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, » 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

-------