DESCRIPTION
Let R be a free draining catchment with negligible detention effects.
R is discretized into m subareas, R., each draining to a nodal point which
is drained by a channel system. The m-subarea link node model resulting
by combining the subarea runoffs for storm i, Q-^t), adding runoff hydro-
graphs at nodal points, and routing through the channel system, is denoted
as Qm1(t). It is assumed that there is only a single rain gage and stream
gage available for data analysis. The rain gage site is monitored for the
'true' effective rainfall distribution, eQ1(t). The stream gage data re-
presents the entire catchment, R, and is denoted by Q ""(t).
y
229
-------
LINEAR EFFECTIVE RAINFALLS FOR SUBAREAS
The effective rainfall distribution (rainfall less losses) in R. is
J
given by e.^t) for storm i where e.^t) is assumed to be linear in e/"(t)
J J y
by
ej1(t) = E Xjk V^"6^ (1)
where xA and el are coefficients and timing offsets, respectively, for
storm i and subarea R.. In Eq. (1), the variations in the effective rain-
J
fall distribution over R due to magnitude and timing are accounted for by
the xA and &A» respectively. The subareas, R., are chosen such that
Eq. (1) is a good approximation for each subarea.
SUBAREA RUNOFF
The storm i subarea runoff from R. is given by Q.^t) where
,t J °
ej1(t~s) *j
ds
(2)
s=0
where 4>. (s) is the subarea unit hydrograph (UH) for storm i such that
Eq. (2) applies. Combining Eqs. (1) and (2) gives
(3)
s=0
Rearranging variables,
(s-e;k)ds
(4)
where throughout this paper, arbitrary function F(s -Z) is notation that
F(s -Z) = 0 for s
-------
LINEAR ROUTING
Let Ij(t) be the inflow hydrograph to a channel flow routing link
(number 1), and 02(t) the outflow hydrograph. A linear routing model of
the unsteady flow routing process is given by
where the a, are coefficients which seem to unity; and the a. are timing
i Ki
offsets. Again, Ij(t -a. ) = 0 for t < a. . Given stream gage data for
*i S
I (t) and 0 (t), the best fit values for the a. and cv can be determined.
Ki Ki
Should the above outflow hydrograph, Oj(t), now be routed through
another link (number 2), then I2(t) = Oj(t) and from the above
°2 2,
L at Z a. I (t -a. -a. )
k2=l K2 kj=l Kl Kl K2
For L links, each with their own respective stream gage routing data,
the above linear routing technique results in the outflow hydrograph for
link number L, 0, (t), being given by
nL Vl n2 na
M*) = I at I at ••• I at I at M*'0^ " ak at * at )
L kL=l KL kL_j=l kL-l k2-l K2 k,-l ki ki KZ KL-1 KL
(7)
Using vector notation, the above 0.(t) is written as
(8)
231
-------
For subarea R., the runoff hydrograph for storm i, Q.^t), flows
J J
through L. links before arriving at the stream gage and contributing to
J
the total measured runoff hydrograph, Q-t). All of the constants
and a1<^> are available on a storm by storm basis. Consequently from the
linearity of the routing technique, the m-subarea link node model is given
by the sum of the m, Q^U) contributions,
J
where each vector . is associated to a R., and all data is defined for
J J
storm i. It is noted that in all cases,
LINK-NODE MODEL, Qm1(t)
For the above linear approximations for storm i, Eqs. (1), (4), and'
(9) can be combined to give the final form for QJU),
m
eJ(t-s) I A.1. «J(s -eA-a1 ) ds (11)
y j" j jK >N^J
Because the measured effective rainfall distribution, eQ1(t), is independent
of the model, Eq. (1) is rewritten in the final form
ft
m . _• •
e (t -s) 7 T a i, y A.. . (s -Q*\ -ot K ) ds (12)
s=0 j
where all parameters are evaluated on a storm by storm basis, i.
232
-------
Equation (12) describes a model which represents the total catchment
runoff response based on variable subarea UH's, ^.""(s); variable effective
J
rainfall distributions on a subarea-by-subarea basis with differences in
magnitude (^-L)» timing (6*L)» and pattern shape (linearity assumption); and
channel flow routing translation and storage effects (parameters a1^ and
MODEL REDUCTION
The m-subarea model of Eq. (12) is directly reduced to the simple
single area UH model (no discretization of R into subareas) given by Qj1
where
eg1(t-s) n1(s) ds (13)
s=0
where ^(s) is the correlation distribution between the data pair
(Qg1(t). eg1(t)>.
