United States
Environmental Protection
Agency
Environmental Research
Laboratory
Athens GA 30613
EPA/600/9-86/023
September 1966
Research and Development
Proceedings of
Stormwater and Water
Quality Model Users
Group Meeting
March 24-25, 1986
Orlando, FL

-------
                                            EPA/600/9-36/023
                                            September  1986
                  PROCEEDINGS
                       OF
       STORMWATER AND WATER QUALITY MODEL
              USERS GROUP MEETING
               March 24-25, 1986
                  Orlando, FL
                   Edited by

            Thomas 0.  Barnwell,  Jr.
       Center for Water Quality  Modeling
       Environmental  Research  Laboratory
               Athens, GA   30613

                 Wayne C. Huber
Department of Environmental  Engineering Sciences
             University of Florida
            Gainesville, FL   32611
        ENVIRONMENTAL RESEARCH  LABORATORY
        OFFICE. OF RESEARCH AND  DEVELOPMENT
       U.S. ENVIRONMENTAL PROTECTION  AGENCY
                ATHENS, GA   30613

-------
                                 DISCLAIMER

     The Information in this document has been funded in part by the United
States Environmental Protection Agency.  Papers describing EPA-sponsored re-
search have been subject to the Agency's peer and administrative review, and
the proceedings have been approved for publication as an EPA document.  Mention
of trade names or commercial products does not constitute endorsement or
recommendation for use by the U.S. Environmental Protection Agency.

-------
                                  FOREWORD

     A major function of  research  and development programs is to effectively
and expeditiously transfer technology developed by those programs to the
user community.  A corollary  function is to provide for the continuing ex-
change of information and ideas  between  researchers and users, and among the
users themselves.  The Stormwater  and Water Quality Model Users Group,
sponsored jointly by the  U.S. Environmental Protection Agency and Environment
Canada/Ontario Ministry of the environment, was established to provide such
a forum.  The group has recently widened its interests to include models
other than t!ie Stormwater Management Model and other aspects of modeling
water quality in  urban and natural waters.  This report, a compendium of
papers presented  at the users group meeting held on March 24-25, 1986 in
Orlando, FL, is published in  the interest of disseminating to a wide audience
the work of group members.


                                      Rosemarie C. Russo, Ph.D.
                                      Director
                                      Environmental Research Laboratory
                                      Athens, Georgia

-------
                                  ABSTRACT

     This proceedings includes 22 papers  on  topics  related to the develop-
ment and application of computer-based mathematical  models for water quantity
and quality management.  The papers were  presented  at the semi-annual meeting
of the Joint U.S.-Canadian Stormwater and Water Quality  Model Users Group
held on March 24-25, 1986, in Orlando, Florida.

     Several papers deal with using stormwater and  water quality models on
microcomputers and interfacing microcomputer software such as spread sheets
and data base managers with these models.  Specific programs discussed
include the Storm Water Management Model, DABRO, HSPF, a simplified water
quality program, HAZPRED, QUAL-TX, and OTTSWMM.

     Other papers discuss statistical  properties of point and nonpoint
pollutant sources, particularly hinhw*« runoff and  the effectiveness of
detention/retention basins for mitigating pollution. Two papers discuss
trophic state models in lakes and
                                     IV

-------
                                  CONTENTS


                                                                     Page
FOREWORD ..............................   iii
ABSTRACT ..............................    iv
ACKNOWLEDGMENT ...........................   vi i

"PORTING MAINFRAME-BASED NUMERICAL MODELS TO MICROCOMPUTERS:   A
  CASE STUDY USING THE EPA STORM WATER  MANAGEMENT  MODEL"  ......     1
  R.M. Baker and K.J. Brazauskas
PORTABILITY, MAINTENANCE AND FORTRAN PROGRAMING STYLE  .......    20
  T.O. Barnwell, Jr. and D.  Disney

CONSIDERATIONS ON USE OF MICROCOMPUTER MODELS FOR STORMWATER
  MANAGEMENT ............................    28
  P. Wisner, D. Consuegra,  H.  Frazer,  and A.  Lam

PC SOFTWARE FOR COMPUTATIONAL  HYDROLOGY  ..............    45
  W. James, M. Robinson, and M.  Stirrup

DABRO:  A BASIC LANGUAGE PROGRAM FOR HYDROGRAPH COMPUTATION   ....    58
  B.L. Golding

APPLICATION OF A LOTUS SPREADSHEET FOR A SWMM PREPROCESSOR  .....    80
  S.W. Miles and J.P. Heaney

IMPACT OF EXTENSIVE IRRIGATION PUMPAGE ON STREAMFLOW  BY  HSPF ....    93
  A.K. Nath

THE USE OF SWMM TO PREDICT RUNOFF FROM NATURAL WATERSHEDS IN
  FLORIDA  .............................   109
  W.C. Downs, J.P. Dobson,  and R.E.  Wiles

A SIMPLIFIED WATER QUALITY COMPUTER  PROGRAM FOR REGIONAL
  STORMWATER MANAGEMENT SITE EVALUATION  ..............   121
  J.M. Crouse and M.H. Helfrich

DEVELOPMENT OF THE HAZPRED MODEL ..................   128
  G. Zukovs, J. Kollar, and M. Shanahan

ALTERNATIVE CALIBRATION OF THE QUAL-TX MODEL FOR THE  UPPER
  TRINITY RIVER ..........................   147
  R. McCarthy

-------
                           CONTENTS (cont'd)

LOGNORMALITY OF POINT AND NON-POINT SOURCE POLLUTANT CONCENTTATIONS.  .   157
  E.D. Driscoll

POLLUTION FROM HIGHWAY RUNOFF—PRELIMINARY RESULTS 	   177
  P.E. Shelley and D.R. Gaboury

EFFECTIVENESS OF DETENTION/RETENTION BASINS FOR REMOVAL OF HEAVY
  METALS IN HIGHWAY RUNOFF 	   193
  H.H. Harper, Y.A. Yousef, and M.P. Wanielista

SIMPLE TROPHIC STATE MODELS AND THEIR USE IN WASTELOAD ALLOCATIONS
  IN FLORIDA	219
  R.W. Ogbunr, P.L. Brezonik, and B.W. Breedlove

MODEL COMPLEXITY FOR TROPHIC STATE SIMULATION IN RESERVOIRS  	   235
  R.A. Ferrara and T.T. Griffin

F.D.O.T. DRAINAGE MANUAL:  WHY "DRAINAGE" IN AN AGE OF "STORMWATER
  MANAGEMENT"	255
  E.G. Ringe

APPLICATION OF THE OTTSWMM MODEL FOR RELIEF SEWER STUDY IN LAVEL,
  QUEBEC	263
  R. Roussel, J.C. Pigeon, and J.R. Noiseux  	

APPLICATION OF INLET CONTROL DEVICES AND DUAL DRAINAGE MODELLING
  FOR NEW SUBDIVISIONS	275
  P. Wisner,  H. Fraser, C. Kochar, and C. Rampersad

USE OF CONTINUOUS SWMM FOR SELECTION OF HISTORICAL RAINFALL DESIGN
  EVENTS IN TALLAHASSEE   	   295
  W.C. Huber, B.A. Cunningham, and K.A. Cavender

AN EXPERT SYSTEM PROTOTYPE FOR RECEIV-II USING TURBO PASCAL  	   322
  R.E. Dickinsonson, I.B. Chou, and F.V. Ramsey

MODELING FLOOD HYDROLOGY USING HYMO   	   326
  J.E. Scholl

LIST OF ATTENDEES	332

-------
                               ACKNOWLEDGMENT

     The Stormwater and Water Quality  Model  Users  Group  relies on  local
hosts to make arrangements for its meetings.   The  hosts  for  the meeting
reported in this proceedings  were  Dr.  Larry  A.  Roesner of  Camp, Dresser,
McKee, Inc. and Dr. Wayne C.  Huber of  the  University  of  Florida.   Dr.  Huber
reviewed abstracts  of papers  and arranged  the  meeting agenda.  Dr. Roesner
made local  arrangements for meeting rooms  and hotel accommodations.

-------
  "Porting Mainframe-Based Numerical Models to Microcomputers:



    A Case Study Using the EPA Storm Water Management Model"







             Richard M. Baker and Karl J.  Brazauskas



         Metcalf & Eddy,  Inc., Wakefield,  Massachusetts







                            ABSTRACT







        This  paper  describes  porting  of  the  United



        States  Environmental  Protection  Agency  Storm



        Water Management Model (SWMM Version 3)  from  an



        IBM   mainframe   CMS   environment   to  a  DOS



        environment  on  an  IBM  PC  AT  microcomputer.



        Subsequent  application  of the micro SWMM model



        during a study of a combined sewer system  in  a



        major New England city is then briefly reviewed.







 INTRODUCTION



       As a result of the rapid evolution of micro hardware  and



recent  advances  in  the  development of micro-resident Fortran



compilers,  porting and application of traditionally  mainframe-



based  numerical  models  on  micros  has become a practical and



extremely cost effective method of computing.  The advantages of




                                1

-------
computing on micros are numerous  and  not  restricted  to  cost
considerations alone.  They include:  easier, more user friendly
access,  possible reduced turn-around times and  integration  of
pre-processing,   numerical  modeling,   post-processing,   word
processing,   graphics  and  software  development  capabilities
within one, relatively compact machine.
       Prior  to porting of the SWMM model to micros,  engineers
at Metcalf & Eddy executed it on a remote  IBM  3033  commercial
time-share  computer.  Input  data sets were developed on an in-
house Digital Equipment PDF 11/70 and transmitted to the  remote
IBM  mainframe  via telephone line.  Following execution of SWMM
on  the  mainframe,  model  results  were  stored  on  disk  and
subsequently  transmitted back to the in-house POP 11/70,  where
printing was accomplished using a line printer.
       Use  of this remote job submittal system was satisfactory
as long as only one or two users were active at  once.  However,
during   periods  of  heavy  use,   turnaround  times  increased
drastically from the usual 30 minutes to as much as three hours.
In addition, if either the in-house POP 11/70,  remote mainframe
or telephone communications were down, computing would come to a
halt.
       It was recognized that several options were available for
decreasing  turn-around  times on the above remote job submittal
system.  However,  these options  would  not  address  cost  and
dependability  issues.  Due  to  several  previous  successes in
porting relatively small numerical models to micros at Metcalf &

-------
Eddy, porting of the SWMM model to in-house micros was initiated
in an attempt to address the issues of cost,  dependability  and
turn-around.
THE PORTING PROCESS
       The  process  of  porting numerical models from mainframe
(or mini) computers to micros can be divided into several  major
tasks, including:
     Obtaining the mainframe source code,
     Transferring source code to the target micro,
     Selecting a micro-resident compiler,
     Modifying and compiling the source code,
     Selecting a micro-resident linkage editor,
     Developing an overlay structure and linking object modules,
     Testing and documenting the ported numerical model,  and
     Applying the model on projects.
       The  above  tasks are more or less the same regardless of
the relative size or  complexity  of  the  mainframe  model.  In
addition,  the  porting process is independent of the high level
language in which  the  model  source  code  is  written,  e.g.,
Fortran,  Pascal  or c.  However,  selection of a micro-resident
compiler and  linkage  editor  and  development  of  an  overlay
structure  are best described using a large,  relatively complex
numerical model like SWMM.  As a result,  the following detailed
description  of  the  porting process for SWMM should serve as a
useful guide to  others  interested  in  utilizing  the  rapidly
expanding  capabilities  of  micros.

-------
OBTAINING MAINFRAME SOURCE CODE



       Mainframe  source code for the latest version of the SWMM



model was obtained on  9-track  magnetic  tape  from  the  USEPA



Environmental Research Laboratory in Athens, Georgia.  This tape



included   the   main  SWMM  program  and  all  subroutines  and



functions.  It also contained test input data sets  and  results



for use in verifying model operation.  A listing of all files on



the  SWMM  tape  was  produced on an in-house line printer.  The



SWMM source code was examined and found to be complete.



       Each subroutine contains its own COMMON BLOCK statements.



However,  it is important to note that many  models  incorporate



COMMON BLOCK   specifications  and  other  source code segments,



e.g.,  PARAMETER statements,  into the  main  body  of  code  at



compile time through the use of INCLUDE statements.  Thus, it is



necessary  to obtain source code for all of these included files



too.



       Model documentation  and  users  manuals  for  SWMM  were



obtained  from Dr.  Wayne C.  Huber at the University of Florida



in Gainesville.  From an examination of these  manuals,  it  was



found  that  the  mainframe  SWMM model utilizes overlays at run



time to reduce the  size  of  the  region  in  which  the  model



executes  to  approximately  400 kilobytes  (kb) of random access



memory  (RAM).  As will be shown later,  use of the above overlay



structure  for  the  micro  SWMM  model  eliminated  the need to



develop one from scratch.



       A description of the job control language (JCL) used with

-------
SWMM on  the IBM mainframe OS/VMS (Release 3.8) operating system



was  also  included  in the documentation.  This information was



useful in identifying input  and  output  (I/O),  e.g.,  scratch



files, used during SWMM model execution.



       Documentation for the  Fortran  compiler  used  with  the



mainframe  SWMM  model  was also obtained.  This information was



used later  in  the  porting  process  during  selection  of  an



appropriate   micro-resident  compiler.   In  general,   if  the



mainframe Fortran compiler used with the model to be ported uses



features not included in the Standard Fortran-77 Language  (ANSI



X3.9-1978),  then the mainframe compiler documentation should be



obtained as an aid in porting to micros.



TRANSFERRING SOURCE CODE TO THE TARGET MICRO



       The 9-track tape containing the SWMM source code and test



data was mounted on an in-house PDF 11/70 tape drive and  copied



to  a  user  area on a peripheral disk.  The numerous SWMM files



were then merged using the PDP  11/70  operating  system  APPEND



utility.



       Transfer of the master SWMM file from the  PDP  11/70  to



the  target  in-house IBM PC AT hard disk was accomplished using



Hayes  Smartcom  communications  software  and   a   micro-based



internal  smart  modeia.  Due to the use of relatively slow (1200



baud-rate) modems at  both  the  PDP  11/70  and  target  micro,



transfer of the master SWMM file required almost four hours.



       Transfer of mainframe source code to target micros may be



accomplished using  several  other  methods,  including:  direct

-------
modem-based communication between mainframes,  minis and micros,



magnetic tape and a nJcro-based tape subsystem or  micros  hard-



wired to minis or mainframes.



       Service  bureaus  may  also  be  used  to  copy data from



magnetic tape to target micro-compatable floppy diskettes.  As a



last resort,  hard copy of the  mainframe  source  code  can  be



entered  manually  into the target micro using a micro- resident



text editor.  When micro-based tape  subsystems  become  cheaper



and more versatile, their use for transferring mainframe data to



micros  and  for  backing  up  micro-based hard disks and floppy



diskettes will likely become commonplace.  It  is  important  to



realize  that  whenever  data is transferred between mainframes,



minis and micros using magnetic tape,  the compatability of tape



read/write utilities on these systems must be addressed.



SELECTION OF A MICRO-RESIDENT COMPILER



       Documentation for the  mainframe  Fortran  compiler  used



with  SWMM  was  examined.  As  expected,  it  was found to be a



superset of the ANSI Fortran-77  standard.  Information  on  the



various  language  extensions  available  with this compiler was



used  later  during  evaluation  of  alternative  micro-resident



compilers.



       The  Microsoft  Fortran  compiler  (Version 3.2) had been



previosly applied successfully at Metcalf & Eddy during  porting



of   several  small  numerical  models  from  Digital  Equipment



minicomputers to Intel 8086 processor based 16-bit micros.  As a



result,  compilation of SWMM was initially attempted using  this

-------
compiler. Unfortunately, Microsoft Fortran was unable to compile



several of the larger SWMM subroutines,  e.g,   RHYDRO and TRANS,



due to its 64 kb  limit  on  program  size.  In  addition,  this



compiler's  memory model forced all local data,  i.e,  variables



used locally within subroutines,  in the entire  SWMM  model  to



reside  within one 64 kb local group (DGROUP)  in micro RAM.  The



64 kb limitation on program size was overcome  by  splitting  up



the  several  very  large SWMM subroutines.  However,  the above



DGROUP restriction resulted in DOS "stack overflow"  and " heap"



errors at model run time.



       Due  to  the  above  limitations of the Microsoft Fortran



compiler,  a search was initiated  for  another  compiler.  This



search  resulted  in  selection  of  the  Ryan-McFarland Fortran



(RM/FORTRAN 2.0) compiler for use in porting  SWMM.  RM/FORTRAN,



which  is  also  marketed  by  IBM  as Professional Fortran,  is



available for either DOS or XENIX operating systems.  It uses  a



memory  model  which  overcomes the 64 kb DGROUP restrictions of



Microsoft  Fortran  and  allows  program  sizes  up  to  sixteen



megabytes.  RM/FORTRAN  is  certified  by  the  General Services



Administration  (GSA) as the only micro-resident Fortran compiler



that meets the full  ANSI  X3.9-1978  Fortran  Standard  without



discrepancies.   RM/FORTRAN   includes   many   frequently  used



mainframe Fortran language extensions,  including those used  in



Digital  Equipment  VAX/Fortran,  IBM  mainframe  VS/Fortran and



Fortran-66.  RM/FORTRAN applications run 30% to 40% faster  than



those  compiled with Microsoft Fortran Version 3.2.  A numerical

-------
coprocessor chip (Intel 8087 for IBM PC or XT  and  Intel  80287



for IBM PC AT) is required when using RM/FORTRAN or when running



RM/FORTRAN compiled numerical models.



MODIFYING AND COMPILING SOURCE CODE



       The  master  SWMM  file  on micro hard disk was separated



into its component files,  i.e.,  data files,  main program  and



subroutines,  using  a  BASIC  language program developed on the



target micro.  This program was also  used  to  convert  several



mainframe  language  extensions  used  in  SWMM,  which  are not



available in RM/FORTRAN,  and to separate out all  COMMON  BLOCK



statements and related PARAMETER statements into INCLUDE  files.



       Use  of   include files often results in considerable time



savings during model testing, debugging and modification because



COMMON BLOCK array dimensions and variable lists modified in one



INCLUDE file result in global changes  to  the  model.  However,



when  splitting  out  COMMON BLOCK statements into INCLUDE files



one must be cognizant of the fact that some  programmers  rename



variables  within  subroutine COMMON BLOCK statements instead of



formally eguivalencing these  variables.  Several  instances  of



this  practice were found in SWMM.  These locally renamed COMMON



BLOCK variables were respecified using  appropriate  EQUIVALENCE



statements.



       An  examination  of  documentation for the mainframe SWMM



model indicated that  control  of  external  data  files,  e.g.,



interface and input data files, is maintained through the use of



IBM Job Control Language (JCL).  In contrast, file control using
                                8

-------
the DOS implementation of RM/FORTRAN is accomplished by Standard



Fortran  OPEN and CLOSE statements within the model source code.



Accordingly,  the SWMM  source  code  was  modified  to  include



statements  necessary  for  creation and control of all external



data files.  Specifications for these files, including:  logical



unit  numbers,  access modes (sequential versus direct),  format



(formatted  versus  unformatted)  and   status   (named   versus



scratch),  were  determined from a detailed examination of their



corresponding  source  code  READ  and  WRITE  statements.    The



example  JCL  specifications  included  in  the  mainframe  SWMM



documentation and user manuals were also useful for establishing



the characteristics of these files.



       All SWMM  subroutines  were  then  successfully  compiled



using RM/Fortran.  This compilation, which was automated through



the use of a DOS batch file, required approximately two hours to



complete.  The  total  size  of the resultant object modules was



over a megabyte.



SELECTING A MICRO-RESIDENT LINKAGE EDITOR



       Selection of an 80286 processor-based linkage  editor  to



be used on the target IBM PC AT involved several considerations.



The requirement of compiler-linker compatability did not present



a  problem,  since  RM/FORTRAN may be used with numerous linkage



editors.  However,  during the initial attempt  at  linking  the



Microsoft  compiled SWMM model,  it was found that the Microsoft



Linker supplied with DOS had two serious deficiencies.



       First,  it was found that the DOS linker can only link up

-------
to  900  kb of object modules into an executable program.  Thus,



if the DOS linker were  to  be  used,  SWMM  would  have  to  be



separated  into  several  blocks and each block would have to be



executed as a separate model.  This would  likely  prove  to  be



very inconvenient during model application.



       Second,  it was found that the DOS linker can not be used



to  build  multiply-nested   overlay   structures.   As   stated



previously,  four  overlay  nesting  levels  were  required  for



execution of the mainframe SWMM model within a 400 kb region  of



RAM.  Since  DOS can only address 640 kb of micro RAM,  the need



for multiply-nested overlays with the micro SWMM model was  seen



as a distinct possibility.  Accordingly,  a search was initiated



for a linkage editor which would overcome  the  overlay  nesting



and object module size restrictions of the DOS linker.



       Based  on the linkage editor needs identified above,  the



Phoenix Software Associates PLINK86  (Version  1.48)  linker  was



selected  for  porting  of the mainframe SWMM model.  PLINK86 is



compatable with RM/FORTRAN,  supports up to 32  overlay  nesting



levels and does not have any practical restrictions on the total



size of objects modules it can link.



       The  mainframe SWMM overlay structure was utilized as-is,



following its  conversion  to  PLINK86  format,  to  create  the



numerous  overlays required for execution of SWMM within the 640



kb DOS addressable RAM partition.  A memory diagram of a portion



of the  SWMM  overlay  structure  is  shown  in  Figure  1.  The



vertical  scale  represents  the  relative  RAM memory addresses
                               10

-------
                     RUNOFF
       TRANSPORT
             ERROR
            QHYDRO
                  CTRAIN
               RHYDRO
                  BUILD
                                     High
                                Memory
                                Address
       Time
                                       During Execution
                        HCURVE
                  WSHED
                  HYDRO
  Low

          FILTH   INITIAL
       INFIL  DWLOAD
   TSTRDT
SLOP
                              PRINTR
                  RUNOFF. RUBLOCK
                                    LOP r
PRINTQ
                                                      PRINTF
                                            TINTRP
                                            TRANS
                                SCALE
                                GRAPH
                                 MAIN
                FIGURE 1. MEMORY DIAGRAM SWMM V3 III
occupied by SWMM subroutine  object code and   data  during   model
executioii.   In  contrast,  the horizontal  scale indicates memory
addresses  (and subroutines)  which are overlayed at various  times
during a model run.  It  is seen that at run   time  many  of  the
memory  addresses  occupied   by code and data of the SWMM Runoff
Block are  replaced,  i.e.,   overlayed,  by code and data of  the
SWMM  Transport  Block.   This  overlaying  occurs at the instant
                                 11

-------
when the SWMM Executive Block  (subroutine  MAIN)  encounters  a



RETURN  from  the  Runoff  Block   (subroutine  RUNOFF)  and then



initiates a CALL to  the  Transport  Block  (subroutine  TRANS).



Also,  while  subroutine QHYDRO of the Runoff Block is executing



in RAM,  code and data for  all  of  its  ancestors,  including:



RHYDRO,  HYDRO,  RUNOFF,  RUBLOCK,  MAIN  and all the RM/FORTRAN



library routines used by SWMM must also be in RAM.



       Linking of the SWMM object  modules  into  a  relocatable



execute  module  required approximately five  minutes.  The SWMM



execute module required 1.4  megabytes  of  hard  disk  storage.



However,  use  of  overlays  resulted  in  a  maximum 507 kb RAM



requirement at run time.



TESTING OF THE MICRO SWMM MODEL



       The micro SWMM  model  was  then  tested  using  the  EPA



supplied  mainframe  test  cases.   Results  were  found  to  be



identical to those produced on  the  mainframe.  Test  case  run



times  varied between approximately two minutes for execution of



the Runoff Block to almost forty minutes for  execution  of  the



Extran  Block.  Results  were printed at the rate of eight pages




per minute on a Hewlett-Packard Laserjet+  Printer  attached  to



the target IBM PC AT.



       Mainframe  models which utilize highly  iterative solution



techniques may exhibit significant computational error  or  even



numerical  instability  when  ported  to micros.  The reason for



this is that round-off errors can  accumulate to a higher  degree



when  using  16-bit   (typical micros) processors than when using
                              12

-------
full 32-bit processors  (mainframes).   This  problem  should  be



assessed  during  model  testing.   If  deemed necessary,  double



precision may be specified for certain  key  variables  used  in



iterative procedures of the micro numerical model.   Test results



for the micro SWMM model suggest that increased precision is not



necessary.



       When  an  overlay structure is developed from scratch for



use with a ported micro model,  it may be desirable to  optimize



that  structure  in  an  effort  to speed up model  execution.   A



general  guideline  to  be  followed  during   development   and



optimization  of  overlay structures is to set up overlays which



contain isolated functional groups of code that will execute  to



completion and then will not be returned to for a long time,  if



ever.  Also,  the simplest overlay  structure  required  to  run



within  addressable  RAM  is  the best.  Stratcom Systems,  Inc.



presently markets a  program  which  creates  optimized  overlay



structures   based   on   subroutine  sizes  and CALL  sequence



information produced by the RM/FORTRAN  compiler.  Use  of  this



porting tool is currently being investigated at Metcalf & Eddy.





        Following the  successful testing of the micro  SWMM model,



 the  SWMM  documentation   was modified  to  provide   users with



 instructions   on  its  hardware/software  requirements  and  its



 execution  on   an  IBM PC XT  or IBM PC AT.   Since  the micro SWMM



 model's   functionality   is    identical   to   the    mainframe



 implementation,    no    additional   modifications   to  the  EPA



 documentation  and user manuals were  required.






                               13

-------
APPLICATION OF THE MICRO SWMM MODEL



      The  micro SWMM  model  was then used to simulate storm-



induced  flooding within a combined sewer system in a  major New



England  city.  Figure  2 shows the general configuration of the



modeled  sewer system,  which was divided into  12 sub-catchment



areas feeding into a sewer pipe system discretized into 12 sub-



systems  with  a total  of approximately  1500  computational



elements.  A schematic of the main sewer system SWMM model given



in  Figure 3 shows the hydraulic connections between the various



sewer sub-systems .



      Single event,  24-hour simulations were  made  using the



SWMM  Runoff  and Transport Blocks and a modified version of the



SWMM  Combine  Block  capable  of  accepting  up to  16   input



hydrograph  files.  A  ten  minute  time  step  was  used in all
                   	^r:~?\
              x-rqT- v:r/'  M-7 r-N^   _	
             /,----.•!••-. M-4-,-! v-----iJ--'-V;v-, —7"   _..... >v

           /V^'c4.v*^^T-K.->.r:.\

           /^g^-y.^^^-^
                                       ^;]  M'12


                                 MAIN WWTP   "S
                  FIGURE 2. MAIN SEWER SYSTEM



                             14

-------





M-2
1
I
*^^













-1
I
L






M-
1
M-:
1
J



^^^





4

3











— ^




f-

l e
1-O



M-
1
M




^^^m






•' -i
!
- !
1
1
1
1
I
1
f




H
n^
|w\

r








I_Q
O

MN
NJP

	











1

M-9
1 1
1 1
1 1
1
1 1
t t
M-11
{
M_4 fl
1U
I



^^^1*
••» «





M-12







!• In Syst
— ^ Normal
   FIGURE 3. SWMM SUBAREA FLOW SCHEMATIC FOR THE MAIN SEWER SYSTEM
simulations and only flow quantity  was  modeled.   Each  of  the



twelve  sub-systems  of  the  total  sewer  system was simulated



during individual SWMM runs.  Batch files  were  constructed  to



control  sub-system model execution and the cascading of outflow



and in-system overflow hydrographs throughout the sewer network.






       The   SWMM  sewer  system  model  was  then  successfully



calibrated and verified using two independent rainfall  and   in-



system  flow  data sets and synoptic field observations of sewer



system depths of flow and surcharging.  The verified  model   was



subsequently   used   to  simulate  in-system  flooding,   i.e.,





                                15

-------
surcharging and street  flooding,  during  10-year  and  50-year
design  storm  events.  Based  on  the locations and severity of
surcharging and flooding during the 10-year storm event, several
sewer relief projects were developed.  These projects were  then
designed and tested using the micro SWMM model.
       Both  the  SWMM  model calibration and verification input
data sets were also run on the remote IBM  mainframe  commercial
time  share  system as a cross check of the micro SWMM model and
to  provide  comparative  run   cost   and   speed   benchmarks.
Statistics  for  the  verification  runs  on  both the micro and
mainframe are presented in Table 1.  Model run time on the micro
was much longer than that experienced on the mainframe computer.
However,  run cost on the micro was only 12% of that incurred on
the mainframe.

                TABLE 1: Benchmark Run Results
            IBM PC AT                     IBM 3033
    RUN TIME     RUN COST          RUN TIME     RUN COST
    2.9 Mrs.       $20.            1.1 Min.       $175.


       The longer run times experienced  when  using  the  micro
SWMM  model were not found to represent the critical path in the
engineering analysis of the above  sewer  system.  In  addition,
use of the in-house micro for SWMM modeling resulted in almost a
90% decrease in computing costs incurred during that project.
                               16

-------
FUTURE DEVELOPMENTS



       The  porting  and  application of traditionally mainframe



based numerical models on  micros  has  some  limitations.  Very



often  model array sizes must be decreased in an effort to "fit"



the model into addressable micro RAM.  For example,   the maximum



number  of  finite  element grid nodes allowed in a  two or three



dimensional hydrodynamic or water quality model may  have  to  be



decreased from 900 to 400.  This may result in a micro model with



rather  limited  applicability  to  the  simulation   of  complex



systems.



       Another limitation of micro models  is  their  relatively



slow speed of execution. In most applications, e.g., simulations



using  any SWMM block except Extran,  the slower execution times



on micros do not represent the critical project  path.  However,



in  certain  applications,   use  of  micros may be prohibitively



slow.  Long-term, continuous-mode simulation of a large combined



sewer system using the SWMM Extran Block is a good example.



       Fortunately, there are several methods for increasing the



potential model  sizes  and  execution  speeds  on  micros.  The



following  discussion  briefly  reviews  two of these methods as



they relate to the micro SWMM model.



       SWMM  makes extensive use of external scratch data files,



both   within   individual   computational   blocks   and    for



communication  between  blocks.  These scratch files,  which may



require over a megabyte of storage, are written to and read from



high access speed  drum  disks  on  mainframe  computers.  These



peripheral disks are also usually controlled by fast,  dedicated



                                17

-------
processors.  In contrast, these I/O activities are controlled by



non-dedicated processors using relatively slow access floppy and



Winchester hard disks on micros.  As a result,  use of hard disk



scratch files by the micro SWMM model is very inefficient and no



doubt contributed to the relatively long run  times  experienced



on the micro.



       Data  transfer  within micro RAM is much faster than data



I/O using hard disks or floppies for storage.  To take advantage



of this fact, several virtual (RAM) disks could be configured on



the target IBM PC AT using  the  DOS  device  driver  VDISK.SYS.



Resident  RAM  on  the  micro  could  then be increased from the



current 640 kb to over three megabytes through the use of  plug-



in  RAM  chips  mounted  on  its  AST  advantage  board  and  an



insertable piggy-back RAM expansion board.



       Following  the  above reconfiguration of DOS and addition



of RAM,  file OPEN and CLOSE statements within the  SWMM  source



code would have to be modified to include the new DOS configured



virtual  disk  drive  specifications for all scratch files.  The



SWMM executable module would then have to be rebuilt.





       The above method could be used to speed up SWMM  runs  by



eliminating  much uneccesary,  relatively slow I/O activity.  It



could potentially be used to increase the maximum size of  sewer



systems modelable on micros using SWMM by allowing the temporary



storage  of  certain  groups  of variable arrays in virtual disk



resident external data files located outside of the 640  kb  DOS



addressable RAM partition.
                               18

-------
       The  same   result  could  be  more  efficiently   achieved

through  the use  of a Unix or Unix-like operating  system instead

of DOS.  Xenix  operating systems are currently available for IBM

PC's and many other 16-bit (and 32-bit) micros.  Unix and  Xenix

systems  do not have the RAM addressability restrictions of DOS.

They are capable  of  addressing  the  entire  amount  of  micro-

resident  RAM.  Thus,   all  the SWMM scratch data  files  could be

stored in RAM and the size of variable arrays,  and hence  model

sizes,  could   be  increased.   It  is important to note  that the

micro-resident  Fortran compiler used  during  model  compilation

would  have to  be compatable with the target micro-resident Unix

or Xenix operating  system.   Fortunately,   a  Xenix  version  of

RM/FORTRAN  has  recently  been  released.  Current  efforts  at

Metcalf & Eddy  involve the migration of the micro SWMM model and

numerous other  ported mainframe  models  to  16-bit  and  32-bit

micro-based Unix  and Xenix operating systems.
 The work described in  this paper was not funded  by the U.S. Environmental
 Protection Agency and  therefore does not necessarily reflect the views of
 the Agency and  no official endorsement should be inferred.
                                19

-------
            Portability*  Maintenance  and FORTRAN Programing Style


                          Thomas 0. Barnwell, ir.
                              David  W. Disney2
Introduction

     At the Center for Water Quality Modeling,  we have evolved a  set of pro-
graming conventions that make  our FORTRAN  programs easily maintainable and
transportable.   These  conventions  have  helped us  1n transporting many of our
supported programs  from  minicomputers and  mainframes to the  microcomputer
environment.  This paper  describes these  programing conventions and develop-
ment tools and discusses how they  have aided 1n  moving  our  programs to micro-
computers while still  providing a  workable environment for the maintenance of
code on different  hardware  and  software systems.

Program Design Considerations

     It 1s  appropriate to  discuss two  aspects  of programing  style — design
considerations and coding  conventions.  Design  relates to  the  overall struc-
ture of the program while  coding  conventions refer to the the actual FORTRAN
Implementation.  Much has been written about structured  programing and soft-
ware design and we  have found  this design concept to be useful.  The follow-
ing paragraphs outline our  approach to  structured programing 1n FORTRAN.

     Possibly the  most Important design consideration 1s modularity of struc-
ture.   Modularity  1s fostered by writing small  subroutines. To borrow a rule
from the writing profession,  be brief and concise.  Treat a subroutine as you
would  a paragraph and confine  it to  a  single thought  —  a subroutine should
only do one thing.  A good rule Is to restrict a subroutine to two pages or
less;  some  prefer to limit a subroutine to  a single page.   But a  single page
may be too restrictive 1f your style uses a lot of open space to Increase
understanding.  Always keep  1n mind that  both you and others will have to
read (and maintain)  what  you  write.

1.  Center  for  Water  Ooiallty Modeling
2.  Computer Sciences Corp.
   Environmental Research Laboratory
   U S EPA, Athens,  Ga         30613
                                     20

-------
     COMMON Blocks are a major part of our design.  Although FORTRAN does not
have code structures like many other languages* code structures such as a
common work or data area can be emulated  with  COMMON Blocks.  We also prefer
that data be passed between subroutines using COMMON  Blocks rather than Argu-
ments* particularly at  higher levels of the program.  Although there are some
arguable advantages to  using arguments for this  purpose  (after  all*  that 1s
their sole ralson d'etre),  they can be troublesome.   As  discussed  later,  we
have found them to be a frequent source  of problems.

     Of course, modular structure 1s much  more  than writing small subroutines
and using COMMON  Blocks.  A clear specification of  objectives and plan for
achieving them 1s at the  heart of  structured programing.  But these princi-
ples are a good beginning.

    Although there may  be  good  reasons for not  considering  1t, portability 1s
another Important  design consideration.  Our  programs are written  to be used
by others,  mostly  on  their own computers and  the  programs have a surprisingly
long lifetime.  Although we cannot claim credit for the original  design,
QUAL2E (Brown  and Barnwell, 1985)  1s based on  a structure set out over 15
years ago.  We use only 1977 ANSI Standard FORTRAN and avoid extensions to
the standard.

     If writing single-purpose  code for a specific application,  the program-
mer Is justified  1n using the FORTRAN extensions available on his computer.
In general,  these extensions are provided to enhance the efficiency of both
the programmer and the program.  But 1f there  1s the slightest  Inkling that
the program will  be used on  different hardware, the portability Issue must be
considered.

     To Insure portability,  1t 1s useful  to  Identify target compilers that
represent a  reasonable  spectrum of computers.   For example, U S EPA staff can
count on the availability  of  at least three kinds of  computer — IBM PC
compatible,  DEC VAX technology,  and IBM 30xx-ser1es  mainframes.  We  run our
code through compilers  on  each of these systems.

     Where hardware-specific code must be used,  the  programmer  must Isolate
and clearly  Identify this dependent  code.  For example, DATE and  TIME subrou-
tine calls are  a  common extension that are not part  of the ANSI-77 standard
but are available on many systems.  Although not hardware-specific,  file
opening and closing  conventions may vary  among  Installations,  even  for the
same type of hardware,  and it 1s good practice  to Isolate these statements 1n
a separate,  clearly Identified  subroutine.

User Interface

     The user  Interface 1s  becoming more  and more Important to  the computer
programmer.   The "friendliness" of programs 1n  the microcomputer environment
such as LOTUS 123* have raised user expectations  so high  that the traditional
batch-oriented  card file  Input 1s fast becoming  unacceptable.   In  fact,  the
Center for Water Quality  Modeling has begun a major effort to  develop a
consistent Interactive  user  Interface for  all our programs.
                                     21

-------
     In designing a user Interface, the programmer should strive to use com-
plete English  sentences  for such things as error messages,  status
diagnostics,  or  run-time  prompts.  Although  Inside  jargon  and  computerease
may be expedient 1n the short  run, the novice user will often be confused  by
such cryptic statements as "PROGRAM STOP  - INPUT ERROR."  Even an experienced
programmer may  wonder  1f this message was generated by the operating  system
or the program  Itself.

     An Important function  of the user Interface  1s verification of all
Inputs.   Although the user Interface  may  not  be  able to catch subtle  errors
1n program  Input,  several standard checks should be made.   Perhaps the most
obvious 1s checking of character versus numeric  Input.  How  many of us have
not spent considerable time tracking  down an  error to  find that Its cause  1s
a letter "0"  where  the  number "0" should be» or vice-versa?

     Another good practice 1s to explicitly  check for real  and  Integer num-
bers and to check that numbers are 1n a  valid range.  Also, default values
should be  stated or displayed explicitly.   Cases  exist where widely used
computer programs  with perhaps Inappropriate default  parameter  values have
Inadvertently established  "standard practice."

     It 1s useful to place Input verification  criteria  and defaults  1n  tables
or external files.  This makes the criteria easy to locate and change.  The
use of an external  file 1s particularly  convenient as changes can be made
without recompiling the program.

     An echo of Inputs  should be provided as an option so that the user can
document Input when desired but switch 1t  off  when It  1s not needed. As most
programs have the  option of  running  either Interactively or batch,  the pro-
grammer should provide program traps  to stop execution when Input, errors are
uncovered rather than let  execution continue.

     And when writing tabular  output  from a program, concise but descriptive
headings should be provided and output formatted for easy reading. Nothing
1s more frustrating than 10 pages of densely packed numbers with no  Identifi-
cation.   Include your documentation on the output, not  In the users  manual.

Documentation

     Documentation 1s critical to a successful computer program.  We think of
documentation targeted  both for the user and  for  the programmer.  User docu-
mentation should  Include a  description of the  application or use of the
computer program.  The exact  equations and solution  techniques used as well
as guidance  for the user on ranges  and  typical  values of Input variables
should be provided.

     The actual  Installation  of the  program  on the users' computer system,
however,  1s often done by  a programmer who 1s Interested 1n quite  different
things.   The programmer 1s Interested In  what we call  "Implementation docu-
mentation" that  contains detailed Instructions on how to Install  and maintain
the program on a user's system.
                                     22

-------
     Both the programmer  and  the user are Interested  1n sample Input and
output to verify that the Installation 1s correct and the program 1s operat-
ing correctly on their system.  The  system  or Implementation documentation
also should  Include a functional organization of  source code files by name as
well as descriptions of program variables (particularly  global variables) In
case the user and his programmer have to troubleshoot the software.  Often*
the user may  want to read the program 1n order to better understand what 1t
1s doing.  This should be made as easy as possible by making available copies
of the source code and Identifying development  and maintenance software.

Coding Conventions

     Comments are an essential  part of the source code and are often the only
source code documentation.   Although pseudocode 1s useful when  designing  a
large program*  we have  found  1t difficult and costly to keep current.   Each
subroutine or program segment should have Introductory comments that clearly
state  Its purpose;  describe Inputs,  outputs and modified variables;  and
Identify  error conditions.  Comments  are  useful to physically  separate
sections of  code that do not justify isolation  in a subroutine and to explain
loops and conditional statements.  These are often the hardest parts of a
FORTRAN program to decipher;  particularly 1f the dreaded GO TO  is used  with
indiscretion.   (We all know the rule  on  60 TO  ~ DON'TIII)

     Complete English  statements should be used; many programmers  use a
shorthand that can  be completely indecipherable.   Others will have to  read
what you write  and base the maintenance and  understanding of the source code
on your  commentary.  When trying  to decipher a program*  remember that
comments are not Infallible —  they can lie.

     Just as  indentation  and  physical  separation are  used to  help delineate
parts of a composition,  these  devices should be used to graphically explain
program logical  flow  and  subordination.  Loops should be Indented and should
end in the same column they begin.  IF statements and IF-THEN-ELSE structures
also should  follow a consistent Indentation  format.  Comments  may be helpful
in separating parts of loops and conditional  statements.

     Indentation 1s useful when it is necessary to continue COMMON,  DATA and
other type and declaration statements beyond a single  line.  For  example,
beginning the continuation lines at,  say, column  10 helps to delineate blocks
of code.  And  if there 1s no alternative  to using a GO TO construct (remember
the old maxim — rules are for privates and lieutenants; experienced program-
mers know when  to break the  rules),   indentation is useful  to help Identify
the continuations.

     Modularity and  structure  in code  can also be enhanced by  placing COMMON
Blocks in separate source code files that may be brought Into other source
code files via FORTRAN'S  INCLUDE statement.  In addition to producing smaller
subroutines,  use of INCLUDE files for COMMON blocks  will  Insure that COMMONS
are consistent across  a  program.   Don't  dimension or  name variables differ-
ently in different subroutines.  If it 1s necessary  to rename variables,  use
the EQUIVALENCE  statement.
                                     23

-------
     It also 1s useful to Include data type specifications 1n INCLUDE  files.
If possible,  group  different type variables  (eg, CHARACTER,  INTEGER and REAL)
Into separate COMMON  blocks.   This practice will  help  avoid  boundary align-
ment problems and enhances the maintainability of a program.   Explicit decla-
rations of INTEGER*2  and INTEGERS data type are also helpful.  Most com-
puters default  to the  INTEGERM  data type but, on  microprocessors 1n parti-
cular, the INTEGER*2  calculations are considerably faster  than INTEGER*4
arithmetic (Microsoft, 1985) 1n addition to saving some  memory.

     We have found  1t  desirable  to Isolate file OPEN and CLOSE statements 1n
a central, clearly Identified routine because, as  mentioned above,  these
statements may  be  Installation-dependent.   It also  Is  good  practice to use
variables  for unit  numbers  1n  Input-output statements and assign  these vari-
ables 1n an "environment" INCLUDE file, or within 1n the file-open subrou-
tine.  Al ways open terminal  Input-output units expl 1c1tly.   Although some
Installations assume units 5  and 6 for terminal  Input and output,  respective-
ly, others will use unit  5  for both and still  others  may use  units 1 and  2.

     Where possible, the user should be allowed to specify file names either
at execution or at  run time  rather than specify the names 1n the compiled
code.  This will allow easier Installation of code.  In addition, the OPEN
statement error return codes should be used to recover open errors grace-
fully, giving explicit error messages using  complete  English statements.

     We favor the use  of the PARAMETER statement  or  a  BLOCK  DATA subroutine
to declare array dimensions for system-wide variables, DO loops, I/O units
and other appropriate  system wide parameters.  Proper  use of the PARAMETER
statement will allow easy modification of program limits  by modifying a
single subroutine or source code file brought  Into  the program  with an
INCLUDE file.  This feature is useful  If one  wishes  to  modify a  program for
different  machine  memory configurations.  For example,  the IBM/PC version of
the QUAL2E program 1s  distributed with array dimensions  set  for a 256K
machine.  These array dimensions can  be changed by  modifying one line  In a
single subroutine.   One  should be careful,  however,  not to use PARAMETER to
set values for  variables.  Remember that the program source code must be
edited,  recompiled,  and  linked, to change any  PARAMETER or a variable Initia-
lized 1n  a BLOCK DATA  subroutine.

     In defining variables,  explicit declarations of variable type should be
used rather than the default  type.  A useful extension  on some compilers 1s
the IMPLICIT NONE statement.   This statement  will  force explicit  declaration
of all variables when  writing code and can be easily  removed  for  the distri-
bution version.  Dimensioning variables 1n INTEGER  and REAL   Statements  was
considered good practice  but  we  have found this to be  a problem  with micro-
computer compilers that only allow dimensioning  1n the DIMENSION and COMMON
statements and require type specification 1n  separate statements.   As
mentioned  above, the same name  should be used everywhere.  This  helps avoid
later confusion when debugging the code.   Prefix/Suffix-Type  variable names
are also  useful  in  helping readability.   And a clear  specification of Initial
value avoids  problems  with  systems where a zero default Initial value 1s not
the convention.
                                     24

-------
Downloading Problems

     In the process of downloading our software to the microcomputer*  we have
encountered several problems that relate to programing style and coding con-
ventions discussed above.   In general*  microcomputer FORTRAN compilers  are
quite strict  1n  following the ANSI standard for  FORTRAN  77.   Few 1f any FOR-
TRAN language extensions  are  available  on  microcomputers and practices that
are acceptable on larger computers are not  acceptable with these compilers.

     For example*  the compilers  are quite  sensitive to  statement  order  and
will not allow mixing of specification* declaration and executable state-
ments.   Variables that are assigned character string values should be defined
as Character  1n  a  type  specification.  Programmers should be careful not to
use variable  names longer than six characters as the more  restrictive com-
pilers  will not  recognize differences beyond the  sixth character.

     A  good practice  1s to separate character  and  numeric data  Into Indivi-
dual COMMON blocks to avoid problems with  boundary alignment.  The same  1s
true of Integer* real and double precision variables.   And  one  should  avoid
mixing these  different type variables 1n executable statements*  always  using
explicit type conversions 1n arithmetic  statements.

     Another  good practice Is to Initialize  variables 1n a BLOCK  DATA subrou-
tine and avoid reinitializing with DATA statements.   On micros*  argument
types must be compatible 1n CALL  and SUBROUTINE statements.  Although com-
pilers on  larger computers may convert between types* the microcomputer
compilers do  not.

     Some compilers will  allow FORMAT statement  field separators to be  omit-
ted — on  micros* this will cause  errors.  ENCODE and  DECODE statements
should  not be  used;  they have been replaced  by  Internal  READ and WRITE
statements 1n FORTRAN 77.

Development Tools

     There are a number of FORTRAN compilers available for personal com-
puters.   Two  recent  reviews  1n PC Tech Journal  (Howard*  1985)  and  Computer
Language  (Bensor et  al.*  1986) discussed nine different FORTRAN compilers
(Table  1)  for the MS-DOS operating system and the Macintosh.

     Several  letters  responding to  the  Howard's  review  are  contained  In  the
March 1986 Issue and  provide additional  Insight  to  the compilers.   Of parti-
cular Interest  1s an editorial 1n the  March  1986  PC Tech Journal  (Fastle*
1986) that observes "the volume of comment on our coverage Indicates consid-
erably  larger Interest  1n FORTRAN than expected ...  confirming  that FORTRAN
1s alive and  well. ... One particularly surprising comment 1s  about FORTRAN'S
portability.  Because 1t  1s not a  systems  programmers  language*  programmers
are not as likely to exploit extensions supplied by a particular  vendor*
opting  Instead for a  more standard approach. This portability extends beyond
the desktop to the minicomputers  and mainframes crunching FORTRAN."
                                     25

-------
                  TABLE 1. REVIEWS OF PC FORTRAN COMPILERS
          Compiler                     PC Tech        Computer
                                      Journal        Language
                                      Oct. 85        Jan. 86
           IBM Professional FORTRAN           X             X
           (a.k.a. Ryan-McFarland)

           Lahey                             X             X

           Microsoft                         X             X

           Digital Research                  X             X

           Intel FORTRAN-86                                X

           Prospero                                        X
           ProFORTRAN

           Supersoft                                       X

           WATFOR-77                                       X

           Microsoft                                       X
           MACFORTRAN
     The Computer  Language article  1s particularly Interesting because 1t
also discusses  a number of add-on tools  that provide graphics, string manipu-
lation,  windows, scientific subroutines,  and productivity enhancements.

     Our experience 1s  limited  to the Microsoft™  and IBM Professional™
(a.k.a. Ryan-McFarland") FORTRAN  (ProFORT) compilers.   Microsoft's compiler
1s flexible 1n that 1t can produce executable code that executes  on  PCs
without the 8087  Math Coprocessor chip, whereas ProFORT requires the  co-
processor.  However,  Microsoft's FORTRAN  syntax*  particularly for INCLUDE
statements,  and use of Metacommands  for compiler options differ from conven-
tional usage.

     We have found that ProFORT 1s  compatible with our primary development
environment, DEC VAX FORTRAN.  ProFORT 1s a  full  Implementation of FORTRAN 77
with few, 1f any,  extensions and  1s  often 100% compatible with other FORTRAN
development  environments.  This compatibility has been a key to our ability
to develop  and maintain software targeted for a wide  variety of computer
systems.
                                    26

-------
     The key to Installing large FORTRAN programs on  the PC lies 1n the link-
age editor.   The LINK command  provided  with  both  the  Microsoft  and ProFORT
compilers will allow only one level of overlay and  will only allow overlay of
code segments.   In trying to  link a large program (25*000 lines of code)  with
these linkage  editors,  we received a message Indicating that code segment
DGROUP exceeded 64K (Parker* 1986, discusses this problem).  Phoenix Soft-
ware's  PLINK86  linkage  editor allowed us to bypass this problem  by  allowing
multiple overlay levels  (up to 32 levels of overlay may be used).  A complex
overlay structure  may be  required  by large program size and  proper  software
design  can make for a much more efficient overlay structure.

Summary

     We have reviewed  programing  design and coding conventions that  have
served  us well  1n maintaining and transporting  out FORTRAN programs  among  a
variety of computers.  The development tools discussed above make 1t  possible
to Install quite large FORTRAN programs on microcomputers  and our experience
1s showing  that the microcomputer  can be an attractive environment  for
running these programs.   When execution times are excessive, the programs can
be easily moved to larger machines to take advantage  of their faster speed.


References

Bensor, R. et a!., "Software Review: FORTRAN on the MICRO", Computer. Lan-
3iLfl£g> Computer Language Publishing Ltd.,  San  Francisco, CA ,Jan 86, pp83-
110.
Brown  L.  C.  and T. 0.  Barnwell, Computer Program Documentation for
Enhanced  Stream Water Quality Model QUAL2E.  EPA/600/3-85/065, U S EPA,
Athens,  GA,  30613,  Aug  1985.

Fastle,  W.,  "Language  Surprises," PC Tech Journal,  Z1ff-Dav1s,  NY,  Mar  1986,
p!2.

Howard,  A.,  "FORTRAN Options," PC Tech  Journal. 21ff-Dav1s, NY, Oct 1985,
pp!49-160.

Parker,  P.,  1n  "Letters",  PC Tech Journal. Z1ff-Dav1s, NY,  Mar  1986, p22.

. _ ,  IBM Personal Computer  Professipna.1 FORTRAN,  by Ryan-McFarl and Corp.,
IBM Corp., NY 1985.

_ , Microsoft FORTRAN  Reference Majiual, Microsoft Corp., Bellevue, WA.,
98009
 The work described  In this paper was funded  by the U.S. Environmental Pro-
 tection Agency and  has been subject to the Agency's peer and administrative
 review.  It has been approved for publication as an EPA document.

                                     27

-------
               CONSIDERATIONS ON USE OF MICROCOMPUTER MODELS
                         FOR STORMWATER MANAGEMENT

               "by:  Wisner, P.
                    Professor, Dept. of Civil Engineering
                    University of Ottawa, Ottawa, Canada, KIN 9B4
                    Consuegra, D.
                    Research Assistant, Dept. of Civil Engineering
                    University of Ottawa, Ottawa, Canada, KIN 9B4
                    Frazer, H.
                    Gumming - Cockburn Assoc.,  1735 Courtwood Crescent,
                    Ottawa, Ontario, K2C 2B4
                    Lam, A.
                    Andrew Brodie Assoc., 3?1 Gilmour, Ottawa, Canada,
                    K2P OR2
                                 ABSTRACT

     Widespread use of microcomputer hydrologic models raises the question
of their optimum design.  Microcomputer models developed in North America
and Europe can "be classified into two categories:

     1.  Models with explicit incorporation of many parameters which can be
         used as default values or modified for calibration purposes if data
         is available.

     2.  Models reducing drastically the number of parameters and incorpora-
         ting default values in the program without giving the user a
         possibility for a change.  In this category rainfall distributions
         are also built in.

     The paper gives examples of possible sources of errors associated with
the second approach.  If relatively sophisticated techniques are used to
simulate rainfall runoff processes, these methods can be incorporated in a
microcomputer user friendly model and still provide enough flexibility to
allow non-modellers to modify default parameters in cooperation with a
specialist.  This also provides a better understanding of hydrologic
principles.  The paper also describes a Canadian package of microcomputer
models based on this principle and some implementation aspects.  A practical
application is also described.  The main runoff simulation techniques
presented in this paper give consistant results with detailed SWNM for urban
areas and with appropriate selection of parameters results can be compatible
with SCS type simulations for rural areas.
                                     28

-------
1.  INTRODUCTION
     In the early 70's the profession started  to  consider  urban hydrologic
models as a desirable replacement for empirical methods  for  runoff  control,
such as the Extended Rational Method.  The use of models was  advocated
mainly because they can be tested against  rainfall-distributions,  real or
synthetic.  This objective can be achieved using  various principles of
rainfall-runoff transformation such as kinematic  routing,  parallel  non-
linear reservoirs, single quasi linear reservoirs,  etc.  A discussion of the
advantages or disadvantages of each type of model and  mainly  of "routing
versus convolution", is beyond the objective of this paper.   What  all these
models have in common is that the user can modify several  parameters for
calibration and verification of applicability  for specific conditions.  This
also applies for rainfall loss models such as  Green-Ampt,  Horton,  Holtan,
etc.

     Flexibility in selection of parameters or use  of  various  storm distri-
butions is an obvious advantage for a hydrologist,  but may be  confusing for
a practising drainage engineer.  Calibration is not always possible because
of lack of data.  Under these conditions two approaches  are used  in practi-
cal applications:

A.  Models In which a relatively large number  of  parameters can be  used
either as default values or modified for calibration purposes.

B.  Simplified models with default values incorporated in  the  program and
very few input parameters.

     Examples of packages of microcomputer programs in the second  category
are those based on the SCS   Curve Number method  with  an empirical  relation
for the initial losses (ie:  la = 0.2S).  Runoff  is computed  by means of the
SCS curvilinear unit hydrograph with empirical relations for  the  response
time incorporated in the program.

     These models can be even simpler if the rainfall  pattern  is  already
built in (Debo, 1985).  Attempts of simplification  are also made  in a more
sophisticated model used in France (Chocat, 1984).  Although  it has interes-
ting features, runoff contribution from the pervious area  in  urban  water-
sheds is neglected, which is not applicable for North  American  conditions.

     Although the response of urban watersheds has  a higher degree  of non-
linearity as compared to rural areas, microcomputer simplified  models based
on SCS techniques, Santa-Barbara method, etc., assume  linearity in  both
cases.

     In most recent microcomputer developments a  significant  effort is made
only towards user friendly structures, attractive output displays  and exten-
sive graphical capabilities.  Although these are  important features they
will not be discussed in this paper, which is  mainly intended  to advocate
the advantages of A (above).  This approach is possible  without reducing
the ease of use.

                                    29

-------
2.  SIMPLICITY versus  ACCURACY
     The pitfalls  of  oversimplifications  can be illustrated by Canadian ex-
perience with  the  SCS methodology (SCS,  1971).

     In the SCS  procedures, relation (1)  in figure 1  between total rainfall
and runoff, has only  two  parameters; soil  storage and initial abstraction,
la.  The initial abstraction is  related  to the  soil storage by the empirical
relation la =  0.2S.

     The SCS unit  hydrograph is  automatically defined by the response time
which is related to  the  time of  concentration by an empirical relation.

     By means  of these simplifications,  the practising engineer has to
specify the same number  of  parameters as  in the Rational Method, namely;

     1.  The soil  storage or the CN number which is equivalent to a runoff
             coefficient.

     2.  The time  of  concentration.

     Comparisons of  simulations  and measurements conducted by the IMPSWM
(Implementation  of Storm Water Management) Program at the University of
Ottawa (Wisner et. al.,  1984)  have shown  that the cascade of linear reser-
voirs unit hydrograph proposed by Nash (Nash, 1957),has an adequate perfor-
mance for small  rural watersheds.  This unit hydrograph is given by a gamma
function defined with two parameters:  n,  the number of reservoirs or shape
factor and t , the time  to  peak  (relation(2), figure 1).  fly proper selection
of n and t  Several  shapes  of  unit hydrographs  can be reproduced.  For ins-
tance the shape  of the SCS  unit  hydrograph is a particular case of the gamma
function for n = 4.7J5.

     Table 1 shows the values  of n determined after calibration of the Nash
model for 6 rural  areas  (Wisner  et. al.,  1984).  These results were obtained
using a modified SCS  curve  number method  in which the initial abstraction is
an input (Ia<>0.2S).   It can be  seen that  the values of the shape factor n
are less than  4.75,  the  value  corresponding to  the SCS unit hydrograph.  For
similar conditions a  default value of n =  3 seems to be more adequate.

     The initial abstraction was also found to  vary from 1 to 4 mm; this is
less, by orders  of magnitude than the value given by the default value,
la = 0.2S.  It was also  verified that for  some  of these watersheds the use
of more sophisticated infiltration methods such as Green-Ampt or Holtan did
not significantly  improve the  results (Consuegra et.  al., 1984).

     The Nash model was  also compared with a more sophisticated variable
unit hydrograph  (VUH,  Ding,  1974).  Results also showed that after calibra-
tion the results with the Nash model and the non linear model were not very
different.  Based  on  the satisfactory performance of the Nash model and of
the modified Curve Number procedure,  it was decided to incorporate this


                                     30

-------
       Table 1.  Comparison of calibrated values of n and tp for 6 rural watersheds with the Nash Model.
                 Comparison of calibrated tp and tp computed using Williams equation.
Watershed
Wixon Creek
Holiday Creek
North Morrow
Clement Menard
Collins Creek
Parimbot
Location
Ontario
Ontario
Ontario
Ontario
Ontario
Switzerland
Area ha
1016
3053
382
96
15500
380
Slope %
1.2
0.7
1.0
0.1
0.15
2.0
Tp Calibrated
hrs
3.7
5.7
6.0
2.0
69.0
1.7
Tp Williams
hrs
1.1*2
2.82
3.31
1.93
11.8
0.61
n
2.6
2.0
2.0
1.85
3-0
3.0
Williams formula:  Tp = 6.5^ A°'39 S

                   Tp = hours

                    A = surface sq. m.

                    S = slope ft/mile

-------
technique in the microcomputer  programs  described  in the  next  sections of
this paper.  The resulting submodel  is called  NASHYD.

     The SCS procedures initially  developed  for  rural  areas  were later
recommended for urban areas  (SCS TR55).   IMPSWM  studies for  urban water-
sheds were conducted by first comparing,  for a hypothetical  watershed,
various linear and quasi  linear models with  EPA.SWMM (Huber  et.  al.,  1982)
simulations under design  conditions.   It  was found that the  performance of
some single linear or quasi  linear reservoirs  could give  results compatible
with EPA.SWMM for a limited  range  of  conditions  (high  imperviousness,  etc.).
Very good consistency with SWMM was  obtained with  a two parallel quasi
linear reservoir   conceptual model  in which the storage  coefficient  is a
function of the time of equilibrium  (P'ng,  1982).   The resulting URBHYD
model based on this principle uses relation  (3)  in figure 1, as  proposed by
Pedersen et. al., (1980).  URBHYD  was also  compared with  measurements  on 4
urban catchments with good results as shown  in table 2. (Wisner  and
Consuegra, 1986).  Flows  shown  in  table  2 were obtained with SWMM default
values.  Later, testing studies on European  watersheds showed  the need for
slight modifications of default parameters  specially for  low rainfall
volumes.

     The selection of meteorological  input  is  also a controversial issue.
It  is useful to compare results from design  storms and series  of historical
storms.  Studies conducted by the  IMPSWM program (Wisner  and Fraser,  1982)
compared Chicago (Keyfer  and Chu,  1957)  and  SCS  (type 2,  24  hours) design
storms with a series of real intense  storms  in the Metro-Toronto area.  Peak
flows were computed with  URBHYD and  NASHYD  for urban and  rural areas  respec-
tively.  For rural areas  the Curve Number (CN) was obtained  from a correla-
tion with API (Antecedent Precipitation  Index).
     For urban areas it was  found  that Chicago design storms generated peak
flows closer to observed  values than the SCS storms (fig. 2).
     For rural areas, flows  are strongly dependent on moisture conditions
and infiltration losses.  With  an  appropriate  selection of CN, the SCS
storms performed better than the Chicago design  storms.  For urban areas and
under design conditions,  peak flows  obtained with  SWMM and URBHYD were found
to  be very sensitive to the  maximum  rainfall intensity.  For the Metro
Toronto region, it was found that  a  time step  of 10 minutes  for  the Chicago
design storm gives realistic peak  flows.

     The above examples show that  microcomputer  models used  for  comparison
of  pre and post-development  conditions can  not be  excessively simple.   If
simplifications are introduced  they  should  be  tested with measurements for a
variety of conditions and their limitations  clearly defined.  As an example,
the default value la = 0.2S  may be acceptable  for  high rainfall  volumes.  A
single quasi linear reservoir can  give good  results for high degrees  of ira-
perviousness.  Simplified traditional methods  can  be applied mainly for
small areas.  Simplified  microcomputer models  should also give the user the
possibility to compare synthetic with real  storms.
                                    32

-------
               HTOROLOCICAL
                 SUBMODEL

                  URBKYD
               Urban Runoff
      METHOD
Borton equation for
losses
                                   Two parallel single
                                   linear reservoirs
                                   for pervious and
                                   Impervious areas
                     EQUATIONS
                                                                  f - Infiltration rate at  tine t

                                                                  f • Initial Infiltration  rate
                                                                   o
                                                                  {  - final  Infiltration rate
                                                                k - dekay factor

                               For both reservoirs, Storage coefficient K
                               K • a.
       T0.6  0.6
       L    U
        0-*  °-3
O)
                            DEFAULT VALUES

                            f  -  76.2 urn hi
                            o
                            f  •  13.2 mm ht

                            k - 4.14
                                                                                                                       pervious   - 0.2S
                                                                  a •  conversion factor     L -  length of vatershed
                                                                  n •  manning surface roughness
                                                                  1 •  maximum rainfall Intensity
                                                                  S •  slope of the watershed
                                                                                     impervious -  0.013

                                                                                    L • (1.5 Area)0'5
00
                  KASHYD
              Urban Runoff
Modified SCS
technique
                                   Nash Cascade of
                                   linear Identical
                                   reservoirs
                         (1)
Q » runoff (mm)      P •> precipitation (nun)

5 • soil water storage
la" initial abstraction (mm)

    25400                ..
C   254+S                U;

CN= modified curve number related to Antecedent
    Precipitation Index (API)

Unit hydrograph defined by two parameters,

n ™ shape parameter

T " time to peak of unit hydtograph
                                                                  ,-  qp
                                                                                                                       Ia<0.2.E.Ia - 3 ma
                                                                                    n • 3
                Fig.  1  Description of  NASHYD and URBHYD
                                                                  qp-  unit  hydrograph  peak

-------
Table 2.  Comparison of observed and simulated peak flows
with
Storm date
05-06-63
10-06-63
14-06-63
20-06-63
01-08-63
14-08-63
02-08-66
21-01-78
16-12-83
30-12-83
01-01-84
22-09-73
23-09-73
31-05-74
21-06-74
13-02-79
22-05-79
18-02-80
lumped OTTHYMO
Observed
(cms)
2.29
2.24
0.87
0.84
2.49
0.98
0.86
2.97
4.62
2.54
4.43
0.92
0.71
0.90
0.58
1.20
1.07
1.59
and detailed SWMM for 4 urban
Gray Haven
peak Simulated peak
OTTHYMO
(cms)
2.05
2.05
0.89
0.90
1.86
1.19
0.93
Hillbrow
3.41
4.53
2.46
4.41
Malvern
0.74
0.46
0.85
0.56
Pine town
1.22
0.74
1.35
catchments
Simulated peak
SWMM
(cms)
2.21
2.21
0.89
0.97
1.79
1.09
0.98
2.78
4.62
2.18
5.38
0.98
0.63
1.11
0.73
1.11
0.72
1.34
                            34

-------
                                                                        RURAL WATERSHED
               URBAN WATERSHED
                                                 100
      D  L«|n*ra*l Distribution
                        r«m hlicocic*! «v«nt«)
SO
40
 O  SCS Da«l|B Se*m Mows

 &  Cble*|o D««lfa Scon Flew*

   • 501

••-— JJX
13
                J.
                3    10    25      100

                HTUWf PEHIOD (y«»rp)
                                                       10
                                                    VI
                                                                  •  PUtting fotUUn {hliterlc  »toi«
                                                                                     flew*)
                                                                  O  U|n*iMl DUtrlfciitUn
                                                                     (Mit«ric •t*r« (Iowa)
                                                                  A  ChlcM* D««l«n Ctera CM*»5J
                                                                  O  ICS D««l|n lt*ra     CH'»55
                                                                          10
15
100
1000
                                                                      •CTUUI
                                                                                       l»ri.)
               Figure 2. Comparison of design storms  and a. series of real  storms for
                         Metro-Toronto conditions

-------
3.  USER FRIENDLINESS  versus  FLEXIBILITY
     An example of a  simple  procedure  for the incorporation of flexibility
in a user friendly program is  the  FASTHYMO model.   FASTHYMO was developed by
a consulting firm participating  in the IMPSWM program (Brodie et. al.,1984)
FASTHYMO can be used  for  simplified lumped watershed simulation with URBHYD
and NASHYD, using very  simple  inputs (figs. 3 and  4).  The user of FASTHYMO
works with a series of  design  storms,  given by municipal criteria or with
real storms.  Once these  storms  are entered,  they  can be permanantly stored
on disk.  The user can  first check list of storms  available on disk.  Prior
to simulations, users can review all default  values stored on disk and
eventually modify and store  a  new  set  of parameters on disk.  Figure 5
displays a typical parameter file.  As an example  for rural areas a default
value of n = 3 and a  value of  la = 1.5 mm were recommended for Metro Toronto
conditions.  The user can also specify n = 4.75 and la = 0.2S to ensure
complete compatibility  with  the  SCS procedures.

     However, the user  must  specify the time  to peak of the Nash unit
hydrograph (fig. 3).  No  attempt was made to  include one of the available
formulas for t  .  It  was  felt  that this may limit  the choices of the user
and promote a given empirical  formula without testing it.  Table 1 shows a
comparison between calibrated  values of tp and computed ones using the well
known Williams formula  (Willians and Hann, 1973) for the same rural areas.
It is evident that differences may be  significant.  Similar discrepancies
may be found using other  relations.  An analysis of travel time (overland
and channel), preferably  by  hand computation, is recommended.  Providing t
as an input also allows for  sensitivity analysis and a better understanding
of the model.  The rest of the input data are very easy to determine
(fig. 3).

     Figure 5 also shows  a set of  default parameters used to compute runoff
from urban areas.  The meaning  of these parameters  was already presented in
fig. 1.  A default relation  between length and area is also provided.  For
maximum flexibility,  the  user  can  also specify particular values for the
storage coefficient and use  the  model  as a linear  one.  If a municipality
intends to standardize  parameters  it can use  values from calibration studies
or those recommended  by a specialist.

     For a complete watershed  analysis including detention reservoirs
channel and conduit routing, users require the more advanced models descri-
bed further on.  The  same principles were maintained for parameters and
design storms.   PLANHYMO  (Brodie and et. al.,1984) is a simple Interactive
microcomputer watershed model  for  simplified  watershed simulation.  The
input for each watershed  is  similar to that of FASTHYMO.  At each junction
the user specifies all  the information related to  the different sub-areas.
He also indicates the type of  routing (if any) to  the next downstream junc-
tion.  PLANHYMO has the following  capabilities:
     1.  Determination  of runoff hydrographs  from  contributing urban and
         rural areas  at each junction  with URBHYD  and NASHYD, respectively.
                                     36

-------
               FASTHYMO..  NASHYD
          NOTE :  PRESS    'RETURN' TO SO BACK. TO MENU
            PLEASE  ENTER  :
                                                                        FASTHYMO.. URBHYD
                                                                 NOTE : PRESS  'RETURN'   TO   60  BACK  TO  MENU
                                                                   PLEASE  ENTER :
U>
1)  UNITS,  
i)  TIME TO PEAK ,   HOURS
7>  OUTPUT  ON PRINTER  , Y/N
•3J  PRINT FINAL HYDROGRAPH,  Y/N
1)   UNITS,   OR , M/I
2)   GET STORM  FROM DRIVE ?  
-)   NUMBER   OF  'STORM' TO  USE ,

4)   DRAINAGE  AREA  (HECTARES),
5)   IMPERVIOUSNESS RATIO,
6)   SLOPE     IN X,
7>   WANT OUTPUT ON PRINTER ,   Y/N
S)   WANT FINAL HYD. PRINTED ,  V.-M
      Figure 3. Window display for NASHYD in FASTHYMO
                                                              Figure  A.  Window display for  URBHYD  in FASTHYMO

-------
                   <  FASTHYMOi  PARAMETERS >

Fo,
Fc ,
DECAY value
ACCUM. inf.
DEP.STO. IMP.
DEP.STO.PERV
STO.COEF. IMP.
STO.CQEF.PERV
MANN. 'n' IMP.
MANN. 'n' PERV

in, mm =
in, mm =
=
in, mm «
in, mm =
i n , mm «
in, fliffi **
=
=
0

3. OOO
0.520
4. 140
O.OOO
O.O62
0.104
0.000
O.OOO
O.013
O.250

76.200
13.208
4.14O
0.000
1.575
4.674
O.OOO
O.OOO
0.013
0.250
           IN.  ABS.        in,mm  =  O.O59
           # of  NASH  RES.        -  3. OOO
1.499
3,000
           > WANT TO  CHANGE SOME VALUES, Y/N:
                  Figure 5. Parameter file in FASTHYMO
     2.  Addition of these hydrographs with other input hydrographs even-
         tually determined from other computations.

     3.  Routing of hydrographs through reservoirs, channels and pipes.

     4.  Storage or printing of summary results.

     Figure 6 shows a typical input display of PLANHYMO.

     More recently, LOTHYMO (Consuegra et. al.» 1986) was developed as an
extension of PLANHYMO.  The structure of LOTHYMO has been developed to meet
the following criteria:  fully interactive input and output processes and
secondly, efficient executional performance.  The first objective was
achieved by interfacing PLANHYMO with the powerful spread-sheet, LOTUS
1-2-3.  Using ready made templates, input files can be easily created.  The
systematic tabulation of all input parameters permits rapid entry, quick
review and .therefore, quick detection of any inconsistencies.  In the out-
put stage, results can be quickly retrieved by LOTUS 1-2-3 for their review,
interpretation and management.  Graphics can also be generated automatical-
ly.            The second objective was satisfied by the development of a
driver program written in BASIC, which reads the input files created with
LOTUS and calls all the appropriate hydrological sub-routines for the entire
drainage network.  At this stage absolutely no user interaction is required.

-------
                                PLANHYMO

 - Press  'RETURN* after your  INPUT

NOTE: Maximum No. of hydrographs  @  one  point is 4.

Please enter the following*         Simply hit 'RETURN' to go back to MENU

*»   ENTER the JUNCTION POINT NO. ,    =
 Any ROUTED HYDROGRAPHS at this junction, (Y/N)   =
  1> How many  UPSTREAM HYDR06RAPH  to be added,   =

  2) How many  RURAL WATERSHEDS <0- >,  de-f«=o     =
  3) How many  URBAN WATERSHEDS <0- ),  def«O     =
  4) How many  INPUT HYDROGRAPH <0- >,  def=O     =

  5) Want the FINAL HYDROGRAPH  to be ROUTED, (Y/N)   =
                        Figure 6. Window display for PLANHYMO


  Figure 7 shows the Highgate Creek (Ontario) watershed where  LOTHYMO  was
  applied.  The total area is 540 ha.  The downstream area  is  urbanized  and
  drained by an open channel with a capacity of 16.4 cms.   In  the  near future,
  the entire watershed will be urbanized.  If no runoff control measure  is
  undertaken, the 100 year post-development flow will raise up to  42.6 cms.
  Three detention reservoirs are proposed to reduce the post-development flow.
  Schematization for the post-development conditions is also shown in  figure
  7.  Figures 8 and 9 show the output obtained with LOTHYMO and the
  corresponding plots generated with LOTUS.


  4.  EXPERIENCE WITH THE MODELS
       FASTHYMO and PLANHYMO are presently used on IBM PC and Apple  He  for
  quick project verification by organizations such as the Department  of
  Engineering of the Town of Markhara, the Department of Public Works  of  the
  City of Scarborough, the Project Review and Approval Group of  the  Ontario
  Ministry of Environment, the Water Resources Service of the Metro  Toronto
  Conservation Authority, etc.

       The following are examples of projects using the flexible microcomputer
  models presented in this paper, illustrating the range of applications by
  consultants:

    i)  Preliminary Master Drainage Plans.  A consulting firm in Montreal
  applied PLANHYMO for an area including  two dry  ponds and  three wet ponds.
  The application was done during a two day training session on  the  job  with
  an engineer without previous computer experience.
        Safety study for a SWM reservoir.  A  consulting  firm in Illinois used
  PLANHYMO for the analysis of a  reservoir during  the  maximum possible storm.
                                       39

-------
                                  urban area
                                R rural area
                                  junction number

                                  junction
                                                                      Dummy junction
                                                                      no effect
 Figure  7.  Highgate Creek watershed and post-development schematization
ill)  Multiple  on-site detention reservoirs for an Industrial pagk.   A con-
sulting  firm  in Ottawa designed a drainage system with  three storages for
large parking lots  and a separate storage for the minor system with
PLANHYMO.   The  objective of this study was to drastically  reduce  peak flows
due to a lack of  downstream capacity.

     Drainage engineers involved in the review of projects presented  above,
did not  have  previous experience with hydrologic modelling and initially
needed indications  regarding the choice of default values.  Two years of ex-
perience with these microcomputer models show that practising engineers can
easily modify the described microcomputer models to suit their own needs.
Some of  them  perform sensitivity analyses on their own  initiative.  Storms
have also been  compared, and this led to useful discussions.
5.  FINAL COMMENTS
    As indicated  in  the  Introduction, there are several models for  rainfall-
runoff transformation,  infiltration losses and design storms.  While  their
advantages and disadvantages  are still discussed by hydrologists, there  is
some agreement,  that  with proper parameters reasonable results can  be
                                    40

-------


UATERSHED
Rural
Urban
Urban
TOTAL"
•ROUTE TO 2

WATEWSKB
U/S JN Ho 1
TOTAL"
to 3
UAT£R«ME1>
Urban
U/S JM No 2
TOTAL-
RffOUTC TO 4

UATERSrCD
U/S JN No 3
TOTAL"
to 3
DV P WIHY uvir
BullAH'T OUTPUT FOR
AREA IHP CN*
na. X
47.21 «"• 41
40.93 27 •*
37.30 70 a*
143.14 »» ••
149.24 ~ ••
BUnTtARV OUTPUT fly*
AJtCA IHP CM*
ha. X
149.24 *• ••
149.24 •• *4
149.24 •• 4»
SUrWARV OUTFVT FOR
MtCA f^^^ CMv
ha. x
213.40 30 ••
149.24 *» ••
371.44 a. a*
370. 44 •« ••
•ur*iAftv OUTPUT FOR
ARCA IHP CN»
na. X
370. *4 »• ••
370.44 ^» ••
370.44 •• •*
^f i rut r^E
JUNCTION
8
X
1.2
1.3
1.4
«
~
JUNCTION
S
X
*«
•«
««
JUNCTION
9
t
I.I
*•
•«
••
JUNCTION
B
X
••
•«
•a
. l««wl«t
POINT NO.
PCAK
cm
U3»
».24
14.27
23.09
4.43
POINT NO.
PCAK
eaa
4.49
4.43
4.44
POINT NO.
ffftf
CM
27.42
4.44
20.»2
10.30
POINT NO.
cam
10.30
10.30
10.30

1
TPEAK
13. OO
11. BO
11.00
11.00
12.00
2
TPCAK
12.00
12.00
13.16
3
rr
11.00
13.14
11. BO
12.00
4
TPEAH
nrm
12.00
12.00
13.24


RUNOFF VOL
42.74
31.43
04.49
40.53
40.92

RUNOFF VOL
40.92
40.92
40.90
•UNQpa VOL
93.93
40.90
94.00
94.00

RUNOFF VOL
•Ml
94.flO
94.00
34.00


WATERSHED
Urban
Urban
U/B JN No 4
TOTAL"

HATER9HCO
U/8 JN No 9
TOTAL-

UATER6HEO
Urban
Urban
U/S JN HO 4
TOTAL-
RftDUTE TO 0

-ATEITBMCB-
U/6 JN No 7
TOTAL"



SUHIIAI1Y OUTPUT FOR
AREA IMP CN.
na. X
26. 7O 3O •*
44.00 30 ••
378.64 •* 4«
44*. 34 a* ••
SUmARV OUTPUT FDR
AREA IHP C»«
444.34 »a •»
44*. 34 •» ••
SunnARV OUTPUT FOR
AREA ln» CN*
na. X
34. BO 30 ••
3». 70 30 •»
44*.34 •• w
923. 04 •• •»
323.04 »• •.
BUrHARV OUTPUT FOR
AKEA TMP en*
ha. t
923.84 »« •»
323.04 •• ••



JUNCTION
fl
X
o.a
0.0
a«
-
JUNCTION
s
X
•«
JUNCTION
•
X
0.0
0.0
•«
• •
• •
JUNCTION
s
X
••
»»



POINT NO.
PEAK
4.44
4.04
10.30
13.08
POINT NO.
PEAK
cm
19.08
19. 0B
POINT NO.
PEAK
cam
3.98
4.2*
13.08
S7.72
19.07
POINT NO.
PEAK1
cm
13.07
I9.O7



3
TPCAK
nrm
11.00
11.80
13.24
11.00
4
TPCAK
11.00
11.00
7
TPCAK
hra
11.00
11.00
11.00
11.00
12. 4O
B
T^CAK
nrm
12. 4O
12. 4O




RUNOFF VOL
93.94
93. V4
34.00
94.34

RUNOFF VOL
94.34
94.34

RUNOFF VOL
a*
93.*4
34.34
34.00
94.00

-OlBMCyE- Hf"f
tnttn^r VDL
•M
9*. 00
96. OO


Figure 8. Output generated with LOTHYMO for  the Highgate  Creek watershed

-------
a)
    Ul


    0
   S.S


    O
          9
11
       19
Time (hrs)
          Figure 9. Plots of inflow and routed hydrograph generated with LOTUS
                    at junction 3 (a) and at junction 7 (b)
          9
                                                           Time  (hrs;
                                        42

-------
 obtained with  several  types  of  models.   A key issue in Storm Water Manage-
 mentis_ whether  the  rapidly  developing  microcomputer software for prac-
 tising drainage  engineers should  or  should not promote a better understan-
 ding of basic  hydrologic principles.
 It is not yet  known  which of  these options will prevail.  The two years of
 experience with  some of the  models discussed  above, proves, that at least
 the first approach is  possible.

      Many agencies require consultants  to use standardized parameters, a
 given unit hydrograph  and a  specific  design storm.   Experience with the
 models proves  that even in these  situations there  is some advantage in com-
 paring results obtained with  standardized parameters and other input
 parameters*

     Information  on all models based  on  the IMPSWM  research is available from
 The IMPSWM group, Department  of Civil Engineering,  University of Ottawa,
 Ottawa, Canada,  telephone (613) 564-3911.
                                REFERENCES
1.   Brodie, A., Lam, A.t Rampersand, C.  Implementation of Microcomputers
     for SWM Studies in Canadian Municipalities.  Proceedings of the
     Conference Stormwater and Water. Quality Management Modelling.
     Burlington, 198^.

2.   Chocat, B.  Conception, evaluation et dessin des reseaux d'egouts
     (CEDRE).  Proceedings of the Third International Conference on Urban
     Storm Drainage.  Vol. 2.  Gotebttrg, Sweden.  1984.

3.   Consuegra, D,,  Jaton, J.F.,  Wisner, P.  Etude comparative de diffe-
     rentes fonctions d*infiltration.  Rapport Ecole Polytechnique Federal e
     de Lausanne.  Institut de Genie rural (IGR).  Lausanne, Suisse.  1985.

4,   Debo, T.  Personal Computers and Stormwater Management Programs.
     Proceedings of the Conference on Computer Application in Water
     Resources.  Buffalo.  1985.

5.   Ding, J.  Variable Unit Hydrograph.  Journal of Hydrology.  Vol. 22.
     1974.  pp. 53-69.

6.   Huber, W.C., Heaney, J.P., Nix, J.J., et al.  Stormwater Management
     Model Users Manual.  Version III. University of Florida, Gainsville,
     Florida.  1982.

7.   Keifer, C.J.,  chu, H.H.  Synthetic Storm Pattern for Drainage Design.
     Journal of the Hydraulics Division.  Proceedings of ASCE.  Vol. 83.
     No. HY4.  August 1957.

8.   Nash, J.E.  The Form of the Instantaneous Unit Hydrograph.  Inter-
     national Association  of Scientific Hydrology.  Publication 14-5.  1957.

                                     43

-------
9.   Pedersen, J.T., Peters, J.C., Helweg, O.J.  Hydrographs by Single
     Linear Reservoir Model.  Journal of the Hydraulics Division ASCE.
     Vol. 106.  No. H5.  Proc Paper 15^30.  1980.  pp. 837-852.

10.  P'ng, C.  Conceptual Hydrologic Modelling for Master Plans in Urban
     Drainage.  MaSc Thesis.  Dept. Civil Engineering, University of Ottawa,
     Ottawa, Ontario, Canada.   1982.

11.  Soil Conservation Service.  National Engineering Handbook, Section 4,
     Hydrology.  United States  Dept. of Agriculture.  U.S. Government
     Printing Office, Washington, D.C.  1971.

12.  Williams, J.R., Haan, R.W.  HYMO A Problem-Oriented Computer Language
     for Hydrologic Modelling.  ARS-5-9,  U.S. Dept. of Agriculture.  1973-

13.  Wisner, P., Frazer, H.  Design Storms for Stormwater Management Studies.
     Part V.  IMPSWM Urban Drainage Modelling Procedures.  2nd Edition.
     Dept. of Civil Engineering, U. of Ottawa, Ottawa, Ontario,  1983-

14.  Wisner, P., Frazer, H., P'ng, C.E., Consuegra, D.  An Investigation of
     the Runoff Components in the HYMO, OTTHYMO and VUH Models.  Heport to
     the Ministry of Natural Resources, Ontario, Canada.  1983-

15.  Wisner, P., Consuegra, D.  Testing of OTTHYMO on Twenty Watersheds.
     Proceedings of the International Conference on Storm Water Modelling.
     Belgrade, Yugoslavia.  1986.

16.  Wisner, P., Consuegra, D., Morse,B.  Utilisation de Lotus 1-2-3 pour
     1'amelioration de la gestion des donnees d'un modele hydrologique.
     First Canadian Conference  on Computer Applications in Civil Engineering/
     Micro-Computer.  McMaster  University.  1986.
The work described  in  this paper was  not  funded by the  U.S.  Environmental
Protection Agency and  therefore does  not  necessarily  reflect the views  of
the Agency and no official endorsement  should  be inferred.


                                     44

-------
                  PC SOFTWARE FOR COMPUTATIONAL HYDROLOGY

                 by:  W. James, M. Robinson and M. Stirrup
                      Computational Hydraulics Group
                      McMaster University
                      Hamilton, Ontario, L8S 4L7
                      Phone:  416/527-6944
                                 ABSTRACT

     The  Computational  Hydraulics Group at  McMaster  University  has  been
developing  and using integrated software for personal computer  systems for
the past 8 years.   The software encompasses data collection,  data transfer.
data  base management,  hydrologic  modelling  (storms,   stormwater,   water
quality,  hydraulics, receiving waters)  and real-time  control.   Most  of the
programs are written in FORTRAN.

     This paper describes the relationship between these activities and the
sequence of data processing necessary for various types of study.   About 50
programs are listed and briefly described - most are geared to  the acquisi-
tion  of  continuous high-resolution rainfall data,  subsequent   data  base
management, continuous stormwater modelling and real-time control.

     Performance of the programs in an IBM-PC compatible hardware  environ-
ment  is indicated in broad terms;  integration of this range of hydrologic
software is now being completed.
                                INTRODUCTION

      Long-term  flow and pollutant  records may  be   synthetically  generated
 using an elaborate,   deterministic  computer  model  such  as  the  U.S.  Environ-
 mental Protection Agency's  Stormwater  Management Model  (SWMM){1).    Version
 3  of  SWMM  has  been adapted by our  group  to  run  on  IBM-PC   compatible
 machines,  so that a complete,  very  large  and intricate  model can now be run
 on  a computer costing less than $1500 US (2).   There  are no  serious draw-
 backs using PCSVMMJ in this environment.   Ve have supplied this package to
 a  significant  number of users, and  the existing large   SWMM  user  group
 infrastructure  assures  wide-ranging, continuing support and interest  in
 this product.  The package  is  being used  world-wide.

      PCSWMM3  has many advanced capabilities not available in  other  models
 in  use  today,  e.g.  menu-driven  design dialogue,  input error  checking,
 diagnostic  messages,  transparent  file  handling,  verification/validation


                                    45

-------
        External
        data sources
        Data decoding
        data transfer
        Time-series
        management
Local
keyboard data
File
management
Input data
files
                                    4-
Program
library
                                                     V
Storm models

Stormwater models

Flood plain models

Receiving waters
             Figure 1.  Hydrologic computing environment.
modules,  continuous modelling,  water quality, diversion structures,  sedi-
mentation and scour,  costing,  screen-oriented graphics,  statistical sum-
maries, to name a few.  Nevertheless the package is easy to use,  especially
because  of the user-friendly documentation,  simple menu-driven input  and
graphical  output.   Probably  the most important attribute  is  continuous
modelling,  since this removes all the intrinsic shortcomings of the design
storm  approach - all flows are ranked directly,  giving correct probabili-
ties and risks (5).  Difficulties due to the management of large amounts of
input  and output data are eliminated by our special data  base  management
software  (CHGTSM),  also specially written for IBM-PC compatible  environ-
ments  (4).  The general computing environment  is shown in Figure 1.

     One  question  that is seldom meticulously addressed is model  valida-
tion.  We have devised a methodology incorporating systematic verification,
sensitivity  analyses,  calibration and validation,  that appears  to  work
smoothly  and  faultlessly (5).   The procedure is set out in  the  PCSWMM3
Workbook  Module.   Of course,  adequate field data is essential,  and  our
inexpensive  raingauges  and  data loggers  (6) render it  easy  to  collect
sufficient  summer  thunderstorm activity in one or two months (in the north
during the period May-October) to satisfy model calibration for most criti-
cal  modelling situations.   The same Instrumentation is used in our  real-
time   controllers   (7).   This points to the advantages of  considering  an
integrated  microcomputer-compatible  system from the  outset.   A  typical
sequence of activities described here is shown schematically in Figure 2.

     There are other important benefits that accrue from a dense network of
field  stations.   Perhaps the best example  occurs in storm  modelling.   It
should be obvious that urban flooding often results from the short,  sharp,
local  convective storms that are so common in warm weather.   These storms
are  much smaller in area than the catchments  and have  surprisingly  short
lives,  typically   50 minutes or so.   Our  software PCRAINPAC analyses data

                                     46

-------
from a network of synchronized raingauges to compute storm  speeds,   direc-
tions,  growth  and decay (8).   Output from PCRAINPAC is fed directly into
PCSWMM3  (9)  to compute flood flows and pollution loads  that  are   demon-
strably more accurate than any other hydrologic package known to us.    This
is to be expected:  the better the Input, the better the results.

     Figure  2  presents  the typical sequence of  activities  involved  in
carrying   out  a  hydrologic  modelling  study  on  a  personal   computer
environment.   Any  such  study requires time series' of rainfall as  input
to the model.  A record of discharge is required in order to calibrate  and
validate  the  model.   These  data are typically  supplied  by  government
agencies such as the Atmospheric Environment Service (AES) and Water Survey
of  Canada   (₯SC) in Canada and the National Weather Service (NWS)  in  the
United States.

      Large   amounts  of data  are often  required,  especially  for  studies
involving continuous simulation.   This data can cover periods in excess of
40  years.    It would not be  feasible to distribute or acquire such data on
media such  as floppy diskette or over telephone lines.   Consequently,  the
most   efficient  method  for  supplying/acquiring this data   is   on  9-track
magnetic tape.   Several commercially available subsystems exist which per-
mit  interfacing of standard,  9-track, mainframe-compatible tape  drives with
IBM-PC  compatible microcomputers.   These  subsystems interface  through   a
             Obtain data
             from archive
             Tape interchange
             package
             Compress data
             Data base
             management
             Hydrologic
             models  (storms,
             floods,  levels,
             receiving waters)
              Final Report
9-track
tape
drive
30 Mb
Hard
disk
area
               Figure 2.  Typical sequence of activities.

                                      47

-------
board  installed  in one of the expansion slots on the PC  bus.   The  tape
interchange  utility supplied with the subsystem downloads the data to  the
user's system.

     Depending  on  the amount of data required for the  study,  the  files
created  from  the  data supplied on tape can range from 1 to 60  Mbyte  in
size.   This necessitates having sufficient storage available on the  users
system.   It  has been our experience that a hard disk drive with 4-0 to  50
Mbyte  capacity  can be acquired fairly inexpensively and will satisfy  the
needs of most studies requiring large data storage.

     In  order  to properly manage,  retrieve and update this  database  of
hydrometeorologlc  data special software is required.   The software  which
has been developed has the capability to take the data in its raw form  and
store  it   in a "compressed" form in which only non-zero data  is  retained
thereby  reducing the data file size.   In an application performed by  our
group, a rainfall database covering the period 1975-1984- with an uncompres-
sed  size   of 52 Mbyte was reduced to 18 Mbyte using the  data  compression
technique.   The  database,  once constructed,  is coherent and can be made
available to all members of a modelling group.   Updating or correction  of
the  database is the responsibility of one person who needs only manipulate
a single   file.   This  ensures that all modellers are using   an  identical
database.    Increased study reliability results.   The data is accessed  by
users  through  menu-driven  retrieval utilities and can  be   automatically
interfaced  to simulation package  such as PCSWMM.

     Such   a hydrologic database management  system when  being used  by   a
group  of modellers  has  been found to function most efficiently in a network
environment.    The   database   and application programs reside  on a  central
network  computer,   or  server.    The modelling group uses a number  of PC's
which  are  also  nodes on the network.  Applications programs run locally and
 independently on  each  node access data  from the  simulation.   Results  can be
 returned to the database for  use  by other  applications.   The   applications
programs are described subsequently.


      Flood and pollution control  may be effected by:

      a)    diverting overflows to storage,   directly  to receiving waters   or
           to some other sewer network,
       *
      b)    pumping or draining overflows from detention storage,

      c)    warning  responsible  authorities  of impending  flooding  and/or
           pollution,  for example for timely closure of swimming beaches  or
           underpasses, and

      d)    computing the extent of the swimming or recreational areas  which
           should be evacuated.

      Real-time control of urban stormwater may be achieved by onsite micro-
 computers  rather than by a single,   central computer.    The  microcomputers
 need  not  be expensive;  in fact,  $30 hand-held Z-80 based  microcomputers


                                     48

-------
having BASIC in ROM are probably sufficient at each control  site (7).    The
site will generally include:

     1)   an  instrument  to  measure flow,   water   level,  rainfall  and/or
          pollutant concentration,

     2)   circuitry  to transform the signal into  information meaningful to
          the microcomputer,

     3)   a small,   replaceable microcomputer running a model derived  from
          statistical  or similar analysis  of data relevant  to the  control
          site,

     O   circuitry to drive  the electrical control mechanism, and

     5)   a  data  logger recording all incoming information,  the  control
          action, and the precise date and  time.

     The microcomputers and associated circuitry,   for onsite control,   and
inexpensive raingauges measuring rainfall intensity,  require large amounts
of software, also developed by the Computational Hydraulics  Group at McMas-
ter University.   The real-time control software (TSDASUTIL) has been based
on transfer-function models (TFM) of synthetic long-term flow and pollution
at  each  control  site.   The TFM is easily coded into  the  data  logging
program,  written in BASIC.   TSDASUTIL is  menu-driven and has, among other
attributes,  data communication software, so that the data may be automati-
cally  transferred to a central database (10) after manually retrieving the
tape.   Other  software that may typically be involved in such a  study  is
depicted in Figure 3.

     The TFM software is derived directly from the synthetic long-term time
series generated by continuous PCSWMM3 at each site.   It predicts expected
flow  and pollution in the vicinity of the control site a few minutes ahead
(enough to complete a control action).   There is a risk,  in some drainage
systems,  that bad timing of diversions could cause flows to coincide down-
stream such that flow conditions become worse than they need have been.  In
these  cases   it  is necessary to run the  continuous  simulation  for  the
overall  drainage system with all local control software at each site simu-
lated.   This  is why an elaborate overall deterministic model is necessary.
PCSWMM3  also  allows the full range of other stormwater management  strate-
gies (separation of roof leaders, tile drains, etc.) to be evaluated (10).

     The   minimum   functions  that must be  supported in a   real-time   flood
control  system are:

     a)   data acquisition,
     b)   data storage/retrieval,
     c)   streamflow forecasting,
      d)   project operation.

     The  software  which might be utilized a typical RTC   application  is
 illustrated  in Figure 3.    These  programs perform the four  tasks  listed
 above.

                                      49

-------
                          TSDASUTIL
                      •4-
      DASDECOD or
      TSDASDECOD
INTERFACE
CHGTSH
      DASTRAN
PCGRAPH3
          RAINPAC
                                                        PCSWMM3
                                                        RTCSIH and
                                                        TS analysis
                                                        TOTSED
        Figure 3.  Typical Application (evaluation of RTC),
     DAS3 and TSDASUTIL collect data from nearly raingauges and store it on
cassette tapes in the field.   TSDASUTIL might also be controlling a diver-
sion  structure,  locally,  based on the incoming real-time rainfall  data.
After  collecting  the cassette tapes and transporting them  to  a  central
site,  the data is decoded and communicated to our local area network (LAN)
of IBM-PC compatibles.  The LAN consists of a 52 Mbyte hard disk, a 9-track
magnetic tape drive,  a Idne printer,  and three microcomputers,  each with
512 kbytes of RAM, 8087 co-processors, and two floppy disk drives.  Program
DD1  decodes the data collected by DAS3,  using a data decoder.   TSDASUTIL
transmits the data via a RS232 interface.   Data is received by DASDECOD or
TSDASDECOD through the serial port of one of the LAN stations,  and  stored
in a disk file.

     The  data can be processed immediately by DASTRAN or  permanently  ar-
chived in the CHGTSM for a variety of applications.  DASTRAN translates the
raw  data file created by DASDECOD or TSDASDECOD into event summaries.   If
desired  the  program also produces two additional  files:   1)  a  PCSWMM3
compatible interface file for input to the PCSTATISTICS3 module, useful for
                                      50

-------
ranking rainfall events,   based on peak intensity,   total volume,  duration,
etc.,  and  2)  a  PCGRAPH compatible data file for   plotting  hyetographs.
INTERFACE  reformats the  raw data for entry to the   CHGTSM,   our  rainfall-
runoff data base,  which  resides on the LAN.   This makes the data available
to  our  group  for a number of applications,   including  data  processing,
graphics, report generation, and hydrologic modelling studies.

     Figure 3,  presents  a typical hydrologic  modelling application,   aimed
at  evaluating  proposed  RTC schemes.   First  RAINPAC is  used  to  prepare
rainfall input for the PCSVMM3 continuous model.  STOVEL is  used to analyse
a number of point rainfall records for direction,  speed,  and peak growth/
decay  mechanisms.   THOR4-DPT  uses the results of the STOVEL  analysis  to
generate  hyetographs for each subcatchment in the   PCRUNOFF3  model.    The
continuous water quantity and quality record generated by PCRUNOFF3 is then
input to the PCTRANSPORT3 model of the diversion structure(s).   The results
of  the  continuous PCRUNOFF3 and PCTRANSPORT3 simulations can be  used  to
develop  a simple rainfall-runoff forecast model which can be evaluated  in
RTCSIM,  and  implemented in TSDASUTIL.   TOTSED models the  deposition  and
scour of sediment at a combined sewer outfall.

     Other  programs  which  might  be included in   such  an  analysis  are
PCSTORAGE3,  CHGQUAL,  OVRFL03, and TOTSED.  At any  point in the hydrologic
modelling,  the rainfall, runoff., and pollutant time series output by these
programs can be stored in the CHGTSM for subsequent  analysis.

     The  whole system may sound unduly complex,  but runs very effectively
on IBM-PC compatibles.   All of the programs are operational;  the software
is listed below.


                          1.  STORMWATER MANAGEMENT
      The  PCSWMM3  package  for  urban  runoff quantity and quality  in  storm  and
 combined  sewer  systems  comprises  11 modules:

      1.    PCRUNOFF3  module  -  simulates overland flow hydrographs and pollu-
           tographs from urban and non-urban watersheds.

      2.    PCTRANSPORT3  module -  generates infiltration,   dry weather  flow
           and  provides hydraulic routing model for flow and pollutants   in
           non-surcharged  situations.    Includes  pipe design   options   and
           various pre-programmed  pipe shapes  and  regulators.

      3.    PCEXTRAN3  module  -  hydraulic routing model for flow  in surcharged
           situations.   Includes  regulators,   pump stations, tides and  some
           pre-programmed  pipe shapes.

      4.    PCSTORAGE3  module  - simulates the  flow routing  and  removal   pro-
           cesses  of  detention ponds and  sewage treatment plants.    Includes
           cost estimation procedures.

      5.    PCSTATISTICS3  module - provides  simple statistical  analysis   and
           frequency   analysis of continuous  simulation results generated by
           other modules.
                                     51

-------
     6.    PCGRAPHICS3  module  - provides   a general  plotting utility   for
          PCSWMM3 results and other  data.

     7.    PCEXECUTIVE3  module  - assists  the user  in data preparation   for
          PCSWMM, controls file handling.

     8.    PCCOMBINE3  module  - provides  a utility for   merging  files   of
          results  from individual PCSWMM3 simulations for input to  subse-
          quent  modules  thereby allowing large areas to be  simulated   in
          segments.

     9-    PCWORKBOOK3  module - leads novice user systematically through  a
          series of verification tests on  several modules of PCSWMM3.

    10.    PCINSTAL3  module - provides an  "official" check on  the  perfor-
          mance of the users hardware and  software system.

    1 1 .    PCSWMM3 - RUNOFFG - a modified version of the  RUNOFF module which
          includes  interflow simulation.    This provides a means for   con-
          tinuous computing of startup values for wet weather events.

    12.    CHGQUAL - modified version of the SVMM3 RUNOFF block water quali-
          ty modelling algorithms, for pollutant buildup and washoff.

    13.    OVRFL03  - given the geometry of the main channel and side weir,
          computes  the  water surface profile along a side weir,  and   the
          over-spill, given incoming flow.

    14.    TOTSED - analyses quasi-transient sediment motion, deposition and
          scour at a combined sewer outfall.

Expert System:

    15.    PCSWMM  A-S  (Auto  Sensitivity) represents a  group  of  program
          modules  forming a shell around the PCSWMM program.   The program
          modules work on an interactive basis with the  user and the RUNOFF
          module of PCSWMM.   They provide sensitivity and uncertainty   in-
          formation  on  the  effects of eleven physical  input  parameters
          required  by  the RUNOFF module.  The  output   generated  through
          PCSWMM  A-S includes a ranking of sensitivity magnitudes for  each
          of  the  tested parameters and the uncertainty,  expressed  as  a
          percentage,  that  is associated with the output from the  RUNOFF
          module.

          The program requires the standard minimum hardware  configuration
          specified  for the PCSWMM3 package.   This includes an IBM-PC  or
          compatible  with at least 512k RAM,  two 360k drives and an  8087
          HDP.   To  operate  the program the user responds to a series  of
          prompts contained  in the batch "shell" and indicates an  existing
          file which is to be analysed by the A-S.  Input required includes
          confidence  levels  in the input parameters and  certain  utility
          data.   Analysed output data  includes information on total volume
          of runoff,  peak flow,  mean  flow,  time to peak flow,  flows ex-

                                     52

-------
          ceeding  prespecifled  limits  and the volume of   any   exceedance.
          Output is directed to the  console as well  as to  files  on  the  disk
          drive.   The time  required for a complete  run  of the package  is  a
          minimum  of  60 to 80 minutes  for a single subcatchment   with  12
          data points for the input  hydrograph.   Internal  processing of the
          data requires approximately 80* of the  total run time.

                            2.  STORM MODELLING
     RAINPAC  is  a  package of 4 programs  which  analyses   and  models   the
kinematic properties of urban rainfall  events:

    16.   STOVEL  - analyses a number of point  rainfall  records  for   direc-
          tion,  speed and peak growth/decay mechanisms

    17.   THOR4DPT  - simulates  the kinematic  properties  of  a storm using
          STOVEL results to generate hyetographs  at point  locations.

    18.   THOR4D - simulates kinematic  and  areal  averages  for  storm  events.

    19.   THOR3D - simulates the areal  average  of a line storm event.


                         3.  FLOOD PLAIN ANALYSIS


    20.   PCHYMO   - hydrologic modelling language for simulating  non-urban
          watersheds  on  a  simple event basis using the SCS  curve  number
          technique.  Methodology not recommended for serious hydrologists.

    21.   HEC-2   - a steady  state backwater curve calculation  model  which
          simulates water surface elevations along open channels.

    22.   FASTHEC-2  -  a preprocessing program for HEC-2 which  assists  in
          data preparation.

                   4.  TIME SERIES MANAGEMENT SYSTEM (TSM)


     The  following  5  programs are used to build and/or update  the  data
base:
     23.   OPENFL - opens a TSS file and optimizes record size and number of
          records in a file.

     24.   OPELBL - opens a data set in the file opened by OPENFL.

     25-   PATCH  - reads  the  contents  of a data  set  opened  by  OPELBL
          (created by user previously) and inserts it into the data base.

     26.   INSERT - similar to PATCH (replaces current data with new data).

     27.   UPDATE - similar to PATCH.

                                    53

-------
The following 5 programs are used to retrieve data.   QUERY and CONTINS are
similar  to  RETRIV,  and could be classed as output handlers for  specific
purposes,

    28.   RETYDIR  - displays information regarding the types of data which
          reside  in the data base,  and selects the appropriate TSS  file,
          containing the data requested by the user.

    29.   RETSSDIR - displays information regarding the locations for which
          data reside in the directory,

    3C.   RETRIV - retrieves data for user specified period (1-365 days) in
          constant or variable timestep format.

    31.   QUERY  - retrieves and displays data for user specified site  and
          period.   A disk  file containing this data, at either constant or
          variable timesteps may be  created.

    32.   CONTINS - similar to QUERY,  but the user may specify the  format
          of  the output file,  and  thus can be used to prepare input   data
          files  for other  programs.   Default format is for PCSVMM3 RUNOFF
          data file (Rainfall data groups).

    33.   INTERFACE  - utility for processing data  (flow,  rainfall)  from
          government agencies such as National Veather Service (US), Atmos-
          pheric Environment Service (Canada),  Vater Survey of Canada,  or
          private  groups such as our group (CHG), into forms suitable  for
          input to PCSVMM3  or the CHG-TSM.

                            A. DATA MANAGEMENT
    34.   INTERP  - processes  rainfall from decoders or  flow  from  chart
          records  into event  summaries.   Also provides input to FASTPLOT.
          Similar version for  Apple lie,  written in Applesoft BASIC.  Also
          FORTRAN version for  APIOS DCPS-D/VPS units.

    35-   DASTRAN  - processes raw rainfall data as created by DASDECOD  or
          TSDASDECOD   into   event  summaries  (similar  to  INTERP).   Also
          creates  PCSVMM3   compatible  interface  file  of  rainfall,  and
          PCGRAPH input file for plotting data.

    36.   TSDASUTIL  - program package performs various tasks  related  to
          data  acquisition  and RTC,   including collecting rainfall data and
          controlling  a  simple  diversion structure  in  real-time.   The
          package also processes the  data, plots hyetographs, and transmits
          the data to  IBM-PC compatibles for further processing; written in
          Sinclair  BASIC for  the Timex/Sinelair 1CB0  microcomputer.   Two
          main  modules handle  these tasks:  1) RTCONTROL - data acquisition
          and RTC,  and 2)  TSDECODER  - data processing, plotting, decoding,
          and communication.
                                      54

-------
         6.  DAS PROGRAMS (DAS = Data Acquisition System)


37.   LPDAS2.V2 - microprocessor controller program for timing and data
      collection for DCPS and lov-pover data logger;  written in macro-
      assembler.

38.   DAS3 - same as above for standard data logger.

39-   TBDAS3 - same as above but for TBRG.

40.   DASMOE1  - same  as DAS3 but also records data from wet/dry  pre-
      cipitation sampler.

                   7.  DAS Data Decoder Programs
41.   DD1 - decodes data collected by DAS on cassette tape,   and Trans-
      mits  it  to computer (IBM-PC compatibles,   TRS-80,   PD?  11/23);
      written in macro assembler.

42.   DD1-APP - same as above but  for Apple lie.

43-   TSDECODER - retrieves data from storage,   converts from  Sinclair
      to  ASCII character set,  and transmits data serially  to  IBM-PC
      compatibles.  Also performs  some simple processing of the data in
      the field,  including simple graphics;  part of TSDASUTIL package
      (see 36).

44.   DASDECOD  - receives data sent by TSDASUTIL  through RS232 device
      and creates a disk file (same as DASDECOD)  on an IBM-PC  compati-
      ble; written in GVBASIC.

45.   TSDASDECOD  - receives data sent by TSDECODER module of TSDASUTIL
      through  RS232 device and creates a disk file (same as  DASDSCOD)
      on an IBM-PC compatible; written in GVBASIC.

46.   DASINTERFACE - receives data sent by decoder (DD1-APP),  creating
      a disk file on Apple He; written in Applesoft BASIC.

                       8.  PLOTTING PROGRAMS
47.   FASTPLOT - plots hyetographs or hydrographs of events, input from
      INTERP program; for use with Houston Instruments plotter.

48.   PCGRAPH  - (same  as GRAPHICS module of PCSWMM3),  used  to  riot
      hyetographs  as created by DASTRAK;  written for IBM-PC  compati-
      bles.
                                  55

-------
                          8.  REAL-TIME CONTROL
   49.   RTCONTROL - see 36.

   50.   BOXJEN  (MITS  originally) - program for the identification of  a
         discrete  linear  Transfer Function Model from input  and  output
         time series;  vritten by the Chemical Engineering Process Control
         Group at McMaster University modified for use on IBM-PC  compati-
         bles.

   51.   RTCSIM - simulates the operation of a simple diversion structure,
         continuously,  in  real-time.   Uses the results of  a  TRANSPORT
         simulation,  and  runoff  forecasts to simulate the diversion  of
         CSO.
                                REFERENCES

1.   Huber,  Wayne C., Heaney, James P., Nix, Stephan J.,  Dickinson,  Robert
     E.  and Polmann,  Donald J.  Stormwater Management Model User's  Manual
     Version III.   U.S. Environmental Protection Agency,  Cincinnati, Ohio,
     1981 .

2.   James, ₯. and Robinson, M.A.  An affordable alternative to a mainframe
     computer  environment for continuous modelling.    In:   Proceedings of
     the Stormwater and Water Quality Modelling Conference,   USEPA,  Gaines-
     ville, Florida, 1985, pp. 13 - 30.

3.   James,  W.  and Robinson, M.  Continuous urban runoff modelling.  Con-
     ference on Urban Drainage Modelling, Dubrovnik,  Yugoslavia, 1986.

4.   Unal,   A.  and James, W.  Distributed continuous hydrologic processing
     using microcomputer networks.   In:  Proceedings of the Stormwater and
     Water   Quality  Modelling Conference,  USEPA and Ontario  Ministry  of
     Environment, Toronto, Ontario, December 5-6, 1985, pp.  169 - 180.

5.   James,  W.  and Robinson,  M.A.  Standard terms of reference to ensure
     satisfactory computer-based urban drainage design  studies.   Canadian
     Journal of Civil Engineering, Vol. 8, No. 3, 1981, pp.  294 - 303-

6.   Haro,   H.,  Kitai,  R.  and James,  V.   Precipitation instrumentation
     package  for sampling of rainfall.   Institute of Electrical and Elec-
     tronics  Engineers (Transactions on Instrumentation and  Measurement),
     Vol.  IM32, No. 3,  1983, pp. 423 - 429.

7.   James,  W.  and Stirrup, M.  Microcomputer-based precipitation  instru-
     mentation.   International  Symposium on Comparison of Urban  Drainage
     Models with Real Catchment Data,  IAHR, IAWPRC, Dubrovnik, Yugoslavia,
     1986.
                                   56

-------
  8.    James,  V.   and  Scheckenberger,  R.   Storm dynamics model for  urban
       runoff.    International Symposium on Urban Hydrology,  Hydraulics  and
       Sediment  Control.   University of Kentucky, Lexington, Kentucky, 1983,
       pp.  11  -  18.

  9.    James,  ₯.   and Robinson, M.  Conversion of the USEPA SWMM3 package to
       microcomputers.   ASCE  Fourth Conference on Computing in Civil  Engi-
       neering,  Boston, Massachusetts, 1986.

  10.   James,  V.  and Unal A.  Evolving data processing environment for compu-
       tational   hydraulics systems.   Canadian Journal of Civil Engineering,
       Vol.  11,  Mo. 2, 1984, pp. 187 - 195.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.

                                     57

-------
                                    DABRO
            A BASIC LANGUAGE PROGRAM FOR HYDROGRAPH COMPUTATION
            by: Bernard L. Golding, P.E.
                Consulting Hydraulics Engineer
                Orlando, Florida  32819
                                  ABSTRACT

      In this  paper  a  hydrologic model designated DABRO for the computation
of  hydrographs  from complex drainage basins is presented and discussed.  In
this  model,  written  in  the  BASIC  language,  rainfal1-excess  increments,
computed  by  the  SCS  runoff   curve  number  procedure,   are  applied  to  a
mathematically  computed  unit  hydrograph of  each  subbasin  to  obtain  the
runoff design hydrograph  from  each subbasin.  These  subbasin hydrographs  are
then  summed and/or  routed downstream  by  the model to obtain the total  basin
hydrograph.  The hydrographs  computed by  the  model  are similar  to  those
computed  by  TR-20  and  HEC-1.  However,  because of  easy  data  entry  and
editing  capabilities, the  model  is  extremely  easy to  use.  Also, the code
may be easily changed by  the user to  fit particular situations.

      Simulations of actual  runoff events  from three  urban  watersheds  and
one large  rural  watershed  by  use of  this  model are also presented  and  the
results discussed.

INTRODUCTION

     DABRO,  in  its  "batch  model   format",   actually  consists   of  three
separate programs or  parts. The first program  (Part  1  - Data Input Files)

                                     58

-------
is  used to input  the  various subbasin properties  (subbasin areas,  runoff
curve numbers,  unit  hydrograph  and routing parameters,  etc.)  on to a disk
file;  the second  program  (Part  2 -  Read  Data Files)  is  used for  reading
and/or  editing  the data  placed  in disk file by the  first  program; and the
third  program  (Part  3  -  Hydrograph  Computation)   computes  the  rainfall-
excess,  unit  hydrograph and  total hydrograph  of each  subbasin,  and sums
and/or  routes  them downstream to  the  basin  outlet.  In this version of the
model,  the total  hydrograph  of  each  branch  is  computed  in  a sequential
manner  and  stored  on disk  file.  Later, during subsequent program  operation
as  the  main channel  hydrographs are  being  computed, these hydrographs are
read  from disk  file by  the  program  in the same order  in  which they were
input.  The  program then adds the branch  hydrographs  to the  main channel
hydrographs   to   produce   the   total  hydrographs  below   the   channel
intersections.  A  definitive  numbering  sequence  is  necessary to  insure
proper  program  operation.

      A  separate program  (RAIN.FIL)   is used  to  input  rainfall increments
comprising the design storm on disk file.

     These  programs,  plus programs   to   perform the   various  operations
separately  (compute hydrographs, sum  hydrographs, flood  route hydrographs),
constitute the  book  "Design Hydrographs by the Drainage Basin Runoff Model
(DABRO),  published by Hilbern Engineering Software.

      In  the model, rainfall-excess increments are computed from successive
rainfall  increments  by a  modification of  the standard  Soil  Conservation
Service   (SCS)  rainfall-runoff  equation  which  modification  accounts  for
runoff  from three  sources  -  from the urban  directly  connected impervious
area;   from  the  urban   pervious   (grassed)   and   nondirectly   connected
impervious area; and from  the rural  portion of each subbasin  or from each
source  only.  The mathematically computed  synthetic  unit hydrograph used  in
the computation of the design  hydrograph of each subbasin, although similar
in  shape  to the unit hydrograph of F.  F.  Snyder, permits  the user to vary
both its peak  flow value (discharge) and the time to peak (lag time) of the
peak  flow. Reservoir  routing  is accomplished  by  the storage-indication
working curve  method and channel  routing by the Muskingum Method.
                                    59

-------
     In addition  to successive  rainfall  increments comprising  the  design

storm,  inputs to  the  various  programs  are the number (designation) of each
subbasin; the properties of each subbasin including:

     1.   the previously computed peak flow (discharge) and time to peak of
          the unit hydrograph of the subbasin,
     2.   the area of the subbasin,
     3.   the  percentages  of  the directly  connected  impervious  area  and
          rural areas of the subbasin,
     4.   the  runoff  curve  number (CN)  for  the  urban  pervious  and  non-
          directly connected impervious area of the subbasin,
     5.   the  runoff  curve  number   (CN)  for  the  rural   portion  of  the
          subbasin,
     6.   the  initial   abstraction  (IA)  for  the  urban  pervious  and  non-
          directly connected and rural portions of the subbasin;


the properties of the various channel  reaches  through  which flood routing

is being accomplished,  including:

     1.   the  length  of the  river  reach  (reach length) through which  the
          subbasin  hydrograph  or combined  subbasin hydrographs are  to  be
          routed,
     2.   the reach velocity,
     3.   the reach routing coefficient (X);


the properties of the storage reservoir  through which  flood routing is to

be accomplished,   including:

     1.   discharge rates  from the reservoir  at  various  elevations  above
          the spillway,
     2.   the storage volume in the reservoir at these same elevations;


and the computational  time interval.


     Outputs  from the  programs  are  a  table  of  the original  successive

rainfall  increments  input  to   the  programs  and  the  resultant  computed

rainfall-excess   increments;  a  table  of  the  computed  unit  hydrograph

ordinates of each  subbasin;  tables  of  the total   combined  (summed  and/or

routed) hydrographs of the various subbasins; and a plot of the  final total

basin  hydrograph. Included with the tables of  hydrograph  ordinates  (unit,

design, summed and routed) are the number of inches of rainfall-excess from

each subbasin (=1.0+" for the unit hydrograph).
                                     60

-------
RAINFALL-EXCESS

     In  the  DABRO  Model,  rainfall-excess  increments  are  computed  by a
modification  of  the   standard   rainfal1-runoff  equation   of   the  Soil
Conservation  Service.(1)    In  this  modification,  as  previously  stated,
direct runoff  from  a drainage basin  is considered  to  come from three  or
less sources  -  from the  urban  directly connect  impervious  area;  from the
urban  pervious   (grassed)  and  non-directly connected  impervious  (urban)
areas  and  from the  rural  area  of the  basin,  each of which  is  considered
separately.

     In  the  original   SCS  procedure,  a  runoff  curve  number   CN  for  a
particular entire subbasin was first computed which enabled the computation
of  the direct  runoff  Q up  to  a certain  time  from the  basin  knowing the
rainfall  P which fell  on the  basin  up to  that  time.   In  the  originally
derived  procedure,   the  following  equations  were  used  to  compute  direct
runoff Q from rainfall:

           Q  =  (P-IA)2
                P-IA+S                                     (1)
           CN =  (1000)
                10+S                                        (2)
where
           Q  =  Direct runoff (inches)
           P  =  Total storm rainfall (inches)
           IA =  Initial  abstraction (inches)
           S  =  Watershed storage factor (inches)
           CN =  Runoff Curve Number (dimensionless parameter)

The  runoff AQ in inches which occurred in a certain time  interval  At was
then equal  to  (Qt~Qt-l)»

     The  Runoff Curve  Numbers for various types of agricultural,  suburban
and urban  land  uses  as originally developed by the SCS for computing  direct
runoff from rainfall  utilizing  these  equations  are given  in  various  SCS
manuals.  As most  engineers know, these Runoff Curve Numbers depend  on  the
                                      61

-------
Hydrologic  Soil  Group,  the  land  use or cover  and  the Antecedent Moisture
Condition for the particular drainage basin.

     For  purposes  of computing  direct  runoff  by  the method  presented  in
this program, the following equation is used:
      AQt =  AP(%DCIA) +AQ,, (100-%DCIA-%RURAL) +AQr  (%RURAL)  (3)
                10~0
where AQt is the  total  direct  runoff from the  basin  in  a particular time
interval  £t  (=Qt-Qt-l); _ A P(%DCIA)   is  the  direct  runoff  from  the
                             IfiO
directly  connected  impervious  area  (DCIA); £QU  (100-%DCIA-XRURAL)  is the
                                                        100
direct  runoff from the pervious areas and non-directly connected impervious
(urban)   areas   in  this   time   interval  as  computed   by  Equation  1,
AQr  (%RURAL)
       100  is the direct  runoff  from the undeveloped  (rural)  areas  in this
time  interval as  also  computed by Equation 1, and AP  is  the  rainfall that
falls  on  the  basin during  this time  interval.  The ability to compute  runoff
from three  separate  type  areas within  a subbasin enables the  evaluation of
land  use  changes  occurring on  the basin.

      In   the  computation  of  direct  runoff   from  the  directly connected
impervious  area  (DCIA) by  this program,  the first 0.1 inch falling  on such
areas  is  assumed to  fill  up  depression  storage  and, therefore,  is not
assumed to  contribute  any  runoff.

      In  the original  procedure  as developed  by  the  SCS for  use  in  rural
areas,  an Initial Abstraction of 0.20S  was used in Equation  1  to complete
direct  runoff.  However,  in the  application  of the  procedure as modified
herein  the  following values of IA can be used as being more representative
of the  initial abstractions  imposed  on  runoff  by  urban  areas (2):
                     Hydro! ogic Soil  Group      IA
                             A               0.075S
                             B                0.10S
                             C                0.15S
                             D               0.20S  (unchanged)
                                     62

-------
Where runoff from undeveloped  (rural) land  is  computed,  the value of 0.20S
(unchanged) should be used.

     In the  DABRO Model,  runoff can be  computed  from completely urbanized
basins in  which  case zero values are input  to the program  for  the Runoff
Curve No.-Rural; the Initial Abs. Coef.-Rural and for the Percent-Rural.  If
the basin  is completely  rural  (no urbanization), zero values  are input  to
the  program  for  the  Runoff Curve  No.-Urban  Pervious &  DCIA;  the  Initial
Abs. Coef.-Urban Pervious & DCIA; and the Percent-DCIA.

     In the  computation  of  runoff  by the  SCS  procedure and  as modified
herein, three  Antecedent Moisture  Conditions  (AMC)  are  defined depending
upon the amount  of  rainfall (antecedent  rainfall)  that   has fallen  on the
basin prior  to  the  storm  from  which direct  runoff is being computed, which
AMC's are defined in Section 4, Hydrology of the SCS's National Engineering
Handbook.

     In the  computation  of a  Runoff  Curve Number  (CN)  for  a  particular
basin  or   portion  thereof,  by  the SCS  procedure,  a weighting process,
depending  on the percentage of the  particular  land  use  cover, is used.  As
previously mentioned,  separate  Runoff  Curve  Numbers for  both  the urban
portion  of  the  Drainage   Basin  not  directly  connected  - the pervious
(grassed)  area  and the non-directly connected  impervious  area  of the urban
portion and  the  rural portion of the subbasins are both program inputs.

     Since the  DABRO  Model  uses the  SCS  Runoff  Curve  Number  to  compute
rainfal1-excess,  it  is,  of  course,  a   single  event model.   However,   as
subsequently  discussed,  it  has  been   and can   be   easily modified  for
continuous analysis  using  an  antecedent precipitation   index  (API) which
relates  rainfall to  S,  the  watershed  storage  factor,   or  CN,  the Runoff
Curve Number.

UNIT HYDROGRAPH

     In the  DABRO Model,  the ordinates of  a synthetic unit hydrograph for
each basin are  computed  by  the  following equation  developed  by James   A.
                                     63

-------
 Constant, formerly  Chief Reservoir  Regulation  Unit, Albuquerque  District,
 Corps of Engineers{3).


                                          (Log t)2

                                              2
                                            2cr

                     Q =  111.8094 A e
                               crt                                      (4)

 where

 Q = Flow Rate  (Discharge) in cfs
 A = Drainage basin  area  in  square miles
 t = Time in hours from beginning of  runoff
 T = Time in hours at which  50%  of flow has passed
   = 10  (In lOcr2 +  Log TmaxQ)
 o" = Standard deviation of the log 10 of  the  values  (see  note  below)


      This equation  produces hydrographs  that generally conform to  the shape
 of the  synthetic unit hydrograph  developed  by  F.  F. Snyder as  presented  in
 Corps  of Engineers  Manual  EM-1110-2-1405 -  the rising  limbs of the  unit
 hydrograph  are  parabolic  and the   recession  curves   are   approximately
 exponential and are similar to  the unit  hydrograph  of the Soil  Conservation
 Service.


       In the  evaluation  of  Equation  4, Q is differentiated with  respect  to

 time  and set  equal  to zero -   standard  procedure  for determining the  peak
 value   of  a   function  - which  results   in  the following equation  (after

 transposing and rearranging):

      -'(In 10)2 o- 2
           2
    e                          - TmaxQ  Qmax        (See  note  below)    (5)
    ff-                            111.8094A
Note:  cr = Greek letter SIGMA; letter S  used  for  SIGMA  in  program  listing.

 (In  10)2  # 2   =  2.650949056 CT2
   2
                                       64

-------
The values  on  the right  side  of Equation  5  are generally easily computed
from standard  equations as  will  be  subsequently  described.  TmaxQ  is the
time (in  hours)  from the  beginning of rainfall-excess  to  the peak of the
unit hydrograph  (TOTAL  TIME  TO  PEAK-HOURS)  which is equal  to the lag time
(normally defined as the time from center of mass of rainfall-excess to the
time of peak) plus one-half of the unit hydrograph duration or tp + D/2.