From Eq. (13) it is seen that the classic single area UH model
represents a highly complex link node modeling structure. For the case
of having available a single rain gage and stream gage for data correlation
purposes, the derived n^s) represents the several effects used in the
development leading to Eq. (12), integrated according to the sample from
the several parameters' respective probability distributions. Because the
simple Qj'U) model structure actually includes most of the effects which
are important in flood control hydrologic response, it can be used to develop
useful probabilistic distributions of modeling output.
233
-------
STORM CLASSIFICATION SYSTEM
To proceed with the analysis, the full domain of effective rainfall dis-
tributions measured at the rain gage site are categorized into storm classes,
<£>• That is, any two elements of a class <£ > would result in nearly
identical effective rainfall distributions at the rain gage site, and hence
one would "expect" nearly identical resulting runoff hydrographs from the
stream gage. Typically, however, the resulting runoff hydrographs differ
and, therefore, the randomness of the effective rainfall distribution over
R results in variations in the modeling "best-fit" parameters in correlating
the available rainfall-runoff data.
More precisely, any element of a specific storm class <£ > has the
effective rainfall distribution, e °(t). In correlating {Q^(t), e °(t)},
a different n^s) results due to the variations in the measured Q-^t) with
respect to the single e °(t).
y
In the predictive mode, where one is given an assumed (or design)
effective rainfall distribution, e (t), to apply at the rain gage site,
the storm class of which e (t) is an element of is identified, »
and the predictive output for the input e (t) must necessarily be the
distribution
CQ,(t)] = egu(t-s) [n(s)]Dds (14)
D
e<
s=0
where [n(s)]D is the distribution of n^s) distributions associated to
storm class [En].
234
-------
Generally, however, there is insufficient rainfall-runoff data to
derive a sufficiently unique set of storm classes, <£ >, and hence additional
assumptions must be used. For example, one may lower the eligibility
standards for each storm class, <£q>» implicitly assuming that
several distributions tnU)Jq are nearly identical; or one may transfer
[n(s)] distributions from another rainfall-runoff data set, implicitly
assuming that the two- catchment data set correlation distributions are
nearly identical. A common occurrence is the case of predicting the
runoff response from a design storm effective rainfall distribution, e (t),
which is not an element of any observed storm class. In this case, another
storm class distribution of [n(s)J must be used which implicitly assumes
that the two sets of correlation distributions are nearly identical.
Consequently for a severe design storm condition, it would be preferable
to develop correlation distributions using the severe historic storms which
have rainfall-runoff data available for analysis.
EFFECTIVE RAINFALL UNCERTAINTY
The paper by Hromadka, (1987)(1), includes brief statements from
several reports which conclude that the variability in the rainfall (and
hence the effective rainfall) over the catchment is a dominant factor in the
development, calibration, and application, of hydrologic models (e.g.,
Schilling and Fuchs, 1986; among others)(2). Including this premise in hydro-
logic studies would indicate that hydrologic model estimates must be functions
of random variables, and hence the estimates are random variables themselves.
235
-------
From Eq. (12), the correlation distribution for storm event i, ^(s),
includes all the uncertainty in the effective rainfall distribution over R,
as well as the uncertainty in the runoff and flow routing processes. That
is, VU) must be an element of the random variable [n(s)] where
i = y y a1 y x1 1(S e1 a1'
J-i j < *J J j
and Eq. (15) applies to a specific storm. For severe storms of flood
control interest, one would be dealing with only a subset of the set of all
storm classes. In a particular storm class, <£Q>, should it be assumed
that the subarea runoff parameters and channel flow routing uncertainties
are minor in comparison to the uncertainties in the effective rainfall dis-
tribution over R (e.g., Schilling and Fuchs, 1986; among others), then
Eq. (15) may be written as
m
= v T * . VTI..IA. /c_rfl..i_«. ^ (16)
0 1 = 1 <-£;> "•'^ J* J J* VRV .
J j J J
where the overbars are notation for mean values of the parameters for storm
class <£0>. Although use of Eq. (14) in deriving the [n(s)j distributions
results in both the uncertainties in both the effective rainfalls and also
the submodel algorithms being integrated, Eq. (16) is useful in motivating
the use of the distribution concept in design and planning studies for all
hydrologic models, based on just the magnitude of the uncertainties in the
effective rainfall distribution over R. That is, although one may argue
that a particular model is "physically based" and represents the "true"
hydraulic response distributed throughout the catchment, the uncertainty
in rainfall still remains and is not reduced by increasing hydraulic
routing modeling complexity.