     In this program for computing unit hydrographs, 
-------
ordinate  of  the design hydrograph.  The  program automatically performs the
required time increment lagging required for hydrograph computation by this
procedure.  Flood routing  through channel  storage is  accomplished  by the
Muskingum  Method  and  flood  routing  through  reservoir  storage  by  the
storage-indication working curve method or Modified Puls  Method.

     In  the  model,  the total  basin  (outflow)  hydrograph  is  computed by
summing  up  the  various  subbasin  design  (and/or  routed)  hydrographs.  The
model automatically performs the time  lagging  required  for  basin hydrograph
computation. As  previously  stated,  the DABRO  Model  will  also  plot  up the
final outflow  hydrograph  from  the total  basin under consideration (time in
hours vs  discharge  in cfs)  if desired. The  final  basin outflow hydrograph
can  also  be  saved  on disk  for possible  further manipulation by the other
programs  in the book.


MODEL VARIATIONS

     Because the DABRO Model  program is written in Microsoft  Basic and the
program can be  listed (not copy protected),  changes can easily be made. For
instance, data  can  be easily  imbedded  into the program  so that manual entry
of subbasin variables does not have  to be input to the model  every time it
is used.  Also,  modifications to allow manual  entry of  a  series  of rainfall
increments so that a  different set of  rainfall  increments  can  be applied to
each subbasin  can easily be done.  As  subsequently described, this enables
the model to be  used  in real-time reservoir  regulation.

MODEL APPLICATION

     General:  The  DABRO Model  has,  to date,  been used to simulate runoff
from four different  basins. Three of  the  basins were  small,  fairly steep,
almost completely urban basins located in the  Tallahassee  area  (hereinafter
referred  to as  Basin  Nos. 1, 2 and 3). The fourth  basin (Basin  No. 4) was  a
large,  fairly   flat,   sandy,  almost  completely  rural  basin,  located  in
Manatee County,  Florida (the Manatee River basin). All  were well documented
as to their area,  physical  properties, land use,  soil  type, etc.,  and all
                                    66

-------
had accurate gaged rainfall-runoff data including information on the amount
of antecedent  rainfall  prior  to the  event  simulated.  In all  cases, the
Theissen  polygon  method was used  to  weigh the  rainfall  increments of the
various storms simulated.

     Urban Basins:  Basin  Nos.  1,  2-and 3 are,  as previously  stated, all
highly  urbanized  basins,  were  gaged  as  part  of  the USGS's  urban   basin
gaging program, conducted  in  the  Leon  County,  Florida area.  Basin No. 1,
the  Frenchtown  Creek basin,  3.29 square  miles  in size,  is  essentially a
completely sewered  basin  consisting of the  Frenchtown Creek  area  and the
Florida State University  (FSU) campus.  The lower portion of the Frenchtown
Creek basin consists of a  1200 feet  long,  8  feet diameter circular conduit
which  carries  Frenchtown  Creek  under  the FSU  campus.  Basin  No.  2, the
Franklin Avenue drain basin, 2.06 square miles in size, is also essentially
a highly urbanized basin.  It consists of two  definitive  portions or parts;
the upper  one-half,  which  consists of a sewered  residential  area,  and the
lower  one-half,  which  consists   of a  portion  of the  downtown   area  of
Tallahassee and the  State  Capitol complex. The  open  channel  draining this
lower  portion  of the  basin is  essentially   retangular  or trapezoidal  in
shape, much  of  which channel  is  lined.  Basin No.  3,  0.21  square  miles in
size, is an essentially mixed  residential  and commercial  area drained by a
fairly steep open channel. Soils  in all three basins were classified by the
USGS as being in Hydrologic Soil  Group B.

     In all,  sixteen rainfall  events  were simulated  by the  DABRO Model;
five  on  Basin  No.  1,  five  on  Basin  No.  2  and six  on  Basin  No.  3. The
results of these  simulations  are shown on  Figures 1 through  16.  In all
cases,  the total  impervious  area and  directly connected  impervious area
were determined from  a  study  of  aerial  photographs supplemented  by   field
inspection.  Runoff  Curve Numbers   for   the  pervious   and  non-directly
connected impervious areas were determined in the  usual  manner by weighing
the grass and impervious areas. In all cases, the time to peak in hours and
peak  flow  rate  of  the  ten-minute  unit  hydrograph  of  each  basin was
determined by the  equations developed  by  Espey, Altman  and  Graves(4). In
accordance with  program requirements that  the time  step  At be equal to the

                                      67

-------
CT>
CO
                                                                                                                                          OAOKO >v\OOEl_

-------
                                                            3O,  198
u.
\J
VJ


UJ

o

oi
u
in
a
                        it          i£          i


                            (UOURS -V  IO AAIM.
                          SAS//^ * 2  STOCM Of JULY II.
                         tL  f
                                     69

-------
 81
4
I
    0:
                      SIMULATION
  t
             DA&RO MODEL.
id£ASUR£Q DI5CHARCC
                                70

-------
                                          ^4,IB8O
PA0RO  MOD£L_
                                   10
TIME.
          -r  IO
          71

-------
 DRAINAGE -BASIN *3  ST&RM or JULV 24,
                   QA/SKO
Zl
    23       2^        I
TIMS. ( HOURS v /O M
-------
  DRAJWAGE. BASIXJ * 3  STORM OP  SEPT 6.
                   DA6RO MODEL
                                 MEA:>URED on CHARGE
fe
       ORA/WAdE BASfM «» 5  S-TOR^N O^ S«^*T El.
                    CAflgO MDDEL.
                         73

-------
SAS1NJ *J 3
OF
                                                                               , /9flO
IZO
                              DAB.^O  MODEL
                                    9           10           u

                                    HOURS 4  IO -V»/MOr£
                   •10
                                          ORAPWA6E 6A51M a
                                          STORM OP MARCH 3O, 1981
                            T(A1£ f UOURS -KO
                                         74

-------
unit hydrograph  duration D, ten-minute  rainfall  increments  of the various
events  simulated were  Inputs  to  the  program.  No  attempt  was made  to
calibrate the model  to improve the simulation results.

     As can be observed  from  these Figures, the results of the simulations
were quite good, except for the rainfall of  March  31,  1981 on Basin No. 1,
shown  on  Figure  5,  in  which  case it  appeared  that the  1200 feet long, 8
feet diameter  conduit  under the  FSU campus was  inadequate  to  convey the
peak runoff rate resulting from that particular rainfall.

      Rural  Basin:  Simulation of  various runoff events from  the  130  square
 mile Manatee  River  basin  (Basin  No. 4) using the  DABRO  Model was done  as
 part  of  a  real-time  reservoir  regulation program   implemented  by  Camp
 Dresser and McKee for  the  Lake  Manatee Dam(5).

      For  the  prior  design  of a new  emergency spillway for the dam and  for
 the purposes  of a  possible  future real-time regulation  program,  the  basin
 had been  previously subdivided  into  five subbasins  and runoff curve  numbers
 based   on  land  use  in  each of  the  subbasins for  the  AMC II  condition
 computed. Also,  as  part  of this  prior  work,  two or  more rainfall  gages with
 telemetering  equipment  had been  installed  in each  of  the  subbasins.

      As part  of the  simulation  work,  the  lag time  of  each subbasin  was
 computed  by Snyder's  Method  using  a Ct=2.2 and the time to  peak Tp  (=0/2
 + tp)  for  the  two-hour  unit hydrographs  for  each subbasin  computed.  The
 peak flow rate  of each of  the two-hour  unit hydrographs  were then  computed
 by the equation  Qp=CA/Tp.  A "C"  value  ranging  from 300 to 425 for each sub-
 basin  was selected  based on  the  characteristics of  the individual subbasin.

      The  results of- simulation of  the June 17-18, 1982  rainfall,  obtained
 by summing  and/or  routing the various  subbasin  hydrographs, are  shown  on
 Figure No.  17.  This was  a  fairly  major rainfall  event  in  that an average of
 approximately 6.6  inches   of rainfall  fell over  the basin in a  38-hour
 period, most  of which  fell  in  a  24-hour  period. This  resulted in 2.96
 inches of  runoff  at  the  dam.  The  prior  dry  period  to this  rainfall  was

                                     75

-------
      o
      ir
IOVAAJ
14000
I200O
10000
8000
6030
4000
2000
0























































1






L















J~-















?'














I/












/
/
/|











i
1

_r











f
'


'





*x









/










0 10 20








_









J---


^






— \








(
N




\









— 1




\
\

"t





RESERVOIR GAT
OPERATION
i


T


^SERVED INFL<
IT RESERVOIR
SIMULATED INFL
AT RESERVOIR


V,
^^





F






S^
\









	


^


^





so



. 0
u


T
X1





VI

OVIL






••


JTFl
ANA

^




Fi
.OW
TEE


--,









gu
AT
DA



•^







_...
re
M



^
1-
L











— t
\7





~,





i


"" *•

40 5<
                                    TIME
reflected by the average computed basinwide CN of 67.  In the computation of
the simulated  inflow,  shown on Figure  17,  each  of the previously computed
subbasin CN's for AMC  II were  adjusted  down three units to account for the
difference  between  the originally computed average  basinwide  CN of 70 for
the AMC  II  condition  and  the  actual average  basi-nwide CN of  67 for this
rainfall event.

      As  can be observed from  Figure 17, reasonably  good  simulation  of  the
June  17-18,  1982  rainfall  was  achieved.

      Verification  of  the  various  parameters input  to the DABRO  Model  and
the model  itself  were  limited  by  the lack of a second major rainfall  event.
However, reasonable verification  was achieved using the Storm  of August  28,
1981  during which an  average  of 2.3 inches of  rainfall fell  over the basin
resulting   in   1.1  inches  of  runoff.   This  large   runoff  volume  (CN=85)
reflected  the  relatively large amount  (3 inches)  of rainfall  which  fell  in
the three  days prior  to the August  28,  1981 event.  Again, the originally
computed  subbasin  CN's for  AMC  II were  adjusted  (this  time  upward)  to
account  for the difference between  the computed  average basinwide CN of  70
for the AMC II condition,  and the  actual  average  basinwide CN of 85. Note
                                     76

-------
that this value of CN=85 is very close to the CN value for AMC  III as given
in Section 4, Hydrology of the SCS's National Engineering Handbook.

     However, because  of  the obvious extreme  importance of the antecedent
moisture  condition  in  this sandy  basin,  the following  equation,  based on
the standard API, was  derived  to  enable adjustment of the CN values in the
Manatee River basin.
     S = 0.48 + 6.72/API
where
     S = average basin storage coefficient

In  the application  of the  adjustment  factor,  the average  basin storage
coefficient, based  on API,  is first computed  by  Equation 6  and  then the
basinwide  runoff curve  number  CN, based  on  the  API, computed by  the
standard SCS equation. A basinwide correction factor, which is  then applied
to  the runoff  curve  number  of  each  subbasin,  is then  computed by  the
following equation:

                 CN  = CNftpi-70

 Simulation  of  the August  28,  1981  event  using these equations  is shown  on
 Figure 18.

      Both of the above equations  have  been integrated into the DABRO Model
 and the Model  revamped {DABRO  1)  such that  inputs  to  the model  are now  only
 rainfall  increments  and API.

 ADVANTAGES  OF THE  DABRO MODEL

      1.  The  DABRO  Model  is a simple,  easy to use interactive model which
          instructs  the  user as to  parameter  input.
                                     77

-------
2.   The  model  has  been  verified by  application to  both  urban  and
     rural  basins.  However, like all models, careful determination  of
     subbasin  parameters  is necessary.

3.   The  model  takes  both the  total  impervious area  and  directly
     connected  impervious  area  into  consideration.

4.   The  unit  hydrograph  parameters  (time  to  peak and peak  flow  rate
     of the various subbasins) can  be  changed by the user  to  fit  the
     local  situation.

5.   The  program,  written in Microsoft  Basic  and not copy  protected,
     is very  easy to modify since  almost any engineer can  understand
     what the  various  lines in the  program do.  For  instance,  A=area,
     V=velocity,  Q=flow rate, etc.

6.   The  program  will  operate  on  any microcomputer with 64K's  of RAM.
     However,  with  a  64K  computer,  the  program  is limited  to  six
     subbasins.
Ul
o
K
10000
8000
6000
400O
2000











=t_










r
v —










_r-










[/
?









,








ULA
RES










TED
ERV


r^







INFl
OIR


=^:







LOW



m^T"
~— ,










^^»
^^.










-L
•*s










^w
••^









.
^^^
N ^









^-
"" V
^ «


Fi






g^






re






18













OBSERVED INFLOW
JVT RESERVOIR






-T_

->_



"0 10 30 30 40 K
                               TIME (hri)
                                78

-------
                           REFERENCES
1.   "SCS National  Engineering  Handbook, Section  4,  Hydrology", Soil
     Conservation Service, U. S. Department of Agriculture, Washington
     D.C., 1972.

2.   Golding, Bernard  L.,  Discussion  of  Paper  Titled  "Runoff Curve
     Numbers  with  Varying  Site  Moisture",  Journal  I  and   D Div.,
     Proceedings ASCE, Vol. 105, IR4, December 1979.

3.   Constant,  James  A.,   "A  Mathematical  Determination   of  the
     Ordinates  of  a Unit  Hydrograph",  Proceedings of  the  Seminar on
     Urban   Hydrology,    Hydrologic    Engineering   Center,    Davis,
     California, September 1970.

4.   Espey, W.  H.,  Jr.,  Altman, D. G.  and  Graves, C. G., Jr., "Nomo-
     graphs   for   Ten-Minute   Unit   Hydrographs   for   Small   Urban
     Watersheds",   ASCE   Urban  Water   Resources   Research  Program,
     Technical Memorandum No. 32,  ASCE New York,  N.Y., December 1977.

5.   Powell,  Robert  A.,  "Lake  Manatee  Reservoir  Regulation  Manual",
     Camp, Dresser and McKee, Inc., November 1983, 120 pp.
     The work  described in  this  paper  was  not  funded  by the  U.  S.
     Environmental  Protection Agency and therefore the contents do not
     necessarily  reflect  the  views  of  the  Agency  and  no  official
     endorsement should be inferred.
                                  79

-------
          APPLICATION OF A LOTUS SPREADSHEET FOR A SWMM PREPROCESSOR
                   by:   S. Wayne Miles and James  P.  Heaney
               Department of Environmental Engineering  Sciences
                            University of Florida
                          Gainesville, Florida  32611
                                   ABSTRACT

     This report describes a preprocessor for the EPA Stormwater Management
Model (SWMM) which is based on a Lotus 1-2-3 spreadsheet.   Currently used
methods to input data into SWMM are described and are compared to  the Lotus-
based preprocessor.   The advantages and disadvantages of Lotus  as  a pre-
processor host are discussed.  The internal structure of the  preprocessor
commands as written  in  Lotus' macro language is described  including the "Help"
command which  allows access to a number of reference files.   Future improve-
ments to the preprocessor are explained and conclusions as  to its  development
are made.


                                 INTRODUCTION

     Large mainframe computers have long been used to run hydrologic  models*
One of the most popular models,  the EPA Stormwater Management Model (SWMM),
has also been recently converted into a personal computer  version  available
from several sources.  This conversion has prompted  the use of  personal com-
puters in many other areas  of data preparation and handling in this field.

     In" order  to fully exploit the potential of the personal  computer,  soft-
ware packages  may be used which create a structured format to store and  handle
data.  This paper will describe one of these software packages, Lotus  1-2-3,
and how it is being used to process input data for  SWMM.   The following
sections in this paper will describe current methods of parameter  entry  into
SWMM and will  compare  them with a Lotus 1-2-3 based  preprocessor.   The  struc-
ture of each section of  the preprocessor will also be described.


                       PARAMETER ENTRY IN ORIGINAL SWMM

     The original method of parameter entry Into SWMM was very tedious.  The
model user had to enter all parameters into the input file in the  correct

                                     80

-------
format.  Each input card had  to be  entered Jn  the correct order and each
parameter  had to  be entered in the  correct space allotments within each card.
The  user was totally dependent on the written  documentation to determine the
correct format as is shown in  the excerpt  from the SWMM  User's Manual (1984)
in Figure  1.   It was easy to make errors  in entering values  and difficult  to
determine  where  format errors had been made.    Even though this method of
parameter  entry  is still being used,  it  is quickly becoming obsolete.


  Card             Card                                           Variable      Dfrfau.lt
  Group   Format   Columns            Description                    Mane-         Val»«


  HI         BF5.0    16-20    Width of subcatchment, ft finj. This tern   WCl)*        Sana
                             actually refers to the physical width.
                             of overland flow ia  the subcatchmeat and   ^3
                             may be obtained as illustrated in the text.

                     21-25    Area of sub.catchment, acres [ha].        WAHEAs*«*(2)*-  Hooe-

                     26-30    Percent imperviousneas of subcatchmeot, ZMH3f**         None-

                     31-35    Ground slope, ft/ft  [diaensionleas].     HSLQBE=WW{4>*- None-

                     36-40    Impervious area. 1                      WW(S)*'         Vane
                                           I Roughness factor,
                     41-45    Pervious area.  j  (Manning's n)         BW(6)*         N6ne>

                     46-50    Impervious area.)                      WSTORE=^W(7)* Vontf
                                           > Depression storage, is_
                     51-55    Pervious area.  I [mm).                 USTORE=WW(8)* Sotic-
             *** Horton equation parameters if INFILM = 0 (Card Bl) **

                     56-60    Maximum (initial) infiltration rate.      WUIAX-WWC*)*  None-
                             in./br [mm/hrj.

                     61-65    Minimum (asymptotic)  infiltration rate-,   WLMIJi=WW(10)* Notre-
                             in./hr [mm/hrj.

             F10.5    66-75    Decay rate of infiltration in Hortoa's    DECAT=WW(ll)* Hone-
                             equation,  I/sec.

             *** Greett-Ampt equation parameters if 1NTILM - 1 (Card Bl)  ***

             2F5.0    56-60    Capillary suction, inches [ma] of vater.   SCCT-WWC9)*'   Nane-

                     61-65    Hydraulic conductivity of soil, in./hr    HT3':ON=UV(lf?)"-None-
                             [tnm/hr].

             F10.5    66-75    Initial moisture deficit for soil,        SttDMAX=W( 11)*None-
                             volume air/volume voids.


  H2                         Blank card (except for identifier)  Co terminate-
                             subcatch/nent  cards:  one card.
      Figure 1.   Example of  parameter  format  description in. SWMM
                   User's Manual  (1984).
                                          81

-------
                      IMPROVED PARAMETER ENTRY INTO SWMM

     Improvements on the manual parameter entry method have been made.  In the
personal computer version of SWMM currently being marketed  by James and
Robinson (1985),  the user is prompted  to enter parameter cards in their cor-
rect order (see Figure  2.).  Prompts for the correct number of each card type
are also included according to previously  entered parameter values. The user
is still responsible, however, for positioning the parameter values within
each card.  The user's  manual  must be  consulted  for each c'ard entry as to the
order in which the parameters  are entered  on each card  and  the default values
for the parameter.  This program is helpful in that the user does not need to
enter correct spacing between each parameter on a card. The values need only
to be separated by commas.  If a card is entered erroneously, however, this
program  relies on outside editing procedures.   Once  the card entry process
begins, it must be completed.  Then the file may be  edited with a standard disk
operating system.  This program is clearly  an Improvement over the original
SWMM  input procedure.


                          MENU-DRIVEN  PARAMETER ENTRY

      Another approach  to this problem is used by the South Florida  Water
 Management District (SFWMD).   They have been distributing a'menu driven pro-
 gram which aids in performing calculations needed in their permitting pro-
 cedures.  The program  is menu driven  in that the user  is offered choices
 throughout  the program.   Initially, the user may choose from the Soil Con-
 servation Method, the  Santa Barbara Method, or  his own method (Mass Route) to
 perform runoff calculations (see  Figure 3). This program is easy to use and
 contains convenient  editing procedures.  The program does lack,  however,
 sufficient  documentation for  use  without  consulting  the reference manual.
            E3.  ENTER SWMM BLOCK TO BE CALLED
             RUNOFF

            SWMM RUNOFF BLOCK REQUESTED

            ENTRY MADE TO SWMM RUNOFF BLOCK INPUT

            Rl.  ENTER FIRST TITLE CARD (DATA GROUP Al)
           A1SFWMD REPORT EXAMPLE

            R2.  ENTER SECOND TITLE CARD (DATA GROUP Al)
           A1RUNOFF MODULE

            R3.  ENTER FIRST CONTROL CARD (DATA GROUP Bl)
           610,0,0,1,0,0,0,12,0,1,7,85,0,0,0

            R4.  ENTER SECOND CONTROL CARD  (DATA GROUP B2)
           B260,60,30,0,0,0

  Figure 2.  Example of parameter entry mode in PCSWMM (James and  Robinson,
             1985).

                                     82

-------
                   DESIRED STRUCTURE  IN A SWMM PREPROCESSOR

     In order to create a  working SWMM input processing program,  several ideas
must be addressed.  Firstly,  all of the problems  previously mentioned  with
current parameter entry methods must  be solved.   The program  should  include a
format with which the user may input  all  parameters directly into a  space
allotted specifically for  that parameter.   This would necessitate  a  program
with clear documentation.   The  user should  also have free  editing  capabilities
within the program.  The user should  be able to move back  and forth  between
cards while entering parameter values, and change any  previously  entered
values while continuing the current parameter entry mode.   Also the  choice  of
the values themselves is important.   The user  must  feel  comfortable  with the
choice of each value.  Therefore, a complete program should include  documen-
tation which aids the user in choosing parameter values.  If  calculations are
needed to choose any value, a format  for these to be performed should  be
included.   Lastly,  this program  should be able to operate  such that  the  begin-
ning user is  comfortably led  step by  step while the experienced user is  not
slowed by  the entry process.


                       OTHER PREPROCESSORS

     A good example of a model  preprocessing program was one  for  the CREAMS
Model  by Dennison  and James  (1985) using  dBASE III. They present  several ideas
which are important in the development of a working preprocessor  for a model:
1) It should  be flexible and  easy  for a beginning user  while  not  hindering
expert users.   2) It  should refer the user  to  the manual when additional


         TYPE  -

        1  - TO EXECUTE "SCS" PROGRAM
        2 - TO EXECUTE "SANTA BARBARA" PROGRAM
        3 - TO EXECUTE "MASS ROUTE" PROGRAM

          SELECT ONE -
        1

        *****  ENTER LEGEND INFORMATION *****

        ENTER  PROJECT NAME  (                           )
        SFWMD PROGRAM EXAMPLE
        ENTER  REVIEWERS NAME  (                           )
        W. E.  COYOTE
       ENTER  PROJECT  AREA (           .0000000 ACRES)
       999
       ENTER  GROUND STORAGE  (           .0000000 INCHES)

       ENTER  TERMINATION  DISCHARGE (          .0000000 CFS)
       10

Figure 3.   Sample from South  Florida  Water  Management  District
           runoff and flow routing program  with menu choices (SFWMD,  1983).

                                   83

-------
information is needed.  3) It should provide a convenient  edit mode.  All of
these ideas are necessary features for a complete preprocessor.

     A choice of calculation procedures for the user is an important aspect of
the South Florida Water Management  District's program (1985).  Their  program
generates and/or routes hydrographs.  The beginning  of  this  program  asks  the
user to choose from the Soil Conservation Service method,  the Santa  Barbara
Unit Hydrograph method, or an alternative method for generating runoff  hydro-
graphs.  This feature allows the user to generate hydrographs by whatever
method he  feels confident and comfortable.  An interesting addition  to  a
program such as this might be provisions for a "knowledge  base"  from  which the
chosen calculation procedure may extract information.  In  such  a program,
however, the user should be able to check  all intermediate calculations  and
assumptions without  having to review program  lines whenever possible.


               LOTUS  1-2-3 AS A PREPROCESSOR HOST

     The idea of using a personal computer  based system as a preprocessor for
the EPA Stormwater Management Model is examined in  this paper.   Whether this
system is  used with  a personal computer version of SWMM or interfaced with the
mainframe  version, the advantages of the personal desktop  computer for  param-
eter entry into the  model are many.  Among  the software packages which are
available  for the personal computer, however, no clear favorite has  been
generally  accepted to host a SWMM preprocessor.  The many software options
available  all have their own advantages and disadvantages  in terms of data
entry, calculation, and documentation  capabilities.   This  report will describe
the mechanics and basic layout of a SWMM preprocessor which is housed on the
Lotus  1-2-3  spreadsheet  software package (Lotus Development Corp.,  1984).

     The Lotus 1-2-3  package was chosen for its flexibility in all of the
areas  needed for developing  the preprocessor.  As a spreadsheet Lotus is,  in a
sense, a large electronic  notebook which has approximately 3250 pages, all  of
which  may  be randomly accessed at any  time.  This feature will allow the user
of the preprocessor free movement during the parameter entry mode and will
ease any editing of  old parameter values which  may  be  needed.  This  feature
will also  allow the  user to access reference files which may be needed during
the parameter entry  mode.

     Lotus 1-2-3 also has its own internal macro language  which allows data
manipulation to be automated.  Since the main feature of the preprocessor is
to format  the model  parameters and  input data such  that they are accepted by
SWMM,  Lotus' simple  but comprehensive  macro language is a convenient tool.
This macro language  allows  for  the  creation of menus which facilitate the
execution  of the macro programs.  These menus create an atmosphere in which
the user may logically and systematically generate a SWMM input file while
randomly accessing reference files, and finally, formatting the  parameters and
exporting  the complete input  file into a file to be read by SWMM.

     Another point of flexibility with the Lotus package is its interaction
with other types of  files in  the personal computer environment.  Lotus can
import other database and word processing files into its spreadsheet  format  as
well as exporting text and numbers  out of  the worksheet.  This feature pro-

                                      84

-------
 motes Lotus as a preprocessor to be interfaced  with other software packages
 and allows the option of postprocessing SWMM output with the use of Lotus'
 graphical and statistical capabilities.  Depending  on the  size of the SWMM
 output file, all or part of the file may be imported onto a Lotus  worksheet to
 be summarized, analyzed, and/or  graphed.

     Unfortunately, the Lotus spreadsheet  also  has a few drawbacks for use as
 a SWMM  preprocessor.  One area which causes problems in the parameter  format-
 ting procedure is Lotus1 policy for placing numbers in  a column.   For  the
 visual purpose of keeping a space between columns  of numbers, Lotrus  1-2-3 does
 not allow an entire column width to be  filled with the digits of a number.  A
 column with a width of  five,  therefore, can only hold a four digit number.  A
 five digit number entered  into this column will cause a row of  asterisks to be
 displayed across the cell width.  This does not cause problems while per-
 forming calculations within Lotus  since the cell will still contain the value
 entered; the value simply cannot be displayed.  This does  cause problems when
 exporting the SWMM file from  the Lotus  worksheet  into an  interface file as is
 done with the preprocessor.  Lotus  will export the display on the screen and,
 therefore, a row of asterisks instead of the parameter value.   It  is often
 necessary when entering SWMM  parameters to use the full number  of  digits
 allotted to a parameter by the SWMM read procedure.  This problem  of space
 allocation may be solved within Lotus by handling  the parameters as  text would
 be handled.  Lotus  1-2-3 allows any text entered into a cell to completely
 fill the cell and even extend beyond. Treating the parameter as text,  how-
 ever, could increase the processing time up to  a  few minutes for a large
 number of subcatchments and could also  permit a number  with too  many digits to
 be entered into a formatted input card  for  SWMM.

     The spreadsheet format also seems,  at  this time, to impose  certain size
 limitations on a preprocessor.  These limitations  will  not come  in terms of
 the number of cards (e.g. subcatchment)  which may  be entered into  a module.
 The number of rain gauges and number of data points on a rainfall  hyetograph
 may also be very extensive.  The  limitation is  encountered because of  the
 manner in which the parameters must be  formatted  with specific  space allot-
 ments on a. card.  These  allotments  are provided by stringing card  formats end
 to end along the top of the spreadsheet  and adjusting column widths to receive
 each parameter.   Cards  are  separated with  different range names.   Since the
 formatted cards span the entire width of the spreadsheet,  this procedure
 limits the spreadsheet to only process  one  module per file. This  in turn
 necessitates that each  module be run singularly and that input,  output,  and
 interface files  be  called and  manipulated manually.  Manual file control is
 not difficult  on  small  SWMM runs, but it can be cumbersome if larger runs
 employing several modules are performed.


                MECHANICS OF A LOTUS PREPROCESSOR

     The example file that  has been created is  a preprocessing package for a
 simplified version of the SWMM Runoff Module.  In  general, this  preprocessor
 creates  a structured work area in which  the user may, in a systematic  and
 orderly manner,  enter the parameters necessary  to  run  the Runoff  Module.  The
 sections and capabilities of this preprocessor are  summarized and  linked with
a menu which was produced using the Lotus  macro language.   The menu  allows the

                                      85

-------
user to easiy move from the parameter entry mode  to  the help/reference mode
and access a number  of help files.  The menu also includes selections to
execute the Lotus macro programs which will format the parameters into a
Runoff input file and  export this  formatted file  to an ASCII  file.   Once the
user has  loaded the  package onto a  Lotus spreadsheet,  the Introduction screen
as shown  in Figure  A will  appear.   By pressing the keys Alt M, a macro program
will begin which  will  display the  menu choices ("Create"  "Help" "Edit"
"Save"    "Print").  Each of these  choices will be  explained individually in
the following  sections.

Create
  The "Create" mode  allows the user to begin a new job file.  The  first page
of this selection will ask for input  such as job  initials and a job  number
which  will  later  be used  to name the formatted file created by  the pre-
processor.   Items such  as  the  number  of subcatchments  and number  of  gutters in
the simulation are  also requested  in order to allocate space within  the work-
sheet.  The user  will next  move to  the parameter  entry section of  the pre-
processor.   As can be  seen in the sample of this  section shown  in Figure 5,
                                                               KEH-J
(CF:EflIE| HELP EDIT  SAVE PRINT
 Creates a n»n input file for a Runoff Module
       Da      DX     DY     DI	Efl_
 42
 43
 44
                                             ED
     46
     47
     48
     49
     50
     5i
     F1
     JX.
     c-^
     JO
     54
     55
     56
     57
     59
                     -RUNOFF MODULE PREPROCESSOR—-
           This Lotus file aids in the creation of an input file which ii
    coapatible with the FCSUM Runoff Module.

           flt any tise,  press flit H to display the eenu choices.

           When entering data, use the arroH keys to uove the cursor into
    the appropriate cell.  Enter additional cards in successive colusns to
    the right of the previously entered cards.

           To begin a new file, press Alt il and select the CREfiTE aoife.
     60 j
     tl '
                                     CttD
                                        ULC
CAPS
        Figure  4.   Introduction screen with menu  choices  displayed.
                                        86

-------
the  parameter  value is entered just  to the right of a  brief description  of the
parameter.  The  SWMM parameter name  is also given at the far  left of the
screen.   In cases  where more than  one of a given card  type may be needed,  such
as with subcatchment  cards, multiple  card values  are entered  directly  to the
right  of the initial card  values.  If the brief  description of  the parameter
given  in this  section is not adequate to choose  a. value  for a  parameter  or if
calculations need  to be performed  to  obtain a value,  the  user  may access the
help/reference section through the main menu as described  in  the next  section.

Help
   At  any time during the  parameter entry mode,  the user  may display the main
menu by pressing Alt M,  and then choose "Help"  (see Figure  6).   The  user will
then be  prompted to enter  a variable  name whereupon   a help file  for that
variable will  be displayed, or to  enter "index"  whereupon the  entire help  file
index  will be  displayed.    From  the help index  the user  will have access to
help files for all parameters  which will include further  documentation,  calcu-
lation aids,  graphs, and any other material with which the  user may make a
logical  and defensible parameter estimate for a specific job  location. A
                  -SUBCATCHHENT DATA-
                   REPEAT GROUP HI  FOR EACH SUBCATCMENT
                   NAXIHUN OF 100 DIFFERENT SUBCATCHHENTB FDR SINGLE EVENT SWHH3,
                   ICRA1N=0, AND 30 FOR CONTINUOUS SHIM3, ICftfllH NOT = 0.

                   A BLANK LINE JS  NEEDED TO TERMINATE SUBCATCKHENT DATA (H2>
            GROUP ID                                                HI
            JK      Hyetograph nunber                                   1
            HAKEH   Subcatchient nuaber  <»x 100)                         10
            NGTO    Gutter or inlet  liianhole) nusber for drainage.            11
            Kim)   Width of subcatchnent, ft.                         5280
            WAREA   Area of subcatchoent, acres.                         640
            HH(3)   1 ioperviousness of  subcatchnent                    30.0
            HSLOPE  Ground slope, ft/ft.                             0.001
            XW.I5)   Iipervious area. Resistance factor,                  0.05
            nU(6)   Pervious area. (Manning's n)                         0.3
            SSTORE  Impervious area.                                 0.05
            WSTORE  Pervious area. Detention storage,  in.                 0.2
            WH(9)   Maxiiun  infiltration or capillary  suction                3
            NNtlO)  fliniaus  infiltration or hydraulic  conductivity          0.3
            HHtil)  Decay rate or initial aoisture deficit               0.0015


            Figure 5.   SWMM Runoff module  preprocessor, parameter
                         ent.ry mode.
                                         87

-------
documentation file may come  directly  from the manual as  is  the file shown in
Figure 7.   Once  the user is finished viewing any  specific help file,  pressing
return will  display  a menu offering the choices of returning to the parameter
entry mode or returning to the help index.   In returning to  the parameter
entry mode,  the cursor is  moved to the  parameter  for which  the last help
screen was viewed.

Edit
   The edit  mode  is  simply a short macro language program that takes  advantage
of Lotus'  capability to  move through the  worksheet.  This mode allows the user
to update an old file  which  has already been created on  the preprocessor, or
to go back to a previous card  on the present file and make  a change.   The user
is able to move directly to  any card  in the module by choosing edit and  moving
the cursor to a  card name  as  in  Figure 8.  A user  more  familiar with  Lotus and
the arrangement of SWMM cards  will most likely be able to use the Lotus "goto"
or "pageup/pagedown" keys  to  access  a  card  more quickly than with  this edit
mode.  The edit mode does, however, provide a  list of  the card types  in their
respective order for the less  experienced user to choose from.
        DT63:                                                       MENU
        CREATE (HELP! EDIT  SAVE  PRINT
        Provides assistance in choosing values to be entered.
              DS                       DT                        DU
        58 —	CONTROL PARAMETERS	
        59
        60 USE ALT M, HELP FOR MORE INFORMATION ON EACH VARIABLE
        61
        62        FIRST CONTROL GROUP (Bl)
        63
        64 GROUP ID                                               B1
        65 ICRAIN  Continuous SHMN parameter, 0=single event,  l-4=contin.      0
        66 rETRIC  Enter 0 for U.S.units,  1 for Hetric units                0
        67 1SKOH   0=no snon, l=single event SUCH, 2=continuous sno*         0
        68 HR3AG   * of hyetographs, iax=6, §ust be  1 for continuous SHMM     1
        69 IMFILM   Infiltration eq., 0 = Horton eq,  1 = Green-Aupt eq.        0
        70 KWALTY  Duality lor erosion) sinulated? 0 = no, 1  = yes           0
        71 WAP    Evaporation data, OsdefaultlO.l'/day), l=read group Fl     0
        72 KSF:    Hour to start stora (24 hr clock, iidnight = 0)           12
        73 Nr!K    ilinute of  hour  to start stor«                         °
        74 NMY    Day of *onth to start situation                        1
        75 f.ONTH   Month  to start  simulation                            7
        76 IYRSTR  Year to start simulation                            85
        77 IRPfiNT  Print  control paraneter, for  ICRAIN = 1 or 4 only         0
                                          CMD      CALC

           Figure 6.   Referencing  help file from parameter entry.

-------
Save
   The "Save" command, along  with the  "Print" command, do not initiate user
convenience modes as did previously described commands, but are formatting
procedures which must  be  performed  to record the latest version of a parameter
file.   The "Save" command  is the most time consuming  macro program in  its
execution since  it calls a  series of macro subroutines which  perform the
formatting procedure.  The  basic idea  of this procedure is to divide the
aforementioned  parameter  cards, which were  strung end  to  end across  the top of
the worksheet, and arrange  them vertically in their correct  order.  As  was
described previously, the parameter  cards were  strung  end to end so  that each
parameter could  be entered  in  its  own column whose  width had  been adjusted to
accept the correct parameter  format.   In order  to stack these parameter cards
while preserving the correct spacing between parameter values,  it was  decided
to combine all values on each card into  a single  worksheet cell.  The  task was
performed by printing each  range containing  a card  type separately into a
temporary ASCII  file, and then Importing these  files  back into the worksheet
in their proper  order while stacking them vertically.  This procedure  is
slightly tedious  and takes  about ninety  seconds for a simple  version of the
Runoff Module.  The  time to perform this  procedure will increase in proportion
        EH22;
        [RETURN j KIBEX
        Returns to variable sntry position.
              EN     EX     EY     EZ
        23
        24
        25
        26
        2?
        26
        29
        3<<
        31
        32
        33
        34
        35
        36
        37
        38
        39
        40
        41
                               FA
FB
FC
                                                    CND HEi.Li
FD
G3i5'*          ICRftIN          continuous Sltfil paraaeter
       continuous Stifltl paraaeter
=0             Single event SHHH, continuous StWN not used

tmmit       Values greater than zero indicate continuous 'SfMIt
=1             Hourly precipitation values read as card iaso-ss
              froB National Heather Service (NWS) tape. Input unit
              is JINU) for NHS tape.
=2             Proceesed hourly precipitation values (and temperature
              if ISNOH =2) are read froa unit NSCfiAT (2).  These
              values were generated and saved froi earlier run when
              ICRflHM or 4.
=3             Read precipitation values froa cards, using o/raups
              El and E2. Not useable Hith snoMielt, i.e., loNQK mist
              equal zero.
=4             Saie as  1CRAIN=1, except that program stops after
              processing precipitation (and temperature) data. The
              onlyy RUNOFF Black input paraneters required are those
              needed for this processing. Input ceases aft er Group
              Dl.
                                        CfiLC
         Figure 7.   Example help file  from manual with  return menu,
                                         89

-------
to the number of card types used, but not to the number of each card type
used.   For  example,  adding erosion  cards  and  snow cards  to  a  Runoff  file would
increase the run time by a few seconds each while adding an additional fifty
subcatchment cards would add a minimal amount of run time.

Print
    While  the "Save" command is the workhorse of the preprocessor, the "Print"
command takes all of the credit.   Once the formatted listing of cards has  been
reviewed on the worksheet  as seen  in  Figure  9, the user may  invoke the "Print"
command which will print the formatted-cards from the worksheet into an ASCII
file which may be accepted by a personal computer version of the Runoff Module
as input.   As mentioned  earlier,  the input  file will be  named using  the job
initials and job number entered at the beginning of the "Create" mode.  Here
again, however, a more experienced user may perform this print step  manually
on Lotus and provide another choice of file name.
              STEPS TOWARD A COMPLETE PREPROCESSOR

     The preprocessor which has been completed does not  yet contain all  cards
which may be run with the Runoff Module.  Though  it can only format the param-
eters for a simple  problem, it does contain all of  the  basic components
                                                              READ*
                    DX
DY
D2
EA
EB
EC
ED
2"?
23
24
25
26
27
25
29
30
31
32
33
T J
OT
35
36
37
38
39
40
4!

***f*i»****i*»»r»«»ft*t EDIT NODE

HOVE CURSOR TO li'SBE OF CARD


INTERFACE
SCRATCH
MODULE
CONTROL ONE
CONTROL TKD
GENERAL SNQM
MONTHLY HIND
IHPERV DEPLETION
PERV DEPLETION
AIR TEMPERATURES
CONTINUOUS DATA
RAINFALL



»tt**t*tt*f »**•>* H*m Httmt

YOU WISH TO EDIT AND PRESS RETURN.


RAINFALL HYETO
EVAPORATION
GUTTER PIPE
SL'BCATCH DATA
SUSCATCH SiiOH
GENERAL SL'ALITY
LAND USE
CONSTITUENT
EROSION CROUPS
SUBCATCH SURF
BUTTER PRINT



          Figure 8.  List of card types in Edit mode.
                                     90

-------
needed in a preprocessor of this type.  Once all of the card  types  have been
introduced into the Runoff preprocessor, the only major  step  is to  write a
macro language program to delete cards that are not necessary for any partic-
ular run.  Lotus provides all  of the logistical functions which  will  be needed
for this task and no problems are  foreseen at  this  point  in time.

     The next step is to create a preprocessor for each SWMM  module.   Here
again returns the limitation that only one  module will be able  to be  intro-
duced on a single worksheet preprocessor.  Each preprocessor, however,  will
have the same basic skeleton as'the Runoff Module preprocessor  and  should
require less effort in its  creation.  Other  improvements in  this type of
preprocessor may include providing a copy of  the entire SWMM  manual to be
referenced through the help index.   This  would in theory free the user  from
leaving the personal computer to access information.  The eventual  interfacing
of this preprocessor with an "expert system" would further  this  idea  by pro-
viding the user  with calculation alternatives  to SWMM which may  feed  from a
common "knowledge base" of Information.  This "knowledge base" may  be  thought
of as a general  source of data from which any model could extract necessary
          ft99:                	
          CREATE HELP EDIT  SflVE IPRINT 1
          Gives a printout of the loraatted Runoff input file cards
              ft   B  C  D  E  F  G  H   I   J  K  L  H  N
                                                             CUD H£«U
0  P  Q
99
100
101
102
103
104
105
106
107
108
109
no
111
112
113
114
115
116
U7
118

0 0
1 2
RUNOFF
ftlSFHMD
ft2RUNOFF
Bl 0 0
D2
El
E2 0.
E2 0,
0.
61
HI 1
0
HI 1
H2 11

0
0


0
0


0
0


0



0 0



0



0



0



0



000000


EXAMPLE
MODULE
0
60
24
11 0.
54 4.
12 0.
0
10
0
1
0
1
60
60
11 0.
28 1.
12 0.
0
0
30

13
09
12
0
11 5280
0
0
0
0

0
0
0

0.13
0.48
0.12

640
0
0
0
0
0

0.14 0
0.3
0
0
30.00.
0
0
0
12
0

.16 0
0.3 0.
0
0
001 0.
0
0
0
0


.2
18
0

05
0
0
0
1


0,2
0.18
0
0
0.3
0
7


0.27
0.18
0

0.05
0
0
0
0
85 0 0 0


0.34
0.18
0
000
0.2 3 0.3 0.0
0 O 0
0000
0000
EHDPRQGR






















              Figure 9.  Formatted version  of  parameter cards.
                                     91

-------
information for its simulation.  Random access to different methods of  param-
eter calculation and years of data history through the help index would all
lead toward rendering the user independent of conventional referencing methods.
                            CONCLUSIONS

 1.   The Lotus 1-2-3 software package is a suitable host for an input
 file/reference file preprocessing package that can be interfaced  with the EPA
 Stormwater Management Model (SWMM).

 2.   The Runoff Module preprocessor which has been created is a simplified
 package, but it reflects well a basic  skeleton which may be used  to create
 preprocessing packages for the remaining modules of SWMM.  No technical dif-
 ficulties In creating these packages on Lotus spreadsheets are foreseen.

 3.   The preprocessor idea may be the first step toward rendering  the personal
 computer user  independent from conventional  referencing  techniques.

 4.   The "Create",  "Help",  and  "Edit" modes  of this preprocessor are suf-
 ficiently documented for easy use by a beginner with a  minimal knowledge of
 Lotus  1-2-3 and SWMM, but do  not  hinder  their use by an  expert.
                            REFERENCES
 1.  Dennlson, K.D. and  James,  W.,  1985, A  Database  Environment  and Sensitivity
 Framework For a Continuous Water Quality Models, In:  Proceedings of Confer-
 ence on Stormwater and Water Quality Management Modeling, McMaster University,
 Hamilton, Ontario.

 2.  Huber, W.C., Heaney,  J.P., Nix, S.J.,  Dickinson, R.E. and Polmann, D.J.,
 1981, "Stormwater Management  Model  User's Manual,  Version III," EPA-600/2-84-
 109a (NTIS PB84-198423), Environmental Protection  Agency, Cincinnati, OH.

 3.  James,  W. and Robinson, M.,  1985,  PCSWMM3.2 User's  Manual,  Hamilton,
 Ontario.

 4.  Lotus Development  Corporation, 1984, Lotus  1-2-3 User's  Manual,  Release
 1A, Cambridge, MA.

 5.  South Florida  Water Management District, 1983,  "Permitting  Information
 Manual - Volume IV, Management and Storage of Surface  Waters," SFWMD,  West
 Palm Beach,  FL.

 6.  South Florida  Water Management District, 1985,  Program for Generating
 and/or Routing Runoff Hydrographs, SFWMD,  West  Palm  Beach, FL.

  The work described in this paper was not funded by the U.S. Environmental
  Protection Agency and therefore does not necessarily reflect the views of
  the Agency and no official endorsement should be inferred.

                                       92

-------
          IMPACT OF EXTENSIVE IRRIGATION PUMPAGE ON STREAMFLOW BY HSPF

                                       By

                                 Ananta K.  Nath
                     Nebraska Natural Resources Commission
                            Lincoln,  Nebraska  68509


                                    ABSTRACT

     The hydrology of a 2,700 square  mile area of the Big Blue River Basin in
central Nebraska was simulated by the continuous process hydrologic simulation
program HSPF to investigate the impacts of extensive groundwater pumpage for
irrigation on streamflows.   A long-term continuous simulation for a 23-year
period (1953-1975) was carried out by modeling water movement and storage
characteristics through the use of the PWATER section of PERLND module and HYDR
section of the RCHRES module of HSPF.  The simulation indicated that surface
runoff and interflow from irrigated lands increase initially.  Actual
evapotranspiration also increases.  As pumping continues and the aquifer water
level drops, groundwater contribution to streamflow decreases.  Finally, a
stable condition is achieved where the water table ceases to reach the stream
bed and measured streamflow is less than before pumping began.  It is concluded
that HSPF can be used with  a certain  degree of success to simulate
stream-aquifer interaction  and to assess possible effects of deep pumping on
surface water hydrology in  a Nebraska basin.  The accuracy of the results,
however, is limited by modeling complications because some algorithims of HSPF
were not specifically designed to simulate withdrawal of groundwater by pumping.
Coordination of some HSPF input parameters from groundwater modeling may produce
better results.


INTRODUCTION

     Development of irrigation using  groundwater started growing rapidly in the
1950's and expanded very rapidly in the 1970's in Nebraska.  By 1980, more than
73 percent of water used for irrigation in the state was pumped from wells.
This development, however,  has not been without consequences.  The impact of
extensive groundwater pumpage for irrigation was noticeable in the flow regimen
of several streams, including the Big Blue River basin in central Nebraska.  As
a part of the Nebraska Natural Resources Commission's (NRC) "Problem Analysis
and Area Planning" activity for the Big Blue River Basin, an attempt was made to
investigate the impacts of  groundwater pumpage on the streamflows of the Big
Blue River and its tributaries using HSPF.  This paper summarizes the
methodologies used in simulating by HSPF the effect of the time history of


                                       93

-------
groundwater pumpage on the hydrology of the basin, arid the results obtained from
the simulation.  It also evaluates the applicability of HSPF in simulating deep
groundwater pumpage and resulting groundwater-surface water interrelationships.

DESCRIPTION OF THE  STUDY  AREA

      The  Big Blue River Basin is  located  in south-central  Nebraska and  northeast
Kansas, and is a part  of  Kansas  River  Basin.   A map  of  the basin  with principal
tributaries is shown  in Figure 1.   The area of the Nebraska portion of  the  basin
totals about 4,600  square miles.   The  topography ranges  from  loess plains with
thick deposits of silt, sand, and  gravel,  to rolling hills and  stream valleys.

      The  principal  aquifer consists  of unconsolidated sand and  gravel layers  of
Pleistocene age.  These layers belong  to  three formations  composed of sand  and
gravel graded upward  from coarser  to finer material.  Soils are of loess origin.
About sixty percent of the soils  in  the basin are of irrigation suitability
type  A {i.e., with  slight limitations  only).

      Land use is primarily agricultural.   About 83 percent of agricultural  lands
are croplands.   Since  the early  1950's the number of irrigated  acres using
groundwater has increased every  year.   By 1980, approximately 1,061,000 acres in
the basin were irrigated.   The greatest concentration of  irrigation wells is  in
four  counties in the  upper part  of the basin.   This  intensive groundwater
development,  coupled  with occassional  drought conditions,  has caused widespread
water level declines.  Water  levels  have  declined as much  as 30 feet in several
areas.

      Upper reaches  of  the streams  are  characterized  by  small meandering channels
with  intermittent flows.   The streamflow  is variable, being primarily derived
from  precipitation  runoff.  Nearly all of the tributaries  are intermittent.
Even  the  main stem  of  the Big Blue River  does not become a perennial stream
until it  drains about  450 square miles of the upper  part of the basin.  Base
flows in  the streams  are,  however,  relatively low.   Return flows  from increased
well  irrigation have  lengthened  the  period of time when  flows are present in
many  of the streams.
SELECTION OF  SIMULATION MODEL

     As a part of  the  problem  analysis  and  area  planning  study of  the basin
water supply, a  two-dimensional  finite  difference  groundwater model covering
most of the basin  was  developed.  A  large data base  including irrigated  lands
identified by remote-sensing,  geologic  data,  and groundwater conditions  was
generated for development  of the  groundwater  model.  However, since the
groundwater model  could not accurately  represent the effects of  the continuous
changes on the surface water hydrology,  it  appeared  that  a  long-term continuous
model was needed to simulate the  hydrologic-hydraulic behavior of  the stream
aquifer system prior to and during the  period of extensive  irrigation pumpage.
A search was made  for  an integrated  and  comprehensive surface water hydrology
model that would allow comparison of water movement, storage characteristics and
hydrographs for conditions before and after development of  irrigation.   Based
upon these requirements, the program package  Hydrologic Simulation

                                     94

-------
IO
en
                                                                         FIGURE 1

                                                                        BIG  BLUE

                                                                       RIVER BASIN

                                                           BASIN  STREAM NETWORK
                                                                           SCALE

-------
Program-Fortran (HSPF), originally developed by Hydrocomp, Inc. and revised by
Anderson-Nichols and Company for the EPA's Environmental Research Laboratory,
was chosen for this project.  The project was started with Release 5.0, and
later updated with Release 7.0.


DATA BASE DEVELOPMENT

      Data for time series storage  management of HSPF consisted of meteorological
data, land data,  channel-flood plain data,  streamflow data,  riverine  structure
data and groundwater pumpage data.   Meteorological data consisted of
precipitation,  air temperature, dew point,  solar radiation and wind movement.
Precipitation data were available  at five hourly and 21 daily recording
stations.  Daily  maximum-minimum temperature data were available at  three
stations.  Daily  evaporation,  solar radiation,  dew point and wind movement data
were available at one station each.   Daily precipitation and temperature  records
were disaggregated to hourly values.  Daily groundwater pumpage data  were
incorporated into the time series  storage from the input data of the  groundwater
model already developed for the basin.  In the groundwater model, the seasonal
Crop Irrigation Requirements (CIR)  for various crops and soil types  in each node
were first computed by the Jensen-Haise methodology.  The CIR was then adjusted
for irrigation efficiency to estimate the seasonal pumpage and distributed over
 the irrigation months.  The acre inches per acre of pumpage from the  GW model
nodes were then disaggregated to daily values over the PERLND segments of the
HSPF model.

      Daily streamflow data for calibration were available at the following USGS
gaging stations:

       1)  Big Blue River at Surprise (April 1964 - present)
       2)  Lincoln Creek near Seward (October 1953 - present)
       3)  Big Blue at Seward (October 1953 - present)
       4)  West Fork Big Blue near  Dorchester (October 1958 - present)
       5)  Big Blue at Crete (October 1954 - Present)

Pervious hydrologic land segments  were selected based on the unique  combination
of meteorological characteristics  identified by a Thiessen polygon network of
weather stations  and land characteristics by soil type, topography and cover.   A
 total of 38 pervious hydrologic land segments were delineated for simulation
 (Figure 2).


SIMULATION ALGORITHMS

      The hydrologic simulation of the basin was conducted by the utilization of
 the PWATER section of the PERLND" module.  Water movement in the PERLND module
was modeled along three flow paths:  overland flow, interflow and groundwater
 flow.  Storage processes that occur on the land surface and in the soil horizons
were simulated in six storage blocks: Interception Storage (CEPS), Surface
Storage (SURS), Interflow Storage  (IFWS), Upper Zone Storage (UZS),  Lower Zone
Storage (LZS),  and Active Groundwater Storage (AGWS).
                                      96

-------
Figure 2
            BIG  BLUZ RIVER BASIN
                                                      HYDAOLOOIC LAND  SEGMENTS
                                                       IHCOftPORATED IM  THE  Bl6 «l_UE
                                                           WATCH SUPPLY  UOOCi.
                                                                  LE6ENO
                                                          WF3 ML I lOtHTVTIMI
                                                            A imtUIFKW MflE
                                                              (CM.IMATIOH
                                   \
                       C    »	«  itil< 1»»HMTO   n    M

-------
     The processes of infiltration and overland flow are interdependent and
occur simultaneously.  Water in surface detention will later infiltrate and
appear as interflow or it can be contained in the upper zone storage.  Water
infiltrating through the surface and percolating from the upper zone storage to
the lower zone storage may flow to active groundwater storage, or may be lost by
deep percolation to inactive groundwater storage.  Active groundwater eventually
reappears as baseflow, but the deep percolation to-inactive groundwater storage
is not accounted for in the simulation.  Lateral external inflows to interflow
and active groundwater storage were also considered in simulation.
Evapotranspiration is simulated in all phases of the storages associated with
the process, i.e. from interception storage, upper and lower zone storages,
active groundwater storage, and directly from baseflow.

     Hydraulic simulation for routing runoff from land surface and discharge
from groundwater though the stream system was performed by utilization of the
HYDR section of the RCHRES module.  Simulation reaches were identified based on
homogeneity of cross-sectional shape, channel slope and channel-floodplain
roughness coefficients.  Detailed surveyed data on channel-floodplain cross
sections and riverine structures on the main stem of Big Blue and West Fork Big
Blue were available from the Natural Resources Commission Flood Plain Studies.
Stage-discharge information computed by HEC-2 program in NRC Flood Plain studies
were also utilized in generating FTABLES.

     In applying the model on this particular project, the basic fluxes without
and with the dynamic effect of groundwater pumpage on streamflow are illustrated
in Figure 3.  Streamflow comes from surface runoff and interflow, controlled by
soil moisture storage, and from groundwater flow.  For the conditions prior to
extensive irrigation development, the solid line linkages of Figure 3 represent
the hydrologic processes that existed in the basin.  When irrigation pumpage was
introduced, another flow path was added, which is shown as the dashed line of
Figure 3.  This flow path can be represented by various ways in HSPF, but the
dashed line in Figure 3 seemed to be a good approximation of actual conditions.
The layout of a typical pervious land segment incorporated into the Big Blue
model to represent the interrelationship of surface and groundwater flow
components under groundwater pumpage condition is illustrated in Figure 4.


METHOD OF  INVESTIGATION

     The primary strategy used in this  study to evaluate the  impact  of deep
groundwater pumpage on strearaflow was  to simulate  the  long-term
hydrologic-hydraulic behavior of  the basin  for the conditions prevailing prior
to the irrigation development conditions, and project  that  simulation  through
the period undergoing  irrigation development.  The hydrologic-hydraulic behavior
of the basin was then  simulated  for  the actual conditions prevailing in the
basin within the period of irrigation  development  by  incorporating  the
groundwater pumpage element  into  the model.  Superimposing  the series  of
hydrographs representing pre-irrigation conditions on  the series  representing
irrigation development conditions would show the continual  effect of streamflow
resulting  from the extensive withdrawal of  groundwater.
                                      98

-------
CALIBRATION

     Marty of the algorithms contained in the model are mathematical
approximations of complex natural phenomena.  Before the model could be used to
reliably simulate streamflow behavior under various conditions, it was necessary
to calibrate the model by comparing simulation results with measured historical
data.  If significant differences were found, parameters were adjusted to
calibrate the model to the specific natural and man-made features of the basin.
The model was calibrated for the conditions prevailing during two periods.
The first was prior to extensive groundwater pumpage, and the second was during
the period of the greatest irrigation development.

     Though the development of irrigation in the basin began to take place in
1948, total development was minor compared to that in the late 1960's and the
1970's.   Also, available continuous streamflow records for almost all of the
gages in the Big Blue basin do not start before water year 1954,  preventing
calibration of the model prior to 1954.   It was, therefore,  decided to use
several periods between 1954 and 1960 for pre-irrigation development conditions,
and periods between 1960 and 1975 for irrigation development conditions to
calibrate the models of different stream networks, depending on the avilability
of measured streamflow records for that  stream.   The stream segment for the Big
Blue at Surprise, however, could not be  calibrated for pre-irrigation
development conditions as recorded streamflow data at this gage are available
from only after October 1964.   The months for which the model was calibrated are
as follows:
     1.    Big Blue at Surprise:
     2.    Big Blue at Seward:
            October 1966-September 1969
            October 1953-September 1955 and
            October 1966-September 1968
           (U
           co
           CO
           Q)
           4-1

           I
           o
  Precipitation
                                             Evaporation
                           Soil Moisture
                            -^-Surface Runoff
                                 Interflow
                                     Infiltration
Groundwater Storage
Groundwater
 to Stream
                                     Deep Percolation
                         Deep Groundwater
Figure 3 - Basic fluxes in HSPF simulation with groundwater pumpage.
                                      99

-------
                                      Figure
                                             PBEOP
                      LAYOUT FOR A TYPICAL PERVIOUS LAND SEGMENT
                     INCORPORATED IN THE HSPF MODEL FOR THE BIG BLUE
       3.   Lincoln Creek at Seward:

       4.   Big Blue near Crete:

       5.   West Fork Big Blue:
October 1953-Septeraber 1955 and
October 1964-September 1967
October 1953-September 1955 and
October 1964-September 1967
October 1958-September 1959 and
October 1964-September 1967
     The first step  in  the  calibration procedure was to match the annual flow
volumes between simulated and  recorded flows.   Parameters related to the storage
capacities of the unsaturated  soil  zones,  evapotranspiration and infiltration
are important determinants  of  runoff and  volumes.   Guidance for initial values
of such parameters as lower zone  moisture storage (LZSN), upper zone nominal
storage (UZSN), infiltration (INFILT), lower zone evapotranspiration (LZET),
interflow  (INTFW), and  groundwater  recession (AGWRC) were taken from the
suggested values in  the Agricultural Runoff Model Users Manual (Donigian et al,
1978).  Some guidance on parameter  values  was  also taken from a concurrent HSPF
project on the Dee Creek Watershed  in southeastern Nebraska conducted by the
Water Resources Research Center of  the University of Nebraska (NWRC, 1982).  The
UZSN, INFILT, and LZET  parameters were varied  monthly.   In calibrating
parameters for annual flow  volumes,  adjustments were made in LZSN and INFILT.
                                      100

-------
Since LZSN and INFILT indirectly affect the actual evapotranspiration rates,
decreasing LZSN and INFILT makes less water available for ET losses and
increases runoff.  In simulating the daily and seasonal low flow volumes,
adjustments were primarily made in INFILT and AGWRC parameters.  AGWRC controls
groundwater outflows, while INFILT controls inflow to the groundwater.  However,
care needs to be taken in adjusting INFILT at this step.  Since INFILT is
indirectly related to LZSN, significant change of INFILT affects the long term
flow volumes calibrated in the previous step.  Consequently, most of the
adjustment needs to be in the AGWRC term.  Once the annual flow volumes and
seasonal low flows are adequately calibrated, the shapes of the hydrographs were
adjusted by varying the interflow inflow parameter (INTFW).   Some of the salient
calibration parameters for each of the stream segments simulated are summarized
in Table 1.

     In calibrating the model for the two periods with and without irrigation
development, consideration was given to the change in evapotranspiration and
interception rates from irrigated and nonirrigated lands.   ET is affected
considerably by the type of plant and plant density.   Also,  loss by interception
depends on the extent of vegetal cover and density of prevailing plants and
trees.   It was, therefore, necessary to numerically represent the difference in
monthly INTERCPT and LZETPARM between irrigated and nonirrigated conditions.
The relationship between water loss and crop yield can range from linear to
curvilinear response functions (ASAE, 1980).  A 1:1 correlation of ET loss to
crop yield was assumed based on a linear relationship of dry matter yield to
relative seasonal evapotranspiration.  Since corn is the major irrigated crop,
the value of MON-LZETPARM and MON-INTERCEP for the months  of June through
September under irrigated conditions were generated by prorating the ratio of
irrigated crop yield to nonirrigated crop yield.   The results of a typical
calibration run are illustrated in Figure 5.
VERIFICATION

     Verification of the hydrologic calibration was carried out by running the
models of each stream for the period October 1953 through September, 1960 for
the pre-irrigation development conditions, and for the period October 1960
through September 1975 for the irrigation development conditions.  Figure 6
illustrates graphically the results of typical verification runs for a simulated
stream.
RESULTS OF THE CALIBRATION AND VERIFICATION PROCESS

     The calibration process was particularly valuable in assigning values to
prominent land parameters which were seen to be dependent upon soil type, land
cover and on regional meteorological characteristics.  The parameter sensitivity
runs indicate that simulated flows were most sensitive to the values of LZSN,
UZSN, and INFILT.  In certain cases, the impacts of the monthly values of UZSN,
MON-LZETPARM, and MON-INTERCEP on monthly distribution of runoff were more
pronounced.   In all the streams simulated, the first attempt to calibrate
included periods of high runoff extending over a period of about three months,
it was then extended to include longer periods of up to three years.  The

                                     101

-------
 simulated monthly runoff volumes varied from the recorded*volumes by 10 to 20
 percent for the wet months and by 10 to 50 percent for the drier months.  For
 the extended three-year calibration periods, the simulated annual runoff volumes
 varied from 7 to 15 percent of the recorded volumes.   The correlations between
 measured and simulated flows were evaluated by plotting double mass curves of
 cumulative runoff volumes and comparing them with a line of one-to-one
 correspondence.  A typical double mass curve of measured and simulated flows is
 illustrated in Figure 7.  The noticeable difference between recorded and
 simulated annual flows resulted from relatively poor simulation of a few months
 in which measured flows in the drier months were lower than simulated flows.
 This is particularly evident during the initial months of simulation in which
 the parameters representing the initial conditions may be responsible for poor

                                    TABLE 1
                    SUMMARY OF CALIBRATED PWATER PARAMETERS
PARAMETER
LZSN
LZS
INFILT
IRC
INFILD
AGWRC
DEEPFR
AGWETP
IFWS
SURS
UZS
UZSN*
INTERCEP*
LZETP*
Big Blue
at Surprise
8.0
4.0
0.02
0.60
2.0
0.99
0.19
0.5
0.01
0.10
4.0
0.13-0.32
0.03-0.30
0.10-0.70
Lincoln Cr.
at Seward
6.0
3.5
0.01
0.60
2.0
0.999
0.19
0.5
0.01
0.10
4.0
0.13-0.29
0.03-0.30
0.05-0.42
Big Blue
at Seward
6.0
3.5
0.01
0.60
2.0
0.999
0.19
0.5
0.01
0.10
4.0
0.13-0.30
0.03-0.30
0.05-0.42
West Fork Big Blue
at Dorchester at Crete
6.0
3.5
0.015
0.40
2.0
0.999
0.15
0.35
0.01
0.10
4.0
0.05-0.15
0.03-0.20
0.03-0.30
4.0
6.0
0.01
0.20
2.0
0.99
0.10
0.30
0.01
0.10
4.0
0.19-0.29
0.01-0.30
0.03-0.70
* Varied monthly, only ranges shown in table.

LZSN - Lower zone nominal storage, inches
LZS = Initial lower zone storage, inches
INFILT = Mean infiltration rate index, in/hr
IRC = Interflow recession rate, per day
INFILD = Ratio of maximum to mininum infiltration rate
AGWRC = Active groundwater recession rate, per day
DEEPER = Percentage of groundwater to deep aquifer
AGWETP = Fraction of ET from active groundwater storage
IFWS - Initial interflow storage, inches
SURS = Initial surface detention storage, inches
UZS = Initial upper zone storage, inches
UZSN = Upper zone nominal storage, inches
INTERCEP = Interception storage capacity, inches
LZETP = Lower zone ET parameter

                                      102

-------
simulation.  In certain cases readjustments of interflow parameters and
roughness coefficients resulted in closer runoff volumes and improved timing of
peaks.  The difference between observed and simulated flows in verification runs
was somewhat greater than in calibration runs because in calibration greater
emphasis was given to higher than normal runoff events, while the verification
period included both high and low periods of flows.   Consequently, the
calibrated parameters were somewhat biased toward high and moderate flow
conditions.

     The timing of snowmelt events in March and April was not well represented.
This is probably due to both the use of several precipitation data series with
data synthesized from other stations to substitute for missing records, and the
inability of the model to accurately represent frozen ground conditions.  The
INFILT and UZSN parameters were reduced during the winter months, and the snow
input (SNOWCF-parameter) was increased to match the volume of measured surface
runoff.  Additional research and model development work is needed to better
understand and represent frozen ground conditions and the timing and volume of
snowmelt runoff.

     Some erractic behavior of the model in simulating runoff for some major
runoff producing storm events is primarily attributed to the characteristics of
those storm events.  Review of precipitation data from several stations revealed
the frequent occurrence of scattered convective storms.  These convective storms

                                     Figure 5

                 RECORDED AND  SIMULATED  HYDROGRAPHS
                     FOR  BIG  BLUE  RIVER AT  SEWARD
                         PRE-IRRICATIOM DEVELOPMENT CONDITIONS CIOSO
          259 B


          22SO


          2080
      CO
      LL.
      O
          use
      CO

      Q   1888


      ^   Tso


          580


          250
         RECORDED •
            SIMULATED •
                                       JlllllllMlllllltl^J/lVl M
18  IS  28  25  30  5

   JUNE
                                     10  IS  2Q  25  SB

                                         JULY
10  IS  20  25  30

    AUGUST
                                      103

-------
                                Figure 6

                 RECORDED  AND  SIMULATED  ANNUAL
                         RUNOFF VOLUMES  FOR
                             BIG BLUE RIVER AT CRETE
                                                             RECORDED •

                                                             SIMULATEDO
          S4 EE  EB  67 68 60  68  61 62 63  64  65 OB 67 88  OB  70 71  72  73  74 75
                                                                     * 4
                                UATER YEAR
result in very nonuniform distribution of rainfall over a relatively  small area.
These types of storms, which produce localized flooding,  are  numerous  in  the Big
Blue basin.  For such storms, adequate representation of the  rainfall-runoff
relationship was not achieved because there were not enough weather stations in
the model area to accurately measure the rainfall.  Correcting  this deficiency
would require more precipitation gages in order to better represent and
understand the rainfall-runoff relationship.

     In simulating deep aquifer pumpage, the available algorithms  of  HSPF can
not directly add or subtract specific volumes of water to or  from  a deep
aquifer.  This is accounted for by the introduction of the parameter  DEEPFR, the
fraction of groundwater lost to deep aquifer.  The value of DEEPFR was
arbitrarily lowered for the pre-irrigation development condition without  the
availability of any physically interpreted value.   Secondly,  the surface
water-groundwater interaction in HSPF is accounted for by the use  of  a time
series of continuous lateral outflow from active groundwater  storage.   A  time
series of active groundwater inflow AGWI was used, but the impact  of  external
lateral groundwater inflow AGWLI,  which could also have continuous infow,  was
not accounted for.  A percentage of the pumpage rate could have been  used  in the
time series to represent the possible increased leakage of water into  the  deep
aquifer.  Consideration should have been given to linking these values of  AGWLI
and DEEPFR with output from the groundwater model.

                                     104

-------
                                 Figure 7
             DOUBLE  MASS  CURVE OF  RECORDED AND
                SIMULATED  ANNUAL  FLOW VOLUMES
      L_
      O
       UJ
       r-
       H
       CO
           i eee
           oee
           sea
           700
           see
           see
           400
           300
           200
            iee
                            LINCOLN CREEK AT SEWARD
                  100  268  300   400  6B8   000  70O   BOO  800   1000

                       RECORDED FLOW  CCFS>
OBSERVATIONS
     Groundwater pumping has indeed had a significant effect on the baseflow
contributions to the streams in  the basin.   As pumping caused the water table to
decline, it decreased the former gradient towards the stream which in turn
decreased the discharge of the aquifer to the stream, or it reversed the water
table gradient between the aquifer and the stream, which induced streamflow to
seep to the underlying aquifer.  These effects did not instantaneously reach the
streams, but rather  lagged behind the operation of pumpage depending upon
aquifer properties and distance  from the wellfield to the stream.
                                  105

-------
     The trend observed in the simulated hydrograph series as illustrated in
Figure 8 shows that the effect of irrigation pumpage was to increase surface
runoff and interflow from irrigated lands initially.  Actual evapotranspiration
also increased.  Initially, the effect of pumping is a net increase in the
quantity of streamflow due to surface runoff contribution from irrigated water.
This is apparent, as indicated by the early stages of the hydrographs, in almost
all of the streams.  As pumping continues and the aquifer water level drops, the
deep percolation from the active groundwater storage that feeds streams
increases and no longer contributes to streamflow.  Finally, a new steady state
balance is achieved where measured streamflow is less than before pumping began.
For instance, in Lincoln Creek, as illustrated in Figure 8, the wasted runoff
from irrigated lands increased the simulated average flows in the months
of August and September in 1970 by approximately 5 cfs.  However, the reduction
in baseflow contribution due to aquifer level draw-down decreased the average
monthly flow by 14 cfs in October and by 2 cfs in November.
CONCLUSION

     An attempt was made in this project to test the applicability of HSPF in
investigating the impacts on streamflow from deep groundwater pumpage and
provide some informative insights and conclusions regarding modeling of surface
and groundwater interactions.  Certain phases of the calibration and
verification process in this project, however, could not produce the desired
results.  Therefore, suggested improvements need to be discussed to provide
guidance for future use or development of the model.  Some of the specific
conclusions derived from this study are:

      (a)  HSPF can be  used with  a certain degree of success to  simulate
stream-aquifer interaction and possible effects of deep pumping on surface water
hydrology in a Nebraska river basin.  The accuracy of the results, however,  is
somewhat limited by the modeling complication because certain algorithms used in
HSPF  are not specificaly designed to  simulate withdrawal from groundwater by
pumping.

      (b)  Since HSPF can not directly add or subtract specific  volumes of water
to or  from  the deep aquifer, accurate estimation of the parameters such as
DEEPFR, IGWI, AGWI, and AGWS, that play dominant roles in accounting  for storage
and movement of groundwater, is  very  important in successful simulation of
groundwater pumpage impacts.  If possible,  the values of such parameters should
be estimated from, or  determined by,  a groundwater model.

      (c)  Meteorological time series  data,  particularly precipitation, is
critical to effective  continuous hydrologic simulation.  A denser network of
precipitation records  is necessary for accurate representation  of the
hydrometeorologic condition of a basin.  Lacking this data, it  is better to
compare duration and frequency information  from the overall simulation rather
than data on an individual storm event.

      (d)  Equally important is the adjustment of infiltration parameters during
winter and early spring to represent  frozen ground conditions for simulation of
                                     106

-------
o
_l
LL.
7S



70 h



6S  -



68 |-



55



50



•15



40



35



30



25



20



15



to



 5



 0
                 CONTINUAL  EFFECT  OF  GROUNDWATER  PUMPAGE

                                ON  STREAMFLOW  1970


                                    LINCOLN CREEK NEAR SEUARO
              WITH CU PUMPAGE  •


              WITHOUT GU PUMPAGE
         Illn ill in'iln ill in ill
          5  10  IS  20  25  30


               AUGUST
                        5 18 15 20 25  30


                            SEPTEMBER
5 te  15 20 25  30


    OCTOBER       1
5 IB IS  20  25 30


    NOVEMBER

-------
snowmelt runoff, and during summer to represent the degree of antecedent
saturation condition for simulation of runoff from intense storm events.
However, it was observed that while trying to adjust the parameters for
reconstitution of high runoff events, some parameters become biased towards high
and moderate flows, and do not respond to low flow events.  Care should be taken
in adjusting such sensitive parameters.

     (e)  Effectiveness of continuous process hydrologic-hydraulic modeling is
affected by the size of the hydrologic land segment simulated.  Due to  the large
size of the Big Blue Basin, the hydrologic land segments had to be relatively
large in order  to reduce computer costs.  It was, therefore, felt that  HSPF
would be more effectively used to model a medium-sized watershed.
                                    REFERENCES
 1.   American Society of Agricultural  Engineers.  Design  and  Operation of Farm
     Irrigation  Systems, ASAE Monograph  Number  3, 1980.

 2.   Donigian, A.S., Jr. and Davis,  H.H.,  Jr.   User's Manual  for Agricultural
     Runoff Management Model, EPA-600/3-78-080,  1978.

 3.   Donigian, A.S., Jr.,  Imhoff,  J.C.,  Bricknell, B.R.,  and  Kittle, J.L.,  Jr.
     Application Guide for  Hydrological  Simulation Program-Fortan,
     EPA-600/3-84-065 ,  1984.

 4.   Johanson, R.J., Imhoff, J.C., Kittle,  J.K., and Donigian, A.S., Jr.
     Hydrologic  Simulation  Program-Fortan:   Users Manual  for  Release 8-0.
     EPA-600/3-84-066, 1984.

 5.   Nebraska Natural Reosurces  Commission.   Report on  Big  and Little Blue  River
     Basin Area  Planning Study,  Technical  Appendix A-Development, Calibration,
     Verification  of Groundwater Models,  1983.

 6.   Nebraska Water Resources Center.  Development of State Water Quality
     Management  Plan for State of  Nebraska,  Institute of  Agriculture and Natural
     Resources,  University  of Nebraska-Lincoln,  1982.
 The work described in this paper was not funded by the U.S. Environmental
 Protection Agency and therefore does not necessarily reflect the views of
 the Agency and no official endorsement should be inferred.
                                     108

-------
     THE USE OF SWUM TO  PREDICT  RUNOFF  FROM NATURAL  WATERSHEDS  IN  FLORIDA
                                     BY

                               Wayne C• Downs

                               Jon  Paul Dobson

                               Raymond E. Wiles
                               GeoScience  Inc.
                             Gainesville,  Florida
                                   ABSTRACT

     A large Florida watershed,  the Deer Prairie  Slough  basin, was  selected
to test the ability  of the EPA Storm Water Management  Model,  Version III,
(SWMM  III), to predict  runoff from natural and undeveloped watersheds.  The
Deer Prairie Slough  watershed was chosen for  calibration because  it has  a
USGS maintained stream gage, nearby National  Weather Service  raingage, and
nearby USGS monitored wells.  Also, the  site  is near to  a GeoScience
project in Sarasota  County which is similar in size.

     The model was calibrated to the measured runoff for the  test area using
site specific hydrologic data, groundwater conditions  at the  time of the storm
event, and hourly rainfall data.  A second recorded  storm event was selected
to verify SWMM's ability to predict measured runoff once calibrated.  The
model configuration for the calibration  simulation  was left unchanged in the
verification procedure  with the  exception of new hourly rainfall  values  and
new antecedent moisture conditions.

     The modeling results show SWMM III to be quite accurate  in predicting
both peak runoff  rates and  runoff volumes.  The relatively  simple modeling
strategy combined with SWMM's powerful runoff, storage,  and routing features
Indicate the model to be a very  effective tool in predicting  runoff rates  and
volumes  in  Florida.
                                    109

-------
                                  INTRODUCTION

      The US Environmental Protection Agency (EPA) Storm Water Management  Model
 (SWMM) is a comprehensive, physically based mathematical model designed to
 evaluate stormwater quantity and quality.   The model has  been  widely used in
 the United States, Canada, and Western Europe.  It is supported by the USEPA,
 and is updated and maintained at the University of Florida,  Department of
 Environmental Engineering  Sciences.

      The principal use of the model has been to evaluate  stormwater runoff
 from urban sewered systems.  However, the  rainfall, infiltration, evaporation,
 overland flow, channel routing, and detention storage algorithms used in SWMM
 are not peculiar to urban sewered catchments, but  are basic  to  any  stormwater
 analysis.  These features, in fact, make SWMM well suited for the evaluation
 of before and after development runoff peak flows  and volumes  as  required by
 development permitting regulations of the  State of Florida.  Flow attenuation
 by detention storage and channel routing through second and  third order net-
 works are two particularly strong  features of the  SWMM TRANSPORT module.

      The purpose of this  study  was to determine SWMM's  modeling accuracy  on
 pre-development sites in Florida.  Of particular interest was  the model's
 ability to simulate large watersheds as  is  commonly required in large develop-
 ment projects and regional or county-wide  drainage studies.


                           SITE AND DATA DESCRIPTION

      For this work, a large undeveloped  watershed  was selected which has  a
 history of recorded stream gage data monitored by the USGS.  This  basin is
 Deer Prairie Slough located in the southern portion of  Manatee County and the
 northern portion of Sarasota County as shown in Figures 1 and  2.  The water-
 shed is composed of 27,016 acres of flat terrain interspersed  with natural
 depressions and wetland areas.   Shallow  groundwater conditions and  sandy  soil
 types are characteristic of the area.

      Rainfall data were taken from the Venice, Florida  station, within  ap-
 proximately 10 miles of the site.   The data are recorded  in  hourly  increments
 beginning in 1942 (NWS).

      From this rain station a single storm event recorded June 17 and 18,
 1982,  was selected for calibration to the site.  This 8.0 inch  event  over  a
 24-hour duration approximates the design storm required by Sarasota county of
 a 9.5 inch, 24-hour event.

      A storm event recorded September 12 through 26, 1982 was  chosen to verify
 the modeling results.   This event occurred  over 109  hours and  contained a
 depth of 6.6  inches (NWS).   The major portion of the rainfall  took  place  over
 the first 26 hours.

     Ground water conditions were simulated from data taken from the Big Slough
Shallow Well, ROMP 19ES, and ROMP 19WS shallow USGS monitoring  wells near the
site (USGS,  I982b).  The well locations are shown in Figure  2.


                                      110

-------
     Data for the June 17 and 18, 1982,  storm event used for the calibration
of the site, showed groundwater  conditions at an average depth of  2.46 feet
below surface as recorded on June 15.  For the verification trial,  groundwater
data measured on September 20, 1982, indicated an average depth of  1.96 feet
below surface.
     Figure 1.  Location of the Deer Prairie Slough Watershed in Florida.

                                       Ill

-------
                                 METHODOLOGY

     Proper simulation of runoff events in much of Florida requires an under-
standing of the effect of the ground water table on surface flow peaks and
volumes.  Large storms, such as those approximating so called "design storms",
generate overland flow by filling the void space in the soil until the water
table rises to the surface.   The hydraulic conductivities of Florida soils in
                                                      -
                                       ; i ^-' •i[~.j>i ||ii.-	i.
                                       *;   I  [l&'^'-lftspiFW-   \
                                           H^rfSB^l&b  '
V-r
 Figure 2.
                  Location of Shallow Wells with respect to
                  Deer Prairie Slough.

                                      112

-------
these high water table areas are greater than the peak intensities of even
design sized storms.  Thus, runoff is generated by excess rainfall after the
water table has risen to the surface, not by high intensity rainfall on soils
of low hydraulic conductivity.

     This condition may be simulated by modeling each sub-basin as a shallow
reservoir which must be filled before outflow, expressed as overland flow, can
occur.  The simulation is done in the SWMM RUNOFF module by setting the Horton
or Green-Ampt infiltration parameters to low values, essentially turning them
off.  Depression storage  over the area  is  then employed as  the reservoir which
must be  filled before overland flow can occur.

     Determination of the proper amount of depression storage is a function of
depth to water table and water storage capacity above the water table.  Depth
to water table information may be obtained from the US Geological Survey Water
Resources Data Publications,  the Water Management Districts or local data
sources.  Soil moisture storage information is available in publications such
as Characterization Data for Selected Florida Soils available from the Insti-
tute of Food  and Agricultural  Sciences at the University of Florida (Carlisle,
et al.,  1981).  The South Florida Water Management  District (SFWMD)  recommends
the following typical values for storage above the water table in most South
Florida  soils (SFWMD,  1984):


       Depth of Water Table (ft)                  Storage (in)
                 1.0                                   0.6
                 2.0                                   2.5
                 3.0                                   6.6
                 4.0                                  10.9
     These SFWMD values were plotted to aid in interpolation of  soil  moisture
storage values (Figure 3).

     The watershed was divided into six sub-catchments along topographic
features and flow patterns.  Figure 4 indicates the sub-basin delineations.
Sub-catchment widths were determined for model input by taking cross  section
widths perpendicular to the direction of flow at the sub-catchment boundary.
Areas, widths and slopes were taken directly from topographic maps.  Depth to
water table was determined by averaging values from wells near or in  the
watershed, and water storage values were determined from the SFWMD curve.
These values were input as pervious depression storage.  Wetland  areas  were
planimetered and input as containing 1.0 inch of depression  storage within
the wet season hydroperlod elevations.

     Flows were routed using the SWMM TRANSPORT module.  Channels were
simulated as broad trapezoidal channels having side slopes typical of the
topography along the natural channels.  A Manning's roughness value of
0.35  was used for overland  flow  to  simulate undeveloped conditions
(Huber et  al.,  1981).  Channel lengths and bottom slopes were taken from
the topographic maps.   The watershed outflow hydrograph was  plotted at


                                     113

-------
one-half hour intervals.  On the same hydrograph for comparison purposes
were plotted the  measured data from  the USGS gauge.
                                   DISCUSSION

     The 8.0 inch, 24 hour storm of 17 June 1982 was used in the calibration
of the outflow from Deer Prairie Slough.  The only calibration parameter
used to adjust the predicted outflow was the sub-catchment width.  The final
values used were similar to measured  topographic values.  The peak outflow
for the storm  was the largest  of the year for the gage and was recorded as
387 cfs, instantaneous  (USGS,  1982a).   The simulated peak was also
calibrated to  387 cfs.   Table  1  lists pertinent data from this calibration
             12.00
             10.00	
              8.00
              6.00
          ct
          I
              4.00
              2.00
              0.00
                  0.00  0.50   1.00  1.50  2.00  2.50  3.00   3,50  4.00  4.50
                                DEPTH TO WATO< TABLE (ft)
       Figure  3.   Soil Moisture Storage as a Function of  Depth  to
                   Water Table (as per SFWMD, 1984).
                                     114

-------
simulation.  The volume calibrations performed within the model indicated a
measured volume of 2.85 inches compared to a predicted volume of 2.51
inches.   The measured and predicted hydrographs are shown on Figure 5.
             Table 1:   DEER PRAIRIE SLOUGH CALIBRATION DATA
               WATERSHED SIZE:
               AVERAGE DEPTH TO WATER TABLE:
               TOTAL PRECIPITATION:
               STORM DURATION:
               PREDICTED PEAK:
               MEASURED PEAK:
               PREDICTED VOLUME:
               MEASURED VOLUME:
27,060 acres
 2.46  feet
 8.0
 24
 387
 387
 2.51
 2.85
inches
hours
cfs
cfs
Inches
inches
     The second Deer Prairie Slough storm occurred 21-25 September 1982, when
 6.6 inches of rain fell at the Venice raingage (NWS).   No  watershed parameters
 were adjusted  for  this  simulation.  The only changes from the calibration
 trial were the hourly rainfall values for  the  storm and the moisture storage
 capacity above the water  table.  The measured and predicted outflow hydro-
 graphs are shown in Figure 6.   Features of the two hydrographs are similar to
 those of the first storm.  The predicted outflow  of  264 cfs  is  1.8% greater
 than the measured  260 cfs which  is a daily  average (USGS, 1982a).  The pre-
 dicted volume  is 9.4% less than the measured.   Table 2 includes the data and
 results.
               Table 2:  DEER PRAIRIE SLOUGH VERIFICATION DATA
                    WATERSHED SIZE:
                    AVERAGE DEPTH TO WATER TABLE;
                    TOTAL PRECIPITATION:
                    STORM DURATION:
                    PREDICTED PEAK:
                    MEASURED PEAK:
                    PREDICTED VOLUME:
                    MEASURED VOLUME:
27,060 acres
1.98
6.6
109
264
260
1.74
1.92
feet
inches
hours
cfs
cfs
inches
inches
     There are several noteworthy features of the hydrographs shown on Figures
5 and 6.  The shapes of both measured and predicted hydrographs correspond
very well  to  one another.  The hydrographs in Figure 5 reflect runoff from
rainfall of varied intensity over 109 hours.  SWMM allows a maximum of 200
rainfall values to be input for a single event  simulation, so the measured
runoff  "plateau" at about  260 hours  which resulted  from  2.90 inches of rain-
fall at that  time  is  not  reflected  in  the predicted hydrograph.
                                     115

-------
     The lag noticed  in  both predicted hydrographs when compared  to the meas-
ured hydrographs  is probably due to more immediate flow resulting from the
water table rising above the channel bed.  The simulation generated flow when
the water table rose  to  the land surface.
                        uf :;:  '\'  ^7«-    »     i   11   •  :   •
               — - ."•'"-f?. - (-^gis^v^j.,  ij	U_;_j  	-i._
               >..,6.'  ^•^df^l^i'^   ».|.;>|
                >-  "; J?'-   ^i^Sil/U^^ ^ -
K-
         Figure 4.  Sub-basin Delineations within Deer  Prairie Slough.

                                     116

-------
IN
CF'a
*uu. uuuuu




330 OOOOO






240 OOOOO





160 OOOOO







SO OOOOO






O 0
0

++ *»*
+ •»+•»• *
•f » + *
*• » + «
•*•»-»•»
4- • + «
* » + *
•«• * * •
1- » fr •
+ • *••
1- » »• »
+ » +*
-»• * I- *
+ » +**
+ * •«•»
+ * »+
+ * ** +
s **£»%
*. »» *•+
4. »» +
4- *» +•*•
I. ** +*
1- ** ++
t »» ++
^ *« +++•
1. *«•* •»•+
v »»»» •••+.>+
«. »*»» +++•+•»•
+. « *»»*» •»•+•••++
«. «. »»»**»» -f+-»-f++
K « *»»»*»»+++*
^ « »*•**
«. *
O 60 O 120. 0 ISO. 0 240. 0 300. 0 360. 0 420. O 480 0 54O 0 600.

































0
                                 TIME OF DAY.  IN HOURS                  PREDICTED=*.  MEASURED=+
      OUTFLOW FROM DEER PRAIRIE SLOUCH                                   LOCATION   65
               Figure 5.  Measured and Predicted  Hydrographs for  the Storm of 17  June 1982.

-------
00
3OO. OOOOO

Z4O. OOOOO





1BO COUOO


FLOU

IN
CFS


120. OOOOO







6O. OOOOO







0 0
0
	 I 	 i 	 .| 	 1 	 1 	 1 	 I 	 1 	 I 	
•»•»»«***
* # ++ «*
+ * •*• »»
•f * -f »
+ » •*• »
* » >•»•»
+ * +++++++++++
* * » » +*•»•
» •»- * *•»•*•
»•* * » •»•
* + * # + +
»» +»» *» +
» + *»
» •+- **
»» •*- *
* + **
» f *»
»* »•£ **
++>. + + «**
•»••»•+» *
+ ++• *
+ *
* *
+ *
+ *
+ »
+ *
* *
•f »
* »
II- »*
I f »»•»»

0 40.0 BO. O 1?0. O 160. O 200.0 240.0 280.0 320.0 360.0 400
                                      TIME OF DAY, IN HOURS
           OUTFLOW FROM DEFR  PRAIRIE SLOUGH
PREDICTGD=*,  MEASURED=+
LOCATION    65
                    Figure  6.   Measured and  Predicted Hydrographs for the Storm of 21-25 September,  1982.

-------
                                  CONCLUSION
     The Storm Water Management Model proved to produce  good  results when
compared with measured data for the site.   The  match of  peak  flows  generated
by the storm events was  in very good agreement with  the measured peaks while
the volumes predicted by the model also  compared favorably with the actual
data.

     Uttle initial calibration was required to match predicted to  measured
peak discharge values, with  the  width parameters being the only input value
that was adjusted.  SWMM is able  to  predict runoff events for very large
watersheds, which  gives  it a distinct advantage over other modeling methods
such as  the SCS and Rational Methods.

     One of the more important factors for accurately modeling the events
appears to be the simulation of the water table depth prior to the storm.
Information about  the water table  depths is necessary in order to  model the
antecedent moisture conditions which are critical to Florida  runoff from large
storms.

     It is also important for larger sites to be broken down into  sub-
catchments.  This allows the TRANSPORT module to account for  flow routing  from
one sub-catchment  to another rather than producing one output hydrograph based
upon homogeneity within  the watershed.

     The use of SWMM allows physically measurable, site specific input to
be modeled rather than parameters derived  from regression analysis  of
regional data.  The model also allows definition of worst case conditions
as well as wet  season averages.

     SWMM's capability to include site-specific data and antecedent moisture
conditions together with its channel  network routing and detention  storage
capabilities make  it an  excellent choice for accurately predicting runoff peak
discharges and volumes on high water table Florida watersheds.
                                   REFERENCES
 1.   Carlisle,  V.W., C.T., Hallmark, F. Sodek, R.E. Caldwell,  L.C. Hammond,
      and V.E.  Berkheiser,  1981,  Characterization Data  for Selected Florida
      Soils, Institute of Food and Agricultural  Sciences,  University of
      Florida,  Gainesville, Florida.

 2.   Huber, W.C., J.P. Heaney, S.J. Nix, R.E. Dickinson, and D. Polmann, 1981,
      Storm Water Management  User's Manual Version  III, Environmental
      Protection Agency,  Project No. CR-805664.

 3.   National Weather Service, Venice Florida Station, Rainfall Tape,  1942-
      1983.


                                     119

-------
4.   South Florida Water Management District, January 1984, Permit Information
     Manual, Volume IV, Management and Storage of Surface Waters, pc-39, West
     Palm Beach,  Florida.

5.   U.S.  Geological  Survey Water-Data Report, 1982(a),  Water Resources Data -
     Florida, Water Year 1982. Vol.  3A.  Southwest Florida Surface Water, FL-
     82-1A.

6.   U.S.  Geological  Survey Water-Data Report, 1982(b),  Water Resource Data -
     Florida, Water Year 1982. Vol.  3B.  Southwest Florida Ground Water, FL-82-
     3B.
 The work described  in  this paper  was  not funded by the  U.S.  Environmental
 Protection Agency and  therefore does  not necessarily reflect the views  of
 the Agency and no official endorsement  should be inferred.

                                     120

-------
                A SIMPLIFIED HATER QUALITY COMPUTER PROGRAM
             FOR "REGIONAL STQRMWATER MANAGEMENT SITE EVALUATION

              by:  John M.  Grouse, P.E.
                   Timothy  C. McCormick
                   Greenhorne A O'Mara,  Inc.
                   9001 Edmonston Road
                   Greenbelt, Maryland 20770

                   Michael  H. He!frich
                   Montgomery County  Stormwater  Management  Hi vision
                   101 Monroe Street
                   Rockville, Maryland 20850

                                  ABSTRACT

    Montgomery County, Maryland,  has  recognized  the advantages of  regional
Stormwater management facilities  since  the late  1970s.   Until  the  last  few
years, the design of SWM facilities was  based  on  reducing the peak discharge
for  the  post-development  conditions  to  the  corresponding  peaks  for  pre-
development conditions for  the 2-, 10-,  and  often  the 100-year storms.   It
was  known  that  SWM facilities  also   provide qualitative benefits.   No
compatible method  was available  to  quantify  these  benefits that  would
require an effort similar to the  work  needed to compute the peak reduction.

    The desire  to produce  a  simple  method to  evaluate the water quality
benefits  of  SWM  facilities  and recent  legislation  concerning  watershed
management policy in  the State of Maryland have led  Montgomery County  and
their  engineer,  Greenhorne &  O'Mara, Inc.,  to  develop a simple  computer
program to  evaluate  the water quality  impacts  of SWM ponds  throughout a
watershed.  The model was  developed to  utilize  data prepared  for USDA  Soil
Conservation Service Technical  Release Number  20  because TR-20  is frequently
applied throughout the State for  quantitative  design of  SWM facilities.

    A rainfall amount-frequency relationship was developed from data collected
by the County.    Pollutant  accumulation,  pollutant washoff, and  pond  trap
efficiency information were obtained from  previous  works.

    The computer  model  developed was applied during  two  watershed studies
conducted for the County by Greenhorne & O'Mara.   The output from the model
aided in the evaluation of alternate SWM  facility sites located throughout the
watershed.  The model  enabled  a   comparison of  pollutant  levels at various
points along the main stem  and the larger tributaries similar to a comparison
of the peak discharges  that had  been  accomplished  for   previous  studies.
Since  this water  quality analysis required little  additional data, the new
analysis was accomplished for a minimal  cost.

                                   121

-------
                                INTRODUCTION


    Regional  stormwater  management  facilities  have been  designed and
constructed  in Montgomery  County, Maryland,  since the  late  1970s.  The
equations that  govern  the quantity of  runoff that  is  stored and  released
have long  been  known.   Methodologies that  utilize the storage and  outflow
relationships have been developed  into  small  dam  and  stormwater  management
pond design procedures.

    During  the  earlier years  of  the  County stormwater management  program,
the  emphasis  had  been  placed  on  pond   release  rates, downstream  channel
velocities, and  other  hydraulic parameters.  However, more recent  concerns
about  the  levels of sediment  and  nutrients being  carried  to the  Patuxent
River, the  Chesapeake Bay, and  other valuable bodies  of water have  modified
the  rationale for the  selection  of  regional  stormwater  management sites
towards sites that improve the quality of the runoff.

    Water quality models have been in use for more than fifteen years.  The
models, such  as STORM, SWMM,  NPS,  WASP, HSP, SEDIMENT, and  DEPOSITS,  were
developed for various  levels  of planning and design  applications.   A  model
was needed  for  which data  could be easily developed and would be  compatible
with the water quantity model most frequently used  in  the  State of  Maryland
for  development impact studies,  the USDA  Soil  Conservation  Service  TR-20
computer program (1).
                         RAINFALL AMOUNT-FREQUENCY
DATA ANALYSIS
    Montgomery  County  has an  on-going  rainfall  data collection program  at
sites located throughout the County.  The length of  record for  each  site  is
approximately the  same.   A site ne^ar the center of  the  County  was  selected
to be representative of the precipitation received  County-wide.

    Daily precipitation data were available from January 1980 through December
1983.  Daily rainfalls  greater  than one-half inch were considered significant
for  pollutant  washoff.   A tabulation  of rainfall  amounts  exceeding this
value was compiled, and the rainfall  amounts  were plotted against frequency.

    Rainfall amounts for  selected  frequencies were  read from the  plot and
are presented in  Table 1.  Although a limited number of years  of  data was
available, the results  obtained were judged  to be reasonable  because  a large
number of the very frequent storms  occurred  during  the period of record.   In
addition, the data  plotted  well  compared to rainfall amounts obtained from
Technical Paper 40 for  1 year  and ?-year return periods (?.).
                                     122

-------
           TABLE 1.   RAINFALL  AMOUNTS FOR  SELECTED RETURN PERIODS
                       IN MONTGOMERY COUNTY, MARYLAND

         Frequency (months)                 Rainfall  Amount  (inches)
24
12
6
3
?.
1
3.2*
2.6*
1.8
1.2
0.95
0.6
*From TP-40 (2)

ANNUAL STORM DISTRIBUTION

    In an average year, the following storms  would  be  observed:

    1.  one storm greater than or equal  to  the  twelve-month  storm;
    2.  two storms greater than or equal  to the six-month  storm;
    3.  four storms greater than or equal to  the three-month  storm;
    4.  six storms greater than or equal  to the two-month  storm;
    5.  twelve storms greater than or equal to  the  one month  storm; and
    6.  a larger number of storms less than the one month  storm.

    The rainfall amount associated with  storms greater than or equal to the
one  month  storm  was  calculated  to be  12.3  inches  based on  the above
information.   Montgomery  County receives  an average annual  rainfall  of
about  39  inches.   The difference  between the  rainfall  values represents
a very large number of events that produce  less than 0.6 inch of rain over  a
twenty-four hour  period.   These storms  were  not considered further because
they produce  insufficient  runoff  to  wash  significant  amounts of pollutants
downstream.-

    Rainfall  in  the  County  is spaced  almost  evenly throughout  the year.
Large storms may and  have  occurred  in every  month.  The  storm distribution
developed  assumed  that  the twelve storms  greater  than  or equal  to the one
month storm were equally  spaced.   Therefore, one storm occurred every 30.4
days.  The  twelve-month storm  and  the  six-month  storm were offset by six
months.  The  two  remaining three-month  storms  were also  placed six months
apart and were offset from the  twelve-month  storm  by  three months.  In this
manner, the storms were distributed evenly  over the twelve-month period.
                                     123

-------
                       POLLUTANT BUILDUP AND WASHOFF
ACCUMULATION RATES
    Daily,  dry  weather  pollutant  accumulation  rates  for  pervious and
impervious fractions  of urban and  rural  land  uses  were  developed  by the
Northern Virginia Planning District Commission  (NVPDC) for the Metropolitan
Washington Council  of Governments (COG)  in 1979  (3).   The  NVPDC accumulation
rates were judged to be acceptable for  Montgomery County  because of  similar
physiography and the proximity of the NVPDC study area and  Montgomery  County.

WASHOFF

    The washoff function of Metcalf &  Eddy,  Inc.,  et  al. was  utilized as the
washoff mechanism (4).  The limiting pollutant load value  was found to  occur
over  a  period  of  time shorter  than  the  30.4  days  between  each  storm of
significance.  Hence, the maximum accumulation  of pollutants was assumed to
be obtained before the beginning of each runoff  event.

    Runoff in the washoff process was determined  for a given soil  type and
land  use through the  runoff curve number  (RCN)  methodology of the USDA Soil
Conservation Service (5).  The RCN methodology  was selected  because  the RCN
values were  be  available  from the data prepared for  the  hydrologic  portion
of the study.

                           POND  TRAP EFFICIENCY


    The principal reason  for  the development of the  computer  model  was to
estimate the removal of pollutants by a regional  SWM facility, which can be
related to  the  trap  efficiency of  the  pond.   In   1975,  Chen  developed   a
series  of  curves  of  trap efficiency  versus  the  ratio   of basin  area to
outflow rate (6).   Chen's work was  based  on previous studies by Camp.  The
curves  reflect  the  size  of  the sediment and the  settling velocity of the
sediment particles.

    Settling  velocity  data for  soils  typically  found  in the metropolitan
Washington area were presented  in  the  NVPDC study (3).   The pond retention
time  and  the settling  velocity  are used to  calculate  the percent   of  each
particle  size  which  settles  in the  pond.    The  computer model  developed
employs  a default  soil  gradation  that  reflects the general  soil  of the
County.  A different gradation curve may be  input  by  the  user.

    NVPDC  (3)  presents  a  table that  assigns  fractions  of the suspended
portion of each pollutant included in the study to sediment  particle sizes.
The pollutants were assumed to settle as their associated  sediment particles
settled.  The  pollutants  remaining  in  the  discharge  from the pond was
assumed to  be  the  dissolved  pollutant plus the portion of the suspended
pollutant associated with the sediment  that  failed to setfle  in the  pond.
                                     124

-------
                              MODEL  OPERATION
    Execution  within  the model  occurs in the following manner:

    1.   The total  accumulation of sediment,  BOD, total  phosphorus, total
        nitrogen, extractable  lead,  and extractable  zinc on  each  subarea
        prior  to  each  storm  is  generated  from impervious  and pervious
        fractions for each  land use;

    2.   The amount of washoff from  each  subarea for each storm  is  calculated
        based  on  Metcalf &  Eddy's washoff function;

    3.   The pollutant washoff  data are  annualized by summing  the data for
        the number of occurances of each storm frequency;

    4.   The pollutant washoff from  each  subarea is added based  on  series and
        parallel  subarea  relationships  from upstream to downstream until  a
        pond is encountered;

    5.   The pond  routine is called, and the amount of each pollutant deposited
        in the pond is calculated.   Basin  area  and outflow rates were input
        based  on  the  SCS TR-20 printout for the storm frequencies studied; and

    fi.   The subarea  addition  and  pond  routines  continue until  the study
        outfall  is reached.

                             MODEL  APPLICATION


    The simple computer model - titled  "WATQUAL"  due to the  lack  of a more
imaginative name - was applied  in two watershed studies in the  County.  The
results for a  portion of  one of the studies, the Little Paint  Branch study,
are included in this  section.

    A  portion  of the Little Paint  Branch  watershed is shown  in  Figure  1.
The anticipated urbanization of the watershed  is  reflected in the increase
in curve  number  from existing  conditions  to ultimate  development shown  on
the figure.

    The  data   input  reflects combinations  of  soil  type  and  land  use for
each subarea.   These data, which  had been  digitized  for  the  TR-20 model,
became the bulk of the input  data  required for WATOUAL.

    Without the  proposed  pond  in  Subarea  7, the  annual   sediment  load  at
Subarea 5 would  increase from  26  tons  to  58 tons from existing to ultimate
conditions.  With the proposed  pond, the  sediment  loM  would be  reduced  to
11 tons as shown on Table  2.
                                     125

-------
 Ex.  8O
 a/t.  do
7a/?6}/ewood
        59
        G3
        76
                                                 "* /COO'
/.   L/W
-------
                  TABLE 2.  EXAMPLE STUDY SEDIMENT LOADS

                            Annual  Sediment  in  Runoff  (tons)
Land Use
Condition
Predeveloped
Existing
Ultimate
Subarea 9
0.39
6.97
13.73
Subarea 7
1.04
24.09
47.92
Subarea 8
0.20
0.92
1.06
Subarea 5
1.42
26.23
58.33
Without Site 1

Ultimate
With Site 1
13.73
0.34
1.06
10.76
                              MODEL LISTING
    The length of the listing of the  Fortran  computer  model  prohibited the
publication of the listing  with this paper.  Those interested in obtaining a
listing may contact one  of  the authors.


                                 REFERENCES
 1.  TR-20, project formulation-hydrology (1982 version) - technical release
     number 20.   USDA,  Soil  Conservation  Service, Lanham, Maryland,  1982.

 2.  Hershfield,  D.M. Rainfall frequency atlas of the United States, technical
     paper number 40. U.S.  Department of Commerce, Weather Bureau, Washington
     D.C., 1961.

 3.  Guidebook for  screening urban, nonpoint pollution management strategies.
     Northern  Virginia   Planning  District Commission,  Annandale,  Virginia
     1979.  160 pp.

 4.  Metcalf  &   Eddy,   Inc.,  et  al .  Stormwater  management  model.   U.S.
     Environmental  Protection Agency, Washington, D.C., 1971.

 5.  National  engineering handbook,  section  4, hydrology.   USDA,  Soil
     Conservation Service,  Washington, D.C., 1972.

 6.  Chen, Charng-Ning.  Design of sediment  retention basins.  Paper presented
     at  1975 National  Symposium on  Urban  Hydrology  and  Sediment  Control
     Lexington, Kentucky.   July 23-31, 1975.

    The work described  in the paper was  not  funded  by  the  U.S.  Environmental
 Protection Agency and therefore the contents do  not necessarily reflect the
 views of the Agency and no official  endorsement should be  inferred.

                                    127

-------
                        DEVELOPMENT OF THE HAZPRED MODEL

                   by:   G.  Zukovs1, J. Kollar1,  M. Shanahan2


                    Canviro Consultants Ltd,  Toronto,  Ontario
                  2
                    Environment Canada, Toronto,  Ontario
                                   ABSTRACT

      HAPZRED is an interactive, microcomputer based model capable of the
 prediction of hazardous contaminant (HO concentrations and loadings in dry
 weather sewage flows., and of runoff and CSO volumes and HC loadings.  HAZPRED
 has been designed to estimate both average quantities (expected values) and
 the probability of specified events.

      The paper presents an outline of the predictive techniques used in
 HAZPRED.  A case study of a combined sewer catchment located in the Toronto
 area of Ontario is also presented.  The case study illustrates model input
 requirements and the nature of model outputs.
                                 INTRODUCTION


     HAZPRED is an interactive microcomputer based model designed to predict
the concentrations and loadings of selected hazardous contaminants (HCs) in
dry weather sewage flow (DWF), urban stormwater runoff, and combined sewer
overflow (CSO).  The specific predictive capabilities included in HAZPRED are
summarized in Table 1.

     The selected hazardous  contaminants  included  in HAZPRED  are  based  upon
the U.S. EPA list of  129  priority  pollutants.   The HCs  which  can  be examined
using HAZPRED  include  compounds falling within  five classifications:  Volatile
Organics,  Semi-Volatile Acid  Extractables,  Semi-Volatile Base/Neutral Extrac-
tables, Pesticides and PCBs,  and Metals and Trace  Elements.
                                     128

-------
             TABLE 1.  SUMMARY OF HAZPRED PREDICTIVE CAPABILITIES

Waste Stream                       Predictions Performed
 Dry Weather
 Sewage Flow
  Stormwater
    Runoff
   Combined
Sewer Overflow
HC concentrations in industrial wastewater
HC loadings in industrial wastewater
HC concentrations in dry weather sewage flow
HC loadings in dry weather sewage flow

Average runoff event volume
Exceedance probability of runoff volumes
Average runoff HC concentrations
Average runoff HC loadings
Exceedance probability of runoff HC loadings
Average annual runoff volume
Average annual runoff HC loading

Average CSO event volume
Exceedance probability of CSO volumes
Average CSO HC concentrations
Average CSO HC loadings
Exceedance probability of CSO HC loadings
Average annual CSO volume
Average annual CSO HC loading
                               MODEL STRUCTURE
     Figure 1 presents the overall structure of HAPZRED.  As is indicated,
HAZPRED supports two levels of sophistication in the dry weather flow (DWF)
model.  The following discussion will focus on the Level II, or more sophisti-
cated approach.  Reference may be made to the HAZPRED documentation and to the
User's Manual (Canviro, 1984, 1985, 1986) for additional details of Level I
procedures.


CATCHMENT - SEWERAGE CONFIGURATION

     The conceptual models of the catchment-drainage system used in HAPZRED
for separate and combined sewerage are shown in Figures 2(a) and 2(b).  A
simple catchment model comprised of three parallel sub-catchments is employed.
Open space areas are not treated per se but are assumed to be included in the
pervious portion of each sub-catchment.

     The parallel catchment model  assumes:
      i) that catchments  are  identically affected by regional climatology, and
     ii) that each of  the three sub-catchments  responds  independently to  the
         same climatological  input  (no correlation  of  sub-catchment  output
         probability  distributions)
                                     129

-------
                                    MAIN PROGRAM
                                      HAZPRED
i
SUBROUTINE
OHF DWF
LEVEL I LEVEL 41
] '


SUBROUTINE
RUNOFF
i


!
*
SUBROUTINE
DATA FILE
EDITOR
t
SUBROUTINE
CSO
         Figure 1.  Schematic diagram showing general structure of HAZPRFD.
Given  these  assumptions,  the  probability distributions  of  catchment  outputs
(flows  and contaminant  loads)  are additive.   Previous analysis  of sewer  flows
by Adams  and Gemmell  (1973  a,  b)  have  shown  that  the above may  be justified
since  the  impact  of  any covariance terms on  probability estimates would  be
quite  small  and hence could be ignored.

     In dry  weather  (no precipitaiion),  all  land  use areas contribute  dry
weather sewage flows which  are transported to treatment by either sanitary or
combined  sewers.  During  periods  of precipitation,  rain falling on the catch-
ment is transformed  to  runoff  after accounting for  losses  due to depression
storage and  infiltration  through  pervious surfaces.  Pollutants residing on
catchment  surfaces are  washed  off and  transported with  the runoff to the
catchment outlet.  In separate sewerage  systems [Figure 2(a)],  stormwater run-
off is directly transmitted to the receiving water  body.   In combined  sewerage
systems [Figure 2(b], or  nominally separate  systems  with a large wet weather
component, stormwater runoff enters the  sewer system and mixes  with  the  sani-
tary wastewater.  Combined  sewage flows  in excess of sewer or treatment  plant
capacity are overflowed at  regulating  structures; in the case of nominally
separate systems, flows are by-passed  at pumping  stations  or at the  treatment
plant.

ESTIMATION OF DRY WEATHER FLOW QUANTITY  AND  QUALITY

     The quantity of dry weather  sewage  flow from a  given  area  is dependent
upon the nature of land use activity within  the area and upon the influx of

                                     130

-------
             PRECIPITATION
      I    I    I    1     I     I
                                             SANITARY
                                             WASTE WATER
                                             TO TREATMENT
                                             STORM RUNOFF
                                             TO  RECEIVER
          A.  SEPARATE  SEWERAGE
             PRECIPITATION
      1    I    1     I     I    I
           B.  COMBINED  SEWERAGE
                                             COMBINED  SEWAGE

                                       REGULATOR

                                                     TO
                                                     TREATMENT
                                            CSO
                                            TO RECEIVER
Figure 2.  Schematic of HAZPRED catchment sewerage configuration.
                            131

-------
groundwater through infiltration.   For each land use,  different  techniques are
required to develop the actual sewage flow components.   A  distinction should
be made between flow estimates normally used for sanitary  sewer  or  treatment
facility design, and those that have application in  a  predictive model.
Design values in general tend to be conservative,  often  producing higher than
observed flow.values and hence should not be employed.

     Figure 3 shows the components making up sanitary  sewage  in  dry weather
and indicates the factors that affect the magnitude  of each component.

Residential Wastewater Flow
     The residential per capita wastewater flow rate is developed  from metered
water use data.  Winter data is used since it reflects most accurately water
                                 Summer time activities such as  lawn  watering,
                                use but not to sewer flows.  Where household
                                use data can be developed through  the instal-
                                                                data from  a
                                                                loss in  dis-
returned to the sanitary sewer.
car washing, etc., add to water
metering is not employed, water
lation of test meters in selected households.   Alternately,  flow
suitable pumping facility can be used.   Some allowance for water
tribution must be applied to pumping station data.  An allowance of 10-15%  is
considered typical (AWWA, 1962) for "unaccounted" water (water lost in  distri-
bution).
      Population  data  for  a  given catchment  is developed from the most recent
 census  figures.   Since  census  and catchment boundaries may not match, the
 catchment  population  is estimated from unit area population of the census
 zone(s)  and  the  catchment area.
  RESIDENTIAL WASTEWATER

     - Serviced
       Population

     • Per Capita
       Water Use

     - Fraction water
       use returned
                                  INDUSTRIAL WASTEWATER

                                    - No.  of
                                      Establishments

                                    • Water Loss
                                      To Cooling,
                                      Process,  Product
DRY WEATHER INFILTRATION
- Sewer Age
- Sewer Network
Density
- Groundwater



CONMERCIAL WASTEWATER
- No. of
Establishments
- Fraction Water
Use Returned
                                                                     DRY
                                                                   WEATHER
                                                                  SANITARY
                                                                   SEWAGE
                                                                    FLOW
       Figure 3.  Components of dry weather sanitary sewage

                                    132

-------
Commercial Wastewater Flow

     In the HAZPRED model, it is assumed that the quantity of wastewater from
commercial establishments is exactly equal to the commercial water usage.
Water usage data are, again, ideally derived from individual metered records.
Since the type of commercial establishments is not differentiated, commercial
water usage is simply summed for a given catchment area.

Industrial Wastewater Flow

     The relationship of water consumed to wastewater generated is consider-
ably more complex for industrial establishments.  Resident industries are
first identified either through listing services (ie. Dunn and Bradstreet) or
by field inspection.  The latter method is more comprehensive, particularly
for small establishments, but can be quite manpower intensive.  Once the
industrial inventory is complete, then water consumption records are accessed
and a total industrial water usage is computed.  The industries listed are now
segregated into sublists based on industry type as defined by the Standard
Industrial Category code (SIC code).  In the event that industries are identi-
fied by field inspection, the SIC category is determined by direct contact
with the industry.  Major industrial water users (>1% of the total industrial
water use for the catchment) are identified in each SIC sub-list and are then
contacted in order to estimate the wastewater flow components of interest.   In
the case of combined sewer catchments, the process and sanitary flows as well
as the cooling water flows are identified.  For separated sewerage systems,
only the process and sanitary component is normally needed.  Although where
municipalities allow cooling water discharge to sanitary sewers, estimation  of
both components will again be necessary.  Occasionally, the wastewater or
cooling water flow estimates will not be readily available from the  industry.
Best estimates of these flow components, based upon previous experience or
knowledge of the industry, are then employed.   Industrial wastewater flows
generated by minor water users are estimated for each SIC category using total
water usage data and a return factor of 0.85.

Dry Weather Infiltration

     Dry weather infiltration is determined most accurately as the difference
between metered dry weather flow data obtained at the catchment outlet, and
the computed residential, commercial, and industrial wastewater flow compo-
nents.  Since monitored data is, however, only  rarely available, DWI rates  are
usually estimated.  Ontario experience  (Cooper, 1984) has shown a reasonable
infiltration allowance to be between 91 and 227 Icpd.  The exact magnitude  is
determined subjectively from knowledge  of the  age of the sewer system and
population.

Estimation of Dry Weather Sewage Hazardous Contaminant Concentrations

     Estimates of HC concentrations  in  the residential  and commercial compo-
nents of wastewater are based on a  1979 U.S. EPA funded study carried out  by
A.D. Little (1979).  The objective  of the Little study was to determine  the
relative  significance of  source  (residential,  commercial or  industrial)  con-
tributions of the EPA designated priority pollutants to publicly owned  treat-
ment works.

                                     133

-------
     For industrial flows, HAZPRED, wherever possible, uses contaminant con-
centrations specific to the categories of industry present in the catchments.
However, only a limited number of  industrial effluents have to date received
characterization.  The most comprehensive industrial effluent data base is the
EPA's Treatability Manual (U.S. EPA, 1980).  The manual  is the product of
extensive monitoring and gives the  incidence and effluent concentration of EPA
priority pollutants for certain types of industry.  Table 2 presents the
industrial categories for which HC  data is presently included in the HAZPRED
model.

     Contaminant concentrations in  wastewater from all other  industries are
based upon average data obtained during the A.D. Little  study.


WET WEATHER  FLOW QUANTITY AND QUALITY

Analytical Models

     Derived  probability  (analytical) models are used  in HAZPRED to describe
the wet weather behaviour of  both  separated  and combined drainage  systems. The
models employ as input  single parameter exponential distributions  to describe
the statistics of  rainfall  (ie. volume, intensity, duration,  inter-event  time)
for a given  locale.  These  distributions are mathematically transformed into
the probability distributions of  interest,  such as runoff  and CSO  volumes, and
pollutant  loadings.  The  derived distributions  are then  employed to estimate
distribution  moments (ie. expected values)  and  the exceedance frequencies of
selected events.

     The theory and development of these models is well  documented.  Reference
should be  made to  Adams  and Bontje (1983) for details.

Runoff Quantity

     A modified version  of  the coefficient  method  is used  to  describe  the
transformation of  rainfall  to runoff.  The  relationship  between  runoff and
rainfall is  expressed as:

     QR =  *  (i - $)                                                        (1)

     where QR = runoff  rate,  mm-ha/hr
           <£  = modified  runoff coefficient, ha
             i  = average  rainfall  intensity, mm/hr
           $  = continuous  abstraction rate, equivalent  to depression  storage,

                mm/hr

     The value of  the modified runoff coefficient, $ , is  determined from the
area weighted runoff coefficient for each sub-catchment. The overall  expres-
sion for   
-------
             TABLE 2.  INDUSTRIAL CATEGORIES INCLUDED IN HAZPRED

SIC DESIGNATION
2200 - 2299
2834

2841



2844



2851


2893


3011 - 3079*

3111


3312


3321 - 3325

3431


3471



3479
             INDUSTRIAL DESCRIPTION

Textile Mill Products
- yarn, thread, webbing manufacture
- fabric manufacture and finishing
2821,. 2823, 2824     Plastics Manufacture
- plastic materials, synthetic resins
- cellulosic man-made fibres and synthetic organic fibres
Pharmaceuticals
- other than perfumes and cosmetics
Soaps and Detergents
-manufacture soap, synthetic*organic detergents, organic
  alkaline detergent, and crude and refined glycerine
Cosmetic Preparations
- preparations which function as skin treatments,
  excludes those used to enhance appearance
Paint Manufacture
- trade-sales paints, chemical  coatings, varnishes,
  lacquers, epoxy coatings, and paint removers
Printing Ink
- letterpress, lithographic, flexographic and gravure
  inks, and varnish
Rubber Processing
- tire arfd tube manufacture and other rubber products
Leather Tanning and Finishing
- conversion of animal skins or hides to leather
Coke By-Products
- manufacture of light oils, tars, phenolates, etc. as
  by-products of coking
Foundries
- melting, moulding or finishing of metals
Procelain Enamelling
- application of glass-like coating to steel, iron,
  aluminum or copper
Electroplating
- application of metallic surface coating by
  electrodeposition
Coil Coating
- painting of coiled sheet metals
        * SIC-3079 included in this group since products are related
                                      135

-------
          3                                                                 (2)
     *    I  = modified runoff coefficient

            . = runoff coefficient for the jth sub-catchment
             J
            A. = area of the jth sub-catchment, ha
             j
     The values of  4>j for each sub-catchment  are calculated from  the  runoff
coefficients for pervious and impervious  surfaces within  the catchment,  and
the relative proportion of these surfaces within each  sub-catchment.

     The continuous  abstraction  rate,  $  , is  related to the depression stor-
age, average  rainfall volume,  and  average rainfall  intensity.   The expression
used to calculate  $  is  :

        =  C_!d                                                              (3)
           B

     where   *  =  continuous abstraction rate,  ntm/hr
             t,  =  reciprocal of average rainfall  volume, I/mm

             S
-------
     The washoff constant, K, takes into account catchment hydrology and sur-
face characteristics, and is calculated for each sub-catchment as follows:

     K = $  A.C./100                                                       (5)
          J  J v'

     where  K = washoff constant, gm/mm
           4>'.  = runoff coefficient for the jth sub-catchment

           A- = area of the jth sub-catchment, ha
            J
           C.j = average concentration of a single HC in runoff originating in
                the jtfl sub-catchment, ug/L


     Average hazardous contaminant concentrations  in  stormwater  runoff  (C-)
were obtained from  an Ontario  study conducted  by Marsalek  and  Greek  (19837  and
from studies undertaken  as  part  of the U.S. EPA Nationwide Urban Runoff
Program (NURP)  (U.S. EPA, 1982).

     The study  conducted  by Marsalek  and Greek (1983) examined the frequency,
concentration and loadings  of  51  selected  persistent  toxic substances  in  urban
runoff in the Niagara River area.  Samples of  urban runoff were  collected  at
sites in Fort Erie,  Niagara Falls and Well and.  Each  site  possessed  a  charac-
teristic land use;  of the eight  permanent  sampling stations, two were  located
in  residential  areas and  six were located  in  industrial  areas  (sampling sta-
tions were not  located in any  commercial areas).

     The preliminary NURP priority pollutant  data base was used  to obtain
information on  contaminant  concentrations  in  runoff from commercial  areas
(U.S. EPA, 1982).   A review of all NURP sampling sites identified five  which
listed 10056 of  land  use  to  be  commercial.  Pollutant  concentrations  reported
for each of these sites were abstracted from  this data base.

CSO Quantity and Quality

     Combined sewage quantity  and quality  are  estimated  in HAZPRED through
simple mass and volume balance relationships  based upon  the  relative propor-
tions of runoff and  dry weather  sewage.  Probability  density functions  derived
from the mass and volume  balances are then used to describe  the  distributions
of  CSO volumes  and  pollutant loads.

HAZPRED IMPLEMENTATION

     The HAZPRED model is written in  BASICA for the IBM  PC microcomputer  and
compatibles, and requires the  following hardware:

     o IBM PC,  PC/XT, PC/AT or compatible  microcomputer MS-DOS  2.1  or  higher
     o 128 kilobytes of  random access memory  (128K of RAM)
     o one double sided  disk drive
     o a keyboard
     o monochrome or colour display monitor
     o graphics adapter  (to display piecharts)


                                      137

-------
      In  addition  to  the  above  requirements,  HAZPRED will  also support  dot-
matrix printers  (EPSON FX-series)  which  can  be used to  generate  hard copies  of
program  text.

      Three  DOS system files  are  required to  support the HAZPRED  program;  these
are BASICA.COM,  COMMAND.COM  and  GRAPHICS.COM.   In  order to  run HAZPRED, the
three system files must  be resident  in the  same location  as the  HAZPRED pro-
gram, either on  the  HAZPfcED  diskette or  in  the DOS sub-directory containing
HAZPRED.
                                  CASE STUDY


     The HAZPRED model was employed to study a combined sewer catchment loca-
ted in the City of York, Ontario  (one of the lower  tier municipalities making
up Metropolitan Toronto).  The catchment is 834 ha  in area  and  is wholly ser-
viced by combined sewers.  The area is approximately 80% devoted to residen-
tial land-use with the remainder  equally shared between industrial and commer-
cial land-use.

DRY WEATHER PREDICTIONS

Dry Weather Wastewater Quantity

     Dry weather flow quantities  for the catchment  required as 'HAZPRED input
are presented in Table 3.
               TABLE 3.  DRY WEATHER FLOW WASTEWATER QUANTITIES
     Total Wastewater Flow                          (I/day)
     Residential Wastewater Flow                    (I/day)
     Commercial Wastewater Flow                     (I/day)
     Industrial Wastewater Flow                     (I/day)
     Dry Weather Infiltration Flow                  (I/day)
     Industrial Cooling Water Flow                  (I/day)
34,118,270
15,116,960
    48,890
   503,220
18,449,200
   334,300
     Details of the required industrial component flows are provided in Table
4.  In this case, the catchment contains only seven significant industries of
which two were represented by specific HAZPRED contaminant concentration data.
(ie. 2851 - paint manufacture, 3011-3079 - rubber processing).

Dry Weather  Wastewater  Contaminant Concentrations and Loadings

      Dry weather  concentrations  of priority pollutants predicted  by HAZPRED
are presented  in  Table  5.   The  < sign indicates  that  the  predicted concentra-
tions  are below  the Method Detection Limit,  implying  that the compounds would
be  detectable  but not quantifiable with  routine  analytical  techniques.
                                     138

-------
                 TABLE 4.   YORK  INDUSTRIAL  FLOW COMPONENTS

                        Industrial  Flow  by  SIC  (I/day)
0
0
0
0
0
10,680
0

3011-3079 : 14,140
3111
3312
3321-3325
3431
3471
3479
0
0
0
0
0
0
Other SIC : 144,100
           2200-2299
           2821-2824
           2834
           2841
           2844
           2851
           2893
     Table 6 presents the predicted contaminant yearly dry weather loadings
indicating both component loadings (ie. residential, commercial, industrial)
and total loadings.

     For the York catchment, depending upon the contaminant, either residen-
tial or industrial sources dominate the total loads.  For example, in the case
of the popular cleaning solvent tetrachloroethene, residential discharges are
predicted as the major source of this substance.  In contrast, acrylonitrile
is predicted as exclusively originating from industrial sources.


WET WEATHER PREDICTIONS

Stormwater Runoff Volumes and Loadings

      Input data needed for  runoff and CSO predictions  pertaining to both
catchment and rainfall characteristics are presented in Table  7.  A noteworthy
feature of the catchment  sewerage is the relatively high ratio of interception
capacity to dry weather flow (6.2:1).