236
-------
DISCRETIZATION ERROR
The need for using the Qx(t) model in studies where detention effects
are minor is made more apparent when examining the effects of discretizing
the model into subareas without the benefit of subarea rainfall-runoff
data.
In the above typical case, the engineer generally assigns the recorded
precipitation from the single available rain gage, P^(t), to occur simul-
taneously over each R.. That is from Eq. (1), the 6/1 = 0 anH the X.I are
J J K J K
set to constants X- which reflect only the variations in loss rate nonhomogeneity.
Hence, the 'true' Qmi(t) model of Eq. (12), (and also Eq. (13)), becomes the
estimator Q.JU) where
,t
j y isi ^c** • i j j T
I j * ^R^J j j
s=o
where hats are notation for estimates. These incorrect assumptions result in
'discretization error1. Indeed, an obvious example of discretization error is
the case where a subarea R. actually receives no rainfall, and yet one assumes
J
that P^Ct) occurs over R. in the discretized model. (It is easily shown
that the Eq. (13) model accommodates this example case.)
DISCRETIZATION CALIBRATION ERROR
A current trend among practitioners is to develop an m-subarea link-node
A i
model estimator Qm (t) such as Eq. (17), and then "calibrate" the model
parameters using the available (single) rain gage and stream gage data pair.
Because subarea rainfall-runoff data are unavailable, necessarily it is
assumed that the random variables associated to the subarea effective rain-
falls are given by
237
-------
(estimator, Qm
(18)
assumptions)
jr. o -
But these assumptions violate the previously stated premise that the
uncertainty in the effective rainfall distribtion over R has a major
effect in hydrologic modeling accuracy. The impact in using Eq. (18)
becomes apparent when calibrating the model to only storms of a single
storm class, <£ >.
Again, for all storms in <£>» the effective rainfall distributions
are all nearly identical and are essentially given by the single e °(t).
But due to the variability in rainfall over R., the associated runoff
J
hydrographs, Q 1(t), differ even though e °(t) is the single model input.
It is recalled that in Eq. (17), the effective rainfall distribution
is now the estimator, e_°(t), which is the true e Mt) modified to best
y *?
correlate {Q n(t), e °(t)}. That is, due to the several assumptions leading
^ T
to Eq. (18) for the discretized model estimator, Q (t), the variations due
to [A..J and [6..] are transferred from the [n(s)3 distribution to the e n(t)
JK JK g
function. For storm class <£ >, the estimator Qm1(t) can be written approxi-
mately from Eqs. (16) and (17) as
.t
5J(t) -
s=0
where in Eq. (19), it is assumed that the variations in model output due to
using mean values (overbar notation) are minor in comparison to the
variations in model output due to [X..J and [9.t]. But then Eq. (19) is
J K JN
another single area UH model,
238
m
) ds (19)
-------
Qj(t) =
eJ(t-s) n(s) ds (20)
g
s=0
x%
where n(s) is an estimated distribution which is essentially 'fixed' for
all storms in a specified storm class <£ >. The n(s) is fixed due to
o
nearly the same input being applied to each subarea for each storm in <£ >.
In calibrating Q J'(t), therefore, the work effort is focused towards
A .
finding the best fit effective rainfall distribution, e n(t), which correlates
the several pairs (Q.^t), n(s)}. That is, the 'true' single e °(t) is
A J • A
forced to be modified to be e (t) in order to correlate the {Q (t), r)(s)},
for each storm, i. This contrasts with finding the best fit r^Cs) which
correlates the pairs, {Q '(t), e °(t)}. It is recalled that from Eqs. (16),
y y
A
(17), and (20), n(s) is "fixed" due to the assumptions of Eq. (18), and
due to using a single storm class, <£>•
Because the effective rainfall submodel used in Q.JU) has a prescribed
structure, it cannot match the best fit e ""(t) for all storms and, conse-
quently, modeling error is introduced into the calibration parameters of
the loss rate submodel in order to (l) modify the true e_°(t) due to the
y
effects of IX-k] and [9,-iJ; (2) the derivation of loss rate parameters
which are not "physically based".
Another error which results due to use of Eq. (18) is that the
estimator modeling distribution [Q (t)] for storm class <£ > will be
imprecise due to the variation in derived loss rate parameters not achieving
the true variation in e '(t).
239
-------
The above results indicate that for the given assumptions, the cali-
bration of a highly discretized catchment model will generally lead to a
model that is no more reliable in the predictive mode than the simple single
area UH model. These results appear to be validated by the open literature
(Hromadka, 1987).
EXPECTED VALUE ESTIMATES
In practice, the single area UH model is used to correlate several
record data pairs {Q^(t), e^(t)} of the same or similar storm class <£ >
to derive the associate correlation distributions, {^(s)}. Although
the (r/Cs)} are often integrated and normalized, and the several normalizing
parameters averaged together, the net effect of all this is finding the
expected value of the distribution of correlation distributions, denoted
by E[n(s)]. Then, the model used for predictive purposes (for storms of the
same class, <£Q> used to develop [n(s)j ) is the expected distribution
given by
eg(t) E[n(s)] ds (21)
where e (t) is a design input in <£0>- From Eqs. (12) and (13),
E[Qj(t)] = E[Qm(t)J which is the 'true1 expected distribution for the
given assumptions leading to Eq. (12).
" i
In comparison, after calibrating the estimator, Q (t), to the available
data, the averaging of parameters results in the model (for storms in <£ >)
ECQL(t)] =
'g
s=0
E[e(t)] n(s) ds (22)
240
-------
where E[e (t)] is the "best fit" to the expected value of the true effective
i ~
rainfalls (needed to correlate the {Q (t),n(s)» using a specified rigid
link-node model structure.
Comparing Eqs. (21) and (22), it is seen that for storm class <£Q>,
Eq. (21) is the 'true' expected value.
VERIFICATION TESTS OF MODELS
From Eqs. (21) and (22), the standard use of verification tests on
the models of ECQa(t)] and E[Qm(t)] simply test the distribution of [Q (t)]
/\
about the mean estimates of ECQa(t)] and E[Qm(t)] for storm class <£ >.
The discrepancies reported in the literature for verification tests
indicates that the natural variance between the e 1(t) and Q^U) is
usually quite large.
CERTAINTY IN FLOOD CONTROL DESIGN
Recalling the premise that the variations in the effective rainfall
distribution over the catchment, R, has a major impact on modeling accuracy,
it may be questioned whether using the expected value of a model output is
the proper use of a probabilistic distribution.
For example, suppose that a rain gage station with an extremely long
record length shows that a severe storm condition occurs fairly frequently
(say, about every 100 years), and each occurrence results in a nearly
identical effective rainfall distribution at the rain gage site. Hence,
a storm class of design interest is well defined, <£D>, where each element
has a nearly identical input, e (t), for any catchment hydrologic model.
Yet the catchment stream gage shows a variation in the runoff hydrographs,
QJCt), for each event of e (t). From this infor
y y
bution is derived from Eqs. (12) and (13) to give
, for each event of e (t). From this information, a model distri-
y
241
-------
CQD(t)]
egD(t-s)[nD(s)] ds
(23)
s=0
Equation (23) is the distribution of hydrologic modeling estimates (see
Figure 1), and is the best estimate available. Given another design storm
event, with the same egD(t) resulting, the best a model can do in estimating
the resulting runoff hydrograph is reflected in Eq. (23), and Figure 1.
TIME
(HOURS)
Figure 1. The Hydrologic Model Distribution (Eq. 23)) for a Predicted
Response, [QD(t)], from Input, e[j(t). Heavy line is the
Expected Distribution, E[QQ(t)]
Should the expected model E[QD(t)] be used for design study purposes,
this expected runoff hydrograph typically would not be the most severe de-
sign condition for flood control facilities. Instead, the true distribution
[Q[j(t)] should be used to evaluate the flood control system performance,
and a level of confidence selected as to the success in predictive design.
That is, using the E[QD(t)] model for design purposes often results in a
design product that has only a 50-percent confidence level of protecting
242
-------
for the specified design event, e (t), given the available rainfall-
y
runoff data. Perhaps a higher level of confidence, such as 85-percent or
95-percent, may be more appropriate in the interest of public safety, and to
reduce the exposure to flood damage liability.