      HAZPRED output for both storm and CSO predictions  includes summary sta-
tistics  as well as expected values and exceedance probabilities of loadings
and volumes.  Table 8 presents  the summary runoff statistics  for the  catch-
ment.

      In  total, 72 of the  89 rainfall events  are of  sufficient  magnitude to
produce  runoff.  The residential  area, as may be  expected,  contributes  most  to
runoff volumes.   In addition to average runoff  characteristics, HAZPRED
provides the exceedance probabilities  of annual runoff volumes, which are pre-
sented in Table  9.

      These data may be generated  for any range  of desired  volumes  and may be
used  in  a number  of ways, including  calculation of  the return period  of
various  volumetric events,  according to the  following  equation:
                                                                            (6)
                  e  PCV >  vn]
                                     139

-------
          TABLE 5.   YORK OVERALL CONTAMINANT CONCENTRATION PREDICTIONS

              Contaminant                           Concentration (ug/1)

              Carbon tetrachloride                <      1.00
              Chloroform                                 1.44
              Toluene                                    2.20
              Acrylonitrile                              2.49
              Benzene                             <      1.00
              Chloroethane                        <      5.00
              Ethyl benzene                        <      1.00
              Methylene chloride                  <      1.00
              Tetrachloroethene                          3.19
              1,1,2-trichloroethane               <      1.00
              1,2-dichloropropane                 <      1.00
              Trichloroethene                     <      1.50
              Phenol                              <     10.00
              Butyl  benzyl phthalate              <     10.00
              Diethyl-phthalate                   <     10.00
              Nitrobenzene                        <     15.00
              1,2-dichlorobenzene                        1.25
              Naphthalene                         <     10.00
              Bis(2-ethylhexyl)phthalate          <     10.00
              Di-N-butyl phthalate                <     10.00
              Antimony                            <      2.00
              Arsenic                             <      3.00
              Cadmium                             <      2.00
              Chromium                            <     33.50
              Copper                                    32.76
              Lead                                      44.96
              Mercury                             <      1.50
              Nickel                              <     15.50
              Selenium                            <      3.00
              Silver                              <      2.00
              Zinc                                     103.52
     where         TR = average return period (years)
                   6  = average number of rainfall events/year

            p[V > V ] = the probability of observing a runoff volume greater
                        than VQ

     For example, for the York catchment, a runoff volume of 10,000 m^ has an
average return period of 0.03 years, or approximately 1.5 weeks.

     HAZPREO also predicts the expected values and exceedance probabilities of
runoff pollutant loadings, both per average event and per average year.

                                     140

-------
     Table 10 gives the expected values of annual runoff  loadings  for  various
contaminants, from the sub-areas of the catchment, and for  the  catchment  as  a
whole.  As was the case with the dry weather sewage, the  different land-use
areas contribute varying amounts of a given contaminant to  the  total annual
mass.

CSO Volumes and Loads

     HAZPRED summary statistics for combined sewer overflow events are presen-
ted in Table 11.  As is evident, the large interception capacity results  in  a
moderate number of annual overflow events.
          TABLE 6.  YORK PREDICTED DRY WEATHER FLOW YEARLY LOADINGS
Contaminant

Carbon tetrachloride
Chloroform
Toluene
Acrylonitrile
Benzene
Chloroethane
Ethylbezene
Methylene chloride
Tetrachloroethene
1,1,2-tr i ch1oroethane
1,2-dichloropropane
trichloroethene
Phenol
Butyl benzyl phthalate
Diethyl-phthalate
Nitrobezene
1,2-dichlorobezene
Naphthalene
Bis(2-ethylhexyl)phthalate
Di-N-butyl phthalate
Antimony
Arsenic
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Selenium
Silver
Zinc
Res
(kg/yr)
0.00
16.55
14.35
0.00
1.10
0.00
2.21
0.00
34.76
0.00
0.00
2.21
32.00
37.52
54.07
0.00
15.45
11.59
37.52
49.66
14.90
26.48
9.93
89.94
397.83
536.87
2.21
23.17
20.97
12.14
1,180.79
Com
(kg/yr)
0.00
0.10
0.16
0.00
0.04
0.00
0.04
0.00
0.31
0.00
0.00
0.19
0.07
0.16
0.08
0.00
0.11
0.04
o.ll
.17
,.00
0.04
0.01
0.83
0.80
0.73
0.01
0.18
0.05
0.04
2.03
Ind
(kg/yr)
1.72
0.84
9.01
21.63
3.42
8.91
6.97
2.22
3.19
0.01
0.03
1.07
5.96
6.18
0.00
0.30
0.00
2.01
2.99
3.17
0.13
0.26
1.36
27.76
6.43
15.48
3.07
5.10
0.07
5.54
73.98
Total
(kg/yr)
1.72
17.49
23.52
21.63
4.57
8.91
9.28
2.22
38.27
0.01
0.03
3.47
38.03
43.86
54.16
0.30
15.56
13.64
40.62
53.00
15.03
26.78
11.30
118.53
405.06
553.08
5.29
28.46
21.08
17.72
1,256.79
                                      141

-------
            TABLE 7.  YORK CATCHMENT AND PRECIPITATION CHARACTERISTICS
  Total Area (ha);  834

  Combined Sewer Interception Capacity (1/d);  209,950,000

  Coefficients
                Pervious:
              Impervious:
Runoff
 0.25
 0.90
Depression Storage
      2.0  mm
      0.25 mm
  Percentage Impervious Area
              Residential (%):  45.5
              Commercial  (%):  71.6
              Industrial  (X):  84.8

  Precipitation Characteristics
              Reciprocal Average Intensity
              Reciprocal Average Volume
              Reciprocal Average Duration
              Average No. of Annual Rainfall Events -
                      (hr/mm):  0.716
                       (I/mm):  0.193
                       (1/hr):  0.288
                              : 89.4
     Both expected values and exceedance probabilities of CSO contaminant
loadings and volumes are available as HAZPRED output options.  Table 12 illus-
trates the average values of annual CSO loads including the runoff and dry
weather flow contributions.

     HAZPRED has, as well, graphic output capability in the form of pie charts
which present for specific contaminants the relative fractions originating
from sewage and runoff.  Figure 4 shows the pie chart for the York catchment
representing total PCBs.
                            TABLE  8.   PREDICTED  YORK  RUNOFF  STATISTICS
         Average Annual  Number  of  Rainfall  Events   :
         Average Annual  Number  of  Runoff  Events     :

         Average Annual  Hours of Runoff             :

         Average Runoff  Volume  per Event
                      Residential Runoff Volume  (m;?)
                      Commercial Runoff Volume   (no
                      Industrial Runoff Volume   (nr)
                        89
                        72

                       251
         Total Annual  Runoff  Volume  (m3)
                                 13,030
                                  2,105
                                  2.096
                      Total  Runoff  Volume  from Catchment  (m3):     17,233
                              1,540,661
                                    142

-------
      TABLE  9.   PREDICTED  YORK ANNUAL  RUNOFF VOLUME  EXCEEDANCE  PROBABILITIES

             Total  Volume (nr*)            Probability  of  Exceedance
                       1
                   1,000
                   2,000
                   3,000
                   4,000
                   5,000
                   6,000
                   7,000
                   8,000
                   9,000
                  10,000
0.81
0.70
0.62
0.57
0.53
0.50
0.47
0.44
0.41
0.39
0.37
          TABLE 10.  PREDICTED YORK ANNUAL RUNOFF POLLUTANT LOADINGS
Contaminant

1,2-Dichlorobenzene
1,2,4-Trichlorobenzene
1,3-Dichlorobenzene
1,4-Dichlorobenzene
Hexachlorobenzene
A-Endosulfan
B-Endosulfan
4,4-DDT
4,4-DDE
Aldrin
A-BHC
Chlordane
Dieldrin
Endrin
Heptachlor
Heptachlor epoxide
Total PCB
Arsenic
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Selenium
Zinc
Res
(gm/y)
30.06
2.56
3.61
5.94
0.93
1.51
1.40
1.28
0.35
0.35
19.22
1.40
0.82
1.05
0.47
1.05
13.86
2,310.47
999.89
1,106.72
10,096.39
5,000.03
6,582.05
6,989.78
1,766.90
31,745.26
Com
(gm/y)
0.00
0.00
9.00
0.00
0.00
4.95
0.00
6.65
0.00
0.00
12.56
6.59
6.46
0.00
0.00
0.00
0.00
1,004.11
0.00
0.00
8,378.14
37,390.96
18.83
596.26
125.58
135,179.80
       Ind
      (gm/y)
         7.91
         0.52
         0.97
         0.86
         0.09
         0.13
         0.28
         0.21
         0.06
         0.06
         2.49
         0.09
         0.11
         0.43
         0.07
         0.17
         2.23
       562.84
       326.44
       433.01
    2,308.96
    2,144.96
    1,159.54
    1,634.76
       661.67
    13,442.51
   Total
    (gm/y)
     37.97
      3.09
      4.59
      6.80
      1.03
      6.60
      1.68
      8.13
      0.41
      0.41
     34.27
      8.08
      7.39
      1.48
      0.54
      1.22
     16.09
  3,877.42
  1,326.33
  1,539.72
 20,783.49
 44,535.95
  7,760.41
  9,220.80
  2,554.14
180,367.60
                                      143

-------
                      TABLE  11.   PREDICTED YORK CSO SUMMARY STATISTICS
       Average Annual  Number of Rainfall  Events
       Average Annual  Number of Runoff Events
       Average Annual  Number of Overflow Events

       Average Annual  Hours of Runoff
       Average Annual  Hours of Overflow
       Average Overflow Volume per Event:
            Contribution from DWF (m3) -
            Contribution from Runoff (nr)

                   Total Overflow Volume (m3)

       Total Annual Overflow Volume  (m3)'
                  5,102

                456,186
           TABLE 12.  PREDICTED YORK ANNUAL CSO POLLUTANT LOADINGS
1,2-dichlorobenzene
1,2,4-trichlorobenzene
1,3-dichlorobenzene
1,4-dichlorobenzene
Hexachlorobenzene
A-endosulfan
B-endosulfan
4,4-DDT
4,4-DDE
Aldrin
A-BHC
Chlordane
Dieldrin
Endrin
Heptachlor
Heptachlor epoxide
Total PCB
Arsenic
Cadmium
Chromium
Copper
Lead
Mercury
Nickel
Selenium
Zinc
0.
0.
0.
1.
DWF Contr.
 (gm/y)

   10.16
    0.16
    0.24
    0.35
    0.05
      ,34
      .09
    0.42
    0.02
      .02
      76
    0.42
    0.38
    0.08
    0.03
    0.06
    0.83
  219.09
   81.33
  298.81
1,283.75
2,585.38
  408.98
  575.87
  151.03
9,955.49
Runoff Contr,
   (gm/y)

    43.85
     0.69
     1.02
     1.51
     0.23
     1.46
     0.37
     1.81
     0.09
     0.09
     7.61
     1.79
     1.64
     0.33
     0.12
     0.27
     3.57
   945.85
   351.10
 1,289.99
 5,542.02
11,161.24
 1,765.57
 2,486,06
   652.03
42,978.51
CSO Loading
   (gm/y)

    54.00
     0.84
     1.25
     1.86
     0.28
     1.80
     0.46
     2.22
     0.11
     0.11
     9.37
     2.21
     2.02
     0.40
     0.15
     0.33
     4.40
 1,164.94
   432.43
 1,588.80
 6,825.76
13,746.62
 2,174.55
 3,061.93
   803.06
52,934.00
                                     144

-------
 HAZPRED STATUS

     At present, the dry weather portion of the model  has  been verified in two
 Ontario catchments.   It  is  planned to carry out further  verification (dry
 weather only) in nine additional Ontario catchments.   Wet  weather data collec-
 tion in three Toronto area  catchments will be completed  this year and will be
 used in connection with  simulation modelling to verify the wet weather model
 alogrithms.
                                YORK
        RUNOFF
                                               DRY  WEATHER
                  TOTAL  PCB   4
      Figure 4.   Total PCB loading distribution for the York catchment.
                                 REFERENCES
Adams,  B.J.,  and Bontje, J.B.  Microcomputer applications of analytical  models
for urban  drainage design.  _[n:  W.  James  (ed.), Emerging Computer Techniques
for Stormwater  and Flood Management.   ASCE, N.Y., 1983. pp. 138-162.
                                   145

-------
Adams, B.J., and Gemmell, R.S.  An analysis of individual and grouped perfor-
mance of regionally-related wastewater treatment plant.  TP-73-06-01, Urban
Systems Engineering Centre, Northwestern University, Evanston, Illinois.
1973a.

Adams, B.J. and Gemmell, R.S.  Performance of regionally-related wastewater
treatment plants.  Journal of the Water Pollution Control Federation.  45:10,
2088, 1973b.       	

American Water Works Association.  A training course in water distribution.
Manual MS, 1962.

Arthur D. Little, Inc.  Sources of toxic pollutants found in influents to
sewage treatment plants:  VI - Integrated interpretations.  EPA 68-01-3857.
U.S. Environmental Protection Agency, Cincinnati, Ohio, 1979.

Canviro Consultants Ltd.  Development of the HAZPRED model phase I.  Prepared
for Ontario Ministry of the Environment, Toronto Area Watershed Management
Study, Toronto, Ontario, 1984.

Canviro Consultants Ltd.  Development of the HAZPRED model phase II.  Prepared
for Environment Canada.  Toronto, Ontario, 1985.

Canviro Consultants Ltd.  HAZPRED - user's manual.  1986.

Cooper, B.O.  Ontario Ministry of the Environment, Toronto, Ontario, 1984.
Personal communication.

Marsalek, J., and Greek, B.  Toxic substances in urban land runoff in the
Niagara River area.  National Water Research Institute.  1983. Draft report.

U.S. EPA.  Treatability manual.  EPA-600/8-80-042b, U.S. Environmental Protec-
tion Agency, Office of Research and Development, Washington, D.C., 1980.

U.S. EPA.  NURP priority pollutant "monitoring program, Volume I: Findings.
U.S. Environmental Protection Agency, Washington, D.C. 1982.

Zison, S.W. Sediment-pollutant relationships in runoff from selected agricul-
tural, suburban and urban watersheds.  EPA-600/3-800-022, U.S. Environmental
Protection Agency, Environmental Research Laboratory, Athens, Georgia, 1980.
 The work described in this paper was not funded by the U.S. Environmental
 Protection Agency and therefore does not necessarily reflect the views of
 the Agency and no official, endorsement should be inferred.

                                      146

-------
             ALTERNATIVE  CALIBRATION  OF  THE  QUAL-TX  MODEL  FOR
                         THE UPPER TRINITY RIVER
                            by Robert McCarthy
                         Dallas Water Utilities
                           Dallas, Texas 75201
                                ABSTRACT
      The  objective of  model  calibration  is  to  establish  the  values  of
free  parameters  by  forcing an  agreement  between  model predictions  and
field  measurements.    In   water  quality   modeling,   the   kinetic  rate
coefficients  are  usually  the  model   parameters   of   concern.   It  is
frequently  the  case,   however,  that  there  are  more  model  parameters
requiring definition than  there  are data sets available.   In this study,
Che QUAL-TX  model,  a version of QUAL-II,  was calibrated  for a  210  mile
portion of  the Upper Trinity River which  flows  through and  south of the
major metropolitan areas of Dallas -and  Fort Worth, Texas,  and  into which
five major wastewater treatment  plants discharge  a  combined average daily
flow of over 400 MGD.

     Only one  set  of field data was  available  that was  satisfactory for
model  calibration,  namely  an intensive survey  performed  by   the  Texas
Water  Commission.   To  emphasize  the  role  of   subjective  judgement  in
choosing coefficients when there is a lack  of direct measurement defining
real   river   processes,  two  distinctly   different   calibrations   were
performed based  upon  the  observed  dissolved oxygen  measurements.   The
first  calibration  was  based  upon  the  postulate  that nitrification  is
largely  responsible  for dissolved oxygen  depletion  in  the river,  and
employed a  carbonaceous decay coefficient  of 0.1 per  day,  a coefficient
of nitrification  ranging from 0.3  to 0.5  per  day,  and sediment oxygen
demand   rates   approximately  0.1   gram/m^/day   throughout  the   river
length.  The  second  assumed that carbonaceous decay and  sediment oxygen
demands are  more important than nitrification.   For  this  calibration  a
carbonaceous  BOD decay coefficient  of  0.3  per  day,   a  coefficient  of
nitrification of  0.1 per  day,  and sediment  oxygen demand  rates  in  the
order  of  2  grams/m2/day applied  principally in the  urban areas,  were
assumed with all  other coefficients the  same as  in the first calibration.
No  significant  difference in   goodness of  fit  between  predicted  and
observed  DO values  was  obtained   between the   two  calibrations.   Both
postulates are plausible; however,  the implications  on  control strategies
of employing one set of coefficients, over  the other are quite different.


                                    147

-------
The  first  set would  imply the  need for  advanced  treatment with  strict
ammonia  removal* while  the  second would  require  strict  processes  for
carbonaceous  BOD and sediment control,  and perhaps even nonpoint  source
controls.
                               INTRODUCTION

The Upper Trinity River  flows  through  and  south of the major metropolitan
areas of Dallas and Fort Worth, Texas.  There  are  four major forks of the
Upper Trinity  River -  all of  which are  highly regulated  by  reservoirs
utilized  for water  supply  by municipalities  in  the Dallas/Fort  Worth
area.  The Clear  Fork of the Trinity  is  regulated by  Benbrook  Lake,  the
West Fork  by Lake Worth  and Eagle Mountain Lake,  the Elm  Fork  by  Lakes
Grapevine  and  Lewisville,  and the  East  Fork  by  Lakes  Ray Hubbard  and
Lavon.   Numerous   tributaries,  a  few  of  which  are  also  controlled  by
reservoirs,  complete  the  drainage basin  system for  the Upper  Trinity.
Figure 1 shows the major features of the Upper Trinity River basin system.

A  water quality  modelling  study  was performed  covering  the   210  mile
portion  of  the  river  from Beach  Street  in Fort Worth to  Highway  31  at
Trinidad,   south   of   Dallas.    Five  major   municipal  treatment   and
approximately  30   minor  municipal   and   industrial   treatment  plants
discharge wastewater  effluent  into  the modelled reach.   In addition two
major municipal plants  discharge  effluent into  the East  Fork which  flows
into  the modelled  reach.   The  major  dischargers  and   their  treatment
capacities are shown  in figure 2.  Figure  2  also shows  the locations  of
United  States  Geological  Survey  flow  gaging   stations  and   existing  and
planned water quality monitors.

                                BACKGROUND

Flows  in  the  Trinity   basin  are  highly  variable.   Under  low  flow
conditions the river  flow can consist completely of wastewater  effluent.
A frequency  analysis  performed  by  the  Texas Department of Water  Resources
determined the 7 day 2 year  low flow,  less  discharger  return flows,  to  be
0.0  (zero) cfs at  USGS gaging  stations  located in Grand Prairie between
Fort Worth and Dallas,  at Commerce Street  in  Dallas,  and at  Highway 34,
78 miles south of Commerce  Street  (1).  At  other times, immediately  after
rainstorms   or   later  when  flood  water   is  released  from   upstream
reservoirs,  wastewater  flows  represent only a  small percent of  the  total
flow in the river.

Periods  of  low flow  have always  caused  water  supply and  water quality
problems for residents of  the  area.  Early residents  in  this region were
compelled to construct impoundments  upstream of the  cities to store  water
to improve water  supply  dependability.  However, these  impoundments  have
further  aggravated  the  water  quality  problems by  reducing the  already
insufficient natural flow in the river.
                                    148

-------
                I   LEWISVILLE
                •      LAKE
                                                      -I
                                                       •
                                                 „ LAKE  I
                                                 RAY HUBBARO
 EAGLE
MOUNTAIN
 LAKE !
                       , LftKE _,.
                       ARLINGTON
           BENBROOK	
 	>	I-KIHT—	-——	T
                                                                CEDAR CREEK
                                                                    LAKE
            FIG. I.    TRINITY  RIVER / RESERVOIR  MAP
Low  flow  summer conditions,  when most  of  the  river  flow  is wastewater
effluent, have  caused  oxygen depletion  in  the  river near the cities for
many  years.   The  return  flows  have  effectively  created  an artificial
hydrological  environment   inviting   fish   to   swim   upstream   in  the
effluent-augmented  streamflows.    Fish  kills  have   occasionally  been
observed downstream after  rise events associated with  rainfall occurring
in the  Dallas/Fort Worth areas.   Much  speculation  over the  cause of the
kills is ongoing  even  now and many  studies are being  planned to address
this  issue.  At the  present time it  appears  that the  kills are  rainfall
related and occur within a certain  range of rising  stage  in the river at
Highway  34  south of Dallas.  The  automated  water  quality  monitors show
slugs  of  water  having extremely low dissolved  oxygen levels traveling
downstream; however,  the source of  the  low  DO,  either from nonpoint  urban
runoff, or  resuspension of  bottom  sediments  from wastewater effluent or
nonpoint  runoff  deposits,  is unknown.   It  is  known that  fish  kills
generally do  not or have  not been known to  occur  for rise events  below
2300  cfs  or  above  10,000  cfs and  generally fall  in  the  2300-3500 cfs
range.

                                     149

-------
  The  water problems  in the  Trinity are  difficult  to  solve.   The  small
  river  is  inadequate to support  the  large population of  the  river basin.
  Water  supplies are  being  supplemented by  transporting water  from  other
  river  basins.  The  resulting return flows further increase the wastewater
  loads  on  the stream.

  In   recognition   of  the  effluent  domination  and high  summer  water
  temperatures,  the stream  standard  for dissolved  oxygen  has been set  at
  1.0  ppm for  flows below 80 cfs above Beach street and  set  at  3.0 ppm for
  flows  above 80 cfs  above  Beach Street.  This standard  changes  to 5.0 ppm
  dissolved oxygen  for all flow conditions at the Trinidad boundary.

  Recognizing  the  major  contribution  that  wastewater  effluent  has  in
  maintaining the base flow  in the Trinity  River,  the major plant operators
  have long been interested  in the water quality  of the river.   In October,
  1975,  the cities  of Dallas  and  Fort Worth, the  Trinity  River Authority,
  and  the  North Texas Municipal  Water  District  formed  the Upper Trinity
  River  Basin  Water   Quality   Compact  for  the  purposes  of carrying  out
  cooperative  programs  related  to  water  quality  in  the  Upper  Trinity
      WISE. CO.
                    OEWTON CO.
                            LEWISVILLE LAX
              EAGLE MOUNTAN  GRAPEVINE
                 LAKE       LAKE
                BENBROOK
                  LAKE
          LEGEND
•^ Waetewater Treatment Plant
 • USGS Flow Gaging Station
 A Automated Quality Monitor
 A Planned Automated Quality  Monitor
       Waitewater Treatment Plants
 1  Ft. Worth 100MGD      4 Dalits Southslde 30MQD
 2  TflA Central 100MOO    6 TRA 10 Mil*  10MQD
 3  Dallas Cantral 150MQD  6 Garland  30MQD
	7 Mosquito 12MOD
     FIG.  2 MAJOR WASTEWATER DISCHARGERS AND
             USGS  MONITORING SITES IN THE
             UPPER TRINITY RIVER BASIN
                                      150

-------
River.  Under  this agreement  four continuous  automated monitors  (CAMS)
were  installed  to  measure select water  quality parameters.  The  Compact
members and  the US Geological  Survey have  had joint  funding  agreements
for these  monitors since 1977.   A fifth  monitor  funded by  the City  of
Dallas and  the USGS was  installed in February 1984.  Arrangements  have
been made between the Compact and USGS  for two additional  monitors  to  be
installed  in spring of  1986,  and  the  City  of Garland  and USGS  should
install an 8th monitor later in 1986.

These  monitors  'continuously  measure   four  water  quality  parameters,
recording  data  at hourly  intervals.   These  parameters   are:  dissolved
oxygen,  water  temperature,  specific  conductance,  and  pH.   Discharge
measurements are also  recorded.   In addition to the  continuous monitors,
the Cities of  Dallas  and  Fort  Worth and the  Trinity  River  Authority have
long  had various individual  sampling  activities within the  modelled reach
in which weekly  or monthly  grab samples  were taken at  different sites  on
the  river.   While  the  above  data are  quite useful  in  showing  that  an
improvement  in overall dissolved oxygen levels and  other water  quality
parameters has  taken  place  in  the Trinity in  recent  years as  the  major
wastewater plants  have  been upgraded, they  are inadequate to  develop  an
understanding of river processes.

In  1983  the Compact  realized  it  was  not enough  just  to  know what  the
status of  certain  water  quality  parameters  was in the Trinity.  Rather,
an  understanding  of river  processes  was  desired.   The Qual-Tx model,  a
version of Qual-II, was  acquired from  the  Texas  Water  Commission  (TWC,
then  the Texas Department of Water  Resources)  and  implemented on the City
of  Dallas  computer system.   While the  aforementioned sampling programs
provided a wealth  of  water quality information, due  to the uncoordinated
grab-sample  nature of  the  surveys  and  the  fact  that no  BOD or  N^-N
data  were  acquired by  the CAM monitors,  inadequate data  existed  for
calibrating  the  model.   There  were   however,   two   intensive  surveys
completed by the TWC at the  time  of our  study (2).   The intensive  surveys
included sampling at 35 sites  from Beach Street in Fort  Worth  to  Highway
31  at  Trinidad  in  addition   to  sampling   the  effluent  at  the  major
treatment plants.  A composite sample was made  from the four  to five grab
samples taken at each site spaced over a  sixteen hour period from predawn
to  after  dark.  In-situ  readings at the  time of  each grab sample  were
also  recorded.  The TWC generously made  these data sets  available  to the
Compact for  use in  studying  the  river.   Under the  guidance of  Dr.  George
Ward  of Espey  Huston  & Associates, the  Compact attempted  to utilize  the
data  sets in calibrating the Qual-Tx model.

                             DATA SET REVIEW

The  first  step  in  employing   these  data   sets  was  a  review  of  the
conditions under which they were  obtained.   The Qual-Tx model is designed
to be applicable only to  steady-state conditions.   This means a condition
whereby all  components  of the  system,  including  streamflow,  wasteloads,
water  temperature, etc.,  are  steady  in time throughout the  modelled
reach, and the  dissolved  oxygen profile has  equilibrated to  these steady
components.   For  the  data  sets  to  be  applicable  for  calibration  and

                                     151

-------
verification  purposes   they  must  represent  a  reasonable  steady  state
condition.

The  first and  easiest  check  for  steady  state  conditions  for  the  two
available data sets was  a review of antecedent streamflows in the river.
If  flow  was  too  highly  variable  immediately  prior  to  or  during  the
intensive  survey  sampling activities,  then  the sampling would  not  have
been  made under  steady  state  conditions  and the data sets would not be
applicable for calibration.   In fact this was  found to be  the  case for
one of the data sets.

Figure  3  shows  the daily  streamflow  values  at  USGS gaging  stations for
the  30 days  prior to  and during  the  sampling for  the September,  1982
survey.   As  it  turns  out  the  survey was  conducted  just  a  few days after
flood  releases from upstream reservoirs were discontinued.  The  declining
flows  shown on figure  3 are characteristic of  gradual decreases  in flood
releases  as  reservoir  levels  approach  the  top of  conservation  storage.
Further   review   showed  that   in  fact  very  high  releases   had  been
continuously  made for  a  full  thirteen months  immediately prior to the
survey.
       8 i
       7 -
 to ~
         13  15  17  19  21  23  26  27  20  31   2   468   10  12
                                       DAY

   D BEACH ST.  GAGE  + COMMERCE GAGE    O  LOOP 12 GAGE    A TRINIDAD GAGE
                         FIG. 3 USGS STREAMFLOW
                             13 AUG TO 13 SEPT 1982

                                     152

-------
 While streamflow had steadied  for  two or  three  days before  the  survey,
 this was  insufficient  to  ensure steady  state hydrological  conditions.
 For  this   to  have  occurred,  steady  flows  should  have  been  achieved
 throughout the model reach  for  a  period at  least  equal  to the travel time
 from the head to the end  of the reach.  This  would  have  been on the order
 of 10 days for  the normal  expected  low flows.   The  September,  1982 data
 set was therefore  judged  to  be  inapplicable  for calibration  purposes.
 Analysis  then  turned  to  the  October,  1982  data  set.    Figure  4  shows
 streamflows for  the  30  days during  and prior to  this survey.   While some
 minor variations  in  flow  were  noticeable,   the  data  set  was  judged
 applicable for calibration  purposes.

                              CALIBRATIONS

 Since there were  more  model  parameters  requiring definition than  data
 sets available  to  ensure  the  assumed  rate coefficients  mirrored  real
 world conditions,  and to emphasize  the  importance  of correct  subjective
 judgement  in choosing from  a  range  of "reasonable"  values,  two sets  of
 rate coefficients,  both  within ranges cited in  the general  literature
      7 -
~     6-
w^
fel
      z -
                                     0  11   13   IS  17   19   21   23  25
       25   27   29   1
                                     DAY
 o BEACH ST. GAGE     + COMMERCE GAGE     O LOOP 12 GAGE
A TRINIDAD GAGE
                       FIG.4 USGS STREAMFLOW
                           25 SEPT TO 25 OCT 1982
                                    153

-------
were utilized  in  developing, two distinctly different  calibrations  based
upon the observed dissolved oxygen measurements.

The  first  calibration assumed,  at  20 C, a  carbonaceous  BOD decay  rate
of 0.1  per  day,  a coefficient  for  nitrification ranging from 0.3 to  0.5
per  day,  and a  benthal  oxygen demand on the  order of 0.1  gram/m2/day.
This  last  coefficient would  assume  that  benthal  oxygen  demand  in  the
Trinity  primarily  originates   with  point  source  dischargers   and  the
advanced waste  treatment permits in  effect  for area  operators  (10  rag/L
BOD, 15 mg/L TSS) keep these demands very low.

The  second  calibration, assumed  a carbonaceous  BOD decay rate of 0.3  per
day, a  coefficient for nitrification of  0.1 per day and a  benthal oxygen
demand  rate  within  the  urban corridors  on  the order  of 2  grams/m^/day.
The  2g/m2/d is a  moderate demand  and  its  use would  assume that  there
were sources of  oxygen demanding deposits other than point  sources.   All
other coefficients were unchanged between the two calibrations.

Figures 5, 6 and 7 display  the  measured  dissolved  oxygen values from  the
October, 1982 intensive survey.  The vertical bars represent  the range of
values  obtained throughout the  survey and the  small circles  indicate  the
arithmetic average of the  readings  at  each  sample station.   The modelled
dissolved oxygen  profiles  from  the two  alternate calibrations  are  also
plotted.  Both profiles could be considered as good calibrations.

The Root Mean Square value,  the square root  of  the average  of the sums of
the  differences  between  the observed and  the  predicted values  squared,
provides one measure  of  the "goodness of fit".  These values were  0.538
and  0.356  respectively  indicating  the  second  calibration was   somewhat
better than the first.
       z
       HI
       o
       X

       Ujg

       o
       
-------
 DISSOLVED OXYGEN
       ppm
tf-i
                                                                KILOMETERS
	CALIBRATION 1
	CALIBRATION 2   FIG. 6  ALTERNATE CALIBRATION RUNS
                           COMPARED TO DATA POINTS
                               CONCLUSIONS

With  two essentially equally  good  calibrations,  which  is  better?   The
answer is that we cannot tell  from  the  available  information.  Of  course
it was known  at  the outset that this was  the  case.   The purpose of  the
exercise then was  not to  have a completely calibrated  model which  one
could confidently employ in answering management  decision questions.   On
the contrary,  the  purpose  was  to show  that  Trinity  River processes  are
far from understood and that decisions  made  from one set of  calibration
assumptions could  be  very  different from  decisions  made from the  other
set of calibration assumptions.

For example, a control strategy decision made from the first  calibration,
would  call  for  advanced  treatment with  strict  ammonia  removal.   In
contrast, a  strategy  based on the  second  calibration,   which  emphasizes
carbonaceous  kinetics  and  sediment demands,  would call   for   strict
carbonaceous BOD and sediment  controls,  and  perhaps  even nonpolnt  source
controls.  Either strategy requires  substantial capital  Investments.   At
this time the major plant  operators  on the  Upper Trinity  are  committed to
achieving 10  mg/L  CBOD,  15 mg/L TSS, and  3 mg/L NH3-N  warm weather,  5
mg/L  NH3-N  cool weather limitations as called  for  in  the  most  recent
wasteload allocation (3).

A  concerted data collection effort  is now  underway to collect additional
data  sets  to  help establish "correct"  coefficients  for  future  modelling
efforts.  Future  calibrations  will  center not  only  on  predicted  versus
actual measurements of dissolved oxygen, but also  on comparisons  of BOD,

                                    155

-------
DISSOLVED OXYGEN
       ppm
tt-

10-
 B-
 7-                                                                	^_	

 *" —•••—-^^-- ..*•-""*>^	v	—	5

 «-                                                                       5
 3-                                                                       £

 '"I                                                                      I
 H                                                                      *



                                  KILOMETERS                             <
                                                                        g

                 FIG. 7 ALTERNATE CALIBRATION RUNS           1
 	SiuS2i™2 I        COMPARED TO DATA POINTS
    PALI On A I IVJN Z


  NH3~N,   NC-3-N+N02~N,   and   perhaps   chlorophyll-a.    Three   intensive
  monitoring surveys  have already been performed by Compact members  under
  fairly   steady   state   flow   conditions.   More  will  be  completed  as
  conditions permit.   Additional  special studies  directed  toward  better
  definition  of  the  nitrification  and   CBOD  decay   processes,  an
  investigation of observed  in-stream nitrogen losses, algal kinetics and
  photosynthesis/respiration,   as  well  as a  reanalysis  of  hydrology and
  hydraulics relationships are also planned.

                                REFERENCES

  1.   Texas Department  of Water  Resources.  Waste Load Evaluation For The
      Upper Trinity   River  System  In  The   Trinity  River  Basin.    Texas
      Department of Water Resources,  June 25, 1985. P.22.

  2.   Davis,  J.  W.   Intensive  Survey of  the  Trinity  River Segment  0805.
      IS-53, Texas Department of  Water Resources, June, 1983. 62pp.

  3.   Texas Department  of Water  Resources.  Waste Load Evaluation For The
      Upper Trinity   River  System  In  The   Trinity  River  Basin.    Texas
      Department of Water Resources,  June 25, 1985. 299pp.
 The work described in this paper was not funded by the U.S.  Environmental
 Protection Agency  and  therefore  the contents do not  necessarily  reflect
 the views of the Agency and no official  endorsement should be inferred.

                                     156

-------
               LOGNORMAL1TY OF POINT AND NON-POINT SOURCE
                         POLLUTANT CONCENTRATIONS
                    by :  Eugene D. Driscoll
                         (Wood-ward Clyde Consultants)
                         Oakland, New Jersey 07436
                                   ABSTRACT


      This paper presents a series of probability plots of water quality data
from a variety of discharge sources. It Is intended to provide a visual display
of the appropriateness of characterizing the variable pollutant concentrations
by a log normal distribution.

      Representative examples of observed data that have been analyzed and
plotted to  test whether they can be treated as lognormally distributed random
variables, are presented for data sets from the following applications:

                     Highway stormwater runoff

                     Combined sewer overflows

                     Urban runoff

                     Point Source discharges from POTW's

                     Agricultural runoff

      Such examination suggests that a lognormal distribution either actually
defines the underlying population of  pollutant concentrations, or is  at the least
a satisfactory approximation for most environmental analyses.


INTRODUCTION

      It has always been  recognized that natural processes are variable. The
value of  quantifying this variability in some appropriate way has Increased

                                     157

-------
substantially in recent years. This is because many of the issues, problems,
situations that the environmental engineering community  is now dealing with
are more effectively addressed in a probaballstlc context.

      A  number of articles have appeared in the technical literature over the
past five years or so, in which the authors report that the particular  data
they are dealing with has a log  normal distribution. Almost invariably, such
statements carry a qualifier. Some examples

      The log normal distribution was found to fit .... most consistently.

      (Three distributions).  . . were found  to be adequate, -with log normal
      preferred because of ease of application.

      Visual examination . .  . indicated that data were best described by a log
      normal distribution.

      Generally log normal distributions -were observed for all but the data
      extremes.

      Overall, the . .  . data tend  to fit a log normal distribution.

      Space constraints  will nearly  always preclude the inclusion of any
significant number of distribution plots in a paper or report, so the reader has
no independent  basis for deciding  how -well,  or how  poorly,  the general
conclusion on lognormality really applies.

      The simple objective of this paper is  to present a set of  probability plots
for -water quality data from  a number of different applications, to provide the
reader -with a visual picture of the extent  to which the sampled  observations
fit a log  normal distribution.

      During the inspection of the probability plots, it will be useful to bear in
mind the following considerations. 1 submit that the important  issue is not
•whether a  specific data set can  be reliably concluded to have a log normal
distribution. The  real issue is -whether it is appropriate or reasonable to infer
that  the underlying population  of events  represented  by  the  sample  of
observations is lognormally distributed.

      Each data set is a  small sample of the much larger population of values
represented by the sample. The particular sample taken will be representative
of the underlying population  to an unknown extent, and may be fairly good or
rather poor. There is no way to  resolve this satisfactorily when it is the only
sample  available  for Inspection.   However, -when similar pollutant data are
available for inspection  from a  large  number of comparable sites, or for a
variety  of  other pollutants  at  the  same  site,  the  information to guide the
desired inferences is  extended.
                                     158

-------
HIGHWAY STORMWATER RUNOFF

      The probability distribution of event mean concentrations (EMC's) of four
pollutants are  shown for each  of four highway sites. Three  of the sites,
Nashville, Milwaukee and Denver are urban highways. The Harrisburg site is
in a rural setting. Figure 1 is for TSS, Figure 2 for  Total N, Figure 3 is for Lead
and Figure A is for Zinc.  These  data are from a study currently  in progress
for the Federal Highway Administration.
COMBINED SEWER OVERFLOWS

      The probability distribution of event mean concentrations (EMC's) of BOD
(Fig 5) and suspended solids (Fig 6) are shown for four CSO sites.  Three of the
sites, are in Richmond VA, the fourth in Toronto.

      Figure 7 shows the distribution of the site median concentrations of BOD
and TSS at the six CSO sites that were monitored in Richmond, and at 13 sites
from 6 cities. All data is from the University of Florida Data Base.
URBAN RUNOFF

      The  probability distribution of  event mean concentrations (EMC's) of
Total P (Fig 8) and COD (Fig 9) are shown for six urban runoff sites.  Figure 10
shows the  distribution of the site median concentrations of Total P at 69 of the
urban runoff sites that •were  monitored under the Nationwide Urban Runoff
Program (NURP)
POINT SOURCE DISCHARGES - POTW's

      There  have been a number of studies over the past 6 or 8 years that
have looked  at  the variability  of  daily average  effluent concentrations  from
municipal  sewage  treatment  plants.  The  authors  have,  as cited earlier,
concluded  that  distributions of  daily  average  concentrations  are at  least
adequately approximated by a log normal distribution. I haven't had access to
such daily values and have no plots to present to illustrate this level of detail.
However  the papers  and reports do  present summaries of the  mean and
coefficient  of variation (COV) of daily values  for a sample that generally covers
one to three years of  plant records.

      Figure 11  shows the  distribution on  mean  influent concentrations  of
Cadmium for a large sample of POTW's.  Figure 12 shows the distribution of
the overall mean effluent BOD concentration, and  the COV of daily values for 66
Conventional Activated Sludge plants. The correlation plot  shows that there is
no relation bet-ween the mean effluent that a particular plant produces and the
degree of  variability of daily  concentration  values  - for all plants in this
process category.
                                     159

-------
  I
    HIGHWAY RUNOFF  -  SITE EMC's   (TSS)
6U6PENDCO SOLIDS  
-------
            HIGHWAY RUNOFF  -   SITE EMC's   (TOTAL N)
TOTAL N  (mg/l)
  10- tT     NA6HVILLC 1-4
  \0'0
    -2
-I
   0

Z SCORE
TOTAL N  (fin/I)
  10* IT      MILWAUKEE HWV4
  10*0
  10'-
    -2
          0

      Z SCORE
                                 MEDIAN >   O.D2
                                 COV 3   0.68

                                  N=2I
                                 MEDIAN 3
                                 COV >

                                  N:23
                            1.38
                            0.64
                                       TOTAL H (mg/t)
                                        10- IT     DENVER HIGHWAV SITE
                                                      10*0
                                                      I DM
                                                         -2
                                       TOTAL N
                                        10'I
                                                      10*0
                                                      10*->
                                     -2
                                                                        MEDIAN =
                                                                        COV s

                                                                         H=I6
                                                                          0.93
                                                                          0.66
                                                    -I        0

                                                           Z SCORE
                                                               HARRIS8UR6 1-81
                                                                 MEDIAN 3
                                                                  COV  >

                                                                  NsIB
                      0.70
                      O.S9
   0

Z SCORE

-------
                  HIGHWAY RUNOFF   -  SITE EMC's   (LEAD)
LEAD 
-------
                       HIGHWAY RUNOFF  -  SITE  EMC's    (ZINC)
CA>
        ZINC  (mg/l)
         10* Or
       NASHVILLE 1-40
        ZINC  (mg/l)
         10* IT      MILWAUKEE  HWV4S
         i
-------
        COMBINED SEWER OVERFLOWS  -  SITE EMC's   (BOD)
eOO-5
 1C' 3
 10'2
  
-------
             COMBINED SEWER OVERFLOWS  -  SITE EMCs   (TSS)
en
       T65 (mq/1)
        10-3i
        I0'2
        10'1
          -2
       TSS  (mo/I)
        10-3
        10*2
        lO'l
          -2
  BLOOD v
-i
        0

     Z SCORE
   0

Z SCORE
                                    O. 80
                                 N s  19
                         TSS «mg/l)
                          10'
                                                10'2
                                                 10* I
                                                  -2
                              T66
                               10'3T
                                                10-1
                                                  -2
                                         -I
                                           0

                                         Z SCORE
                                     BAST YofcK
                                                                         N a 14
                                                                         N a 22
-I
   0

Z SCORE

-------
 10*
   -2
TSS (mg/l)
 10'
 10-2
   -2
           COMBINED SEWER OVERFLOWS  -  SITE MEDIANS
                              57
                              0.33
                          N a 6
                  0

                Z SCORE
EULHMOOD tfA
             0.34-
                          N 3 6
•1      0


     Z SCORE
                                          BOD-S  (mg/l)
                                           10-2T
                                           10*1
                               -2
                              TGB  (mg/l)
                                1C'3r
                                           KT2
                                             -2
                                       13 CSO
                                       «a (, CITIES
                                                   MLDMAJ  5?
                                                     N a 13
                                                 0

                                              Z SCORE
IS C5o  SITES /u 6 CITIES
                                                                   N a 13
                                       I       0

                                           Z SCORE

-------
                               URBAN RUNOFF
                             DRTR  FROM NURP STUDY

           EVENT MEAN CONCENTRATIONS  OF  CHEMICAL OXYGEN DEMAND
                                AT  SIX  STUDY  SITES
COD Owj/ll
 ID" 5     CO I  NORTH AVE DRAIN
 ID'?
 10'I
                             WOIAM t  279 22
                              CDV >   0.74
    -2
            •I        0

                  I SCORE
coo (ma/it
 IQ-3     m I  PARKING LOT
 10-2
  ro-i
                             MtDlAHt   7933
                              COT .   0.72

                              N.JJ
    2
            -I
                     0

                  t SCORE
COP  (roj/0
  10- 3      HI I  WAVIRIY H11L6 1W.ET
  ID1 2
  10-
  lO'O
                              MEDIAN t   '4973
                              CUV  •    O.M

                               Ht27
    -2
-t        0
      t scon
                                18-3™*    WAI UK H1UI
                                             10-2
                                             10-1
                               COO  (mt/0
                                 10- 3      TI I HA8I IANE
                                              10-2
                                              10-1
                                                             KDIANt   SII8
                                                             COV  i    08J

                                                              Mill
                                    2
                                                    0

                                                  I SCORE
                                COD
                                 10-J
                                              10-2
                                              to-1
                                                      DC I WtCUtIGM IM.ET
                                                             KDIAN »   45 02
                                                              COV •   0,«
                                                 2
                                                               t ectvtf
                                         167

-------
                               URBAN  RUNOFF

                             DBTfl  FROM  NURP  STUDY

                    EVENT MEAN CONCENTRATIONS  OF  TOTAL P

                                AT  SIX  STUDY SITES
TOTAL P (mo/I)
 tO" 1     CO t  NORTH AVt MAIN
  10" 0
  10'-
    -I        -I
                      0

                   I ECORt
TOTAL P (mj/l)
  1C' 0»     NH I PARttrHGLOT
  10'-
  10'-7
                              HEDIAN'    017
                               COV  •    1.21
    -1        -1
                   I  SCOOT
101 Al P (iM/1)
  tO'P      HI I  WAVIRLY H1UC INLtT
  W-l
  IO'-3
                               MEDIAN.    017
                               COV  •    0.64

                                NiJS
    -i
• 1        «

      t KDM
                                TOTAL P  |mf/D
                                  10-1,     WA I
                                               10-0
                                               lO'-J
                                                           LAKE HILtS
                                                                           fUDIMI.   020
                                                                            COV •   0.01

                                                                            H.I127
                                          -2     -1
                                                            I     2
                                                                t KOBE
                                 TOTAL P 
-------
  11
        URBAN RUNOFF

       DflTB FROM NUHP STUDY
SITE MEDIAN CONCENTRATIONS OF TOTAL P
  FOR  69 URBAN RUNOFF STUDY SITES
                      LOO MEAN
                      LOO SIGMA

                       MEAN
                       SIGMA
                       MEDIAN
                     COEF VAR
                       5,787
                       400.143
                       284.545
                       324.092
                        0.711
            ****#***»**»*********»**»*»»»##**»»»*»*»
TOTAL P (ug / I)
  10* 4T
  NT 3
  IOA2
   ICT I
                                               N  =   69

               -2
          "H        o~~

                Z  SCORE
                               169

-------
99.99

99.95
99.9
 0.01
                          I  I I Illlll—TTT
        o.ooor        o.oo i          001
                                    i
                   Log Concentration CO (mg/IJ
	    to*
r i i nun
         1.Q
                     POTLJ- INFLUENT
                              MEDIHN
                    CHOMIUM CONCENTRRTIONS
                                                                         DATA BASE FOR INFLUENT
                                                                         HEAVY METAL CONCENTRATORS
                                                                         IN POTW3

                                                                         EPA 600/52-81-220
                                                                         2 YEAR STUDY
                                                                         DATA FROM 239 PLANTS
                                                                         'Generally, log normal distributions
                                                                         war* observed for all but ine data
                                                                         txtrvmts.*

                                                                         •Overall, tho Individual plant m*an
                                                                         and median metal concentration
                                                                         data tend to nt a log normal distribution."
FIGURE SHOWS

LOG PROBABILITY PLOT
for reported
MEDIAN CADMIUM CONCENTRATIONS

-------
 UJ
f^^
^%
rv =
1 1 1 1
; **(a
, 1
-3 -2 -19 1 2 3
Z SCORE
*n &. -
O
o. -
   -3   -2
-10    1
 Z SCORE
                                            POTUI

                                     EFFLUENT CHRRflCTERISTICS
                                            BOD
                                           .*-
                                          o  r
                                                                r - o. o i
                                                      PLT MERNS

-------
 AGRICULTURAL RUNOFF

        Figure 13  shows  the  distribution  of the  EMC's  of three  pollutants
 (ammonia, nitrate and TKN) at two study sites in Watkinsville^Georgia.

        The ammonia nitrogen EMC's for agricultural runoff from  four study
 sites in the Four Mile Creek watershed in Iowa  are presented in Figure  14.
 Total  Phosphorus  EMC's at two of these  sites are  shown  by Figure  15. The
 second of the plots  for  each site  excludes  4  observations  during  one of the
 •winter periods when  manure was applied on frozen ground and not worked
 into the soil.

        Figure 16  displays the distribution  of  annual  average values  for the
 concentration of  soluble phosphorus and nitrogen  in both surface  runoff and
 subsurface  storm-water discharges  at  the site of  a long  term  study  at a
 different location in  Iowa.  The  final  plots describe the  variability of the
  concentration ratio between subsurface and surface discharges.
  CONCLUSIONS

        1.   Whether or  not it Is rigorously true that the distribution  of the
  underlying population of pollutant concentrations from the indicated sources ar
  log normal (as suggested by  the material  presented), it  appears  to  be an
  acceptable  approximation that will be adequate for many of  the uses to which
  it  may be put.

        2.  Lognormality can be assumed for the  distribution of EMC's at a site,
  or for the distribution of mean or median concentrations from a number  of
  different sites that are in  the same category.
     The work described in this paper was not funded by the U. S. Environmental
Protection Agency and therefore the contents do not necessarly reflect the views of
the Agency and no official endorsement should  be inferred.

                                       172

-------
                  AGRICULTURAL NPS RUNOFF
                          UlflTKINSUILLE 6fl STUDY
               EVENT MEAN CONCENTRATIONS OF NITROGEN FORMS
                        ( NH4 - N ,  N03 - N  ,  TKN )
    STATION P2
NH4-N (mg/1)
 to-1
 10* 0
 lO'-
 Itr-
H03-H
 10' 7
 io- 1
 to-o
 10
,o-r
 TWI
 10'
  lo-i
  (0 0
  ID'
                _-
              ,^r
            •<•
   -2      -»
               Z SCQRt
                 -»-
           -»      0
               Z SCORE
                             1.93
                         N • 3»
                          Pz.
                       MEM/»»J
                              l.fcl
                         H • 39
                          P2.
                          B . 37
                  -»-
         •I       b
              Z SCORE
                                                      STATION P4
                                   NH4-N (mg/1)
                                      10'0
                                      10-
                                      IT-3
                                                              N i  M
                                               -I
                                                      -H
                                                     «
                                                  Z SLORL
H03-N (mg/l)
  '
 P4-
                                      to-
                                      ID-
                                                                 I- 50
                                                              H • IT
                                              -I      0
                                                   Z-SCORE
TKN (rog/1)
 lO"
P4-
                                      10'0
                                       10'-
                                                               N • If
                                         I      -I
                                                       -t-
                                                       0
                                                     Z SCOKE
                                  173

-------
              AGRICULTURAL  NPS  RUNOFF

                   FOUR MILE CREEK  IOUIR STUDY
          EVENT MEAN CONCENTRATIONS OF AMMONIA NITROGEN

                   AT FOUR MONITORING STATIONS
NH4-N
 10- 1
 10' 0
 IC'-
 W-}
 10-
    1
NH4-N
  In 1
  10' •>
  w-
  w-
                •+-
                 0

              Z SCORE
    j
-+ —
 o
               Z SCOKE
NH4-N
 lO'l
                                   lo-o
                                   lO'-
                                   w-
                                      2
 NH4-N (mg/1)
 10-0
                                    10'-
                                    In-.
                                                  STAft
                                                         » • M
         -1      0      1

             Z SCORE
                           0.148
                          0.61

                       h t tt
          I       •*

              Z SCORE
                                                          I
                                174

-------
  fl
AGRICULTURAL NPS RUNOFF
     FOUB MILE CREEK lOUJfl  STUDY
         EVENT MEAN CONCENTRATIONS OF DISSOLVED PHOSPHORUS
                     AT TWO MONITORING STATIONS
     STATION 1                                       STATION 2
OISSPWO f (mg/l>
 10-1
 10'0
 10'-
 w-i
   -i
  4-Mil£. CfUEJZK.
     ITff-life
                                    10'1
                                    10'0
                                    to-i
                                    ID-:
                                                 t STOW
biss P UHI wmrtR ISBQ)
 10'0
 10'-
 10'-
          -t
                 4 Mile
                 0
               1 ttORt
                       - r?flo
                   DISS f iKHf WIHItR 801
                    10'0
                                    10'-
                                    4MI e/Z£e£
                       -J
                               175

-------
 u
       SURFACE AND SUBSURFACE RUNOFF

              CONVENTIONAL  TILLAGE
           ALBERTS md SPOONER. J SOIL AND WATER CONSERVATION. FED 1905


       ANNUAL AVERAGE CONCENTRATIONS OF SOLUBLE POLLUTANTS

                                               NITROGEN (N03)
PHOSPHORUS (P04)
 IP .'
      BO ( pt/l )
 I0'7
rp.i '.out
 10-I
 tfO
 lO'O
 w-
                      B t 10
         -I     «


             i
               /W*ju*t.
               CTOM
                          r.el
                      N « ID
         -I      0


              i ecm
                              mu cone-sum Act m
                               to-1
                                 w-
                                           2.44
                                   -J     '-I
                                           t stoet
                              IW CVR CUPflfRf TO nif/l

                                '
                                 10' I
                                 .10-
                                            5. IS
                                   2     -I
                                             I SCOfit
                                o.ipnrt/cu-r tot-c«*no
                                 10-1
                                   -j
                                            I tton
                             176
                                                      H . 10
                                                      N • 10

-------
           POLLUTIOM FROM HIGHWAY RUNOFF—PRELIMINARY RESULTS

             by:    Philip E. Shelley
                    EG&Q Washington Analytical Services Center, Inc.
                    Rockville, Maryland  20850
             and
                    David R. Gaboury
                    Woodward-Clyde Consultants
                    Walnut Creek, California  94596
                                   ABSTRACT

      This paper presents the preliminary results of a project aimed at developing
models that can be used by planners and highway engineers for predicting pollutant
runoff from highways.  It is a part of an ongoing research program being conducted
by the Federal Highway Administration of the U. S. Department of Transportation to
characterize stormwater runoff from highways, assess its potential impacts on re-
ceiving waters, and determine the effectiveness of various control measures.

      A brief review  of  different  approaches to predicting pollutant runoff  loads
from highways is followed by a description  of the data base being assembled as a
part of this project.  The probabilistic data analysis methodology that is being used
to characterize highway storm-water runoff  is then described in some detail.  Pre-
liminary analytical results are presented in two main areas—rainfall and runoff
data and water quality data.

      For the sites examined  so  far, it is shown that taking the percent impervi-
ousness of an unmonltored site  as Its runoff coefficient, with an  upper bound of
around  0.9  to  account  for initial abstraction,  offers a  reasonable first estimate.
Thus, the runoff quantity from  an ungaged  highway site can be estimated from a
simple analysis of rainfall records from a gage near the site.

      Based upon the  their overall characteristics, the highway sites  in  the data
base have  been classified  as being either "urban" or "rural."   Best  estimates  for
pollutant concentrations for each of six pollutants (SS, COD, TKN, TPO4,  Pb, and Zn)
for highway sites are presented and compared with corresponding values for urban
runoff.  Lacking site specific data, they represent reasonable first estimates of high-
way stormwater runoff  quality. It  is also  shown that suspended solids can be a
reasonable indicator for other pollutant concentrations in highway runoff.
                                     177

-------
                                INTRODUCTION

     For over ten years the Federal Highway Administration (FHwA) of the United
States  Department  of Transportation has  been  engaged in a research program to
characterize stormwater runoff from  highways, assess Its impacts  on receiving
waters, and determine the effectiveness of various  control measures for  possible
Instances where their use to mitigate impairment of designated beneficial uses of
receiving waters might be required.   As a part of this research program, a con-
tractor team headed by Woodward-Clyde Consultants, and including EG&G Washing-
ton Analytical Services Center, Inc. and the University of Florida,  has been tasked
•with developing models that could be used by planners and highway engineers for
predicting pollutant runoff loads from highways.

     Approaches to  predicting pollutant runoff  loads  from high-ways  can  be
generally grouped into one of two classes;

            • those that are based on regression equations, and

            • those that use some form of simulation model.

Approaches that use simulation models  can be further characterized mechanistical-
ly into three types:

            • those employing rating curve methodologies—actually
              an extension of regression analysis based on a power
              law  relationship between flow and sediment load or
              concentration;

            • those where pollutant build-up and -wash-off
              relationships are used as, for example, in SWMM,
              STORM, HSPF, and the FHwA Urban Highway Storm
              Drainage Model; and

            • probabilistic based approaches, wherein probability
              density functions are assigned to  runoff flows and
              concentrations, and the resulting  pollutant loads are
              characterized  statistically, as was done, for example, in
              the Environmental Protection Agency's Nationwide
              Urban Runoff Program.

Thus,  prediction of highway stormwater runoff quality may be performed using a
number of  procedures that  range from the very simple to detailed  deterministic
simulation models,  and a large body of  literature exists describing them.  However,
all of the approaches have one common aspect—their predictive capabilities tend to
be rather poor -without suitable site-specific data for calibration.  For this reason,
one of the first tasks in the  present effort was  to assemble a data  base of highway
runoff characteristics.  In the following sections of this paper, that data base will
be described, the data will be summarized, and  best preliminary estimates of high-
way stormwater runoff will be made.
                                      178

-------
                 HIGHWAY RUNOFF DATA BASE CHARACTERISTICS

     The data in the highway stormwater data base is taken  from monitoring
projects supported by the FHwA and  several states.  At the present time, there is
coverage of twelve sites, with several more to be added as the data become avail-
able.  There are  from seven to 139 monitored events at each  site.  The data base
consists of three categories of data;

           •  rainfall and runoff,

           •  water quality, and

           •  fixed site.

Each of these categories will be discussed in turn.

RAINFALL AND RUNOFF DATA

     Rainfall  data were taken with  recording raingages.  For each rainfall event
there is  a start  time and date, five-minute  raingage  readings that  describe the
hyetograph, and  a  stop  time and date.   In  a similar  fashion, runoff data were
taken from flowmeters and consist of a start time and date,  (typically) five-minute
instantaneous  flow readings that describe the  hydrograph, and  a stop time  and
date.  The lag  between the time rainfall starts and the time runoff starts, although
highly variable for a  given site due to antecedent conditions, provides an indication
of basin response time or "flashiness."

     Taken together, the rainfall and runoff  data describe  the  water quantity
characteristics of the site. The rainfall data can be analyzed  to provide informa-
tion on rainfall frequency (i.e., the time between storms), intensity, duration, and
total amount.  For example,  the total quantity of rainfall for  the event is simply
the difference  between the raingage readings at the beginning and  end  of the event.
The flowmeter data can be analyzed  to determine peak  flow,  time to peak, total
quantity discharged, etc.  The total quantity of runoff discharged  during the event
is computed by integrating the flow rate record over the time period of the event.
Of particular interest in  this preliminary analysis are the total rainfall and runoff
quantities.

WATER QUALITY DATA

     The highway runoff quality data consist of the results of laboratory analyses
of sequential discrete samples  that were taken at recorded times throughout the
runoff event,  plus the results of  laboratory analyses  of  flow-weighted composite
samples taken over the entire runoff  event.

     Although over 24 analytical determinations were made for the different sites
contained in the data base so far, there was consistent coverage  of up to eighteen
pollutants, and these were selected for this preliminary analysis.   By category,
they are:

           •  Solids                                   TS, TSS,  VSS

           •  Physical Parameters                     pH, Cl~

                                     179

-------
            •  Oxygen Demanding Substances            BOD, COD, TOC

            •  Nutrients                                TKN, N02+N03, TP04

            •  Metals (Total)                            Cu, Pb, Zn, Fe, Cr, Cd

            •  Hydrocarbons                            Oil and Grease

FIXED SITE DATA

     The fixed site  data, so called because they tend to be fixed rather than var-
iable on an event to event basis, fall into three general categories. They are:


     •  Highway Site Data

            — configuration (e.g., elevated, ground level, depressed)

            — pavement composition, quantity, and condition

            — design, geometries, cross-sections

            — vegetation types on right-of-way

            — drainage features

     •  Operational  Characteristics

            — traffic characteristics (e.g., density, speed, braking)

            — vehicle characteristics (e.g., type, age, maintenance)

            — maintenance practices (e.g., sweeping, mowing, weed control)

            — institutional   characteristics   (e.g.,   litter   laws,   speed   limit
              enforcement, vehicle emission regulations)

     •  Surrounding Land Use Characteristics

            — land  use  type  (residential,  commercial,  industrial,  agricultural,
              forest)

            — geologic  features  (relief,  soil  types  and   horizons,  groundwater
              characteristics)

            — agricultural practices (e.g., tillage, irrigation, and cropping practices)

In this preliminary  analysis, the chief fixed site data to be used are  the size of the
drainage area, the average traffic density expressed as vehicles per day, the per-
centage of the basin area  that is impervious, and  three roadway features—the
surface type,  number of lanes, and the presence of curbs.
                                      180

-------
                         DATA ANALYSIS METHODOLOGY

     Pollutant runoff from highway sites (as well as from other types of nonpoint
source sites) is  quite  variable, especially as  compared to  discharges from point
sources.  Therefore, new methodological frameworks are required for analysis of
data from nonpoint sources in general and highway runoff in particular, especially
with respect to  characterization  of  pollutant loads  and their possible  effects upon
receiving -waters.   New approaches  to water quality criteria  and standards  also
seem warranted when one is dealing with nonpoint sources, due to the high vari-
ability and intermittent nature of the latter.

     When one  has flow and polutant concentration data for runoff from a moni-
tored site, there is  a need for a sensible way to reduce the data to summary form
in a way that will be useful to decision makers.   For this task, it was  decided to
follow  the general data analysis approach developed for the  Environmental Protec-
tion Agency's Nationwide Urban Runoff Program (NURP).