USING THE HYDROLOGIC MODEL DISTRIBUTION [Q^tJ]
From the development leading to the model of Eq. (12), use of the
standard single area UH model of Eq. (13) has a powerful representation
of the catchment response including: random variations in the effective
rainfall distribution pattern shape, magnitude, and timing, on an arbitrarily
discretized subarea basis; variations in the subarea runoff response and
channel flow routing effects on a storm by storm basis; storage effects
in channel routing; among others. Calibration of the Qj(t) model to
rainfall-runoff data on a storm class basis results in a distribution
of correlation distributions, [n(s)], which reflects the natural variance
between the record data. The resulting model distribution, [Qa(t)],
reflects the natural variance in predicting runoff quantities for storms
of the same class used to derive [n(s)].
^ i
The link node model estimator, Qm (t), however, cannot achieve the true
distribution of [Q^t)]. Only if rainfall-runoff data were available in each
subarea (in order to determine the xA and 6.^ on a storm by storm basis)
would the model parameters (e.g., the loss rate model parameters be properly
calibrated and'the variance due to the rainfall effects (i.e., [Xjk3 and [0^]
in Eq. (12)) be properly reflected. Consequently, [Qj(t)] should be used.
/\
The distribution [Qm(t)], developed by varying the loss rate parameters (as t
routing parameters are nearly invariant for storms of the same class), cannot
243
-------
achieve the true variance between rainfall-runoff due to the loss rate algorithm
«• ,•
structure. If Qm (t) were supplied subarea rainfall-runoff data, and stream
/s. • • •
gage data to evaluate all routing parameters, then Q (t) = Q (t) = Q^t).
That is, given enough runoff data to evaluate all model parameters on a subarea
and link basis, the link node model will achieve the distribution variance
between model output and the given rainfall data as achieved by the classic
single area UH model.
APPLICATION: DETENTION BASIN VOLUME SIZING
The. above developments are now applied to a simple application. A
catchment of 1,800-acres is studied to size a detention basin. The design
objective is to protect for a historic design storm. Based on the available
stream gage and rain gage data, a class of severe storms, 0>» is developed
and the Q^U) model is calibrated for each element of - The resulting
[n(s)J distribution is shown in mass curve form, [M(s)], where
,t
n1(x) dx (24)
x=0
M1(s) =
A frequency distribution for [M(s)] is shown in Fig. 2.
Using [M(s)], the [n(s)J is found by differentiation and the model
distribution, [QD(t)], is given by Eq. (23) and shown in Fig. 1. Routing
the [QD(t)] through the detention basin resulted in the volume requirement
distribution shown in Fig. 3. Shown in the figure is the expected volume
requirement using E[QD(t)J, and also the 50-percent and 85-percent confidence
estimates. Note that in this case, the "expected" volume requirement derived
by using E[Q(t)J (such as done in usual practice) is slightly less than the
50-percent confidence estimate.
244
-------
PROBABILITY
Ln
ro
rl-
O
3
co
Q>
3
O
ro
ro
m
c:
-s
ro
ro
x -s
-a ro
ro .a
o c
rt- ro
ro 3
to -J.
<•+ w
-S c+
-•• -s
cr -*•
c cr
m -h
i—i o
2 -S
PROBABILITY
ro
o>
ro
ro
-------
CONCLUSIONS
The classic single area unit hydrograph approach to modeling runoff
response from a free draining catchment is shown to represent several
important modeling considerations including, (i) subarea runoff response
(in a discretized model), (ii) the subarea effective rainfall distribution
including variations in magnitude, timing, and storm pattern shape, (iii)
channel flow routing translation and storage effects, using the linear
routing technique, (iv) subarea runoff hydrograph addition, among other
factors. Because the UH method correlates the effective rainfall distri-
bution to the runoff hydrograph distribution, the resulting catchment UH
should be considered a correlation distribution in a probabilistic sense.
Should the uncertainty in rainfall over the catchment be a major concern in
modeling reliability, then the UH output in the predictive setting must be
considered to be a random variable. In this paper, the UH method is shown
to have a rational modeling structure for free-draining catchments. The
correlations represented by the class of UH's derived from similarly
categorized storms, properly reproduces the natural variance between
the effective rainfall and runoff hydrograph. By using the full set of
observed UH's (from the same storm category), a design product can be
developed which accommodates modeling uncertainty due to the uncertainty
in rainfall and other factors. The resulting UH model is then interpreted
to be a probabilistic distribution, in which a flood control design needs to
be tested by probabilistic simulation, varying the UH according to its
frequency distribution. As a case study, a distribution of runoff hydro-
graphs is used to estimate multi-outlet retarding basin design volume
requirements.