     In NURP, flow and concentration data for a site were first analyzed to deter-
mine event mean concentrations  (EMCs).  The EMC is defined as the concentration
that would result if the entire storm event discharge were collected in a  container,
and its concentration determined; i.e.,  it is  the total  mass of pollutant discharged
during the event  divided  by the total  quantity  of water  discharged during the
event.   In practice the EMC is simply taken to  be the concentration of a flow-
weighted composite sample  collected over the duration of  the runoff  event.  If
sequential discrete  samples are taken over the period of  the runoff event, an ac-
ceptable approximation to the EMC can be formed by manually compositing aliquots
that are sized In proportion to flow (manual flow-weighted  composite sample)  and
performing analytical  determinations on the resulting composite sample.   If separ-
ate analytical determinations  have  been performed  on the individual  sequential
discrete samples, an acceptable approximation to the EMC often can be  computed
from the individual concentration values and the corresponding flow values.

     The main  point is that, In dealing with nonpoint sources, the  basic unit of
•water  quality information is the  average concentration of each pollutant  of interest
in the total volume of runoff produced by each  individual storm event.  Thus,
within-storm fluctuations  in pollutant  concentration  are  completely  ignored.
Hence, for a particular pollutant and a particular site, there will be one EMC value
for each storm event  monitored, and the  EMC  is considered  to  be  the random
variable.

     Given a set of nonpoint source  monitoring data for a site,  the first step is to
determine the EMC values for the pollutants of interest. This set of EMC values is
then analyzed to compute the statistics of runoff quality for the site.  However, in
order  to properly  interpret the  site statistics, it  is usually convenient to know
something about the nature of the underlying  distribution function (or, equivalent-
ly, the underlying  probability density function).   From physical considerations,  it
can be shown that a normal distribution cannot be exactly correct for EMCs (e.g.,
pollutant concentrations can never be negative, they are always skewed, etc.).  It
can also be shown, by appeal to the central limit theorem, that when a  series of
multiplicative processes are involved (as  is  the case  with  nonpoint sources), the
resulting distribution function 'will  be asymptotically lognormal.  The adequacy of
this lognormality  assumption has  been checked  with  data  from urban runoff,
highway stormwater  runoff, combined sewer overflows, and limited sets of  agri-

                                      181

-------
cultural runoff data.  In no case has the lognormal distribution been clearly Inap-
propriate,  and in most  cases  it seems to  fit the data better than other possible
distributions.

      Under the asumptlon that the lognormal distribution would be an adequate
representation for the highway runoff data sets in question, it remained to com-
pute the EMC values for each  event  at a site, transform these values into the log-
arithmic domain, compute  the mean and variance in log-space, and then properly
transform these values into the appropriate statistics in arithmetic space (typically
the median, mean, and coefficient of variation).  For surface runoff data, the two-
parameter lognormal distribution is  adequate.  It is completely specified by  a cen-
tral tendency parameter (say, the median) and a dispersion parameter (say the
coefficient  of variation).

      Having determined the two EMC statistics for a pollutant at a site (say, the
median and coefficient of variation), they  completely describe the variable  runoff
characteristics for that  site.  They can then be compared with similar data from
other sites to evaluate similarities and differences in the context of  physical site
characteristics.   This amounts to comparing the underlying probability distribu-
tions for the  sites in question.  The main point  is that the foregoing methodology
recognizes  that the natural processes that lead to highway runoff are highly vari-
able and provides an appropriate way of quantifying this variability.

      Each data set is a  small  sample of the much larger  population of values rep-
resented by the sample.   The particular sample taken will be representative of the
unknown population to an unknown extent, and may be fairly good or rather poor.
There is no way to resolve  this satisfactorily when one  has  only  one  sample
available for inspection.  However, when similar pollutant data  sets are available
for inspection from a large number of comparable  sites, or  for a variety of pol-
lutants at  the same site, the information to guide the desired inferences is extend-
ed.  One way to shed further light on the  representativeness of  a nonpoint  source
data set is to compare the rainfall statistics for the monitored events with similar
statistics computed from long-term  rainfall records for  the site.  If the average
rainfall volume for monitored storms is, for example, 0.80" while the long-term
average for all storms is only 0.40,.", one can be reasonably  certain that the mon-
itored storms are not representative of the total population of storms at that site.

      For reasons just alluded to, as well as the fact  that an integral part  of the
assessment of the impact of storm loads on receiving water quality is the  statis-
tical evaluation of rainfall records, a program to summarize the important rainfall
variables was set up on FHwA's computer  as a part of this effort. The purpose of
the Synoptic Rainfall Data Analysis Program (SYNOP) is to provide the user with a
tool that can be used to summarize and statistically characterize a  rainfall  record
in terms of its important variables (volume, intensity, duration,  and time between
storms).  Since hourly rainfall records of many years duration  are cumbersome
and difficult to analyze, SYNOP provides an easy to use tool  to facilitate the  deter-
mination of, for  example, seasonal trends  important to the assesment  of impacts
and selection of control alternatives for storm-related loads.

      SYNOP  summarizes the hourly rainfall data by storm events, each -with an
associated  volume (inches), duration (hours), average intensity (inches/hour), time
since the previous storm (hours) as measured from the end of the previous  storm,
the antecedent rainfall (inches) for the last 24, 48, 72, and 168 hours,  the hours of

                                      182

-------
missing data, and the hours that the gage did not read.   A storm definition, or
interevent time, must be established  to determine  when, in the hourly record, a
storm  begins and ends.   SYNOP  delineates  storm  events as  rainfall  periods
separated by  a minimum number of consecutive hours without rainfall, a user
supplied  number.   After the entire record has been read, SYNOP  computes the
statistics  of relevant storm parameters by  month and year and  for  the entire
period of record.


                             PRELIMINARY RESULTS

RAINFALL AND RUNOFF DATA


     Let  us  first  consider the  relationship of runoff to rainfall.   Figure 1 Is a
scatter plot of the rainfall and  runoff expressed in inches  for  a  site on Highway
Number 45 in Milwaukee, Wisconsin.  There  it can be seen  that, although there is
some scatter, there does tend to be a rather strong linear  correlation,  supporting
the notion that a runoff  coefficient  might adequately characterize the  site and
allow runoff quantity to  be predicted from rainfall data with a known degree  of
uncertainty.   To examine this notion further, a scatter plot of runoff  coefficient
(Rv = runoff/rainfall) versus rainfall is presented  in Figure 2.   The lack of any
correlation further supports the notion of a single runoff coefficient.

     Although the claim for a single runoff coefficient is Indicated in the foregoing,
it is less  than totally compelling.   One of the reasons for the scatter might lie  in
the limited sample size for that site.  We now turn to a  site  on Highway 94  in
Milwaukee for which there are 137 monitored events.  The corresponding scatter
plots are  presented as Figures  3  and 4.  As can be seen,  the  argument becomes
much more compelling.

     A  summary of the highway site rainfall and runoff data is presented  in
Table 1.    The computed  statistics (mean, median,  coefficient  of  variation—the
standard  deviation divided by the mean and  abbreviated as COV,  and the number
of samples N) are  indicated for rainfall quantity, storm duration, runoff quantity,
and runoff coefficient—the runoff divided by rainfall and abbreviated as Rv.  The
results of regression analyses of the form

           Runoff = A x Rain +  B         and          Rv = A  x Rain + B

are presented in Table 2.   In the first  case, the value of  A is  the  desired runoff
coefficient, and one wants the residual term (B) to be small.  In the second case,
the value of  B Is the runoff coefficient,  and one wants the  value  of A to be small.
As can be seen from Table 2,  the foregoing tends to be the case  for all but the last
two sites.  The mean, median, and coefficient of variation for the runoff coefficient
for each site  are presented for comparison purposes in Table 2.

     Since one would expect that the  percent impervious area will affect the run-
off coefficient for a site, this is examined in Figures  5  and  6. Taken together, these
figures suggest that, lacking any other information, it is not unreasonable to assign
a runoff coefficient equal to the percent impervlousness for a site,  with an upper
bound of around 0.9 or so to account for initial  abstraction.


                                     183

-------
RO
0.90-r
0.80
0.70-
0.60
0.50
0.404
0.30
0.204   .   .
0.10-
   0.00
              .•  •  *
                 H	1-
                                    _,	^
      0.00  0.20  0.40   0.60  0.80   1.00   1.20  1.40   1.60
                              Rainfall
   Figure 1. Runoff (RO) versus rainfall for Milwaukee Highway 45
1.60
1.40
1.20
1.00
Rv 0.80-
0.60
0.40

0.20-
O OA-

*

.
* *
*
* •• *. * *
^ ^
t •

         0.00  0.20   0.40  0.60   0.80  1.00   1.20   1.40  1.60
                                Rainfall
Figure 2. Runoff coefficient (Rv) versus rainfall for Milwaukee Highway 45
                              184

-------
RO
1.40
1.20
1.00
0.80
0.60
0.40
0.20
                        •  **
   0.00 44MC^	«-
       0.00   0.20   0.40   0.60    0.80    1.00    1.20    1.40
                                Rainfall
    Figure 3. Runoff (R0> versus rainfall for Milwaukee Highway 94
      1.60-
      1.40 •
      1.20
         1.00 • •«••»
      Rv
         0.80
         0.60
         0.40
         0.20
                «*   •
            *
            •
            •«
.*  *•    *•   • .*•
»    *****     *
    •  * •     «
                                                •    «
         0.00    0.20   0.40   0.60   0.80
                               Rainfall
                                               1.00    1.20    1.40
   Figure 4. Runoff coefficient (Rv) versus rainfall for Milwaukee Highway 94
                                 185

-------
              TABLE 1.  SUMMARY HIGHWAY SITE RAINFALL/RUNOFF DATA
          RAIN
          On.)
        DUUTMX
          (hr)
         OJNOFF
          (In.]

nun
•tadm
COV
N
ritm
1Mi«n
COV
N
rw,
tadltn
COV
N
Mtm
tedltn
COV
N
Dffivtr r
1-25
042
027
1,14
16
523
1.93
251
16
0.194
0.066
2.01
16
0.36
0.31
054
16
III*****
1-94
0.27
0.14
1.64
137




024
0.11
169
139
0.84
0.81
0.28
137
NHwlItt Si
MO
0.63
0.44
i.se
31
5.34
2.76
1.66
31
0.34
0.16
1.97
31
0.41
045
0.58
31
crimmto IH1wM*M n
Kwy.SO Hwy.45
052
0.32
Ut
34




0.45
026
1.43
34
042
0.91
0.18
34
0.54
0.37
1.0S
29
729
3.34
1.94
29
0.23
0.13
1.53
29
0.42
0.35
0.60
29
ltlwiuk«« KKTlibirg
I-79S 1-flHPrUI)
0.63
0.39
126
35
556
328
1.36
35
0.53
0.32
1.30
34
0.66
0.65
0.14
35
0.71
0.49
1.07
21




O.IB
0.02
6.63
21
0.11
0.04
2.32
21
Eflind Htrrbturg
UK l_Al /Ok t\
i u«j 1^0 1 \KH.U
0.86
0.66
0.64
36




0.64
0.33
2.33
39
0*7
0.47
1.03
36
0.76
0.61
0.74
23
6.14
4.35
1.58
23
0.40
0.16
2.03
23
0.44
0.31
1.72
23
RUNOFF QUALITY DATA

      To compare the runoff data from the highway sites in the data base, data for
six of the eighteen water quality parameters are presented In Table 3.  Since data
from NURP and other nonpoint source studies have indicated that data  from indivi-
dual  sites tends to be  lognormally  distributed,  the summary statistics given  in
Table 3 were computed  in the same  way as the  summary statistics for an indivi-
dual  site.  In order to provide a  basis for comparison, the median values computed
from the NURP data base are also presented in Table 3, along  with  the ratio of the
highway median values to them.  Thus, it can be seen that  the concentrations of
suspended solids and phosphorus in  highway runoff  are  virtually the same  as
those in urban  runoff,  concentrations of COD and zinc  are slightly  higher In  high-
way  runoff, and concentrations of lead  and TKN are significantly  higher in  high-
way  runoff.  In reviewing these  numbers, it  is  important to remember that we
                      TABLE 2.  RUNOFF/RAINFALL REGRESSIONS
          Milwaukee. 1-94      Runoff-

          Milwaukee, 1-795     Runoff-

          Socremen (o. Hwy 50    Runoff-

          Milwaukee, Hwy 45     Runoff-

          Nashville, 1-40      Runoff-

          Oanver, 1-25        Runoff-

          Hvrljfigro Ph II. 1-81   Runofr-

          EHond, 1-85         Runoff-

          Harrlsourg (>h 1.1-81   Runoff-
0.86 RAIN- 0.00  Rv-  0.04 RAIN » 0.83

0.84 RAIN • O.Ot  Rv-  0.06 RAIN * 0.80

0.88 RAIN- 0.02  Rv-  0.06 RAIN » 0.79

0.45 RAIN- 0.02  Rv- -0.02 RAIN » 0.43

0.33 RAIN- 0.04  Rv- -0.02 RAIN* 0.41

0.60 RAIN- 0.06  Rv-  0.37 RAIN » 0.21

0.29 RAIN- 0.08  Rv-  0.IS RAIN- 0.01

0.94 RAIN- 0.20  Rv-  0.20 RAIN » 0.41

0.84 RAIN- 0.2S  Rv-  0.39 RAIN » 0.12
Maen
0.84
0.66
0.82
0.42
0.42
0.36
0.11
0.67
0.44
Median
0.61
0.85
0.81
0.35
0.35
0.31
0.04
0.47
0.31
COV
0.26
0.14
0.16
0.66
0.58
0.54
2.32
1.03
1.72
                                        186

-------
                   0.00  0.10  0.20  0.30  0.40  0.50  0.60  0.70  0.80  0.90  1.00
                                    Runoff Coefficient

            Figure 5.  Percent imperviosness versus mean runoff coefficient
arc talking about concentrations and not loads.  On a load basis, highways would
be much smaller contributers, because they account for only a small percentage of
the land in an urban area.

      Ratio  techniques are frequently used to facilitate comparison among similar
data  sets,  since they  provide a simplistic form of normalization.  In the present
case we have divided the median value for a pollutant at a sjte  by the median
value for all  of the sites  to form the ratio.  A value  greater than unity indicates
that that particular site is "dirtier" than average, while a value less than unity
indicates that the site is  "cleaner" than  average.  The results  are presented  In
              0.00  O.!0 0.20 0.30 0.40 0.50  0.60  0.70  0.80  0.90
                              Runoff Coefficient
1.00
        Figure 6,  Percent imperviousness versus median runoff coefficient

                                       187

-------
            TABLE 3.  SITE MEDIAN DATA SUMMARY AND COMPARISON

                         SITE        SS   COD  TKN  TP04   Pb   Zn
                                   (mg/l)(mg7l)(mg/l)(mg/l)(mg/l)(mg/l)

                   DENVER            410  289 3.38  0.82  0.68  0.62
                   MILWAUKEE HWY 795   183  130 2.52  0.38  2.03  0.46
                   LOS ANGELES        172  196 4.22  0.45  0.99  0.55
                   MILWAUKEE HWY 45    343  134 2.76  0.45  0.88  0.44
                   NASHVILLE          190  113 1.90  1.69  0.41  0.26
                   MILWAUKEE HWY 94    161  122 3.20  0.30  0.90  0.52
                   WALNUT CREEK       224  120 224  0.41  0.75  0.30
                   HARRISBURG (Ph. II)   184   34 2.20  1.08  0.03  0.17
                   SACRAMENTO         90   51 1.90  0.12  0.28  0.27
                   HARRISBURG (Ph. I)     31   31 1.20  0.29  0.10  0.06
                   EFLANO             19   49 250  0.13  0.01  0.06
                   BROWARD COUNTY      9   38 0.68  0.06  0.23  0.07

                   Mean              221  113 2.46  0.55  1.03  0.35
                   Median            108   86 2.18  0.35  0.31  0.24
                   COY               1.78  0.85 0.52  1.21  3.21  1.09

                   NURP MEDIAN        100   65 0.68  0.33  0.14  0.16

                   HIGHWAY/NURP      1.08  1.32 3.21  1.07  2.18  1.48
Table 4.   The last column in Table  4 (labeled mean) is a sort of site cleanliness
index, formed by simply averaging the ratio values for each of the six pollutants at
the site.

      Based upon data in the fixed site data base, we have chosen to simplistically
group the highway sites into one of  two categories, urban and rural,  and repeat
the ratio  analysis just  described.  The  results  are presented in Tables 5 and 6.
From these data, it  appears  that median pollutant concentrations in runoff from
urban highway  sites are from two to four times greater than those found in  ur-
ban runoff, while those from rural highway sites tend to be around one quarter to
a little over one half of that found in urban runoff.   TKN is  an  exception to  this
last observation, and the data will be examined  more carefully to attempt to  find
an explanation.

      Many workers  have tried to relate pollutant concentrations  in runoff to sus-
pended solids values, attempting to use the latter as an indicator parameter.  We
have  done this on an event basis, and sample results for the  Denver highway site
are presented in Table 7.  The relatively low coefficients of variation suggest that,
for this site at  least, such an approach  has some validity.  Summary results for
all of the urban highway sites are given in Table 8.  Only the coefficient of varia-
tion  for lead  seems  surprisingly high.  The two suspect values  are those  for the
Milwaukee High-way 795 and the Harrisburg Phase II sites,  and they will be exam-
ined critically in subsequent analyses.  Similar results for  the rural  highway sites
are given in Table 9.
                                       188

-------
    TABLE 4.  SITE MEDIAN RATIO SUMMARY AND COMPARISON

        SITE           SS    COD  TKN  TP04   Pb    Zn  MEAN

DENVER               3.78  3.36  1.55  2.32  2.23  2.62  2.64
MILWAUKEE HWY 795   1.69  1.51  1.16   1.08  6.65  1.94  2.34
LOS ANGELES          1.59  2.28  1.94   1.27  3.24  2.32  2.11
MILWAUKEE HWY 45     3.17  1.55  1.27   1.26  2.87  1.86  2.00
NASHVILLE            1.75  1.31  0.87  4.77  1.34  1.10  1.86
MILWAUKEE HWY 94     1.49  1.42  1.47  0.85  2.95  2.20  1.73
WALNUT CREEK        2.07  1.40  1.03   1.16  2.46  1.27  1.56
HARRISBUR6CPh.il)    1.70  0.40  1.01  3.06  0.09  0.71  1.16
SACRAMENTO          0.83  0.59  0.87  0.34  0.92  1.14  0.78
HARRISBURG (Ph. I)     0.28  0.36  0.55  0.82  0.32  0.26  0.43
EFLAND               0.18  0.57  1.15  0.37  0.04  0.25  0.42
BROWARD COUNTY      0.08  0.44  0.31  0.17  0.75  0.30  0.34
 TABLE 5. URBAN SITE MEDIAN DATA SUMMARY AND COMPARISON
         SITE          SS   COD  TKN  TP04   Pb   Zn
                     (mg/1) (mg/l) (mg/1) (mg/l) (mg/1) (mg/1)

  DENVER               410   289  3.38  0.82  0.68  0.62
  MILWAUKEE HWY 795    183   130  2.52  0.38  2.03  0.46
  LOS ANGELES           172   196  4.22  0.45  0.99  0.55
  MILWAUKEE HWY 45     343   134  2.76  0.45  0.88  0.44
  NASHVILLE             190   113  1.90  1.69  0.41  0.26
  MILWAUKEE HWY 94      161   122  3.20  0.30  0.90  0.52
  WALNUT CREEK        224   120  2.24  0.41   0.75  0.30
  HARRISBURG (Ph. II)      184    34  2.20  1.08  0.03  0.17

  Mean                 234   149  2.81  0.70  1.31  0.42
  Median                220   124  2.72  0.59  0.55  0.38
  COV                  0.36   0.67  0.27  0.66  2.14  0.47

  NURP MEDIAN          100    65  0.68  0.33  0.14  0.16

  HIGHWAY/NURP        2.20   1.91  4.00  1.78  3.94  2.40
                            189

-------
   TABLE 6.  RURAL SITE MEDIAN SUMMARY AND COMPARISON

        SITE          SS   COD  TKN  TP04  Pb    Zn
                    (mg/l) (mg/l) (mg/1)(mg/l)(mg/l)(mg/l)

SACRAMENTO            90    51  1.90   0.-12  0.28  0.27
HARRISBURGCPh. I)       31    31  1.20   0.29  0.10  0.06
EFLAND                  19    49  2.50   0.13  0.01  0.06
BROWARD COUNTY         9    38  0.68   0.06  0.23  0.07

Mean                    42    43  1.65   0.16  0.26  0.12
Median                  26    41  1.40   0.13  0.09  0.09
COV                  1.24   0.23  0.62   0.71  2.64  0.85

NURP MEDIAN           100    65  0.68   0.33  0.14-0.16

HI6HWAY/NURP         0.26   0.64  2.06   0.39  0.67  0.56
       TABLE 7.  DENVER SITE SOLIDS RATIO RESULTS

      EVENT SS/SS COD/SS 100TKN/SS 100TP04/SS lOOPb/SS lOOZn/SS
1
3
4
5
6
7
6
9
10
11
12
13
H
15
16
Mean
Median
COV
N
1.00
1.00
1.00
1. 00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1. 00
0.00
15
0.84
0.56
0.42
0.56
0.72
0.62
0.60
0.52
0.63
1.52
1.29
0.43
0.67
0.88
0.58
0.75
0.70
0.37
15
1.27
1.12
0.82
0.29
0.59
0.42
0.61
1.36
0.34
0.95
3.20
089
082
1.05
1.03
0.99
0.82
0.66
15
0.29
023
0.19
0.12
0.17
0.15
0.17
0.14
0.19
0.19
0.48
0.19
0.21
0.20
0.25
0.21
0.20
0.34
15
0.18
0.19
0.21
0.14
0.14
0.16
0.18
0.17
0.16
0.17
0.16
0.09
0.19
0.22
0.15
0.17
0.16
0.22
15
021
0.16
0.13
0.11
0.13
0.12
0.12
0.15
0.15
0.22
0.25
0.13
0.15
0.21
0.13
0.16
0.15
0.26
15
                            190

-------
             TABLE 8.  URBAN SITE SOLIDS RATIO RESULTS

        SITE          SS/SS   COD/SS 100TKN/SS IOOTP04/SS lOOPb/SS  lOOZn/SS

DENVER                1.00      0.70       0.82        0.20      0.17      0.15
MILWAUKEE HWY 795     1.00      0.71        1.38        0.2!      1.11      0.25
LOS ANGELES            1.00      1.14       2.45        0.26      0.58      0.32
MILWAUKEE HWY 45      1.00      0.39       0.80        0.13      0.26      0.13
NASHVILLE              1.00      0.59       1.00        0.89      0.22      0.14
MILWAUKEE HWY 94      1.00      0.76       1.99        0.19      0.56      0.32
WALNUT CREEK          1.00      0.54       1.00        0.18      0.33      0.13
HARRISBURG (Ph. II)      1.00     0.18       1.20        0.59      0.01      0.09

Mean                   1.00     0.65       1.34        0.33      0.59      0.19
Median                 1.00     0.56        1.23        0.27      0.25      0.17
COV                    0.00     0.59       0.42        0.73      2.15      0.50
               TABLE 9.  RURAL SITE SOLIDS RATIO RESULTS

          SITE        SS/SS COD/SS 10OTKN/SS 10OTP04/SSI OOPb/SS 10OZn/SS

   SACRAMENTO         1.00   0.57      2.11        0.13     0.31      0.30
   HARRISBURG (Ph. I)    1.00    1.02      1.33       0.32     0.11      0.07
   EFLAND              1.00   2.58      2.78       0.14     0.01      0.06
   BROWARO COUNTY     1.00   4.22      0.76       0.07     0.26     0.08

   Mean                1.00   2.38      1.83       0.18     0.29     0.13
   Median              1.00    1.58      1.56       0.14     0.10     0.10
   COV                 0.00    1.12      0.62       0.71     2.64     0.85
      TABLE 10.  URBAN SITE COEFFICIENT OF VARIATION RESULTS

              SITE             SS    COD    TKN   TP04     Pb     Zn

       DENVER               0.60    0.71    0.76    0.51    0.56    0.59
       MILWAUKEE HWY 795    1.13    1.55    0.87    0.77    0.87    0.94
       MILWAUKEE HWY 45     0.59    0.58    0.64    0.60    0.62    0.51
       NASHVILLE            0.57    0.60    1.02    0.56    0.90    0.46
       MILWAUKEE HWY 94     1.04    0.74    0.50    0.63    1.30    0.84
       HARRISBURG (Ph. 11)     1.16    0.40    0.51    0.52    3.93    0.51

       Mean
       Median
       COY

                                   191
0.86
0.81
0.36
0.77
0.70
0.47
0.72
0.69
0.29
0.60
0.59
0.15
1.36
1.06
0.81
0.65
0.62
0.30

-------
      Thus far  we have only been dealing with  pollutant concentration  median
values.  The same types of analyses could be performed on the site coefficients of
variation, and an example for the urban highway sites is given in Table 10. Here
again, the  lead values for the Harrisburg Phase II site seem suspect and warrant
further  examination.   Otherwise, the results seem  quite  reasonable  and fairly
consistent.

      The "bottom line" of this preliminary data analysis is presented in Table 11.
If a planner has to estimate the  pollutant load coming from an ungaged highway
site,  the values for concentration and coefficient  of  variation for each pollutant
that are given there represent the best initial estimates for screening purposes.

      As our work continues, we plan to refine the foregoing  analyses  and to add
the remaining twelve pollutants  as well as other highway sites.  We also plan a
rather extensive data quality assurance and quality control effort (unscreened raw
data  have been used up to this point).  Hopefully, this will reduce the number of
seeming anomalies pointed out earlier.

                   TABLE 11. HIGHWAY RUNOFF CHARACTERISTICS


                           URBAN     RURAL      ALL    COEFFICIENT
                            SITES     SITES     SITES   OF VARIATION

         SS    (mg/1)       220        26        108      0.8-1.0

         COD   (mg/1)        124        41         86       0.5-0.8

         TKN   (mg/1)       2.72       1.4       2.18      0.7-0.9

         TP04 (mg/1)       0.59      0.13       0.35      0.6-0.9

         Pb    (mg/1)        0.55       0.09       0.31      0.7-1.4

         Zn    (mg/1)        0.38       0.09       0.24      0.6-0.7
     The work, described in this paper was not funded by the U. S. Environmental
Protection Agency and therefore the contents do not necessarly reflect the views of
the Agency and no official endorsement should be inferred.
                                      192

-------
               EFFECTIVENESS OF DETENTION/RETENTION BASINS FOR
                  REMOVAL OF HEAVY METALS IN HIGHWAY RUNOFF

               by:  Harvey H. Harper, Yousef A.  Yousef,  and Martin
                    P. Wanielista
                    Department of Civil Engineering and  Environmental
                    Sciences, University of Central Florida
                    Orlando, Florida  32816
                                   ABSTRACT
     The movement and fate of heavy metal inputs  (Cd,  Zn, Mn, Cu, Alr  Fe,  Pb,
Ni and Cr)  from highway runoff were investigated in a three-year study on 5
1.3 hectare retention facility near the Maitland  Interchange on Interstate 4,
north of Orlando, Florida.   Stormwater  characteristics were  compared with
average  retention  pond water quality  to determine  removal efficiencies  for
heavy metals  within  the  pond.   A  total of 138  sediment  core samples were
collected in the pond over a  three-year period to investigate  the horizontal
and vertical migrations of heavy metals within the pond.  Core samples were
also carried through a series of  sequential extraction procedures to  examine
the  type of  chemical associations and stability  of each  metal  in  the
sediments.  An apparatus  was built which allowed sediments to be  incubated
under various conditions of redox potential and pH to  investigate the  effects
of  changes  in  sediment   conditions   on  the  stability   of   metal-sediment
associations.

                                 INTRODUCTION


     Within the past decade,  a substantial amount of research has accumulated
relating to  the water pollution caused  by the operation of motor  vehicles.
This concern  is based largely on  the potential aquatic  toxicity of heavy
metals such as  lead, zinc, and chromium.   Heavy metals have been proposed by
several  researchers as  the major toxicant present in highway  runoff  samples
(Shaheen, 1975; Winters and Gidley,  1980).  Many  heavy metals are known to be
toxic in high concentrations  to a wide  variety of aquatic  plants and  animals
{Wilber and Hunter, 1977).

     On  a nationwide  basis, the  two most commonly used techniques  for
management  of  highway  runoff are  roadside swales and detention/retention
facilities.   As  these facilities  receive  continual inputs  of Stormwater

                                     193

-------
containing  heavy  metals,   processes  such  as  precipitation/  coagulation,
settling, and  biological uptake will result in  a large percentage of  the
input  mass  being deposited  into the sediments.   However,  no previous
definitive studies  have  been conducted to determine the  fate of toxic
species, especially heavy metals, in these stormwater management systems.   In
particular, no  studies  have been  conducted  to investigate if  physical  and
chemical  changes  which  may  occur  in these  systems  over time  may  mobilize
certain species from the sediment phase back into  the water phase.

     The purpose of this research was  to investigate the fate of heavy metals
within   stormwater  management  systems.    The  site  selected  for  these
investigations was a  series of stormwater management  facilities  located  at
the  Maitland Interchange on  Interstate 4 north of  the  city of Orlando,
Florida.   A retention pond  (West  Pond)  with relatively defined  inputs  and
outputs was chosen as the primary study site.

                               SITE DESCRIPTION
     The site selected for this investigation is the Maitland Interchange on
Interstate 4.   This interchange,  located north of the city  of  Orlando,  was
constructed in  1976 (Figure  1).  Three borrow pits dug  to provide  fill  for
constructing the  overpass  remain  to serve as stormwater detention/retention
facilities.    The  ponds  are  interconnected  by  large  culverts  and  the
northwestern  (Pond  B)  has the  capability  to discharge to the  southwestern
(referred to hereafter as the West Pond) when design elevations are exceeded.
However, under normal conditions,  the  only  input into the West Pond is by  way
of  a  45 cm concrete culvert that drains much of the  Maitland Boulevard
overpass.   Discharge from the  West Pond travels  to  Lake Lucien through  a
large culvert.   A  flashboard riser system regulates the  water  level  in  the
Wast Pond, and a discharge rarely  occurs to Lake Lucien.

     The  West Pond has an approximate surface  area of  1.3 ha and an average
depth of  1.5  m.   The pond  maintains  a  large standing  crop of filamentous
algae,  particularly Chara,  virtually  year  round.   Because  of  the  shallow
water depth and large amount  of  algal  production, the pond waters remain in a
well oxygenated state.  The  sediment material is predominately sand which is
covered by a 1 cm layer of organic matter.

     Maitland  Boulevard crosses over Interstate 4 by  means  of a bridge
overpass created during construction of the  Interchange. The Maitland
Boulevard bridge  consists  of two sections,  one carrying two lanes  of
eastbound traffic and an exit lane,  the other carrying two lanes  of westbound
traffic  and  another  exit lane.   Traffic volume  on  Maitland  Boulevard  is
approximately  12,000  average  daily   traffic  (ADT)  eastbound   and  11,000
westbound.   Traffic volume on  1-4  through the  Maitland  Interchange  is
approximately 42,000 ADT eastbound and westbound each.
                                     194

-------
I Wait land Blvd
                          West Pond  ^

                                It2
                   Exit Culvert to
                   Lake  Lucien
                 Lake Lucien
    SCALE: icm«50m
          Figure I.  Study Site at Malt!and Interchange.


                            195

-------
                             FIELD INVESTIGATIONS


     Field  investigations  conducted during 1982  and  1983 at  the  West Pond
were divided  into  the  following tasks:   1) determination of the quantity of
heavy  metals entering  the West Pond  by way of stormwater runoff;  (2)
determination  of the  average  heavy  metal concentrations in  the  retention
basin  water;  (3)  assessment  of the accumulation of  heavy metals  in the
sediments of  the pond; and  (4)  monitoring of heavy metal concentrations in
ground  waters beneath the retention basin.   To  determine  the  quantity of
heavy  metals  entering the West Pond by  way  of  stormwater  runoff,  an Isco
automatic sampler was  installed  on  the 45  on  stormsewer line.  Flow-weighted
composite samples were collected over a  1 year period for 16 separate storm
events  representing a  wide range of rainfall intensities and antecedent dry
periods.  Samples were analyzed  for heavy  metals  using argon plasma emission
spectroscopy.

     Water  samples were collected on a biweekly basis for 1 year in the West
Pond to document  average  retention pond  water quality.   Each of the five
samples was analyzed separately for the  heavy metals listed, and an average
value was calculated for each metal on each sampling date.

     The  accumulation and vertical distribution of  heavy metals  in the
sediments was examined by collection of  a  series of 2.5  cm diameter core
samples to a depth of  15  cm.  Forty-three  separate  core samples were
collected  in  the  1.3  ha  West  Pond,  and metal  concentrations  in sediment
layers  0-1  cm, 1-3.5  cm,  3.5-6.0  on,  6.0-8.5 cm and 8.5-13.0 were measured
for each core sample.  Metal concentrations in  the 0-1  on layer were used to
investigate horizontal movement  of  heavy metals from the point of discharge
into the pond.  Average metal concentrations in  each of  the sediment  layers
were used to determine the extent of vertical  migration.

     Multiport monitoring  wells  were  installed at the locations indicated in
Figure  1  to   investigate  the  possibility of groundwater  contamination by
leaching of heavy  metals  from the stormwater management  system.  Two of the
monitoring wells  were installed  at the edges  of the West Pond with the
remaining  three  installed  at  various locations  surrounding  the  stormwater
management system.  The wells  were designed so that all  of the sample ports
were housed  in  a  single  casing to minimize soil disturbance and  reduce
recovery time for obtaining representative groundwater samples  compared to
other monitoring well designs such as cluster wells.

     All wells were installed to a depth  of 6  meters with  sample ports  at 0.1
m, 0.5  m, 1.0 m, 3.0 m, and 6.0 m below the average water table depth  in the
area of the well.   Groundwater samples were  collected from each sample port
on  a  monthly  basis  using a peristaltic pump.    Approximately  10  liters of
groundwater were pumped and discarded from  each port before a  sample was
collected.  Collected  groundwater samples were  analyzed for heavy metals as
described previously.
                                     196

-------
            CHARACTERISTICS OF HIGHWAY RUNOFF AT THE MAITLAND SITE


     A  total of 16  storm events/  including a  total of 150  separate  runoff
samples were collected and analyzed over this period for dissolved  and total
heavy metals.   Total rainfall amounts  for sampled storm events ranged fron
0.33  to 3.23  inches with antecedent dry  periods of 0.24 to 25.4 days.
Average flow rates  in  the 45 cm storrasewer, an  indirect measure of rainfall
intensity/ ranged from 0.085 to 59.4 liters/sec.

     A  summary of mean flow-weighted heavy metal concentrations measured  at
the Maitland site is  given in Table  1.   Measured concentrations  of  heavy
metals in highway runoff collected at the Maitland Interchange during 1983-84
showed considerable variability between storm events as well  as during storm
events.   Average dissolved concentrations  of  all  heavy metals,  with the
exceptions of iron and  aluminum,  were  less  than 70 ug/1, with nickel,
chronium,  manganese,  and   cadmium  less   than  3  ug/1.    Measured  mean
concentrations of total metal species were  in  excess of  100 ug/1  for lead,
aluminum and  iron,  while  nickel,  chromium,  manganese,  and cadmium were all
present in average total  metal concentrations of 10 ug/1 or less.   In
general, the variability of mean flow-weighted dissolved metal concentrations
appears to  be  much  less  than that  observed for mean total  concentrations,
with most  dissolved species exhibiting a  five-fold  difference  in  range  of
concentrations, while total concentrations  exhibited over a ten-fold range  in
most cases.

     Of  the heavy  metals which  were measured,  the following  orders  were
observed for mean concentrations of dissolved and total  metal species:

     Dissolved:  Al > Fe > Zn > Pb > Cu > Mn > Ni =  Cr > Cd
     Total:      Al > Fe > Pb > Zn > Cu > Mn > Cr >  Ni > Cd

However,  the   metal  species  aluminum,   iron   and  manganese  are   common
constitutents  of soils and may  not be correlated with  vehicle  usage and
highway operation, as would  be expected for lead,  nickel, chromium,  copper,
zinc, and cadmium.   Therefore, the most common vehicle related heavy metals
found in highway runoff at the Maitland site were  lead, zinc, and  copper  in
ratios  of  4.70:1.91:1.0,  respectively, for  total concentrations, and ratios
of 0.85:1.04:1.0, respectively, for dissolved species.  Together these three
metals  accounted for approximately  91 percent of the dissolved heavy metals
present and 94 percent of the total metal concentrations, excluding  aluminum,
iron and manganese.

     With the  exceptions  of lead,  iron, manganese,  and aluminum,   all heavy
metals  in the  highway runoff samples  appeared to be present  predominantly  in
a dissolved form.  Cadmium,  nickel,  and copper  were all present  in  dissolved
fractions which were near 75  percent of  the total metal  measured.   On the
other  extreme,  lead,  iron,  manganese,   and  aluminum were predominantly
particulate  in  nature  with  dissolved fractions of  only approximately  20
percent.  Zinc and  chromium appeared to be  approximately  equal  in  dissolved
and particulate  forms.
                                     197

-------
                    TABLE 1.  HEAVY METAL CONCENTRATIONS IN SEQUENTIAL HIGHWAY RUNOFF SAMPLES
                               COLLECTED AT THE MAITLAND WEST POND DURING 1983-84
00
HEAVY
METAL
LEAD
ZINC
COPPER
NICKEL
CHROMIUM
IRON
ALUMINUM
MANGANESE
CADMIUM
METAL
SPECIES
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
Dissolved
Total
NUMBER
OF
SAMPLES
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
150
STANDARD
MEAN DEVIATION
(ug/1)
33.0
181
40.0
73.9
28.6
38.6
2.5
3.4
2.5
4.2
77.9
378
125
561
2.70
9.53
1.7
2.2
40.2
331
42.6
71.2
24.7
28.8
2.4
2.8
2.2
3.2
59.7
354
124
563
6.7
10.8
2.0
2.4
RANGE OF
VALUES
(ug/1)
7.0 -
11.0 -
1.0 -
5.0 -
6.0 -
6.0 -
0.5 -
0.5 -
0.5 -
0.5 -
11.0 -
44.0 -
19.0 -
53.0 -
a -
<1 -
<1 -
<1 -
413
3,596
324
372
175
176
15
18
16
18
466
2,172
832
3,499
59
62
12
12
PERCENT
DISSOLVED
(%)
18.2
54.1
74.1
73.5
59.5
20.6
22.3
28.3
77.3
           pH
101
6.30
0.88
4.95 -  8.49

-------
     Probability   distributions   of   mean    flow-weighted    heavy   metal
concentrations  in  the  16  measured storm  events were  examined by  plotting
flow-weighted  mean   metal  concentrations  for  each  measured  event   on
probability paper.   Total  concentrations  of heavy metals  appeared to best
approximate a  straight line relationship  with  a log-normal distribution  of
concentrations  for  the events measured.   Dissolved concentrations  of  heavy
metals also seem to approximate a log-normal distribution.  However/ cadmium
appears to exhibit a convex curvilinear relationship.

     A "first  flush"  effect was observed  for  total concentrations of  lead,
zinc, iron, and aluminum.   In general,  approximately 50 percent of the  total
mass of these metals was found to be transported during the first  quarter  of
a  storm  event, 25 percent during  the  second quarter,  and  the  remaining  25
percent divided between the third and  fourth quarters.   This trend was not
observed for total concentrations of the other metal species or for dissolved
species of any measured metals.

               FATE OF HEAVY METALS  ENTERING THE MAITLAND  POND
     Although  stormwater inputs  into the pond were  characterized by a
considerable degree  of variability, heavy metal  concentrations measured in
the pond were, in general,  relatively consistent and low  in value.   Dissolved
concentrations of all metal species in the pond  with the  general exception of
aluminum, were never found  to exceed 50 ug/1 with  dissolved concentrations of
cadmium,  zinc,  manganese,   nickel,  and  chromium  rarely exceeding  10 ug/1.
Total metal  concentrations followed a  similar  pattern with only manganese,
aluminum, iron,  and  on one  occasion lead,  exceeding  100  ug/1  on any given
sample day  at any of  the  five  sampling stations  within the retention pond.
The mean pH value of the retention pond water was  7.46 with  a  range of
6.62-8.46.   Measurements  of dissolved oxygen  indicated   an  aerobic water
column on all sample  dates.  The mean value for  dissolved oxygen was 5.6 mg/1
with measurements of ORP generally in excess of  500  mv.

     The  Maitland pond was  found  to be very effective  in removal  of heavy
metal inputs  from highway runoff.   A  comparison of summary statistics for
highway runoff  and  Maitland pond  water is given in Table  2.   Heavy metal
concentrations in this table have  been given  in  terms of  the concentrations
of dissolved and particulate species rather than dissolved  and total species.
This was done so  that  the  removal  characteristics of soluble and  particulate
species could be examined separately.

     Upon  entering  the Maitland  retention  pond,  chemical,  physical,  and
biological processes begin  to occur which,  for most  metal species, results in
substantial reductions  in  concentrations.  The  most noticeable removals for
heavy metals occurs for the particulate species.  Particulate species  of lead
and zinc are reduced in excess of 95  percent,  cadmium and iron near 85
percent, with copper  and aluminum averaging  near 75 percent.  Reductions of
particulate nickel and  chromium, however, were  much less,  with a removal of
only  25-35  percent.    The  order for  reduction  of particulate metal  species
upon entering the West Pond is:
                                     199

-------
                        TABLE 2.   COMPARISON OP SUMMARY STATISTICS  FOR HIGHWAY  RUNOFF
                                      AND MAITLAND RETENTION  POND MATER
HEAVY
METAL
LEAD
ZINC
COPPER
NICKEL
CHROMIUM
IRON
ALUMINUM
MANGANESE
CADMIUM
MKTAL
SPECIES
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
Dissolved
Particulate
STORMHATER
MEAN
(ug/1)
33.0
148
40.0
33.9
28.6
10.0
2.5
0.9
2.5
1.7
77.9
300
125
436
2.7
6.8
1.7
0.5
RUNOFF1
PERCENT
OF TOTAL
18.2
81.8
54.1
45.9
74.1
25.9
73.5
26.5
59.5
40.5
20.6
79.4
22.3
77.7
28.3
71.7
77.3
22.7
RETENTION
MEAN
(ug/1)
15.0
7.2
4.7
1.3
14.4
2.3
1.6
0.6
2.2
1.3
18.4
44.7
58.0
98.0
4.5
12.6
0.73
0.09
POND WATER2
PERCENT
OF TOTAL
67.6
32.4
78.3
21.7
86.2
13.8
72.7
27.3
62.9
37.1
29.2
70.8
37.1
62.9
26.3
73.7
89.0
11.0
PERCEWI1
REMOVAL
IN POND
54.5
95.1
88.3
96.2
49.7
77.0
36.0
33.0
12.0
23.5
76.4
85.1
53.6
77.5
-66.7*
-85. 3 J
57.1
82.0
STATE OF FLA.
CLASS III
VOTERS CRITERIA
(2/1/83)
30 (Total)
30 (Total)
30 (Total)
100 (Total)
50 (Total)
1000 (Total
None
None
0.8 (Total)
NOTES:      1. Number of observations = 150
            2. Number of observations =  30
            3. Denotes an increase

-------
                 Zn > Pb > Fe > Cd > Al > Cu > Ni > Cr » Mn

     Dissolved forms of  zinc were  removed to the  greatest degree with  an
average  removal  of almost  90 percent.   Dissolved  iron was  removed at  an
efficiency of  75 percent,  followed by lead,  copper,  aluminum, and  cadmium
with  removals   of   dissolved  species  ranging  50-60  percent.     Removal
efficiencies for dissolved nickel and chromium were  very poor,  with removals
of only 36 and 12 percent respectively.  The order for reduction of dissolved
runoff species upon entering the West Pond isi

                 Zn > Fe > Cd > Pb = Al > Cu > Ni > Cr » Mn

     Studies by Yousef  et al.  (1985)  as well  as observations during  this
research suggest that the removal of dissolved metal species is rapid with  as
much  as 90 percent removal  occurring  in four days.  In  the research  by
Yousef,  et  al., isolation chambers were  placed  in a newly constructed
retention pond near Epcot Center,  the isolation chambers were constructed  of
inverted  polyethylene 200-liter  barrels placed  on  the sediments - which
isolated  a 0.25 m   area of  the sediment  and the  overlying water  column.
Chambers were  constructed  with both open and sealed bottoms to investigate
the effects  of sediments on  heavy  metal  concentrations.  The  chambers  were
first installed then dosed with a solution of heavy metals.   Periodic  samples
were collected and  analyzed for metal content.   A summary of  their work  is
presented in Table 3.

     Soluble concentrations of copper, zinc, iron,  and lead  were added to two
of  the  test chambers in concentrations between  0.5 and 1  mg/1 on  4/1/83.
However, when  the  next  sample was  collected on  4/4/83,  concentrations  of
copper,  zinc,  and  lead had  been substantially reduced  by  an average of  90
percent.  By 4/18/83  (the next sample collection date)  concentrations  in the
closed  chamber were  indistinguishable from  the control which received  no
metal additions.   No change was noted either with or without sediment  contact
in these metal concentrations throughout the test period, even when anaerobic
conditions were established.

                ACCUMULATION OF HEAVY METALS IN THE SEDIMENTS
HORIZONTAL DISTRIBUTION OF HEAVY METALS

     Distributions of  heavy metals in the top  1  cm of of the Maitland  pond
sediments suggest that  upon entering the receiving water body,  the majority
of heavy metals associated  with highway runoff settle out and are deposited
near the point of input for the runoff.   The distributions of selected heavy
metals as a function of distance from the 45  cm RCP inlet are shown in Figure
2.  This tendency was most obvious for lead and zinc which peaked in sediment
concentrations at a distance of only  15 m from the inlet followed by a rapid
decline in  concentrations with increasing distance.  Deposition patterns of
the other metals measured were much less pronounced than  those  observed for
lead and zinc.  Chromium appeared  to reach peak sediment concentrations at a
distance of 30 m from the inlet with increases and decreases much less rapid
than those observed for lead and zinc.  Copper, nickel,  and manganese did not

                                     201

-------
                              TABLE 3.   UPTAKE AND RELEASE OF HEAVY METALS INSIDE ISOLATION
                                                CHAMBERS AT EPCOT POND
ro
o
ro
       CHAMBER
                                             TOTAL METALS CONCENTRATION BY DATE IN 1983 (Ug/1)
DESIGNATION
Sediment
Contact-
Control (no
metals added)
Sediment
Contact
(metals
added)
No Sediment
Contact
(metal s
added)
Pond

METAL
Cu
Zn
Fe
Pb
Cu
Zn
Fe
Pb
Cu
Zn
Fe
Pb
Cu
Zn
Fe
Pb

4-1
23
7
596
23
21
14
744
24
23
13
401
27
22
12
6U3
28
< 	
4-1*
AFTER
-
683
8b7
790
904
590
749
468
724
-
-_ni f f nc«
4-4
15
9
614
32
71
82
648
93
61
50
720
56
3b
4
855
52

4-18.
17
5
455
26
17
10
499
23
19
3
617
30
16
0
454
21
	 •> 
                                     was Supplied
     *
       After addition of nutrient and heavy metal solution.

     SOURCE:  Yousef et al. (1985)

-------
         250
ro
O
CO
                                 30
   45         60        75

DISTANCE FROM OUTFALL (m)
9O
105
120
            Figure   2.   Sediment Concentrations of Selected Heavy Metals In the Top 1  cm of the Maltland
            West  Pond as a Function of Distance from the Outfall.

-------
ro
                         20
 30     40      50     60     70      80    90
SEDIMENT CONCENTRATION (ug/g DRY WT.)
100
110
           Figure 3.   Attenuation Of Heavy Metals  in the Bottom Sediments of the Haitland Pond for
           Samples Collected 10/15/82.

-------
TABLE 4.  StMlARY OF BACKGROUND AND RUNOFF RELATED MEEAL CONCENTRATIONS
                  IN THE SEDIMENTS OF THE MAITLAND POND

u



1



3



b



H



SEDIMENT
DEPTH
- 1 cm:
Mean Cone.
Background
Added Cone.
- 3.5 cm:
Mean Cone.
Background
Added Cone.
.5 - 6 cm:
Mean Cone.
Background
Added Cone.
- b.S cm:
Mean Cone.
Background
Added Cone.
.5 - 13 cm:
Mean Cone.
Background
Added Cone.
MEAN SEDIMENT METAL
Cd

2.2U
U.4U
1.80

0.89
0.4U
0.49

U.56
0.40
0.16

U.45
0.40
U.US

U.66
0.40
U.26
Zn

45.4
L.b
43.9

10.7
1.5
9.2

6.49
1.50
4.99

4.64
1.50
3.14

3.16
l.bO
1.66
Cu

19.1
2.0
17.1

9.44
2.UU
7.44

7.50
2.00
b.iO

5.07
2.00
3.07

4.16
2.00
2.16
A1

49760
12058
37702

24574
1205B
12516

18256
12U5H
6198

1720b
12058
5147

12485
12058
427
CONCENTRATION (ug/g DRY NT.)
Fc

4554
654
3900

1761
654
1107

1303
654
649

874
6S4
22U

482
654
0
Pb

112.7
7.9
104.8

37.6
7.9
29.7

24.5
7.9
16.6

17.0
7.9
9.1

13.5
7.9
5.6
N1

16.5
1.98
14.5

6.49
1.9«
4.51

4.15
1. 98
2.17

4.03
1.98
2.05

3.31
1.98
1.33
Cr

44.7
7.63
37.1

19.3
7.63
11.7

15.4
7.63
7.77

10.8
7.63
3.17

4.01
7.63
0
Mn

43.9
1.6S
42.3

12.4
1.65
10.8

6.86
1.65
5,21

5.55
1.65
3.90

4.96
1.65
3.31
NO. OF
OBS.

138
6


138
6


138
6


138
6


138
6


-------
appear to exhibit pronounced peaks  in sediment  concentrations,  but  seemed  to
settle out  over a  longer  flow path length.   However, in  spite of the
differences  in behavior, most  of  the metals in  the runoff water entering the
Maitland pond were retained  in the  pond sediments within a distance of 60-90
m from the stormwater inlet.

     Of  the four metal species which exhibited the most  rapid settling
characteristics   (lead,   zinc,  iron,  and  aluminum),  all  but   zinc had
particulate  fractions in runoff which were near 80 percent of the total metal
measured.  The remaining metal species (nickel,  chromium,  and cadmium)  which
did  not exhibit  pronounced settling characteristics, were  all present  in
highway runoff at  the  Maitland site predominately in a dissolved form with a
small fraction of particulate species.

     The results of- the horizontal analyses of heavy metals suggest important
design  parameters for use  in the  design of  retention  basins to  optimize
removal  of  heavy  metals.    Designs  should provide physical  configurations
where the  flow velocity  becomes  very  small  to  aid in sedimentation  of
particles.   The distance from points of input to the discharge point from the
pond   should be maximized,  and the  design should minimize the possibility  of
short circuiting and avoid hydraulically dead zones.

VERTICAL DISTRIBUTIONS OF HEAVY METALS

     The vertical distribution of heavy metals in the sediments of the
•Maitland West  Pond was characterized by analysis of  average sediment  metal
concentrations by layer on  each of the three sample dates.   The  vertical
distributions of selected heavy metals in the sediments of the Maitland pond
are given  in Figure  3.  Aluminum was the most  abundant metal present  in the
Maitland pond sediments at all depths, followed by iron.   Lead was the third
most abundant heavy metal present^,  followed by  zinc and chromium,  copper and
nickel, and  finally  cadmium.   Concentrations of  cadmium were generally very
small  with many measured values,  especially in  the  lower sediment  depths,
approaching  the limits of detection.  Measured  concentrations of total heavy
metals in the sediments of the Maitland pond exhibited highest concentrations
in  the surface layer with a  rapid decline  in  concentration  with  increasing
depth.

     Background  soil concentrations  of heavy metals  in  the retention pond
area  were  estimated  from mean soil  metal concentrations  in  core  samples
collected  at depths of  3 m or greater during  drilling of monitoring  wells
beneath the  pond.   These background  concentrations were subtracted  from the
total  sediment  metal  concentrations  to  provide  an  estimate  of  the  added
accumulations as a result of inputs of highway runoff.   A  summary  of
background and runoff  related metal  concentrations  in the sediments of the
Maitland pond is given in Table 4.

     Sediment  concentrations  of  runoff  related  heavy  metals were  also
observed to decline rapidly  with  increasing depth.   The rapid decline  in
concentrations was  found to observe an exponential decay relationship with
values of R-square in most cases  in excess of 0.90 when fitted to the model:
In  (C/C ) =  -K x  (depth).  A summary of the regression of statistics for the

                                     206

-------
serai-log model is given in Table 5.  Values of K, which are a measure of the
rate of attenuation in sediment metal concentrations, indicated the following
order of attenuation of total  heavy metal content in the sediment layers:

    Most Rapid                                                Least Rapid
    Attenuation:   Fe < Zn < Cd < Pb < Cr < Mn < Al < Ni < Cu:  Attenuation

     The   calculated  regression  equations   for   runoff  related   metal
accumulations were used to estimate the extent of metal migration from runoff
related sources  by estimation of  the depths  necessary  to reduce runoff
accumulations by  90 percent and 99 percent to values which are 10 percent and
1   percent  above  estimated  background   levels.     All   runoff  related
accumulations were reduced in concentration  by 90 percent in the first 10 cm
or less.

     Although the substantial majority  of  metal species were  attenuated in
the  first  5.0 cm  of  sediments, the  depths  necessary to achieve  99  percent
reductions in runoff accumulations,  based on the calculated regression
equations,   suggest that  sediment  concentrations of certain metals may be
slowly  migrating  to  lower  depths.    However,  the  vertical  extent of  this
sediment-associated migration appears to be limited  since  all  metal  species
were reduced  in  concentration by 99 percent  within 20  cm or less.   These
calculations suggest  a strong stability of the  metal  sediment associations
since,  after eight years  of  metal accumulations in  the  Maitland  pond,  most
metals associated with sediments have remained  in top  10 cm of the sediment
layer.

              CURRENT STABILITY OF METAL-SEDIMENT ASSOCIATIONS
     The  stability of metal-sediment  associations was  evaluated from  the
results  of several  different analyses.  First,  a sequential extraction
procedure was used to determine metal speciation in composite samples of each
of the  five vertical core layers.   Metal speciations were  divided  into
soluble, exchangeable, carbonate  bound,  bound to Fe/ton  oxides,  and  organic
bound  fractions.   It is  generally  believed that  the stability of  the
metal-sediment associations  increases  in  the following order:   soluble  <
exchangeable < bound  to  carbonates <  bound  to iron and manganese oxides  <
bound to organic  matter.

     Fractional distributions of the total extracted heavy metals for each of
the  five  extracted species are  presented in  Table 6,   Most of the  metal
species tested, with the  exceptions of lead,  iron,  and cadmium, appear to be
predominantly associated  with only one major fraction.  For most metals, the
dominant  fraction is the one which  is bound to Fe/Mn oxides.  However,
cadmium is  predominantly associated with  the exchangeable fraction.   Lead
also has a major association with this fraction.  Alminum and iron appear to
have significant  fractions  with organic particles.   Very few of  the  heavy
metals  present  in the  sediments appear to be present in  a  soluble or
carbonate form although  cadmium,  zinc and  nickel had soluble  fractions of
approximately  10  percent.    It appears  certain that  iron, manganese,  and
                                     207

-------
 TABLE  5.  SUMMARY OF REGRESSION STATISTICS FOR ATTENUATION OF
        RUNOFF RELATED HEAVY METALS IN THE TOP 13 CM OF THE
        THE MAITLAND POND FOR A SEMI-LOG RELATIONSHIP FOR
            ALL THREE SAMPLE  DATES COMBINED
HEAVY
METAL
Cd
Zn
Cu
AT
Fe
Pb
Ni
Cr
Mn
Organ i c
•Content
NO. OF
OBS.
9
11
12
11
10
12
12
9
9
12
VALUE OF K FOR
"BEST- FIT" EQUATION OF
THE FORM:*
In (C/Co) = -KZ
In (Cd) =
In (Zn) =
In (Cu) =
In (Al) =
In (Fe) =
In (Pb) =
In (Ni) =
In (Cr) =
In (Mn) =
In (Org.)
-0.374 (Z)
-0.398 (Z)
-0.286 (Z)
-0.311 (Z)
-0.549 (Z)
-0.368 (Z)
-0.304 (Z)
-0.346 (Z)
-0.327 (Z)
= -0.241 (Z)
VALUE
OF
R-SQUARE
0.821
0.898
0.877
0.902
0.821
0.926
0.898
0.913
0.895
0.850
* Metal concentrations in units of ug/g; organic content in
  percent; and depth (Z) in units of cm.
                             208

-------
TABLE 6.  SPECIATION OF TOTAL HEAVY METAL CONCENTRATIONS IN THE SEDIMENTS
  OP THE MAITLAND WEST POND AS A FRACTION OF THE TOTAL METAL PRESENT
HEAVY METAL
Cadmium
Zinc
ro
S Copper
Aluminum
Iron
Lead
Nickel
Chromium
Manganese
SOLUBLE
15
4
1
<1
<1
1
4
2
1
PERCENT OF TOTAL
EXCHANGABLE
52
1
3
<1
5
44
8
5
9
EXTRACTED METAL
CARBONATE
12
4
1
<1
<1
1
<1
1
1
CONCENTRATION
BOUND TO
Fe/Mn OX.
10
81
89
74
52
52
82
73
86
BOUND TO
ORGAN1CS
11
10
7
26
43
2
6
19
3
TOTAL
100
100
100
100
100
100
100
100
100

-------
organic  content play  dominant roles  in regulating  the mobility  of  metal
species in the sediments of the Maitland pond.

     In  addition  to the  speciation  analyses/  an  incubation  apparatus  was
constructed which  allowed  simultaneous  control  of pH and redox potential in
sediment suspensions to simulate metal adsorption or desorption under various
environmental conditions.   Sediment suspensions were incubated at pH values
of 5.0, 6.5, and 7.5-8.0 which simulated current conditions of sediment pH in
the  pond,  and  at  redox potentials of  -250  mv, 0.0 mv, 250 mv  and 500 mv,
ranging from highly reduced to highly oxidized.  A summary of metal release
under these conditions for selected  heavy metals  is given in Figures 4
through  6.   In general,  release of  most heavy metals was less  than  a few
percent of  the total sediment content under  current conditions  of pH
(7.5-8.0).   The influence  of pH was  found  to be  much more  important than
redox potential in regulating metal release for most metal species.

     The results from the  speciation  and  redox  experiments combined with the
analyses  of  the sediment  metal  concentrations  presents evidence that under
the  current conditions of  redox potential and pH  within the sediments of the
Maitland  pond,  metal  species, with  the exceptions  of cadmium,  lead,  and
manganese, are  stable  and  exist  in relatively  immobile  associations with
Fe/Mn oxides and organic matter.   Lead and  cadmium are apparently held to a
large degree in strong exchangeable associations.   Changes in redox potential
from strongly   oxidized  to  strongly  reduced conditions did  not  appear to
affect the  release of metals  from  the  sediments under current pH values of
7.5-8.5.  The release of most metals, except cadmium and manganese, from the
sediment phase to the water phase was substantially less than 1% of the total
metal present even after several  weeks of incubation.  However, cadmium and
manganese appear  to be  less tightly bound  to  sediments than  other metals.
The  release  of both cadmium and manganese  into  solution  from the sediment
phase  during  incubation is  equal  to  approximately  5% of the  total  metal
present.

           EFFECTS OF THE MAITLAND POND ON UNDERLYING GROUNDWATERS


     A comparison of dissolved concentrations of heavy metals in the Maitland
pond water with groundwater  collected beneath the pond, represented by wells
2 and  3  combined,   is given in Table 7.   In general, concentrations of all
heavy metals  measured, except copper,  were greater beneath  the  pond than
within the pond.  For certain heavy metals such as zinc, manganese, aluminum,
and  iron, measured concentrations in groundwaters were from 5 to 75 times as
great as measured concentrations in the pond water.
                    K
     Analysis  of  variance  procedures  were  used  to estimate  the  vertical
extent  of the  migration  of  heavy metals in  the aqueous phase.   Zinc,
manganese,  aluminum, and  iron were significantly  higher in groundwater
beneath  the pond  than in the pond  at all depths  tested.  The extent of
significantly higher concentrations of  lead extended to the 0.5-1.0 m range.
Copper  was found  to be significantly higher in the  pond water  than in
groundwater in  all  analyses.  Average concentrations  of  zinc,  manganese,
aluminum,  and  iron were  found  to be 4, 12,  8,  and 50 times  greater

                                     210

-------
  40
   30-
  20
O
UJ
   10
***

§ o
UJ
cr
o
o
              CADMIUM
          ±=^:
                  ZINC
                                               pH-7.5-8.5
oioo
   80
o

g
Q.
40



20


 0
                           MANGANESE
                         pH-6.5
        -100
               I    1
     100       200      300

REDOX POTENTIAL(mv)
                                            pH=75-8.0
                                                      400
Figure 4.    Fraction of Total Sediment Metal Concentrations of
Cadmium, Zinc, and Manganese Released at Various Values of Redox

Potential and pH.
                                                            500
                                211

-------
       -100
0        100     200       300
   REOOX POTENTIAL (mv)
4OO      500
Figure  5.    Fractions of Total Sediment  Concentrations  of Copper,
Aluminum, and Iron Released at Various  Values of Redox Potential
and pH.