246
-------
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. Hromadka II, T.V. Hydrologic Models - Where Are We Now?, submitted to
A.S.C.E. Journal of The Hydraulics Division, 1987.
2. Schilling, W. and Fuchs, L. Errors in Stonnwater Modeling - A Quantita-
tive Assessment, A.S.C.E. Journal of The Hydraulics Division, Vol. 112,
No. 2, Feb. 1986.
247
-------
ATTENDEES
at the
STORMWATER AND WATER QUALITY MODEL USERS GROUP MEETING
DENVER, MARCH 23-24, 1987
NAME
John A. Aldrich
Dennis Anderson
Guy Apicella, P.E.
Raffi Bedrosyan
Ed Bennet
Terry L. Borders
Fred Bromberger
Rob Brown
Kelly A. Cave
Amy Conklin
Brett C. Cunningham
Roy R. Detweiler
Robert E. Dickinson
Donald Distante
Garth Englund
Ron Etzel
John Farrow
David H. Foster
Allan Goyen
Paul R. Grover
James Guo
John M. Hamilton
James S. High
Main Hutcheson
M.H. Jackson
William James
Jeff L. Lamberts, P.E.
Roger McCoy
S. Wayne Miles
REPRESENTING
Camp, Dresser & McKee, Annandale, VA
Colorado Dept. of Health, Denver, CO
Lawler, Matusky & Skelly Engrs., Pearl River, NY
City of North York, P.W.D., Willowdale, ON
City of Sante Fe, Sante Fe, NM
Water Quality Stand. & Mod. Sec., Columbia, SC
City of Littleton, Littleton, CO
U.S.G.S. W.R.D., St. Paul, MN
Camp, Dresser & McKee, Annandale, VA
Denver Regional Council of Govts, Denver, CO
Univ. of Florida, Dept. Env. Eng'g Sci., Gainesville, FL
Chadds Ford Enterprises Inc., Chadds Ford, PA
Univ. of Florida, Dept. Env. Eng'g Sci., Gainesville, FL
Lawler, Matusky & Skelly Engrs., Pearl River, NY
Colorado Dept. of Highways, Denver, CO
Anne Arundel Cty P.W.D., Annapolis, MD
Colorado Dept. of Health, Denver, CO
University of Wyoming, Laramie, MY
Pharmacy House, 44 Thesiger Cr., Curtin, ACT
Micor Engineering Inc., Winnipeg, MB
University of Colorado at Denver, Denver, CO
Muller Engineering Co., Lakewood, CO
ESCOM, Boksburg, Transvaal, South Africa
Oklahoma Water Resources Bd., Oklahoma City, OK
Bovay Northwest Engineers & Archts., Spokane, WA
Univ. of Alabama, Dept. of Civil Eng'g, Tuscaloosa, AL
Bovay Northwest Engineers & Archts., Spokane, WA
Univ. of Utah, Dept. Geography, Salt Lake City, UT
Univ. of Florida, Dept. Env. Eng'g Sci., Gainesville, FL
248
-------
Alan B. Morrice
Ivan Muzik
Jeffrey Needle
Calvin Neidrauer
Jay Nelson
Thomas G. Potter
R.A. Reich
Jorge Rivera
Earl H. Smith
Kevin Stewart
William C. Taggart
Ravindra K. Talwalker
Geoff Thompson
Edwin A. Toms
Steven R. Touchi
Ben Urbonas
El Paso County D.O.T., Colorado Springs, CO
Univ. of Calgary, Dept. Civil Eng'g, Calgary, AB
S. Florida Water Management Dist., West Palm Beach, FL
S. Florida Water Management Dist., West Palm Beach, FL
Hydro-Triad Ltd., Lake wood, CO
Univ. of Florida, Dept. Env. Eng'g Sci., Gainesville, FL
DuPont Company, Engineering Dept., Newark, DE
Boulder, CO
City of Greenwood Village, Greenwood Village, CO
Urban Drainage & Flood Control District, Denver, CO
McLaughlin Water Engineers Ltd., Denver, CO
City of Milwaukee, Bureau Engineers, Milwaukee,WI
Pharmacy House, 44 Thesiger Cr., Curtin, ACT
Hydrologic Consulting Engineers, Boulder, CO
Metcalf & Eddy Inc., Palo Alto, CA
Urban Drainage & Flood Control Dist. 69, Denver, CO
*U.S.GOVERNMENTPRINTINGOFFICE:1987.748.12V 67025
249
------- |