                              212

-------
   8
   6
  4
tu
Gl
UJ
I'
           LEAD
                                                pH= 6.5
                                             pH=7.5-8.5
'    '    '     '
                                '    '     '
I     I    I
UJ 2

1
CO
UJ
CL
       NICKEL
                                    PH= 7.5-8.5
                  CHROMIUM
                       pH=6.5
                                = 5.0
                                              PH = 75-8.5
                          -t-
        -IOO      0       100      200      300      4OO      500
                     REDOX  POTENTIAL (mv)

  Figure  6.     Fractions of Total Sediment  Concentrations of Lead,
  Nickel, and  Chromium Released at Various Values of Redox Potential
  and pH.
                              213

-------
                TABLE  7.  COMPARISON OF DISSOLVED CONCENTRATIONS OF HEAVY METALS IN THE POND WATER
                            WITH GRDUNDWATER COLLECTED BENEATH THE POND IN WELLS 2 AND 3
rv>
HEAVY
MFTAI
"ID 1 ML-

Cd
Zn
Mn
Cu
Al
Fe
Pb
Ni
Cr
pH
AVERAGE AVG. CONC.
DISSOLVED IN 0-1 cm
CONCENTRATION SEDIMENT
IN POND
(ug/D
0.73
4.82
4.47
14.4
57.9
18.4
15.0
1.62
2.18
7.46
LAYER
(ng/D*
1,015
20,938
20,246
8,809
22,949,100
2,100,285
51,977
7,610
20,615


0.1 m
1.48
22.9
53,3
9.40
709
1354
24.4
2.88
3.02
5.75
AVERAGE CONC. IN GROUNDWATER
BENEATH THE POND

0.5 m
2.06
18.5
44.2
10.3
742
834
24.8
2.63
4.15
5.92

1.0 m
1.50
18.8
23.3
9.12
192
766
20.5
1.98
2.60
5.17

3.0 m
1.17
19.1
78.9
11.7
89.2
797
11.2
1.94
1.29
4.86

6.0 m
1.00
19.5
70.2
13.1
543
1912
10.9
2.13
1.54
4.56
RATIO OF
G.W. CONC.
AT 0.1 m
TO AVG.
POND WATER
2.03
4.75
11.92
0.65
12.25
73.6
1.63
1.78
1.39

            * Average  sediment  concentration  per  liter of  sediment.

-------
respectively,  in  groundwater  than in  the pond  water.    Concentrations  of
cadmium and iron  in  groundwaters  exceeded water  quality  criteria  for Class
III waters specified  in Chapter 17-3 of the Florida Administrative Code.

     Concentration ratios between the pond water and  the groundwater indicate
that all  metal species,  except copper,  have a greater affinity for  the
groundwater phase than the pond phase and are leaching into groundwaters to
some degree.   The  order of release potential of heavy metals into groundwater
was estimated  to be:

         Least                                               Most
         Mobile:   Cu  < Cr < Pb < Ni < Cd < Zn  < Mn =  Al <  Fe:  Mobile

and was found to be  inversely related to the order of attenuation for metal
species  in  the  sediment  phase.    The  magnitude  of  the  release  into
groundwaters  was  found to  closely correspond to the order  of release
predicted by the incubation experiments conducted  under natural conditions of
pH  (7.5-8.0).  However,  in  spite of  the increased metal concentrations
beneath the pond, the  sediments are  clearly the  primary sink for heavy
metals.

                TRANSPORT OF HEAVY METALS IN GROUNCWATER FLOW


     One  of the objectives  of this research was  to monitor groundwater
concentrations and flow  patterns and to detect, if possible, the movement of
heavy  metals  which  leach into groundwaters.  To  aid  in this detection,
piezometers were installed  at  each well and  a record of pie zone trie surface
was made approximately on a bi-weekly basis during 1983.   The average
measurements are given in Table 8.

TABLE 8.  AVERAGE  MEASUREMENTS OF PIEZOMETRIC  SURFACE AT MONITORING
               WELLS  AT THE MAITLAND INTERCHANGE DURING 1983
                Location          Piezometric Surface  (m, MSL)
Well 1
Well 2
Well 3
Well 4
Wall 5
Pond
27.35
27.38
27.37
27.36
26.87
27.38
                                    215

-------
     As  indicated  in  Table  8,  little  variation  was  measured  in  the
piezometric surface between each well.   As a result of this very small
hydraulic gradient, horizontal movement  of groundwater in the Maitland area
was calculated  to  be  less than 10 m per year.   It appears that vertical up
and down movement with changes in seasonal water table may be more important
than horizontal movement.   As a result,  metal contamination of groundwaters
appears to be very localized.  Since the  hydrologic conditions present at the
Maitland site  are similar to conditions at many other sites in  Central
Florida,  it  appears  very unlikely that heavy metals  from  retention ponds
along  highway  systems  in  the  Central  Florida  area  will pose  a  pollution
hazard to nearby surface and groundwaters.

              POTENTIAL FOR FUTURE MOBILIZATION OF  HEAVY METALS


     Natural aging processes  within  retention ponds as well as lakes result
in  the increased deposition of  organic matter  to  the bottom  sediments
primarily as a result of the death and decay of both plant and animal matter.
As  these  processes occur,  it has  often  been observed that sediments become
more  reduced  and decrease in pH.   Although the incubation experiments
indicated that  most metal species are stable and  tightly bound to sediments
under  current  conditions  of  redox potential  and  pH, decreases in  pH were
found  to increase the   solubility  of all heavy metals  tested.   Changes in
redox potential produced no significant changes in  release rates.

     The  results suggest that as the Maitland pond ages and accumulations of
organic matter in the sediments begin  to cause sediment  pH values to
decrease, mobilization of  all metal species  tested will  increase and  release
to  groundwaters  may occur.   Although all metals  were  found  to increase in
solubility with decreases  in pH,  the release was  in general,  only a small
fraction  of  the total  sediment  metals  present.    For zinc,  iron,  aluminum,
copper,  and  chromium,  the maximum release  was  less  than 3 percent of the
total  acid-extracted  metal  in  the sediments, even at  the most  extreme pH
value  tested of 5.0.  For nickel and  lead,  the  release extended as high as
6-7 percent  at a pH  of 5.0.  However,  as  seen  in Figure 4,  the release of
cadmium and  manganese into groundwaters can be  expected to increase as the
sediments become more acid.   Manganese and cadmium were  found to increase in
solubility substantially  as sediment pH  decreases with almost total  release
of managanese and 25-35 percent release  of cadmium at  a  pH of 5.0.  Releases
of   this  magnitude  may  produce  measurable   increases   in  groundwater
concentrations beneath the pond.  In the  case of  cadmium,  a health hazard may
be present under these extreme conditions.

     The results suggest that maintenance procedures may be necessary  after a
period of time to remove  the accumulated  sediment deposits which may cause
conditions of low pH and release of metals.

                                 CONCLUSIONS


     From the results obtained in these investigations the following specific
conclusions were reached:


                                     216

-------
     1.  Measured concentrations of heavy metals  in highway  runoff  collected
at  the Maitland Interchange during  1983-84 showed considerable  variability
between storm events as well as during storm events.

     2.  With the  exceptions of copper  and cadmium,  the  majority of  metal
species were present in a particulate  form.  The  most  common vehicle  related
heavy  metals  found in highway runoff  at the Maitland site were  lead,  zinc,
and  copper which  together  accounted  for approximately  91   percent  of  the
dissolved heavy metals and 94 percent of the total metal concentrations.

     3.  The Maitland pond was found  to be very effective in  removal of  heavy
metal  inputs  from highway runoff.   Particulate species of most  metals were
removed in the range of 75-95 percent  with most of this mass retained in the
pond sediments within  a distance of 60-90 m from the stormwater inlet.   In
general, dissolved forms of  heavy metals were removed to a  lesser  degree than
particulate inputs with efficiencies  near 50 percent  for raost metals.

     4.  Mean  concentrations of  heavy metals  within  the pond were within
water   quality   criteria  established   in   Chapter   17-3  of  the   Florida
Administrative Code (F.A.C.)  for Class III (recreational) waters.

     5.  Measured  concentrations of  total heavy  metals in the sediments  of
the Maitland pond exhibited highest concentrations in  the  surface layer with
a rapid decline in concentration with increasing depth.

     6.  After eight years of accumulations in the Maitland pond,  most metals
associated with  sediments have  remained in the  top  10 cm  of the sediment
layer.

     7.  Under  current  conditions  of redox  potential  and pH  within  the
sediments of the  Maitland pond,  metal species, with the possible exceptions
of  cadmium and  manganese,  are stable and exist in relatively  immobile
associations with Fe/Mn oxides and organic matter.

     8.  In general, mean concentrations of all heavy metals  measured, except
copper, were greater  in groundwaters beneath the pond than  within  the  pond.
Average concentrations of zinc, manganese,  aluminum,  and iron were found to
be  4,  12,  8,  and 50 times grater, respectively,  in  shallow  groundwater than
in  the  pond  water.    The  extent   of  significantly higher  groundwater
concentrations of nickel, cadmium,  and chromium extended to depths of  0.5-1.0
m,  lead extended to the  1-3 m range, while zinc, aluminum, manganese,  and
iron were elevated in concentrations  past the 6 m sample depth.

     9.  Violations of Class III water quality criteria were  present for both
cadmium and iron in groundwaters beneath the pond.

    10.  The horizontal movement of  groundwaters  in the study area was less
than  10 m/year and as  a result, the  influence of the pond  on groundwaters
appears to be extremely localized.

    11.  As  sediment  accumulation occurs  in retention ponds over  time,  the
corresponding decreases in  pH  and ORP  of  the sediments  will increase the

                                     2V

-------
release of metal ions into groundwaters.   The effect of reductions  in pH were
found to  be  more important than reductions  in ORP  in  regulating the release
of metal species.

     The   work  described  in  this  paper  was  not   funded  by  the  U.S.
Environmental Protection Agency and therefore the contents do not necessarily
reflect the  views of  the Agency and no official endorsement  should be
inferred.

                                  REFERENCES
Shaheen,  D.G.   Contributions  of  Urban  Roadway Usage  to Water  Pollution.
     EPA-600/2-75-004,  U.S.  Environmental  Protection  Agency,  Washington,
     B.C., 1975.

Wilber, W.G. and Hunter, J.V.  Aquatic Transport of Heavy Metals  in the  Urban
     Environment.  Water Resources Bulletin, 13:721, 1977.

Winters, G.L.  and  Gidley,  J.L.  Effects  of  Roadway Runoff on Algae.  Report
     #FHWA/CA/TL-80/24•   Federal  Highway Administration, Washington,  D.C.,
     1980.

Yousef, Y.A.;  Harper, H.H.;  Wiseman, L.P.; and  Bateman,  J.M.  Consequential
     Species of Heavy Metals in Highway Runoff.  Final Report Nos.  99700-7255
     and 99700-7272, Tallahassee, FL, 1985.
                                      218

-------
                 SIMPLE TROEHEC STATE MODELS AND THEIR USE
                    IN WAgEELQAD AIIOCaTIONS IN FLORIDA.

           by:   R.  W.  Ogburn,  P. L. Brezonik1 and B. W. Breedlove
                Breedlove,  Dennis  & Associates Inc., Orlando, FL 32807
                ^University of Minnesota, Minneapolis, MN 55455
                                  ABSTRACT


     Florida's Department of  Environmental Regulation  (FDER)  uses a  simple
input-output trophic state model to establish wasteload allocation limits in
cases involving wastewater discharge with a lake as the impacted waterbody.
The model,  SIMLAK,  is based  on in-lake chlorophyll a  as  predicted  from
loadings  of nitrogen  and phosphorus.  A wasteload  allocation performed  by
FDER for five potential dischargers to the Reedy Creek/lake Russell system in
Polk  and  Osceola  Counties  in  central  Florida   resulted   in  phosphorus
concentration  limits  much  lower  than  levels  achievable  with  available
technology.

     An  independent analysis  of the Reedy  Creek/lake Russell system  was
performed to examine  the applicability  of  SIMLAK  and other trophic  state
models,  and to explore the potential  iinpacts of various nutrient  loading
scenarios  on Lake  Russell.  SIMLAK  did  not  yield  accurate  predictions  of
chlorophyll a for Lake Russell based on present loadings, possibly because of
the lake's highly colored water. The FDER allocation scenarios also required
the  nitrogen-to-phosphorus ratio  in  the lake  to  remain constant  at  the
present  ratio instead  of controlling the limiting  nutrient.  Our  analysis
showed that variation  of nitrogen-to-phosphorus ratios in the model  could
allow slight  decreases in nitrogen concentration limits to offset increases
in  phosphorus   oancentrations   (to   achievable  levels)   with  the   same
chlorophyll a prediction.

     Alternative trophic state models based on one limiting nutrient did not
yield  better predictions  of present  conditions in Lake  Russell.  However,
they  did  indicate  that  wastewater  discharges  at  achievable  phosphorus
concentrations would not cause chlorophyll a in Lake Russell to exceed FDER's
allowable limit.
                                     219

-------
     The initial  FDER allocation of  21.5 ragd would  have required the  five
potential dischargers  to meet a  concentration limit of 2.15 mg/L for total
nitrogen (IN), but total phosphorus  (TP)  limits ranged  from 0.1 to 0.38 mg/L
depending on distance  from lake Russell.  A final  allocation agreed to by all
parties allowed discharge  of 22.5 mgd with concentration limits of 2.0 mg/L
for TN and 0.5 mg/L for TP.
                                INTRODUCTION
REEDY CREEK/LAKE RUSSELL SYSTEM


     The  Reedy Creek drainage  basin in  central Florida  extends  from the
southwest  corner  of Orange  County,  through  Osceola County  and  into  Polk
County,  where it  joins the Kissimmee River system  (Figure 1).  Reedy Creek
originates  near the Disney World  complex  and flows southeastward through
Reedy  Creek Swamp into Lake Russell. Below Lake Russell  the stream splits
with approximately 30 % of the  flow going to Cypress lake and 70 % to Lake
Hatchineha  through  the Dead  River.  Flow from  these two lakes ultimately
reaches Lake Kissimmee. The upper third of Reedy Creek has been channelized
and modified  by  control  structures,  but  the remainder of the creek  flows
through a natural  stream channel.

     lake Russell  is a highly colored lake with  a surface area of about 300
ha, a  mean depth of 1.83 m and a  maximum depth of 2.6 m  (Table 1). The
annual hydraulic residence time is 30 days, and the range  of retention  times
based  on monthly  flows is between 15 and 56 days.  High levels of organic
color  limit Secchi  depth  to  about 0.5 m and  also appear to limit levels of
chlorophyll a in  -Uie lake. Color  ranged  from about  300  to 500 PCU between
 1979 and 1983, and chlorophyll  a generally was less  than  1 ug/L during the
 same  period,  except  for one  value  of  16  ug/L  in  July,  1980  and one
measurement of 13.5 ug/L in July,  1981.


WASTELOAD ALLOCATION PARTIES


      Irr 1982 the City of Kissimmee was ordered by the State to eliminate its
wastewater discharge to Lake Tohopekaliga; later in 1982  the  city initiated
discussions with  the Florida Department  of Environmental Regulation  (FDER)
concerning the possibility of  discharge  to Reedy Creek.  At that time  Reedy
 Creek Utilities was the only discharger to the creek, but several additional
wastewater  treatment  plants  also  requested  consideration  from  FDER for
discharge   capacity  to   Reedy  Creek.   The   five  utilities   requesting
consideration in  the allocation included  Reedy Creek Utilities, the City of
Kissimmee,   Central  Florida Utilities,   Poinciana  Utilities,  and Osceola
 Services Company.  The  FDER  conducted a  one-year water  quality survey and
                                     220

-------
                   WALT DISNEY
                     WORLD
                    REEDY
                    CREEK
                           JL. _QRANGECO_._
                               OSCEotA CO.
                                                           APPROXIMATE

                                                           PROJECT

                                                           AREA
 LEGEND

 — —— MAJOR THOROUGHFARES

 	COUNTY LINES

       LAKE AND CREEK BOUNDARIES
                                                                \
Figure 1.  Reedy CreeVLake Russell study area.
                                       221

-------
 Russell/ estimate the assimilative capacity of the system,  and allocate that
 capacity among the five utilities.


 WASTEIDAD ALLDCATICN RESUUTS
      After  FDER's examination  of the  Reedy Creek/Lake  Russell system  and
 numerous  modeling scenarios using the lake model SIMIAK, FDER published a
 notice  of  its  intent  to adopt an  agency  order  concerning a  wasteload
 allocation  for the Reedy  Creek basin  (2).  The proposed  order included  two
 allocation  scenarios  with total point source discharges of 21.5 ragd and 11.0
 mgd (Table  2).  Both scenarios included equal TN concentration limits for all
 five dischargers but  they  allowed  higher  TP  concentration  limits  for
 utilities with  points  of discharge  (POD)  farther from Lake Russell.  In
 addition, the allowable TP concentrations decreased every five years through
 the year  2005,  and the proposed order stated that design levels of treatment
 for TN  and  TP removal were to be based on limits for the year 2005.

       The  TP  limits  established for  the year 2005 were lower than levels
 achievable  with  existing technology for  all   five  dischargers  in  both
Table 1.    Physical and chemical characteristics of Lake Russell.
Parameter                          SECRET^          FDER1


Conductivity  (umho/cm)                                             95
pH                                   6.2             6.7
BODs  (mg/L)                          1.7             1.0            6.7
Secchi depth  (m)                     0.47            0.4            0.6
Color (PCU)                        343             404            500
Ammonia N  (mg/L)                     0.15            0.36           0.07
Nitrate & Nitrite N  (mg/L)           0.13            0.06           0.06
Total N (mg/L)                       1.74            1.97           1.79
Total P (mg/L)                       0.07            0.11           0.08
Chlorophyll a  (ug/L)                 5.5             0.02           0.2

Surface area  (ha)                                  300
Mean depth (m)                                       1.83
Maximum depth  (m)                                                   2.7
Residence time  (d)                                  30


1 Reference 1
2 Field and laboratory results form Breedlove, Dennis & Associates, Inc. 1983
  - 1984.

                                      222

-------
Table 2.  Discharge limits of total phosphorus  (mg/L) and total nitrogen
       (mg/L) in FDER notice of Intent to Adopt Agency Order  (2).
               Flow                                    TP
Discharger     (mgd)       TN       1985     1990     1995     2000     2005


ROJ             7.5        2.15     0.65     0.57     0.51     0.45     0.38
Kissimmee       6.0        2.15     0.17     0.15     0.13     0.12     0.10
CEU             3.0        2.15     0.29     0.25     0.23     0.20     0.17
Poinciana       3.0        2.15     0.17     0.15     0.13     0.12     0.10
OSC             2.0        2.15     0.41     0.36     0.32     0.28     0.24
               21.5
ECU
Kissianmee
CFU
Poinciana
OSC

4.0
4.0
1.0
1.0
1.0
11.0
2.54
2.54
2.54
2.54
2.54

0.62
0.16
0.28
0.16
0.40

0.44
0.12
0.20
0.12
0.28

0.32
0.08
0.15
0.08
0.21

0.20
0.05
0.09
0.05
0.13

0.08
0.02
0.04
0.02
0.05


scenarios. Poinciana  Utilities,  Inc. subsequently  filed a petition for  an
Administrative Hearing  before the  Florida Department  of Administration  to
review FDER's proposed  order and to request  relief from the impacts of the
proposed order (3).


INDEPENDENT EVALUATION


     Poinciana Utilities, Inc. contracted with the authors of this  paper to
analyze the  FDER wasteload allocation procedures  as applied to Reedy  Creek
and  Lake  Russell, and to conduct an independent  analysis of the  potential
impacts of wastewater discharge on lake  Russell and downstream laloes.  This
analysis  included a thorough review of SIMLAK and the wasteload  allocation
process,  modifications  of  SIMLAK,  alternative  trophic  state  models,  and
additional data  related to  potential  impacts  of wastewater  discharge  on
fisheries, water clarity and nuisance aquatic plants.
                                     223

-------
                         FDER WRSTEIDAD ALLOC3VTICN
SIMLAK
     SBHAK  is a  simple mathematical  model that  can be used to predict
concentrations of IN, IP and chlorophyll a in a lake based on Inputs of N and
P, and physical characteristics of the lake such as hydraulic  retention and
basin morphometry  (Figure  2).  The  S3MIAK equations  were developed using
                                         REEDY CREEK
r
i
i
i



~ .^^— - ~
N
LOADING
~4 	
LEO 2 |

"~

	



P
LOADING
	 f~-
1 EQ 3 1

TN

1
TP
                                     I

                             CHLOROPHYLL  A
                                        LAKE  RUSSELL
               EQ  1:    R = 0.482  -  0.112 In
               EQ  2: TN=(1-R)



               EQ  3: TP=(1-R)
                N LOADING
                  FLOW

                P LOADING
                                  FLOW
EQ  4: CHLA=35.95
                                            VOLUME
                                    ~+TP) J
                                  LENGTH
                              (WIDTHXDEPTH)
            Figure 2.   Block diagram of SIMLAK model.
                                 224

-------
regression analysis of data  from 33 Florida lakes  included in the National
Eutrophication Survey  (4).  Equation 1  is a nutrient  retention factor  that
predicts the fractions of IN and TP retained in a lake as a function of its
flushing rate  (Figure 2). Equations 2  and 3 give predictions  of in-lake TN
and TP based on loadings, the retention factor, and the water  flow rate into
the lake. Equation 4 predicts chlorophyll a in the lake from the predicted
values of  TN and TP,  and a  shape term derived from  the length, width and
average depth of the lake.


WASTELCftD ALLOCATION PROCESS


     The first step  in the wasteload  allocation process  is  to verify that
SIMIAK works for the particular lake in question  by  using data on  present
inputs to predict existing concentrations of TN, TP and chlorophyll a in the
lake.  However, FDER  has not established criteria for determining  whether
SIMIAK predictions are acceptable.

     If  the  model  produces  "acceptable11  results,  the  next  step  is  to
calculate  allowable  increases  in nutrient loading to the  lake. FDER does
this by  estimating non-point source (NPS) flows and inputs of nitrogen and
phosphorus, and using SIMIAK to predict the amount of chlorophyll that should
be  produced by NPS inputs  alone. A maximum increase  in chlorophyll a  of 2
ug/L is allowed if the predicted  NPS chlorophyll a  is  less than 60 ug/L. In-
lake  concentrations  of TN and  TP  are allowed  to  increase by the  same
percentage as the increase in chlorophyll a.

     Equations 2 and 3 are re-arranged to yield nitrogen and phosphorus loads
that would produce the TN and TP levels  obtained  in the previous step. New
values of  flow (Q)  and retention (R) are used to reflect the  combined total
of NPS and design PS flows. When  present  NPS loads  are subtracted from those
total loads, the remainders are the allowable point source loads.

     Of  this  allowable  increase,  only  75%  is  allocated  among  the  PS
dischargers,  and 25% is set aside for increased NPS  loads and as a safety
factor.  The allocatable  load   to  the  lake  is  divided  among  potential
dischargers  in the same proportion as their  contributions  to the total PS
flow  rate.  The  procedure described  thus  far produces an  allocation of
nitrogen,  phosphorus  and flow rates at the lake itself.  FDER's next step is
to  extend  the  allocation to  obtain  limitations  on discharge flows and
nitrogen and phosphorus concentrations  at the  individual points of discharge
to  the system.

     In  the Lake Russell wasteload allocation, FDER concluded that nitrogen
removal  does not occur in Reedy  Creek. Therefore  the nitrogen concentration
and flow limits  calculated at the entrance to the  lake also  were applied
without  change to  the points  of  discharge,  and  all  potential dischargers
received the same effluent TN concentration limits  (Table  2).
                                     225

-------
     However,  PEER did conclude  that phosphorus is  removed during flow  in
Reedy Creek,  and they calculated an average rate of phosphorus  removal per
mile  of creek channel.  This meant that higher phosphorus  concentrations
could be discharged farther upstream of lake Russell  and still theoretically
meet the  loading limits at the lake.  The last column in Table 2 shows that
using this method,  potential dischargers with POD's far from Lake  Russell
would receive  less  stringent phosphorus concentration limits than dischargers
whose assumed  POD was closer to lake Russell.

     The  final step FDER performed in the Reedy Creek  wasteload allocation
was  an attempt to  project NPS  changes through the year  2005. Ihey used
information from the East Central Florida Regional  Planning Council and from
Osceola County  to estimate  the  effect of  future  growth on NPS inputs  of
nitrogen  and  phosphorus  to the  Reedy Creek basin.  FDER concluded that  NPS
loads  of nitrogen  will  increase  by 7%, and that NPS phosphorus loads will
increase by 29% by the year 2005.

     They then compared the projected  NFS  increases  to  the  25%  reserve
loadings  set  aside in  the  WLA  process. In the  case  of  nitrogen,  future
increases did not  exceed the reserve amount. However,  FDER's projected  NPS
phosphorus increases did exceed the 25% phosphorus reserve amount. When that
occurred,  FDER  subtracted  the additional NPS phosphorus increases  from  the
amount  originally allocated to point sources,  thus reducing phosphorus limits
as in the scenarios included in the FDER Notice of Intent.
                            INDEPENDENT EVALUATION
 CRITICISMS OF SIMIAK


      Our  analysis  of the  Reedy  Creek/Lake  Russell WIA  led  to  several
 conclusions regarding the model itself and its use  for  Lake Russell. Most
 nutrient loading models use different retention factors for TN and TP because
 of observations that they are retained to different  extents in lakes where
 nutrient  mass balances  have  been  measured.  Lakes  generally  retain more
 phosphorus than  nitrogen,  and thus  it is  inappropriate to  use the same
 equation for both. In addition, the SIMIAK retention equation originally was
 developed for phosphorus  retention in  northern lakes,  (5)  and  it does not
 give good predictions for Florida lakes.

      SIMIAK also did  not  accurately predict chlorophyll  a based on present
 nutrient  inputs  to Lake  Russell.  The  model was  designed to predict mean
 annual chlorophyll a, which has ranged from about 5 ug/L  to  <1 ug/L for Lake
 Russell.  SIMIAK  predicted 20.7 ug/L of chlorophyll a, but FDER considered
 that an acceptable  fit because of the  maximum values observed  in 1980  (16
 ug/L)  and 1981 (13.5 ug/L).
                                    226

-------
     Ihe WIA precedure allows a maximum  increase  of 2 ug/L of chlorophyll  a
over the  NPS chlorophyll a  level in a  late, but  SIMLAK  is not capable of
predicating  chlorophyll  a  values with  that  degree  of  accuracy  (4).  Ihe
average error  of SIMIAK chlorophyll  predictions  for  the  data  set used to
develop  the model was  +21 ug/L.  In  addition  to the  problem  of model
accuracy,  it would  be very difficult  to measure a 2  ug/L  increase in
chlorophyll. For example,  mean  chlorophyll a  in Lake Kissiiranee was 20.68
ug/L in 1979. Ihe winTimim  statistically  valid increase (at  a 90% confidence
level) with a monthly sampling program would  be 4.0 ug/L.  Even with a daily
sampling schedule, 2.9 ug/L would be the  minimum increase detectable at a 90%
confidence level.

     SIMLAK  uses both  nitrogen  and phosphorus  to predict chlorophyll a.
Other  nutrient  loading models recognize that only  one nutrient at  a  time
limits  the  production of   chlorophyll  and  they  therefore have separate
chlorophyll  equations,  one  for  phosphorus-limited lakes  and  another for
nitrogen-limited  lakes. However,  because SIMIAK  predicts  chlorophyll as an
additive  function of  IN and TP,  the WIA procedure  should allow  for  "trade-
offs"  between  the two  nutrients such that  an increase  in one (above its
"allowable" values) could be compensated for in SIMLAK by a decrease in the
other  (below its  "allowable" value),  the net  effect of which  would be to
keep chlorophyll  a levels predicted by SIMIAK within the allowable  increase.
Thus the WIA procedure does not allow the  flexibility in  computing nutrient
loads that is inherent in the assumptions of the SIMIAK model.

     According  to the SIMIAK model, a  small  decrease  in  nitrogen  loading
could  result in  an increase in  phosphorus loading to Lake Russell with  no
change in predicted chlorophyll a  for the lake.  Ihe validity of the SIMIAK
results  should  not be affected as long as N to P ratios remained within the
range  of N  to P ratios found  in the lake data used to  develop  the  model
equations.  Nitrogen to phosphorus  ratios vary widely in Florida  lakes and
they can change seasonally  within a particular lake. There is thus no reason
to require that N to P ratios in point source loadings should be the sane  as
N to P ratios in NFS loads.

     FDER evaluated discharge  scenarios by  comparing nutrient  loadings  to
calculated loading limits rather than considering  whether the  chlorophyll a
limit would be exceeded.  When FDER projected the Lake Russell  allocation to
the year 2005,  they added their predictions of future increases in NFS loads
to  present and  proposed   loads.  After  the  projected  increases  in  NFS
phosphorus   loads had equalled  the 25%  reserve  phosphorus  load,   FDER
subtracted subsequent  increases from the amount  allocated to dischargers
in order to maintain the total phosphorus load within the WIA limit.

      Instead of  using that approach,  we asked the  question: If  the  PS
phosphorus load were not decreased,  and the total  phosphorus  load exceeded
FDER's WIA limit, what would be the effect on Lake Pussell's chlorophyll a?
Using FDER's projections of future NFS  loads  and  24.5 ngd of  PS discharge
with  IN =  2.15  mg/L and TP  =  0.5  mg/L,  we  calculated that  allowable
chlorophyll a  levels would  not be exceeded through the year 2005 (Figure 3).
                                     227

-------
                                                                             WLA
                                                                             LIMIT
Q

3i

if
II
V)
O
a.
                 I
                 c
                 O

                 u
                 O
                 w
                 o
                 O
                 111
                 DC
                 0.
                           1984
                           1984
                   1965      1990      1995      2000      2005
                                                                             WLA
                                                                             LIMIT
                   1985
                                             1990      1995      2000      2005
                                                                             WLA
                                                                             LIMIT
                           1984
                                    1985
                                             1990      1995      2000      2005
Figure  3.   Predicted chlorophyll a in Lake Russell using SIMIAK equations with
            24.5 rogd point source flow (IN =  2.15 mg/L, TP = 0.5 mg/L) and DER
            projections of future NPS loads.
                                        22b

-------
The TP loading would  exceed FDER's  phosphorus limit throughout that time
frame, but it would be compensated for by the fact that IN loads would remain
less than the nitrogen limit.

     In addition,  FDER's  estimate of  future  changes in NPS  loading in  the
Reedy  Creek basin did not  include  the effect of  surface water management
requirements on nutrient loadings. A smilar analysis was  performed  (Harper,
personal communication) using recent data on reduction of nutrient  loadings
in stormwater management  systems  similar to those that will be required in
future  developments.   These  results  indicated that  the land use  changes
projected by FDER would not cause increases in NFS  loadings of nitrogen or
phosphorus, and that both might actually decrease slightly. In such a case,
predicted chlorophyll a in lake Pussell in the year 2005 would  be at least 1
ug/L lower than FDER's wasteload allocation limit.


ALTERNATIVE MDDEIS


     Numerous  models  are  available  for  evaluating  potential effects  of PS
discharges  on  lake trophic state and  water quality. Some of the best-known
models are accepted  as  reasonable  empirical  approaches to predicting  the
trophic state  of temperate  lakes (7,8)  and  several more recent models  are
modifications  of  their basic  approach.  The models  have been  adapted to
Florida lakes by using a  statistical approach  to  optimize equations that
relate nutrient loading rates to lake response data  (6). When results from
those equations were  compared to the SIMLAK  equations  we  found  that  the
modified Dillon and Vollenweider equations generally gave better predictions
of nutrient and  chlorophyll concentrations  in Florida  lakes  than  did  the
SIMLAK equations  (Table 3).

      The alternative equations also yielded over-predictions of chlorophyll a
for lake Russell based on present loading  rates   (Table 4),  but  they  all
predicted rather small increases in chlorophyll, N and  P in the lake as a
result of increased PS flows.  We did not identify the reason that all  the
models predict higher  chlorophyll levels than those observed in lake Russell,
but the high  levels  of  color  limit light penetration and may  limit algal
productivity.  Additional possibilities include a limiting micronutrient  such
as molybdenum or zinc,  the short water residence time, or high  rates of
zooplankton grazing.  Nevertheless,  the models all  indicated that achievable
concentrations of phosphorus in proposed discharges would have very small
impacts on lake Russell.

      Additional   regression models  were  used to   evaluate the iiipact of
increased PS  loads on user-perceived values  in  Lake Russell. One  equation
relates Secchi depth in Florida lakes  to  chlorophyll and color  (8):

           In Secchi depth = 2.01  - 0.370  In Oil a -  0.278  In Color

The model indicated that predicted increases  in chlorophyll a in  lake Russell
would result in decreases in water clarity too  small to be observable to  lake
users.
                                     229

-------
Table 3.  Predictive equations from SIMLAK and other Florida nutrient loading
          models.
     Source
     Equation
Coefficient of
Determination
Nutrient Retention
     SIMLAK
      Baker et al.
R = 0.482 - 0.112 In Q
                     V
    (for nitrogen)
    (for phosphorus)
   = -0.010 + 0.597 log
                        Q
Rp = -0.056 +  1.40
              1 +  W
     = 0.10
     = 0.15


     = 0.51


     = 0.88
 Chlorophyll Prediction

     SIMAK


     Baker,  et al.
CHIA =f 35.95/1N ^rp\ JL
            ^3     J NWZ
CHLA = 8.30



CHIA = 73.4
                          71
/           \ ^

(o.65 Z + Q j

/     Lp     \0.667

lo.65 Z + Q I
  r2 - 0.52



  r2 = 0.69



  r2 = 0.60
                                      230-

-------
Table 4.  Summary of model predictions for Lake Russell with various
          discharge scenarios, (TP = 0.5 mg/L, IN = 2.15 rag/L).
Predictions
Point Source Loadings (mgd)

        Present          18.5
21.5
24.5
Total Nitrogen
   (rog/L)
Observed
SIMLftK
Baker et al.
1.96
1.95
1.64
-• -
2.02
1.63
— —
2.02
1.63
— —
2.02
1.63

Total Phosphorus
     (mg/L)
Observed
SIMLAK
Baker et al.
0.110
0.068
0.073
	
0.070
0.078
	
0.074
0.078

0.077
0.079

Chlorophyll a
    (ug/L)
Observed
SIMLAK
Baker
Baker
et
et
al.
al.
N
P
3
20
34
13
.0 (16.0
.7
.8
.7
Max) —
21
35
14
^•^B
.5
.3
.2

21.
35.
14.

6
3
2
^»«M4
21
35
14
^ft^M
.7
.2
.6

Nutrient
Retention Factor
SIMIAK
Baker et
Baker et

al.
al.

Rn
Rp
0.
0.
0.
1996
2496
2532
0
0
0
.1709
.2408
.2234
0
0
0
.1658
.2395
.2184
0.
0.
0.
1610
2383
2137
                                     231

-------
     Regression   models   also   were   constructed   to   relate  nutrient
concentrations to late coverage by nuisance aquatic weeds from data contained
in the FIADAB late data base developed at the University of Florida  (9). No
significant  relation was  found  between nutrient  concentrations and  areal
coverage or  percent  coverage by water hyacinths or hydrilla.  The lack of a
nutrient - aquatic weed relationship and the snail changes  predicted  in late
Russell nutrient levels suggested that proposed PS loadings were  unlikely to
affect the status of aquatic weeds in Late Russell.

     Finally, we used equations relating fish biomass  and percent biomass of
sport fish,  forage fish and rough fish to the chlorophyll trophic  state index
in  Florida lakes of varying trophic  state (10).  We  found that  sport  fish
biomass expressed  as a percent of total biomass decreases as trophic state
index increases  (Figure 4).  However,  the predicted change  is caused by an
increase in  rough fish biomass, and the actual biomass of sport fish (kg/ha)
appears to be insensitive to changes in trophic state index.


                          SUMMARY AND CONCXIJSIONS
     •Hie  wasteload allocation procedure applied by FDER resulted  in overly
restrictive discharge  limits  to Reedy  Creek  because  FDER did  not  take
advantage of the ability of SIMLAK to vary N to P ratios and because of the
25%  reserves   of N and  P.  Alternate  models  and  the  SIMLAK  equations
indicated that point source flows of  24.5  mgd with  nutrient concentration
limits of  0.5  mg/L for TP  and 2.15 mg/L for TN  should cause only slight
increases  in  nutrient concentrations  in late Russell  that would not  be
noticeable to users  of  the late.

     The  Administrative  Hearing was  settled  by  a consent  agreement  that
allowed 22.5 mgd of  PS  flows to Reedy Creek with TP of 0.5 mg/L and 2.0 mg/L
of TN. The wasteload allocation procedures  have  been modified recently by
FDER to allow for more  outside involvement, but the SIMLAK model has not been
changed.

     Based on the Reedy Creek/Late Russell wasteload allocation, we recommend
that FDER develop several models better suited to different types of Florida
lakes   (N  limited,  P limited,  highly  colored lakes).  Their  approach should
allow  -reasonable variations  in late conditions such as N to P ratios in order
to attain  allocations  with achievable nutrient concentration  limits.  FDER
also should develop procedures for model calibration and criteria for judging
the acceptability of model predictions.  Their models should be  capable of
predicting changes  in  late conditions that  are equivalent  to  the increase
allowed by the WIA process.

     The   work  described  in  this  paper   was   not  funded  by  the  U.S.
Environmental  Protection Agency and therefore the contents do not necessarily
reflect  the views  of the Agency  and no   official  endorsement  should be
inferred.
                                     232

-------
             20 -
             10 -
                50
            300
        o»
        X
        o
        m
                50
                  TOTAL
Figure 4.  Fish-Trophic state relationship in Florida lakes  (from Bays and
           Crisman 1983; Dashed lines represent approximate  range of TSI for
           Lake Russell based on S3MLAK predictions of chlorophyll).
                                      233

-------
1.   Magley, W.  Reed/ Creek/Lake  Russell (Osceola County)  Wasteload Study.
     Water  Qjaality  Technical  Series,   2   (82),   Florida   Department  of
     Environmental Regulation, 1984.  279 pp.

2.   Florida  Department  of Environmental Regulation.  Notice of Intent  to
     Adopt  an Agency Position Concerning Wasteload Allocations  in the Reedy
     Creek Basin, July 11,  1984.

3.   Poinciana  Utilities,   Inc.   Petition  for  Administrative  Proceeding,
     August 24, 1984.

4.   Hand,  J. and IfcClelland,  S.  Ihe lake Model  "SDCAK" Users Guide. Water
     Quality  Technical Series,  3  (3),  Florida Department  of Environmental
     Regulation,   1979.   24 pp.

5.   Kirchner,  W.  B.  and  Dillon,  P.  J.  An empirical method of estimating
     the  retention  of phosphorus in  lakes.   Water  Resour.  Res.  11:  182,
     1975.

6.   Baker,  L.  A.,  Brezonik,  P.   L.,  and Kratzer,  C.  R.  Nutrient  loading-
     trophic  state relationships  in Florida lakes.  Water Resources Res. Ctr.
     Pub. No. 56, Univ. of  Florida,  1981.

7.   Vollenweider,  R. 'A.  Irqput-output models with  specific reference to the
     phosphorus  loading  concept  in limnology.  Schweiz,  Z. Hydrol.  37: 53,
     1975.

8.   Dillon,  P.J. The  FC>4  budget  of  Cameron lake, Ontario:  The importance
     of flushing rate to the degree of  eutrophy of lakes. Limnol. Oceanogr.
     20:  28, 1975.

9.   Canfield,  D.E. and  Hodgson,  L.H.  Prediction  of Secchi  disc depths in
     Florida  lakes: Impact of Algal biomass and organic color. Hydrobiologia
     99:51, 1983.

10   Huber,  W.C.,  Brezonik, P.L.,  Heaney,  J.P.,  Dickinson,  R.E.,  Preston,
     S.D.,  Dwornik,  D.S.,   and Demaio,  M.A.  A  classification  of  Florida
     lakes.  Water Reources Research  Center  Pub.  No.   72,  University of
     Florida, 1982

11.  Bays,   J.S.   and   Crisman,   T.L.   Zooplankton   and   trophic   state
     relationships in  Florida lakes.  Can.  J.  Fish Aquat.  Sci.  40:  1813,
     1983.
                                      234

-------
          MODEL COMPLEXITY FOR TROPHIC STATE SIMULATION IN RESERVOIRS

               by:  Raymond A. Ferrara
                    Assistant Professor, Department of Civil Engineering
                    Lafayette College
                    Easton, PA  18042

                    Thomas T. Griffin
                    Najarian and Associates, Inc.
                    Eatontown, NJ 07724
                                ABSTRACT

     Frequently model users are faced with choosing an appropriate model
structure and level of complexity for a particular application.   Increased
model complexity generally implies incorporation of more portions or a
larger portion of the real world, but also generally implies an increased
number of rate coefficients and input parameters, each of which has some
associated uncertainty.  This paper examines this issue through mathematical
simulation of trophic state in reservoirs with two models of different levels
of complexity.  Analysis and comparison of the models' output is accomplished
through various statistical measures and techniques including root mean
squared error, regression analysis with student's t test for slope and inter-
cept, and difference of means.  This study revealed that for predicting
steady-state total phosphorus concentration, no advantage to using the more
complex model over the simpler model was observed.  For time-variable simu-
lation, it was determined that the models were substantially in agreement
for long term averaging (i.e. greater than 10 years), but that such models
do not agree for short term simulation nor even for the frequency distribu-
tion of phosphorus concentrations within that simulation period.  In less
biologically productive oligotrophic systems, the level of model complexity
appeared to b.e of lesser importance than for more biologically productive
eutrophic systems.  The nonlinearity of the phosphorus uptake relationship
is demonstrated to be more pronounced in eutrophic than in oligotrophic
systems, and explains the discrepancy between the models' predictions.


INTRODUCTION

     Frequently model users are faced with choosing an appropriate model for
their particular application.  For trophic state determination in impound-
ments, the choices range from the steady state total phosphorus loading dia-
grams (1, 2) to the time-variable ecosystem models (3, 4, 5, 6, 7, 8).  In
many cases a user may assume that the more complex the model, the more accu-


                                      235

-------
rate or reliable the simulation.  This may not always be true as even the
most complex models are never mathematical clones of real physical systems.
Thomaim (9) aptly states, "A mathematical model is simply an analytical ab-
straction of the real world.  As such, it does not pretend to incorporate all
phenomena but rather abstracts only those portions of the real world that are
relevant to the problem under consideration."

     Although increased complexity, generally implies incorporation of more
portions or a larger portion of the real world, it must be realized that in-
creased complexity also generally implies an increased number of rate co-
efficients and input parameters to utilize the model.  Since the numerical
value of each of these has some associated uncertainty, then it is entirely
possible that uncertainty in the model output may also increase with complex-
ity.  The parabolic curve of Figure 1 represents a feasible picture of the
relationship between model complexity and model uncertainty.  Of course, as
the state of our knowledge of individual biological, chemical, and physical
processes improves throughout time, so too will our ability to mathematically
describe these processes and hence to build more certain complex models.

     This paper does not attempt to resolve the entire dilemma, but rather to
examine it through mathematical simulation of trophic state in reservoirs by
comparing the output of two models of different complexity.  The comparisons
are made under steady-state and time-variable conditions.  For the latter,
two types of analyses are conducted:   (1) Direct comparison of the models'
predicted concentrations at corresponding times, and  (2) Comparison of the
frequency distribution of the models' predicted concentrations over the en-
tire simulation period.  The methodology and conclusions are applicable to
modeling water quality constituents in general.
    Model
 Uncertainty
                                                  future
                                                {e.g. 2000)
                                                       Model Complexity
           Figure  1.   A Relationship  Between  Model  Complexity  and  Uncertainty

                                      236

-------
METHODOLOGY

     Two contrasting models of phosphorus dynamics in reservoirs are con-
sidered.  Both models account for time-variable reservoir volume, influent
and effluent flow rate, and influent phosphorus concentrations.   The first
model (10) represents a reservoir as a single fully mixed system and models
a single water quality variable,  total phosphorus (TP).   The model is de-
picted in Figure 2.  The change in TP with time is modeled as the sum of the
influent load, minus the effluent load minus a net decay of phosphorus (e.g.
due to sedimentation), according to the following equation:
      d(V'TP)                    vs
        at    =  VTpi  ~  VTP " T  'TP'V
                                                                  (1)
 where
V
t
        Qe
       TP±
        TP
        VS
         H
reservoir volume
time
influent flow rate
effluent flow rate
influent phosphorus concentration
phosphorus concentration in the reservoir and
in the effluent
apparent settling velocity
mean depth of reservoir
 Typical  values  for  vg reported in the literature are 10 to 16 m/yr   (1, 11,
 12).   The form  of the model  is similar to those previously presented by
 Vollenweider (1) and Chapra  (13).   It will be referred to hereafter as the
 Total Phosphorus Model,  TPM.

     The  second  model (14)  represents a reservoir as having two fully mixed
 layers,  an epilimnion and  a  hypolimnion, during periods of thermal stratifi-
 cation.   When temperature  gradients are small or absent, for example during
 overturn, the reservoir  is represented as a  single fully mixed layer.  Three
 forms of phosphorus - organic phosphorus (OP), dissolved inorganic phosphorus
 (DIP), and particulate inorganic phosphorus  (PIP) - and dissolved oxygen are
 modeled  in each layer.   The  model is depicted in Figure 3.  The model equa-
                TPi
                                 V, TP
                                   1
                                             Qe, TP
                 Figure 2.  The Total Phosphorus Model  (TPM)

                                    237

-------
                 Stratified Model
                Fully Mixed Model
Figure 3.  The Multi-Component Phosphorus Model (MCPM)
                        238

-------
 tions have been presented previously in Griffin and Ferrara (14) .   The model
 will hereafter be referred to as the Multi-Component Phosphorus Model (MCPM) .

     TPM may be considered very nearly the simplest of model structures, i.e.
 a low level of complexity.  MCPM provides a higher level of complexity in-
 corporating characteristics of nutrient cycling  and biological transforma-
 tions.  However, it is not as complex as some of the ecosystem models cited
 above.  MCPM was developed to simulate seasonal changes in water quality,
 yet still be simple and economical enough for long term simulation on the
 order of decades.  TPM was also derived for long term simulation,  yet be-
 cause of its simple structure, would not be expected to consistently predict
 seasonal changes in water quality accurately.

     TPM requires specification of one rate coefficient, i.e.  vs.   The quo-
 tient ^ is sometimes reported as a constant,  a, a first-order net sedimenta-
 tion  H  rate coefficient.  At steady-state,  the solution to  TPM is
            TP .          TP .
     TP = - = - —                                              /2\
         1  + ^s  V    1 +  at
              3  Q

 where t  is  the detention  time,  Q-.  Vollenweider (15)  has reported, based on
 data from many lakes, that
Hence, knowledge of the detention time of a particular waterbody leads to an
estimate of the single reaction rate coefficient necessary to utilize the
total phosphorus model.

STATISTICAL MEASURES

     Various statistical techniques exist for comparing data sets whether the
data are actual observations or the results of model simulation.  Statistical
measures used in this paper include the following (16, 17).

1.  Confidence Interval about Mean

     For a limited number of data points, X, one can identify an interval
within which it can be said, with a certain amount of confidence, that the
true mean of X, yx, will reside.  This interval is given as  follows:


     5 '

where  X = the sample mean
       S = the sample standard deviation
       t = the percentile of the student t distribution
           with N-l degrees of freedom and y equal to
           100 minus the % confidence level.

                                     239

-------
When comparing two data sets, if their intervals overlap, then the sample
means are not significantly different, and vice versa.

2.  Root Mean Squared Error

     A convenient method of comparing the difference between two data sets
is to compute the root mean squared error, R, where
     R =
N                •
z <*ii - V
         	                                               (5)
               N
The subscripts 1 and 2 refer to data sets 1 and 2 respectively.  The signif-
icance of R is pictured in Figure 4 where it is shown to represent the de-
viation from perfect agreement between two data sets.  Since R has the same
units as X (e.g., concentration), it is useful for comparison to express a
relative error, RE, where


     RE = -J-                                                           (6)
           A

Note that either XT or X2 could be used in the denominator of equation (6).

3.  Regression Analysis

     A further test of agreement can be made by regressing the data as pre-
sented in Figure 4  and comparing the intercept, a, and slope, b, of the re-
gressed line with the theoretical values of a and $, respectively.   A two-
tailed t test with N-2 degrees of freedom and the following statistics is
conducted,


                            '-                                         <»
where Sfc and Sa are the standard deviation of the slope and intercept, re-
spectively.  For t, leads to the following statistic
                                                                         (8)
                                     240

-------
                        ROOT MEAN SQUARED ERROR
              T2
              x2
                R =
                       1 - Xj)2 + (Y2  - X2)2 +  ... (YN - XM)
                          REGRESS[ON ANALYSIS
                                                      LINE OF PERFECT
                                                      AGREEMENT
                                                           LINE OF
                                                           BEST FIT
      a=0
Figure A.   Root  Mean  Squared Error and Regression  Analysis
                                  241

-------
where    6 = true  difference between data  set means
        Sd = the standard  deviation of   d  which  is
            given by  a pooled  estimate  of the  population
            variance.

If the  variance of  the two data  sets are  assumed equal, then
                 d  -  6             K*2  (N1+N2-2) I                         (9)

           I	2	2~>      « + N,
           1(^-1) S^  +  (N2-1)S2    ll      1     2


 Note  that  for  perfect  agreement  between data set means,  6=0.  Then a value
 of  t{j
-------
TABLE 1,  STEADY STATE SIMULATION RESULTS
Test
I
II
III
IV
V
VI
VII
VIII
IX
X
XI
XII
XIII
XIV
XV
XVI
XVII
XVIII
Mean
Standard Deviation
t(yr)
0.06
0.10
0.12
0.20
0.21
0.27
0.34
0.35
0.41
0.43
0.48
0.62
0.75
0.82
1.37
1.92
2.47
3.01


a(eq.3)
4.13
3.12
2.92
2.26
2.21
1.91
1.71
1.68
1.56
1.52
1.44
1.27
1.15
1.10
0.85
0.72
0.64
0.58


Coefficient -of Variation
*'TpM»"9' *
48.3
45.4
44.7
41.6
41.3
39.4
37.9
37.7
36.6
36.2
35.5
33.6
32.1
31.5
27.6
25.2
23.3
21.9
35.5
7.59
0.21
MCPM(ug/1)
43.0
39.1
41.1
41.1
37.6
41.2
37.7
41.6
30.1
42.1
37.9
38.5
39.0
29.0
29.0
29.3
29.9
30.5
36.5
5.26
0.14
                     243

-------
diet water quality changes from one year to the next, but rather to predict
the frequency distribution of water quality (as measured by total phosphorus
concentration) due to some reservoir management scenario.  MCPM was intended
for simulation of seasonal changes in water quality, yet be economical enough
for long term simulation as well.

     It might be expected from the steady-state analysis presented previously
in this paper and from the objectives of the two models that their predic-
tions for long term average concentrations should be in agreement.  In such
cases it would be expedient to use TPM.  However, it might further be ex-
pected that as the period of simulation decreases, the model predictions
would begin to diverge.  At that stage, the model user must choose between
TPM and MCPM, and would likely accept the latter.  The critical question
then becomes how to identify this point of divergence and how different the
model predictions really are.  This may be accomplished through comparison
of the model predictions for simulations of varying time periods.

     MCPM was run for a fifty-seven year simulation utilizing the rate co-
efficients previously defined  (Figure 5).  From_the model output a mean
total phosphorus concentration was calculated, TP^cpM = 27.2 ug/1.  TPM was
then run for the same conditions.  However, it was first necessary to choose
a value for the sedimentation rate coefficient.  Equation (3) is inappro-
priate for two reasons, (1) it was derived for steady-state analysis, and
(2) the concept of detention time becomes meaningless under time-variable
simulation unless of course a temporally averaged flow rate and reservoir
volume are used to compute it.  Values reported for vs in the literature
were similarly determined from steady-state data.  Chapra (13) utilized a
to
=>
OL
O
X
Q.
in
o
       50

       45

       4O

       35

       30

       25

       20

        15

        10

         5

         0
              50
               100  150  200 250  300   350  400  450  500  550  600  650  700
                                TIME   I MONTHS)

                  Figure  5.  Tine Variable  Simulation  -  Caso  1

            (dashed  line  represents TPM;  solid  line  represents MCPM)

                                     244

-------
value of v_ = 16 m/yr in a time-variable total phosphorus model of the Great
Lakes.  Since there is   no a priori method for choosing a value for vs and
a sufficiently large data base for model calibration did not exist_,_ it was
decided for this study to adapt a value which results in TP-ppM = TPjiCPM
for the 57 year simulation.  This of- course forces the two models to agree
in the long term average but does not prejudice their relative agreement or
disagreement for shorter periods or for the frequency distribution of concen-
trations over the_long term.  A value of vs = 9 m/yr results in TP~TPM approx-
imately equal to WMCPM (Table 2).  This value is indeed close to the re-
ported steady-state range of 10 to 16 m/yr and the time variable value
utilized by Chapra (13).

     The 57 year simulation may be broken down into various averaging periods
for comparison of the two models.  For any averaging period of simulation
years there will be (57-x+l) values of TPjpM and TPMCpM for comparison.  For
example, for a 57 year averaging period there is one value each for TF-pp^
and TPjicPM' i-e-» 27-5 and 27.2 ug/1 respectively.  For a 25 year averaging
period, there will be 57-25+1 = 33 values for each model.  For monthly
averaging periods of course there will be 57x12 = 684 total phosphorus con-
centrations for each model.

     The root mean squared error for various averaging periods is provided
in Table 3.  Note that the relative error or disagreement between the models
increases up to 19% when monthly average TP concentrations are compared,
whereas the disagreement is only 5% for the 25 year average concentrations.
The results of the regression analysis are also presented in Table 3.  Slopes
and intercepts are significantly different than one and zero for the monthly
and 1 year groups but not for the 4, 10 and 25 year groups.  This statistical
analysis verifies the hypothesis that the degree of agreement between the two
models does decrease significantly as the time period of simulation for com-
paring average total phosphorus concentrations is also decreased.

     An analysis of the frequency distribution of predicted monthly average
total phosphorus concentrations for each model was also conducted.  Although
it is shown above that the monthly average predictions in individual months
do not agree, it may be that the distributions of those predictions will
agree.  This may be satisfactory for certain management decisions where for
               TABLE 2.  TP    AS A FUNCTION OF v  - CASE  I
                           1 r M                   5
                Vs                  TPTpM(ug/])

                   5                          34.5
                   7                          30.6
                   8                          29.0
                   9                          27.5
                  10                          26.2
                  12                          23.9
                  15                          21.1
                                    245

-------
      TABLE 3.  STATISTICAL ANALYSIS OF TIME VARIABLE SIMULATION - CASE I
Averaging
Group
Monthly
1 year
4 year
10 year
25 year
57 year

R
5.05
4.66
3.84
2.19
1.48
0.30

RE
0.19
0.17
0.14
0.08
0.05
0.01

Slope
1.36
1.74
1.75*
1.23*
1.54*
-

Intercept
-9.32
-19.86
-20.11*
-5.18*
-13.51*
-
     *Indicates value not significantly different  than  1.0  for  slope or 0.0
      for intercept at 95% confidence level.

example  it  is  only necessary to know the percent of time concentration ex-
ceeds a  certain standard.  The data  were grouped into  intervals of 0.5 ug/1,
(Figures 6.a and 6.b).   Note that the  data are approximately normally dis-
tributed but that  the  characteristics  of the two models are interestingly
different.  As in  the  steady-state simulations, although both models have
essentially the same mean value, TPM is characterized  by a larger standard
deviation and  hence  coefficient of variation.  Consequently, the distribution
of predicted monthly average total phosphorus concentrations appears to be
different.  This was tested by conducting a regression analysis comparing
the  frequency  of occurrence in respective concentration intervals.  The re-
sults of this  analysis reveal a slope  and intercept of 0.32 and 7.43 re-
spectively, values which were further  determined to be indeed statistically
significantly  different  than the theoretical values of 1.0 and 0.0 re-
spectively.  Therefore,  in addition  to the fact that the two models are pre-
dicting  significantly  different concentrations in corresponding individual
months,  they are also  predicting significantly different distributions of
concentrations over  the 57 year simulation period.

TIME VARIABLE  SIMULATION - CASE II

     The above time-variable simulation was conducted  on an impoundment which
might be "considered  eutrophic, i.e.  total phosphorus concentrations generally
greater  than 20 ug/1.   In order to compare the models  in a less nutrient-rich
system,  the same simulations were conducted with the influent phosphorus con-
centration  reduced to  one-fourth of  its previous value.  The model simula-
tions are presented  in Figure 7.  A  value of TPtfcPM =  iO-07 "S/1 (very nearly
oligotrophic conditions) was obtained.  Interestingly, the calibrated value
for  vs was  found to  be 3 m/yr (Table 4).  Because of the reduced influent
phosphorus  load, this  system is far  less biologically  active.  Since the role
of biological  uptake and deposition  has diminished, the system must be char-
acterized by a  smaller net sedimentation/decay rate coefficient.  In the pre-
vious system,  a value  of vs = 9.0 m/yr was obtained and was in close agree-


                                     246-

-------
   32
   24-
o

Ul

O
Ul
£E
U.
    8-
o
2
Ld
Z)
O
UJ
(E
   84
   63-
42-
   21-
                       I
               10         20        30

              TOTAL  PHOSPHORUS  (ug/l)
                                         |k
                                          40
                                          TPM
50
                                      llVlt  1-1 r.
               10         20        30        40

              TOTAL  PHOSPHORUS  (ug/l)-MCPM
                                                     5O
      Figure 6a.  Frequency Distribution for Tine-Variable Simulation
                Case I
                             247

-------
 (Jj
     700-t
     600-
 3   500-
     400-
O
Q
2

Al

£   300-
     ZOO'
 u.
 0   100
                                                MCPM
                                                       ~PM
10      15     ZO     Z5     30     35
     TOTAL  PHOSPHORUS   (ug/1)
                                                            40
                                                                   45
     Figure  6b.   Cumulative  Frequency Distribution for Time Variable
                 Simulation  - Case  I
merit with the range of  10 - 16 m/yr, a range which was determined from data
on predominantly biologically productive eutrophic lakes.  This study re-
veals that for a less productive oligotrophic/mesotrophic impoundment, vs
should be substantially reduced.

     Results of the statistical analysis for this case are presented in
Table 5, and lead to conclusions similar to those drawn from Case I and
Table 3.  The distribution of predicted monthly average concentrations is
presented in Figures 8.a and 8.V.  Again the data are approximately normally
distributed, but TPM is characterized by a greater degree of variation as in
the previous case.  The regression analysis again revealed that the models
are predicting significantly different distributions.

DISCUSSION
     The analyses presented above raise an important question as to the ap-
nlicabilityof a simple total phosphorus model for predicting trophic state
in impoundments.  Models which represent the influence of biological activity
on the cycling of phosphorus between organic and inorganic forms include
functional relationships for their reaction rate coefficients which express
a dependence on various internal or external system parameters.  An example
                                     248

-------
               TABLE 4 - TPTpM AS A FUNCTION OF vs  - CASE II
                                           TP
                                             TPM
                      1
                      2
                      3
                      5
                      8
                     10
                                          11.2
                                          10.6
                                          10.0
                                           8.7
                                           7.3
                                           6.6
of this in the multi-component phosphorus model is the organic phosphorus
growth rate term which is a function of nutrient availability.  If nutrient
concentrations are high, biological uptake and deposition will also be high.
The converse is also true.  The total phosphorus model with its single re-
action rate coefficient does not include this flexibility, and hence as
demonstrated^ two distinctly different values for vg were required to obtain
the appropriate model response under the two distinctly different phosphorus
loading scenarios.

     This phenomenon is clearly demonstrated in Figure 9 which is a plot of
the time-variable nutrient limitation term.  Note that in Case I, values are
typically greater  than those in Case II corresponding to the reduction in vg
from 9 to 3 m/yr respectively.  To further verify this, a third simulation
V)
D
tE
O
I
35

30-

25

20

15

10

 5

 0
                   •MCPM
                                                     TPM
        (    50   100  (50  20O  250  300  350  400  450  500  550  600  650  TOO

                                TIME {MONTHS)

                        Figure  7.  Time Variable Simulation - Case  II

                  (dashed line  represents TPM; solid line represents MCPM)
                                     249

-------
was conducted, utilizing conditions identical to Case I,  however,  the  half-

saturation constant was increased by a factor of three for demonstration

purposes.  TPMcpM was determined to be 37.4 ug/1, indicating fairly  enriched

conditions, yet the appropriate value of vg was found to  be 4 m/year.  The

nutrient limitation term illustrated in Figure 9 provides the explanation.

Here again the net organic phosphorus growth rate terra is reduced  preventing
      132-
       99-
    O
    UJ
    tr.
    u.
       33-
                      1
                   IO         20         30        4O


                    TOTAL  PHOSPHORUS (ug/l)-TPM
50
     424-
      318-
    z
    Ul 212-
    r>
    o
    uj
    tr
    u.
      IO6-
                   10         20         30        40

                    TOTAL PHOSPHORUS (ug/1)-MCPM
50
          Figure  8a.  Frequency Distribution for Time-Variable Simulation

                     Case  II
                                    250

-------
        700
        6OO-
     UJ

     i
     Q
     id
soo-
     u
     O  400
     z

     A|
     Q.
        300-
        200-
        100
                                               MCPM
                                                          TPM
                                                 10
                                                         12
                                TOTAL  PHOSPHORUS
                                      (ug/l)

         Figure 8b.   Cumulative Frequency Distribution for Time Variable
                     Simulation - Case II
the cycling of phosphorus into the organic form with subsequent deposition to
the sediments.  This study demonstrates the importance of including the
strong influence of biological activity to represent the net sedimentation of
phosphorus in a model for trophic state simulation.  A simpler version does
not respond accordingly for time-variable simulation, unless of course vs is
adjusted as demonstrated to be required above.

     The nutrient limitation term is also the key to the observed difference
between relative error and correlation coefficients for the two cases as re-
ported in Tables 3 and 5.  There was significant improvement in Case II as a
result of the more constant nature of the nutrient limitation term.  This re-
duces the nonlinearity of the multicomponent phosphorus model and provides
better agreement with the linear total phosphorus model.

     Another difference which becomes apparent from this study is the con-
sistently higher coefficient of variation observed with the total phosphorus
                                    251

-------
   TABLE 5 - STATISTICAL ANALYSIS OF TIME-VARIABLE SIMULATION -  CASE  II
Averaging
Group
Monthly
1 year
4 year
10 year
25 year
57 year

R
0.979
0.949
0.776
0.374
0.197
0.05

RE
0.098
0.095
0.078
0.037
0.020
0.005

Slope
1.55
1.7
1.64*
1.54*
1.96


Intercept
-5.53
-6.9
-6.48*
-5.34*
-9.53

Indicates value not significantly different than 1.0 for slope or 0.0 for
intercept at 95% confidence level.
model in both steady-state and time-variable simulations.   This results from
the fact that TPM because of its simple biochemical structure, is more
strongly dominated*by input-output forces, whereas MCPM represents an effec-
tive ecological buffering capacity which reacts to changes in external
forces.  This difference in variance is most clearly demonstrated in the fre-
quency distribution histograms of Figure 6 and 8.

CONCLUSIONS

     Several conclusions have been developed from this study.  For steady-
state simulation, the two models do not differ appreciably in their predic-
tion of total phosphorus.  However, for time-variable simulation, the models
     0.6 -
    o.o
                                                        LEGEND
                                                             CASE I
                                                     	  CASE 2
                                                     	  CASE 3
                100      200      3OO      400      SCO

                                TIME (MONTHS)
600
70O
              Figure 9,  Time-Variable Nutrient Limitation Term

                                     252

-------
do predict significantly different concentrations.   Furthermore, the fre-
quency distribution of those concentrations even over very long term simula-
tion periods also differ.  For a given simulation,  the total phosphorus model
will predict a wider range of concentrations than the multi-component phos-
phorus model.

     It is important to include a functional relationship describing the in-
fluence of biological activity in the cycling of phosphorus through its
various organic and inorganic forms thereby determining the net sedimentation
rate for total phosphorus.  The value for vs is not constant across trophic
states.  The nonlinearity of the nutrient limitation term is responsible for
the difference between model predictions.  The relative error between the
models decreases as the degree of variation in this term decreases.
                                REFERENCES

 1.  Vollenweider,  R.A.   Input Output Models with Special Reference to the
     Phosphorus Loading Concept in Limnology.  Schweiz.  A.  Hydrol.  37:
     1975.

 2.  Dillon,  P.J.   The  Phosphorus  Budget  of  Cameron Lake, Ontario:  The
     Importance of  Flushing Rate to  the Degree  of Eutrophy  in Lakes.
     LjLmnol.  and Oceanogr.  19:28,  1975

 3.  Jorgensen, S.E., Ecological Modeling of Lakes  Chapter 9 In: G.T.
     Orlob (ed.), Mathematical Modeling of Water  Quality: Streams,  Lakes,
     and Reservoirs, Wiley-Interscience,  N.Y.,  N.Y.,  1983

 4.  Jorgensen, S.E.  A Eutrophication  Model for  a Lake.  Ecological
     Modelling. 2:  147,  1976

 5.  Thomann,  R.V.,  et  al.   Mathematical  Modeling of  Phytoplankton  in  Lake
     Ontario,  1.  Model  Development and  Verification.  EPA-660/3-75-005,
     U.S.  Environmental  Protection Agency, Corvallis, Oregon, 1975.  177  pp.

 6.  DiToro,  D.M.,  and  Matystik, W.F.   Mathematical Models  of Water Quality
     in Large Lakes, Part I:  Lake  Huron and  Saginaw Bay.  EPA-600/3-80-059,
     U.S.  Environmental  Protection Agency, 1980a.

 7.  DiToro,  D.M. and Connolly,  J.P.  Mathematical Models of Water  Quality
     in Large Lakes, Part 2:  Lake  Erie. EPA-600/3-80-065, U.S.  Environ-
     mental Protection  Agency, 1980b.

 8.  Chen, C.W. and Orlob,  G.T.   Ecologic Simulation  for Aquatic  Environ-
     ments In; B.C.  Patten (ed.)  Systems Analysis and Simulation in Ecolo-
     j»y_ Vol.  3, Academic Press,  1975.

 9.  Thomann,  R.V.,  Systems Analysis and  Water  Quality Management,  Environ-
     mental Science Services Division,  Environmental  Research and Appli-
     cations,  Inc.  N.Y., N.Y. 1972.  286 pp.
                                     253

-------
10.  Ferrara, R.A., T.S. Nadbielny, T.T. Griffin.  Reservoir Management for
     Water Supply and Water Quality Objectives. Current Practices in Envir-
     onmental Science and Engineering. 2:1, 1986

11.  Dillon, P.J. and Kirchner, W.B.  Reply to Comment by S.C. Chapra  Wa-
     ter Resources Research. 11:6, 1035, 1975

12.  Chapra, S.C.  Comment on An Empirical Method of Estimating the Reten-
     tion of Phosphorus in Lakes. Water Resources Research, 11:6, 1033, 1975

13.  Chapra, S.C.  Total Phosphorus Model for the Great Lakes.  JEEP, ASCE
     103: EE2, 142, 1977

14.  Griffin, T.T. and Ferrara, R.A.  A Multi-Component Model of
     Phosphorus  Dynamics in Reservoirs.  Water Resources Bulletin,
     20:5,  777,  1984

15.  Vollenweider, R.A.  Advances in  Defining Critical Loading Le-
     vels for Phosphorus in Lake Eutrophication.  Mem. 1st Ital.
     Idrobio. 33:1976

16.  Thoraann, R.V.  Measures of Verification In: R.V. Thomann and
     T.O. Barnwell (ed.), Workshop on Verification of Water Quality
     Models.  EPA-600/9-80-016, U.S.  Environmental Protection Agency,
     Athens, Georgia, 1980. 258 pp.

17.  Miller, I.  and Freund, J.E.  Probability and Statistics for
     Engineers 2nd edition, Prentice  Hall, Englewood Cliffs, N.J.
     1977.  529 pp.
 Disclaimer

      The work described in this paper was not funded by the U.S.  Environ-
 mental Protection Agency and therefore the contents do not necessarily
 reflect the views of tfie Agency and no official endorsement should be
 inferred.
                                     234

-------
 F.D.O.T. Drainage Manual:Why "Drainage"1n an Age of "Stormwater Management"?

                by:  Edward G. Rings
                     Florida Department of Transportation
                     Tallahassee, Florida 32301
                                  ABSTRACT
     In 1963, the Florida State Road Department issued the first "Drainage
Manual".  New policy and procedures were added by memorandum and design
directives.  In the intervening 23 + years, enviromental awareness in general
and in Florida in particular increased dramatically.  The intervening
development of Florida stormwater regulation is traced.  When in 1985 a
replacement manual was begun, the decision was made to retain the title
"Drainage Manual" rather than changing the title to "Stormwater Management
Manual".  This was not done to indicate that stormwater management was to be
disregarded, but to emphasize that drainage, which is critical to
transportation, must be provided.  An overview of the draft Drainage Manual
is presented which demonstrates that proper drainage is good stormwater
management.  Ineffective or inadequate drainage is inappropriate stormwater
management.

                          1963 FSRD DRAINAGE MANUAL

     In Florida the "state of the art" for highway drainage as it existed 25
years ago can be seen by reviewing the 1963 (revised 1967) Florida State Road
Department Drainage Manual.  A listing of the chapter headings is provided in
Table 1.  The 1963 manual has only a few tables, equations and figures which
are identified in Table 2.  The manual relied on the user having a good
working knowledge of hydrology and hydraulics and a fundamental
hydrologic/hydraulic library upon which to draw.  The manual supplied the
necessary information to enable the user to utilize reference information to
accomplish highway drainage to Department standards.


     What was the thrust of the manual?  In two words it was runoff
conveyance.   The methods were simple and straight forward, and not
surprisingly, they worked.  This can be explained because the manual was

                                    255

-------
TABLE 1.         STATE ROAD DEPARTMENT DRAINAGE MANUAL 1963
                             TABLE OF CONTENTS
   Chapter              Chapter Heading                           Pages
1
2
3
4
5
6
7
8
9
10
11
12
13

Introduction
The Drainage Section
Engineering Law
Policies
Survey and Field Information
Rainfall and Runoff
Hydraulic Design of Culverts
Ditches and Canals
Hydraulic Design of Bridges (never completed)
Subgrade Drainage
High Water Clearance for Pavement Grades
Storm Sewers
Checking Plans
TOTAL
1
1
10
5
3
15
4
5
0
2
1
9
6
62
written by District and Central Office drainage engineers reflecting 35 +
year record of the highway drainage experience in Florida.

     By 1963, Florida had begun to stir and start the transition from a
mostly rural state with a few predominately coastal urban centers into one of
the fastest growth states in the U.S.  It is safe to categorize the 1960's
Florida drainage law as "common law" which permitted upper owners to increase
the rate (quantity) of discharge provided the increase did not cause damage.
The standard applied was one of reasonableness.  This has been described as a
modified "look out below" drainage law.  By the late 1970's most major urban
areas developed regulations and ordinances which limited and defined maximum
allowable peak discharges, since the cumulative effect of many "resonable"
increases is not necessarily reasonable.  The special problem of total volume
amount in landlocked watersheds prevalent in north and central Florida's
karst topography was not addressed.

     The first Drainage Manual required the highway drainage designer to
consider future development within a 20 year period but did not allow private
connections to the Department's drainage facilities.  However, connections
were routinely granted provided the applicant demonstrated that the discharge
.does not come from divirted areas and that the Department's system was
designed for the level of discharge.  The conveyance system below the
Department's outlet has always been evaluated with the extent of the
evaluation  increasing in recent times.  Quality was addressed in general
terms.  Quality requirements were broad, requiring dischargers to meet State
water quality standards.  A general interpretation of the earlier requirement
was "do not contaminate the stormwater".

                                     256

-------
 TABLE 2.          STATE ROAD DEPARTMENT DRAINAGE MANUAL 1963
                        TABLES ,  EQUATIONS AND FIGURES


         Tables                 Equations               Figures


  Storm Design Frequency     Rational             Rainfall  Stations
  Runoff Coefficient         Talbot's  Formula     Intensity-Duration  Curves
  Minimum Culvert Sizes                          Time  of Concentration  for
  Cleaning Velocities                                   Overland  flow
  Base Clearances                                Talbot's  Discharge  Curve
  Manning's "n" Values                            Storm Sewer Tabulation Form
  Maximum Manhole spacing                         Highway Section Check  List
                    THE INTERVENING PERIOD - 1963 to 1985

     Prior to 1972, water management legislation had developed on a piecemeal
basis.  In that year, a comprehensive law has enacted to provide extensive
protection and management of water resources throughout the state.

     The Florida 1972 Water Resources Act, Chapter 373, Florida Statues,
provides a two-tiered administrative structure presently headed at the state
level by the Department of Environmental Regulation (DER).  DER is the chief
pollution control agency in the state and was created from parts of several
agencies through Legislative reorganization in 1975.  DER's jurisdiction over
water pollution control extends to "waters of the state".  Practically all
dredge and fill activities conducted in areas either in or connected to
waters of the state are required to comply with water quality standards.  DER
regulates discharge or untreated stormwater that could be a potential source
of pollution to the state.  This regulatory scheme is predominantly
qualitative.  All stormwater discharges must meet the water quality standards
of the class of water body the stormwater actually reaches.  Additionally,
DER regulates stormwater by requiring retention or retention with filtration
systems that allow separation of polluting substances by percolating the
water into the ground.  DER supervises five regional Water Management
Districts designed to provide the diverse types of regulation needed in
different areas of the state.  These include the previously existing Central
and Southern Florida Flood Control District, renamed the South Florida Water
Management District, and the Southwest Florida Water Management District.
The three new districts established under the Act were the Suwannee River,
St. Johns River, and Northwest Florida Water Management Districts.

     Most of the thrust of DERs and the Water Managment Districts efforts
have been towards water quality.  Efforts to address stormwater quantity
control were mainly left to local government.  Under the present law,
municipalities have authority to brovide for drainage of city streets and
reclamation of wet, low, or overflowed lands within their jurisdiction.  They
may construct sewers and drains and may levy special assessments on benefited
property owners to pay all or part of the costs of such works.  Additionally,

                                     257

-------
municipalities have the power of eminent domain to condemn property for these
purposes.  Thus, they have the means to deal directly with stormwater and
surface-water runoff problems.  Municipalities may enact floodplain zoning
ordinances.  Such ordinances may simply require compliance with special
building regulations or may exclude certain types of development in a
designated floodplain.

     Most counties and municipalities have a drainage plan ordinance that
requires submittal of a drainage plan for proposed developments.  In
addition, they commonly require that a drainage impact assessment be prepared
and submitted if there is to be a change in the development site.  Many local
governments have ordinances restricting the amount of surface-water runoff
that may be carried by a particular drainage system, or the amount of
sediment transported by the runoff.  Many local ordinances also incorporate a
flood plain regulation element or minimum elevations.

     Subdivision regulations relating to surface-water runoff control tend to
be more detailed than local government ordinances, and often require
submittal of a comprehensive drainage plan, approval of which is often a
prerequisite for plat approval.  Some regulations include runoff and rainfall
criteria to which the proposed drainage system must conform, while others
indicate permitted or preferred surface-water runoff control structures and
techniques.  Other provisions found in subdivision regulations include:  a
requirement that runoff from paved areas meet certain water quality
standards; the encouragement or requirement of onsite retention of runoff;
the regulation of grading and erosion control methods; and a monitoring
requirement for the discharge of surface-water runoff into lakes, streams,
and canals.

     The Department of Community Affairs has recently been actively charged
with the responsiblity of coordinating growth management in the State, which
will reflect on drai/iage facilities and projected areas growth.

     To provide adequate drainage and stormwater management, the regional
Water Management Districts work tn conjunction with the Departments of
Community Affairs and Environmental Regulation to obtain the necessary master
planning and overall coordination.  The transportation facilities overseen by
the Department of Transportation must be recognized and included in the
planning and implementation strategy of these conjoining authorities.

                       1986 FOOT DRAINAGE MANUAL

     The Florida Transportation Code establishes the responsibilities of the
State, counties, and municipalities for the planning and development of the
transportation systems with the objective of assuring development of an
integrated, balanced statewide system.  The Code's purpose is to protect the
safety and general welfare of the people of the State and tq preserve and
improve all transportation facilities in Florida.  The Code sets forth the
powers and duties of the Department of Transportation to develop and adopt
uniform minimum standards and criteria for the design, construction,
maintenance, and operation of public roads.
                                    258

-------
     The policy set forth In the manual  provides guidelines for the planning
and design of drainage and stormwater management control features for the
Department.  The guidelines, while positive and effective, are intentionally
broad, since no policy statement can cover all contingencies.   Its primary
purpose is to establish uniform practices and to ensure that the best
information available is applied to project conditions.  It neither replaces
the need for professional engineering judgement nor precludes  the use of
information not presented in the manual.

     Whether the Department must comply with local rules and programs for
stormwater management is a question that generates great doubt and confusion.
As a general rule, the Department should cooperate and comply  with local
regulations where such compliance would not be detrimental to  the
Department's interests or its ability to carry out its responsibilities.

     Stormwater management must be a marriage of both quantity control and
quality control.  This is the direction taken by the proposed  Drainage
Manual.  The manual is presented in three volumes: Policy, Procedures, and
Theory.

     Volume One sets forth the prevailing policies as they pertain to
collecting, receiving and passing stormwater runoff from and through the
rights of way.  The policies reflect the Department's responsibility to act
on the public's behalf in matters of safety, economics, effectiveness, and
environmental concerns.  A listing of chapters is contained in Table 3(a).

     Volume Two provides the most common or usually appropriate procedures
required to design the accordance with the policy.  Volume Two includes
frequently used charts, equations, computational forms and figures extracted
from numerous technical publications.  Textual material from these
publications is not duplicated in its entirety.  Discussions on applying the
procedures focus on step-by-step calculations which assume that the reader
has a working knowledge of theoretical fundamentals and terminology.  Desktop
technical procedures are presented in the manual.  Computer programs of the
Department's mainframe computer are summarized.  Tables and figures are
located at the end of each chapter.  A listing of chapters for Volume Two is
presented in Table 3(b).
  TABLE 3(a)     1986 FOOT DRAINAGE MANUAL VOLUME 1 - POLICY
     Chapter              Chapter Heading                           Pages
1 Introduction
2 Organization
3 Drainage Law
4 Policy Guidelines
Appendix

2
8
18
21
9
TOTAL "58
                                      259

-------
TABLE 3(b)
Chapter
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

1986 FOOT DRAINAGE MANUAL VOLUME 2 - PROCEDURES
Chapter Heading
Introduction
Planning and Facility Location
Flood Plain Encroachment and Risk Evaluation
Data Collection
Stormwater Hydrology
Roadway Grades
Open Channel Hydraulics
Culvert Hydraulics
Bridge and Bridge Culvert Hydraulics
Storm Sewer
Culvert Material Selection
Gutter, Inlet and Pavement Hydraulics
Retention/Detention Hydraulics
Stormwater Pumping
Subsurface Drainage
Shore and Bank Protection
Erosion and Sediment
Computer Programs
Standard Indexes and Special Details
Permitting
TOTAL

Pages
3
9
25
39
68
5
40
76
25
40
23
32
20
12
32
21
30
14
30
39
573
     Volume Three provides background information on drainage fundamentals,
supplementing the procedures presented in Volume Two.  The focus of the
contents is on theoretical background.  Not all chapters contained in Volume
Two are included.  A listing of the chapters is presented in Table 3(c).

     The manual has increased in size due mainly to an attempt to provide
most of the more widely utilized engineering materials.  Stormwater quality
management requirements and recommended procedures are given chapters rather
than lines.  There is also a difference in quantity requirements that must be
emphasized.  In recoginzation of recent adverse court decisions the
Department has altered its design concepts significantly.  Increases in peak
Stormwater discharge rates by upland owners will not be accepted or designed
for unless it is part of or provided for in a comprehensive local plan which
is substantially complete.

     Volume One, Chapter 4 includes permitting both by and with the
Department.  A discussion on the consideration of future development in the
design is presented.

     Volume Two, Chapter 2 deals with floodplain encroachments, establishing
the level of impact, detailing agency coordination, planning and permitting
at the location planning phase.  Chapter 3 identifies various highway
                                    260

-------
 TABLE 3(c)   1986 FOOT DRAINAGE MANUAL VOLUME 3 - THEORY	


     Chapter             Chapter Title                             Pages
1
2
3
4
5
6
7
8
9
10
11



Introduction
Surface Hydrology
Subsurface Drainage
Open Channel Hydraulics
Culvert Hydraulics
Bridge and Bridge Culvert Hydraulics
Storm Sewer Hydraulics and Hydrology
Gutter, Inlet and Pavement Hydraulics
Retention/Detention Hydraulics
Erosion and Sediment Control
Shore and Bank Protection
References
Glossary
TOTAL
3
45
18
23
14
24
8
20
12
17
19
11
25
239
activities in floodplains givinr tht level  of studies required by Drainage
for each activity.   Chapter 5 discusses various hydrologic procedures, often
required to satisfy permit requirements of other agencies as well as
Department design needs.   Chapters 6,7 and 8 are hydraulic procedures used to
establish highway grades, open channel flow and culvert requirements.
Chapter 9 provides information required for bridges and bridge culverts,
including coordination requirements with the local  communities when Federal
Emergency Management Agency (FEMA) regulated floodways are involved.  Chapter
13 provides design procedures for. retention and detention design both to
reduce outfall requirements and to conform with permit requirements.  Flood
routing procedures, filtration calculations for detention with filtration and
landlocked basin retention requirements are presented.  Chapter 15 provides
procedures for subsurface drainage, including exfiltration (french) drains
and drainage well designs.  Chapter 16 provides information for shore and
bank protection.  Chapter 17 contains erosion and sediment predictive methods
and control procedures both temporary and permanent.  Standard details and
indexes used by Drainage are explained in Chapter 19.  Permitting activities
with regulatory agencies as well as permits issued by the Department in
drainage or stormwater related matters are explained in Chapter 20.

     Most of the procedures in the manual are obtained from widely available
sources, particularity the U.S. Department of Transportation, Federal Highway
Adminstration (FHWA).

     The manual will be updated annually.  Updated material will be furnished
to purchasers of the manual at no charge provided that the manual owner's
address is current.  The manual is being given a final internal  review prior

                                     261

-------
 to printing.  It is anticipated to be available by Fall 1986.  Its purchase
 price has not been established.  If you would like to get on a "reserved"
 pre-publication list you may do so by either writing or telephoning and
 requesting to be put on the list.  Requests should be directed to the State
 Drainage Engineer, Florida Department of Transportation, Mail Station 32, 605
 Suwannee Street, Tallahassee, Florida  32301; Telephone (904) 487-1700.
 Persons who are on the list will be advised of the price and availability and
 would be guaranteed a copy from the initial printing.

      I hope the preceeding presentation has provided insight into the
 changing needs which the new Drainage Manual is addressing.  The need for
 safe, well-drained and available transportation facilities has not changed,
 but the technique required to meet a new awareness of the need for quanity
 and quality stormwater management has changed significantly.  The Department
 is dedicated to preserving and enhancing the unique environmental
 geographical area known as Florida while providing for its transportation
 needs.

    The Department should be considered as a "developer" of transportation
facilities much the same as a residential site developer or a commercial site
developer.  As such the Department should support and conform to regional and
local comprehensive stormwater plans rather than establish such plans.  The
primary need of the Department remains adequate drainage.  Required stormwater
management practices should reflect the procedures developed and required by
those agencies primarily responsible for the entire basin.

                 The work described is this paper was not funded
                 by the U.S. Environmental Protection Agency and
                 therefore the contents do not necessarily reflect
                 the views of the agency and no official endorsement
                 should be inferred.
                                     262

-------
                       APPLICATION OF THE OTTSBMM MODEL

                    FOR A  RELIEF  SEWER STUDY IN LAVAL,  QUEBEC

                                      by

                               R. Roussel   ( 1)
                               J.C.  Pigeon  (2)
                               J.R.  Noiseux (3)

                                   ABSTRACT

      The  OTTWSMM (Ottawa University Storm Water Management Model) model  is
 an  expanded version  of  EPA  - SWMM  which accounts for overland  flow,  inlet
 restriction and  surface  ponding.   The  model was recently applied  in  the re-
 lief sewer  study on an area with an  old  combined sewer  in Laval,  a municipa-
 lity in the Montreal  Urban Conmunity.

      The  first part of the paper describes briefly the area, flooding pro-
 blems and constraints related to the interceptor and high levels  in  the re-
 ceiving water body.

      The second  part  of  the  paper  describes the solution developed by  means
 of the inlet control  analysis and gives  an economic comparison with more ad-
 ditional alternatives.

      The last part of the paper presents some considerations on  the use  of
 the model and its  interface with a simple screening technique.


                                 INTRODUCTION

      Severe flooding  problems  in the  city of Laval  prompted  the municipal
 authorities to request a full  scale investigation.   This  was undertaken  by
 Les Consultants  Dessau Inc.  and the proposed design is presented in this pa-
 cer.
(1)   Drainage Engineer, Les Consultants Dessau Inc.,
     1200 ouest,  boul.  St-Martin, Laval (Quebec) H7S 2E4

(2)   Director, Drainage Dept.  City of Laval,
     3, Place Laval,  suite 300, Laval (Quebec) H7N 1A2

(3)   Manager, Environmental Division, Les Consultants Dessau Inc.,
     1200 ouest,  boul.  St-Martin, Laval (Quebec) H7S 2E4

                                     263

-------
     The city of Laval  is the  second  largest  city in  the province  of  Quebec/
Canada  and has a population  of 300,000.  It  is  situated  on an island  sur-
rounded by  Des  Mille-Iles  and Des Prairies rivers.   The undertaken investi-
gation  deal with  a  particular watershed currently served by a  combined  sys-
tem.

                          DEFINITION  OF  THE PROBLEM

FLOODING PROBLEMS

     The Montmorency-Brien  watershed, located in Laval, Quebec, has  a drai-
nage area  of 102  hectares.   A  single  family  housing development has led  to
an  imperviousness of  35%.   This  urbanized  area  is  currently  drained by  a
combined sewer  system that appears to be insufficient to  handle  a  10  year
flow.

     In order to  determine  the extent of flooding, a study was performed  in
which citizen complaints were  evaluated.   The following trends  were evident:

        Flooding  occurs all over  the  watershed  as a result of insufficient
pipe capacity  (collector and  laterals)

        Spring  and  Summer  are the critical flooding periods with  lower  ele-
vation  areas being the  most  affected.   The  frequency of flooding has  been
known to  range from none to  the three times  per year.   The average return
period  of  flooding  for  the  whole watershed is around  two years.

        Most complaints were  for  basement flooding  as a result of combined
sewer surcharge backing up  into the houses.

     It should  be pointed  out that the study  was based on  actual  complaints
registered  at  the City Hall.   In reality, the number of flooding incidents
is somewhat higher.

PRELIMINARY INVESTIGATIONS

     A  computer simulation of  the existing pipe network indicated that 90%
of the  existing collector pipes  as well as  60% of the lateral pipes do not
meet  the   up-dated  design  criteria   required  by  the City  of  Laval.   The
collector  average capacity  corresponds to a two  year  storm  while the  lateral
pipes indicated a very diverse capacity to  handle  storm  generated  flow  (3
times in a  year to  one  in 25 years).   These are  definitely  not  within the  10
year return period  design restraints  as  specified by  the city.

     In an attempt to  better  comprehend the  flooding problems,  Spring and
summer  conditions  were individually  simulated.   Spring conditions are  cha-
racterized  by  a lower rainfall  (3 times less than summer conditions) and  a
high water  level of  the receiving  body as  a result  of  snowmelt.   It  is
indicative  that the outfall pipe is  found to  be  completely submerged.  As  a
result of the higher water  level  in the  receiving water, flooding  conditions
are more severe in  the depressed  areas,  which have a lower potential ener-
gy.   This  was  reflected  in  the citizen  complaints.   During  the  summer

                                    264

-------
months, the water  level  is at the half-point  mark  in the outfall pipe even
though the combined sewer is subject to intense rainfall.

     As a  result  of the preliminary  analysis, it was felt  that  due to the
existing circumstances, all pipes should be individually analysed.  For the-
se purposes,  the  OTTSWMM  model  (Ottawa storm water management  model)  was
used  in  connection with  the  surcharge sub-model  EXTRAN  (Extended  trans-
port) .  In this way, the  influence  of the  variations of the river level and
the very low slope of the existing combined collector were accounted for.

                              OPTIMUM SOLUTION

DESCRIPTION

     The traditional approach to rectify the  flooding problems of the Mont-
morency-Brien watershed would have called  for the  twinning of 60%  of the
lateral pipes.  Due to space limitation twinning could not be applied to the
collector pipes hence almost 90% of them would have to be replaced by higher
capacity pipes.  The  implementation of this programm would have  led to ex-
tensive excavations in the majority of the  existing roads.  Furthermore, due
to topographic  peculiarities, a combined  sewer  of  3,000 mm  diameter  at a
depth of 9 meters would have to be installed at a location.   Apart from the
obvious economic and severe  inconvenience  repercussions  this approach would
involve, it would provide only a limited protection  against flooding.  (1 in
10 years)

     A solution was  developed in order to  reduce the twinning of the exis-
ting sewer system  and  also provides  for  a higher protection  against floo-
ding.  This system utilizes  inlet  control  techniques to reduce the flow en-
tering in the existing system (as indicated in figure 1).  Calibrated orifi-
ces are installed in the catchbasins in order  to restrict the water flow en-
tering in the combined sewer.  This keeps the hydraulic grade line below ba-
sement  levels and  consequently,  avoids the  possibility of  basement floo-
ding.  This solution uses the streets  to carry the water flow exceeding the
pipe capacity.  Hence  the streets are  actually acting as open channels, in
rare storm events, transporting the excess  flow which is subsequently captu-
red by  new storm sewers.   Figures 2  to 5 outline  typical  cases of actual
conditions and how these will be altered by the implementation of the propo-
sed solution.

The present lateral pipe  and collector could not accomodate a  10 year  flow.
In the proposed system (figure  2b), use of  inlet control devices allow  for
surface runoff  to be  collected  at  a lower point by a proposed storm  sewer
collector.  This method avoids twinning of  the lateral pipe.

In  figure 3a,  a  low point  is  located upstream  of  the existing  collector.
Street ponding  was  not a  suitable solution due to excessive volume of  flow
and  traffic  restrictions.   In  the proposed  system, inlet  restrictions  are
installed  in  the latter part of  the street.  A storm collector is  installed
to accomodate the increased  flow to the low point.   This  eliminates  twinning
of the latter part of  the  lateral pipe.
                                    26b

-------
                                  CATCH8A5IN ORATING

                                 COU-tCTlHO C4TCHB01N
                                 INLET CONTROL
            FIGURE'!
A  similar problem  as the one  shown in figure  3.   The solution  (figure  4b)
involved the installation of an additional  lateral  pipe and collector nearer
to the  ground surface.  Thus by this technique  overloading of the downstream
system  was avoided.

In a particular  location,  it was deemed necessary to install  restrictions  in
the catchbasins  to direct the  flow in a  shallow storage.   Storm water was
stored  and gradually  released  into the  system at  a rate which allowed for
its accomodation by the existing  network.  While this approach allowed for
keeping  the  lateral pipe and collector intact/  it  was difficult to  install
the storage due  to  espace  limitations.
                                     266

-------
                   1
LOW POINT
                                  FIGURE 2o
                             EXISTING SYSTEM
                                                           EXISTING
                                                           COMBINED
                                                           COLLECTOR
              |«mrol
                                  FIGURE '2b
                           PROPOSED   SYSTEM
                                                      LOW POINT
                                                           EXISTING  I
                                                           COMBINED j-
                                                           COUECTOR
DATA INPUT AND  ANALYSIS

     In order to implement  the proposed  solution, the following input para-
meters are required:

        Tributary area

        Elevation of high and  low points

     -  Street  longitudinal inclination
                                    267

-------
     -  Curb or sidewalk elevation

        Catchbasin grating specification (figure 6)

        Actual pipe capacity

     It is pointed out that flow generated by impervious  areas  directly con-
nected to the system will  not  flow  on  street.   This  was  incorporated  in all
subsequent analysis.
                        HIOH  POINT
LOW POINT
•::=^:
— —
— 1
- —
I
L-
\

<
!
•

I
LOW POINT
\
rr^ 	
r
EXISTING
COMBINED
COLLECTOR
S*\
1 1-r 	 -,
PRQTFCT.^ |QYFJPe 	 -»-• 	 •
1 ^1^ — PROTECTION f invr... .

/
/
                                 FIGURE-3o
                            EXISTING SYSTEM
                        HIGH  POINT
      LOW  POINT
                                                      LOW  POINT
                                                           PROPOSED
                                                           STORM
                                                           COLLECTOR]

                                                           EXISTIN6
                                                           COMBINED
                                                           .COLLECTOR
                                 FIGURE = 3b
                          PROPOSED   SYSTEM
                                     268

-------
                 LOW POINT
                                               HIGH POINT
                                                         EXISTING
                                                         COMBWEO
                                                         COLLECTOR
                               FIGURE 4o
                           EXISTING SYSTEM
                                                         PROPOSED
                                                         STORM
                                                         COLLECTOR
                                FIGURE '• 4b
                         PROPOSED   SYSTEM
     A detail  analysis  was  undertaken to  predict  the extent  of  necessary
measures to alleviate the flooding problem  in  the  Montmorency-Brien water-
shed.  It was  necessary  to  perform  this  analysis pipe by pipe, street seg-
ment by  street  segment  and  catchbasin by  catchbasin.   The  following were
evaluated:

        Flow in the existing combined  sewer

        Flow in the proposed storm sewer

     -  Flow on all the street segments and corresponding depths of flow in
        order to avoid flooding  in shallow  curb areas.

                                    269

-------
     The purpose  of the above-mentioned  detailed  hand-analysis  was the de-
termination of the  optimum inlet control location  and  the development of a
preliminary "sizing"  of the pipes.  We  used the OTTSWMM model (Ottawa Uni-
versity  storm water  management  model)   with  the  results of the detailed
hand-analysis as  input parameters.  The  model  was manipulated to  generated
and  "cut"  hydrographs which correspond to existing and proposed systems (as
showm  on  figure  7).   In  order to compute the surcharge imposed on the sys-
tems,  the model  EXTRAN  (Extended  transport)  was  used in  connection with
OTTSWMM.   This was  necessary so as to  accomodate the fluctuations  of the ri-
ver  flow as well  as low gradient  of  the existing combined  collector.
       1
                   LOW POINT
HIGH POINT
                                                             EXISTING
                                                             COMBINED
                                                             COLLECTOR
                                   FIGURE'So
                              EXISTING  SYSTEM
                    LOW POMT
                          PROPOSED
                          STORAGE
                                         ~l
                                                   HIGH POINT
                                                     UH8T
                                                             EXISTING
                                                             COMBINED
                                                             COLLECTOR
                               PROTECTION 5 IOYPAP
-------
°DXX)   OX)1
O.O2   OX»  O.O4   OJ05   O.06  O.O7  ODB   OO9   0.1O

                0 APPROACH (cmi)
                  FIGURE 6a
   A             ^^    ^—IAPPROACH FLOW

  T_           /   \  ^—ICARRY-OVER
                                                               OJ2
                                      CAPTURED
                                      BY ORATING
                                      WITH ORIFICE
                                      RESTRICTION
                             TIME
                         FIGURE'-6b
      HYDRAULIC PERFORMANCE  OF CATCHBASIN
                             271

-------
     It is important to  follow up this analysis by monitoring of the field
implementation of the recommended alterations  to the system.

GLOBAL IMPLEMENTATION

     The global  implementation of the  suggested solution required several
alterations to the existing combined system.  A shallow storm sewer collec-
tor of 1,650 mm diameter at the outlet was proposed.   For  space limitations,
some wide rectangular sections were used.   It  was thought  appropriate to se-
lect a shallow  depth so as to minimize excavation  costs  and disruption of
other utilities systems.

     It was necessary to completely block  88  catchbasins  and to install in-
let controls  in 40 of them.   In  the City  of  Laval, catchbasins are located
at almost twice the frequency compared to a neighbouring municipality.  Hen-
ce , it was possible  to block as  many without  serious repercussions.   Two
hundred and  fifty-four  (254)  catchbasins  of  the existing system  had  to be
cut-off and  reconnected to the  proposed  storm  system.  The reason for the
large  number  of reconnected catchbasins  is that the  proposed laterals and
collector pipes intersect  low points  where a large number of these catch-
basins were  situated.   A larger  than  usual special  inlet was installed at
low points  locations in order to  discourage  surface ponding.  The new gra-
ting was directly  connected to the new storm system.
                                FIGURE:/
                    HYDROGRAPHS SIMULATED

                                    272

-------
     As a result of  the  detailed  analysis,  four street intersections had to
be reprofiled and sidewalk lengths totaling about  60  m had to be built.  In
this way, runoff flow was prevented  from intruding into garages located be-
low the  street  levels.   Furthermore,  three  (3) new  combined lateral piped
had to be installed so as to accomodate the flow generated by the impervious
areas directly connected.

                               MODEL SELECTION

     It is evident that  there  is  a need for selecting the most appropriate
model for a  given set of  conditions.    The  opinion of the  authors is that
simple models should be  used in order  to better evaluate  the prevalent con-
ditions and  allow  for the  selection  of the most  appropriate sophisticated
model for the final  design.   This approach carries the benefit of reducing
the necessary man-hour and computer time.

     In this study, the IMPRAM model (Improved rational method) was employed
to  define  the  actual  problems.    ^Preliminary  results prompted the  use of
OTTSWMM and EXTRAN models for the final analysis.
                          DISCUSSION AND CONCLUSION

     A need to improve the current combined system serving the Laval munici-
pality prompted this  study.   A novel  technique-was  used by which the exis-
ting system was relieved by means  of  storm sewers and catchbasin inlet con-
trols.  This led to the development of a mostly separated system with consi-
derable advantages:

     a) Reduction of  the  length of proposed new pipes from 10 km projected
by traditional design techniques to 2,8 km.

     b) The installation of the shallow storm  sewer  will not interfere with
other utility  connections  and will be  less  prone to • river  fluctuation in-
fluences .

     c) As the proposed system is  not directly connected to the residences,
flooding by surcharge will not be a problem even if the hydraulic grade line
reaches the ground surface.

     d) Inlet  controls  keep  the hydraulic grade  line  (of the existing sys-
tem) below  basement  levels while  still use the existing collector  to its
full capacity.

     e) Implementation  of  the proposed  system will  restrict  the  volume as
well as the frequency of combined overflow into the river by about 65%.

     f) A traditional approach to  this  problem was estimated to cost around
7.5 million $.  This proposed solution is projected at 4 million $.
                                   273

-------
                              ACKNOWLEDGEMENTS

     The work presented  in this paper was performed by Les  Consultants Des-
sau Inc.   The  investigation was commissioned by  the City of Laval,  Quebec,
which the authors wish to  thank for permission  to publish  this paper.

Note:   The work described in this paper  was not  funded by the U.S. Environ-
        mental  Protection  Agency and therefore the contents do not necessa-
        rily  reflect the  views of the  Agency and  no official  endorsement
        should  be inferred.
                                     274

-------
                         APPLICATION OF  INLET  CONTROL
                       DEVICES AND DUAL DRAINAGE  MODELLING
                             FOR NEW SUBDIVISIONS
                                      by
         Paul Wisner, PHD, P. Eng.
         Dept. of Civil Engineering
         University of Ottawa
         Ottawa, Ontario
C. Kochar, MASc, P. Eng.
Canada Mortgage &
Housing Corporation
Ottawa, Ontario
         Hugh Fraser, MASc, P. Eng.*
         Novatech Engineering
         Ottawa, Ontario
C. Rampersad, P. Eng.
Andrew Brodie Associates
Thornhill, Ontario
                                   ABSTRACT

Concern over basement flooding has led to the application of a new technique
in stormwater management practice, known as inlet control.  Major storms have
often caused basement flooding when the sewer pipes have surcharged.  A
traditional, but costly remedy to this problem involves resizing the sewer
pipes to handle the higher flows.  A new method is to control the flow into
the sewer system and thus reduce the occurrence of basement flooding.  This is
achieved by applying the dual drainage concept and installing inlet control
devices in the catchbasins.

This paper summarizes the results of a study conducted for the Canada Mortgage
and Housing Corporation that evaluated the performance of three commercially
avaialable inlet control devices.  The hydraulic operation of the units was
tested with a physical model using debris laden water.  The study recommended
that a minimum size for the orifice type flow regulators be between 14 and 20
Ips.

Numerous applications of this concept have been made in the United States and
Canada in connection with relief sewer studies.  In addition, the concept is
being applied in new subdivisions where significant cost savings may be
obtained by designing the pipes to be compatible with the level of inlet
control.
*presently with CCL Consultants Inc.
                                      275

-------
INLET CONTROL DEVICES AGAINST SEWER SURCHARGES

     The initial objective of storm sewers, traffic convenience, has been
replaced by property protection against flooding.  This was assured by a
judgemental selection of design storm return periods that vary from 5 years
(Metro-Toronto) to 100 years (Dallas).  After a major storm that caused
damage by basement flooding, the traditional solution was to Increase the
design return period or modify the inlet time in the rational method.

     The choice of the municipal engineer used to be between an expensive
increase of storm sewer sizes or risk flooding during the life span of a
dwelling.  In the recent past the problem has become more critical with
basements being used for family or-recreation rooms in modern homes, damage
from flooding is more severe.

     One of the most promising solutions to this dilemma was found to be the
application of the dual drainage principle in conjunction with inlet control
devices (Wisner and Hawdur, 1984).  The principle of this solution, which
eliminates surcharge without pipe size increases, is shown in Figs. 1 and 2
for new developments and Fig. 3 for old developments with points.  This
solution is now mandatory in many Canadian municipality where it is designed
with the OTTSWMM model.  OTTSWMM (The Ottawa University SWM Model) is a
modified version of the EPA SWMM-EXTRAN model which accounts not only for
conduit but also for overland flow, ponding and inlet controls.  A schematic
of the models operation is shown in Fig. 4 (Kassera, 1983, Wisner, 1982).  The
Appendix gives a brief description of OTTSWMM which is available for mainframe
and microcomputers.  Development of OTTSWMM was made by the IMPSWM
(Implementation of Storm Water Management) program at the University of
Ottawa.  This program is based on a co-operation between IMPSWM researchers
and participants from engineering firms and municipalities.  The research
varies from  modelling and experimental studies to review of policies, public
attitudes and economics of storrawater management.

     Many projects in Southern Ontario have used the dual drainge-inlet
control principle in new subdivisions.  In these cases the maximum flow
accepted in the storm sewer and controlled by ICD's (inlet control devices)
varies from 28 Ips to 40 Ips per catchbasin (Table 1).  These limits were
selected to eliminate sewer surcharge for a 100 year storm with sewer sizes
designed for free surface flow at a 5-year design storm.  One of the main
reasons for this ICD limit was the lack of information on the performance of
various available types of ICD's and concerns regarding clogging.  Relief
sewer projects in the U.S.A. for older systems used lower limits such as 10
Ips per catchbasin or even less.  Some of these designs were done with a
proprietory model (Donahue 1982), others have been developed without a
detailed analysis of street flow depths (Pisano, 1982).

     A study  co-ordinated  by NOVATECH  ENGINEERING with  co-operation  from
ANDREW BRODIE ASSOCIATES and IMPSWM  in  1985,  examined  the  operation  of ICD's
and the economic and operational  implications  of  various  levels  of  ICD
control.  The study was sponsored  by the Canadian Housing  and Mortgage
Corporation.  This paper presents  briefly  some  of its  findings and
conclusions.

                                      276

-------
               ORIFICE
                PLATE
                           O
                           J
                                    TIME
                TO STORM
                 SEWER
                                     APPROACH FLOW
                                     CARRY-OVER
                                      CAPTURED BY
                                      GRATING
                                      WITH ORIFICE
                                      RESTRICTION

                                        APPROACH FLOW
  Severe Storm

   STREET  PROFILE
                     .CARRY OVER FLOW
                                                   »	^
   STREET   SECTION
   MAXO
K*^ 1
•15 mj	
                                                         SIDEWALK
                    INLET CONTROL DEVICE LIMITS FLOW
                                        STORM SEWER
Figure 1    Principle of Inlet Control
                                 277

-------
FIG.  2a
LOCATION OF LOW POINTS AND LANDSCAPING OF (DEPRESSED)
PARK AREA FOR RUNOFF  DETENTION.
                             FISHBONE COVER
                             Inlet Control  t.Scfs
                             Inlat Control  tdi
                             EDMONTON COVER
                       8    10    12
                      0 APPROACH (Cfs)
14
IB
18
RG. 2b  HYDRAULIC PERFORMANCE OF CATCHBASIN COVERS
                          278

-------
EXPERIMENTAL RESULTS

     A comparison of several commercial  types  of  ICD's was conducted in the
field in Skokie, Illinois (Donahue,  1984).   Field observations may confirm if
an operation is acceptable or not,  but will  not explain the hydraulic
performance and clogging mechanism.   Critical  rainfall events are rare, and
difficult to monitor.  Therefore,  some field observations used an artificial
catchbasin loading from fire hydrants (Pisano, 1985).  For this reason the
present study compared four ICD's  on a hydraulic  model at 1:1 scale in the
hydraulic laboratory at the University of Ottawa  (Fig. 5).  ICD performance
was observed in a plexiglass catchbasin  and  flow  measurements were made with a
triangular weir.

     The three ICD types commercially available and  tested at the University
of Ottawa are:

A.   SCEPTER - an orifice type of  ICD with a self-cleaning notch (Townsend,
     1984).

B.   CROMAC - an orifice type with a vcariable area  slot ICD.
Figure 3a  Inlet Control  to reduce the  Sewer  Reconstruction  Cost
  PARKING  LOT
                                           STREET
 Figure 3b    Parking  Lot  Inlet  Control
                                    279

-------
C.   HYUROVEX - a vortex ICD, representing an improved German version of
     HYDROBRAKE orifice.


     The forth type called the HANGING  TRAP is a self made ICD proposed in the
Skokie studies. The schematics of different devices are shown in Fig. 6.

     Head-Discharge curves were determined and found to be practically the
same as those indicated by the manufacturers.  The discharge coefficients are
relatively the same for the first two devices, but are much small for
HYDROVEX.
                                      RAINFALL
                RUNOFF
             TIME-
          MAJOR
          SYSTEM
        SUB-MODEL
                               SURFACE
                                RUNOFF
                             SUB. MODEL
                                 INLET
                             SUB.MODEL
A»*«3ACMnav
CMHT-CVCK
 awnmeo IT
 •xrtTH oNmci
 ~~iTir •"•
  *«
                               T1MC
   MINOR
  SYSTEM
SUB.MODEL
        MAJOR SYSTEM
          STORAGE
         SU8-MOOEL
                MINOR SYSTEM
                  STORAGE
                SUB-MODEL
                               OUTPUT
     FIG.  4  : AN OVERVIEW OF THE UNIVERSITY OF OTTAWA

             STORMWATER MANAGEMENT MODEL (OTTSWMM)
                                280

-------
                              TABLE  1  - EXAMPLES OF CASE STUDIES USING INLET CONTROL DEVICES
CASE
STUDY
A
B
ro
co
1-1 c
D
^
AREA
HA.
465
117
10
32
44
LAND
USE
RESID
RESID
RESID
RESID
RESID
AV.IMP.
00
31
30
25
40
35
TOTAL NO. INLET Z INLETS
OF INLETS DENSITY WITH
(CB/ha) RESTRICTORS
MDP
320
32
100
166
STUDY
2.73
3.2
3.1
3.7
25
75
50
51
LEVEL OF SURCHARGE
INLET ELIMINATED
CONTROL (LPS)
5 YEAR
42
42
34 9 28
16 0 42
28
CONTROL AVOIDED
Limited &
Acceptable
Negligible
Ipa Limited &
Ips Acceptable
Limited 6
Acceptable
MAX. STREET
PLOW DEPTH
(cm)
28
20
24
26
PARK STORAGE
USED
YES
YES
NO
NO
YES
EXTERNAL COMMENTS
AREAS
(ha)
- A detailed inlet
control analysis
was not conducted.
ICD's used mainly
to avoid surcharg-
ing the sewers
during rare storms.
15.4
118    RESID   35
                                            100
28
ELIMINATED    37
YES
90.8   - Severe Inlet con-
         trols were required
         because of the
         status of the pro-
         ject at time of
         analysis
 12    RESID   25      50
                                   42
                                             68
28       Limited 4
         Acceptable
              12
                                                                                               NO

-------
   MULTIPORT  OIFFUSER

   DIFFUSING TANK
                                     CATCH  BASIN TANK
                               L_T1_
           (UDNGITUDINAL STREET SLOPE  I -5 %)
   ENTRANCE  BAFFLE
   HEADER  TANK
                  GUTTER
                                                    |	
                  CURB
                        4-27m
                              PLAN
                                 J
GRATE
•8% GUTTER SLOPE
     >/0  STREET  SLOPE
                 PLEXIGLASS  CATCHBAStN

                         400mm  PLEXIGLASS LEADER PIPE
                                    DEBRIS  TRAP
                                    RECTANGULAR  NOTCH
                                    CATCH BASIN  TANK
                         SECTION  A-A
  Figure   s      The  Physical   Model
                           282

-------
  SIZE
  CONSTANT
                  A)  Schematic  of the Scepter
                        Inlet Control Device
                    Low flow
                    free discharge
                    (D.W.F.)
    Controlled flow
        B)  Schematic of the Hydrovex linit in Operation
                                    ORIFICE
C) Schematic  of  the Cromac ICD
D) Banging Trap ICD
Figures Inlet Control Devices
                               283

-------
     Clogging  experiments  were  in  general  conducted  in  a  conservative mode as
compared  to  natural  occurences.  The  limit found  for operation without
permanent  clogging was  20  Ips for  CROMAC and  14  Ips  for SCEPTER.  HYDROVEX was
tested  and operated  well down to 8.5  Ips,  although it may function adequately
for even  lower flows than  in the test.  For all  ICD's it  was  found that
certain combinations of leaves  and branches loading  may temporarily plug the
orifice.   Consequently, it is considered  that periodic  visual inspection and
cleaning  is  necessary.

     Hydraulic laboratory  findings were compared with field experience  in
Markham and  Scarborough, municipalities near  Toronto, where ICD's have  been
successfully used for several years.   A group of municipal engineers  attended
demonstration  tests  and presented  useful  comments.   As  a  result  of these
discussions, another factor introduced in the selection process  is the  effect
of  protruding  ICD's  on current  catchbasin cleaning practice.  Based on  these
considerations,  for  typical Canadian  separated sewer catchbasins the  HANGING
TRAP device  was  not  recommended.   The cost of HYDROVEX  is at  present  higher
than that  of the simpler orifice  type ICD's and it is therefore  useful  only if
a  low  level  of control is  required.  This, however,  has other implications,
which  can  be examined with OTTSWMM.
 SELECTION OF AN INLET CONTROL DEVICE

      As  indicated above,  the present practice in new Canadian subdivisions  is
 to  use a minimum control  level of 28 Ips.   This limit can be reduced  as  low as
 14  Ips and orifice ICD's  (SCEPTER) could still be used.   For a lower  level,
 the HYDROVEX is the only  recommended ICD.   Operations and cost implications
 were analyzed with the OTTSWMM model for five levels of  control in a  42  ha
 typical  subdivision,  shown in Fig. 7.  The OTTSWMM model sized conduits  to
 avoid surcharge for the 100-year storm (Keifer and Chu distribution in
 Metro-Toronto).

      Reducing the control level of ICD's gives pipe flow equivalent to a
 traditional  design for a  more frequent storm.  As an example a 2-year storm,
 corresponds  for Metro-Toronto conditions to a 20 Ips ICD.

      Results are summarized in Table 2.   It was found for example  that for  a
 20  Ips ICD,  park storages for overland flow would operate about twice per year
 instead  of once every five years as  for  the 42 Ips ICD currently used.   The
 depth in the parks for this more frequent  flooding would, however,  be
 relatively small.   The increase in maximum street flow depth at the gutters
 would be very small as compared to the present practice  (1-4 cm for a 100-year
 storm).

      Table 2 is an example of information  available to a decision  maker  from
 the OTTSWMM  model.   It is found that if  the level of control is lowered  to
 28  from  42 Ips  currently  used in Ontario for ICD's In new subdivision, the
 savings  may  be  significant.   For example,  in a typical residential  area, inlet
control  at 20  Ips  per  catchbasin can be  achieved with the less expensive
orifice  type  ICD's.   The  corresponding saving is $5,500/ha.  Of course
municipalities  now using  a  10-year or larger storms can  achieve greater  cost
reductions.

                                     284

-------
r\j
00
                                                                                                                 MINOR SYSTEM
                                                                                                                 MAJOR  SYSTEM
                                                                                                                 CATCH  BASINS
       Figure   7      Studied  System

-------
   Table 2
   COMPARISON OF  ECONOMIC AND OPERATIONAL  CHARACTERISTICS
Area of
Park
Approximative Storage
Level Savings in Maximum Gutter % of
of $l,000/ha Depth (cm) Total Area
Control 100-Year 5-Year 100-Year 100-year
(Ips) 7.3 ha .42 ha 7 ha 42 ha 7 ha 42 ha 42 ha
42 0 0 7.4 14 13.7 27 1.1
28 2.9 3.2 8.4 16 15.0 30 1.4
20 4.6 5.5 9.6 18 15.7 31 1.8
14.0 4.7 6.6 10.2 20 16.2 33 2.2
8.5 6.4 8.6 10.9 22 16.5 34 2.7
Average
Depth
in Park
for a
5-Year
Storm
(cm)
18
17
18
20
23
     The main limitation in achieving higher savings with inlet controls of
HYDROVEX type, is the concern related to higher  street flow depth during major
storms and increased frequency of park flooding.  These effects vary with
contributing areas, imperviousness,  slopes, -etc. and can be assessed using
detailed OTTSWMM computations for the specific conditions of each project.

     The sewer network or minor system can follow either a dendritic or looped
pattern.  When a new system is being designed, the pipes should follow a
dendritic pattern.  More sophisticated analysis  of looped or surcharged pipes
is possible with the EXTRAN (Extended Transport) submodel.  The pipe slopes
control the flow direction in the minor system.  Water enters the minor system
through storm inlets which are connected  to  manholes at sewer junctions.  At
                                    286

-------
each junction the flow from upstream pipes is added to the street inlet flow
giving the total flow to be routed to the next segment.  Pipes are sized for
the peak flow at each junction.


MODEL OPERATION

     The model is composed of four main submodels,  a surface runoff submodel,
inlet submodel, minor system submodel and the major system submodel.

     The input data consisting of subareas, street segments, pipe segments,
and storages are read and connectivity matrices set up.  The input data order
is shown in Figure C.I.  Computations are done in a number of steps.  First
the runoff for each subarea is computed using the Runoff Block routine
borrowed from EPA-SWM Model.  This is followed by the major system routing.
Starting at an upstream major system segment subarea runoff is routed down the
street segment with any upsteam carryover flow.  In conjunction with the major
system routing, inlet flows are determined.  Any flow captured by the inlets
are stored for minor system analysis.  Excess flow not caputred by the inlets
form carryover flow to be routed down the following major system segment.  The
street segment routing and inlet capture are continued until all of the street
segments have been considered.

     The computations for the minor system components are performed next.  The
user selected input determines how the pipe system is analyzed.  Design of a
new system starts with the most upstream sewer segment and proceeds
downstream. The pipe flow is the total of inlet capture flow and the upstream
pipe flow.  Design or resizing of individual pipes is performed with the MINOR
submodel.  When the pipes are sized, free surface and a dendritic sewer system
is assumed. Pipes are selected based on the input slopes, pipe roughness and
computed peak flow, using the Manning equation.  The model selects the
smallest commercial sewer size that will maintain free surface flow.

     If the pipe surcharge analysis option is selected the inlet flows are
saved in a separate file, to be used by EXTRAN subroutine.  Analysis with
EXTRAN can be conducted in looped pipe systems and in systems where pipes have
insufficient capacity for free surface flow.  Some small surcharging may be
desirable during major storm events, to prevent resizing of surcharged pipes.
Surcharge levels must be kept below foundations, to prevent basement flooding
damage.  The EXTRAN model must be used to determine if the surcharge levels
are acceptable.

ARE MODELS SAVING MONEY?

     The change of a well entrenched methodology requires not only studies but
also pilot projects and innovative municipal engineers.  The first ICD-dual
drainage project in Markhara, Ontario was proposed  in 1978.  At present the
system is used in other Metro-Toronto municipalities (Vaughan, Scarborough)
and is recommended by criteria in other smaller cities (Oakville, Barrie).   It
was also applied in various  projects in Nepean, Cumberland, etc.  Some initial
applications were  implemented after  flooding experiences.
                                     287

-------
      The proposed reduction of pipe size's, is a next step and may require a
 dialogue with decision makers and developers.  The use of OTTSWMM for new
 subdivisions showns that modelling can be used for a more economic drainage
 design.

      A previous IMPSWM study (Wisner, Cheung 1982) developed an improved
 rational method in which the runoff coefficient is a function of
 iraperviousness and the inlet time varies with rainfall intensity.  This
 method, available for microcomputers under the name IMPRAM gives the peak
 flows in close agreement with SWMM model (Fig. 8).

      It follows that for a conventional design the use of a more sophisticated
 model does not offer an advantage as compared to a slightly improved
 traditional computation.  In the case however, of the sophisticated dual
 drainage-inlet control-park storage design, modelling is justified by
 significant savings and evaluation of a more realistic level of protection.

      Experience shows that the OTTSWMM model is very simple to use, even if
 consultants are not familiar with the SWMM model.  Huitt-Zollars, a consulting
 firm in Dallas, used OTTSWMM without previous SWMM experience on a relatively
 complex relief sewer study for a large downtown area in Dallas (Urable 1986).
 The study found that the role of gratings even without ICD's is very
 important.  OTTSWMM users have support through the IMPSWM (Implementation of
 Storm Water Management) program at the University of Ottawa which operates a
 hot-line (613-564-39U or 613-564-7022).

      The main difficulties are of a different type:

      0  acceptance of the reality of overland flow by decision makers.

      0  integration of shallow storage parks for overland flows at strategic
         locations by planners.

      0  convincing consultants for developers to pay more attention to street
         grading and avoidance of low points.

      If a minimum $2,000/acre saving is not an adequate incentive to overcome
 these difficulties, one should also consider the considerable damages caused
 by basement flooding in areas where foundation drainage is connected with
 storm sewers (Wisner, Hawdur, 1983).

                               ACKNOWLEDGEMENTS
     Mr. D. Keliar, Director  of Engineering,  Town  of Markham, Mr. Jean Claude
Pigeon, Drainage-Director  in  Laval, Quebec  and  other engineers  in Canada made
a significant contribution  to the  implementation of OTTSWMM  based
applications of ICD's and dual drainage.  This  was seminal in generating
further applications and particularly  this  study.  The Ontario  Ministry of
Environment included the ICD-dual  drainage  concepts in its recent criteria and
provided assistance for OTTSWMM.
                                     288

-------
                         AREA Ma.
                         SUB AREA BOUNDARY
                         UINOM SYSTEM
 Figure 8  Comparison of Peak Flows done  with
SWMM and IMPRAM* for the example area

Cumulative area
(fla)
3; 27
4.52
5.90
8.94
10.52
12.42
13.96
15.72
13.37
(ti «
SWMM
(cms)
0.32
0.44
0.55
0.83
1.00
1.19
1.34
1.34
1.78
• 10 min.)
IMPRAM
(cms)
0.3
0.4
0.5
0.7
0.8
0.9
1.1
1.5
2.0
* IMPRAM (Improved Rational Method) by IMPSWM available
  for microcomputers.  •
                          289

-------
      Mr. John Meunier, distributor of HYDROVEX, Mr. Bob Crosmas,  developer of
 the CROMAC device attended the experiments and provided ICD devices.   The
 SCEPTER ICD was also made available by the Manufacturerers.  Drs.  Stuart
 Welsh from Donahue Associates and W. Pisano provided documentation on ICD
 applications on existing sewer systems.

      The OTTSWMM model was used in a microcomputer version of the IMPSWM
 mainframe program, developed for Andrew Brodie Associates by Mr.  Kassera.
 Review of the study and comments from Mr. Martin Hawdur, P. Eng., President,
 Novatech Engineering and advice and discussion with Dr. Ron Townsend,
 Professor at the University of Ottawa, on the methodology of experiments and
 data from Mr. Andrew Brodie on his numerous applications of ICD's were very
 useful.  All these contributions are gratefully acknowledged.

                                   REFERENCES
 1.    Donahue  and  Assoc.  Inc.,  "Flow Regulator  Pilot  Study Runoff Control
      Program  Howard Street  Sewer  District",  Village  of  Skokie, Illinois, March
      1984.

 2.    Kassem,  A.,  "Development  and Application  of  Simultaneous Routing Model
      for  Dual Drainage",  Ph.D.  thesis,  Department  of Civil Engineering,
      University of  Ottawa,  1982.

 3.    Ontario  Ministry  of  the Environment,  "Ontario Drainage Design
      Guidelines", Draft  1983.

 4.    Pisano W.C., "An  Overview  of Four  Inlet Control Studies for Mitigating
      Basement Street Flooding  in  Cleveland and Chicago  Areas", Presented to
      ASCE Luncheon  Meeting, Cleveland,  November 1982.

 5.    Townsend R.D.  Wisner P.,. and Moss  D., "Inlet  Control Devices for
      Stormsewer Catchbasins:  A Laboratory Study", Canadian Hydrology
      Symposium, Toronto,  May 1980.

 6.    Townsend R.D.,  "A Novel Inlet  Control Device  for Storm Sewer Systems",
      Proceedings, International Symposium  on Urban Hydrology, Hydraulics and
      Sediment Control, Lexington,  Kentucky,  July  1984.

 7.    U.S. Department of Transportion, "Design  of Urban  Highway Drainage - The
      State of  the Art", Federal Highway Admninistration, Office of Research
      and  Development,  (FHWS-TS-79-226), August 1979.

8.   Wisner P., "IMPSWM - Implementation of  Storrawater  Management Procedures
      for  Urban Drainage Modelling -  2nd Edition",  University of Ottawa,
     February  1983.

9.   Wisner P. and Hawdur M., "Evaluation  of Urban Drainage Methods for
     Basement Flooding Proofing".  A report  prepared by Novatech for Canada
     Mortgage  and Housing Corporation,  October 1983.


                                      290

-------
10.  Wright-McLaughlin Engineers,  "Urban Storm Drainage Criteria Manual",
     Denver Regional Council of Governments,  Denver,  Colorado, 1968.

11.  Wisner P. Eraser H,  Hawdur M  and Rarapersad C "Evaluation of Inlet Control
     in Dual Drainage Systems" A report by Novatech for Canada Mortgage and
     Housing Corporation, 1985.
                                   APPENDIX
DUAL DRAINGE COMPUTER MODEL - OTTSWMM
     Most urban storm drainge models assume that all the catchment runoff is
transferred directly into the minor system.  They have been developed for
designing or analyzing systems with low return period rainfall events.
However, if the dual drainage conept is to be employed, drainage systems must
be designed for events with a high return period.  In this case, all the storm
water runoff is not captured by catchbasins and transferred into the pipe
system.  A portion of the street flow is captured and transferred to the pipe
system while the remaining carry over flow is transported by the streets.

     The OTTSWMM model was designed specifically for analyzing dual drainage
systems.  The program has the capability of determining the surface flow, the
hydraulic capture by catchbasin inlets and the pipe flow.  It can be used in
four modes:

i)   to determine pipe sizes for free surface flow;
ii)  to analyze an existing system or proposed design and resize pipes  to
     maintain free surface flow;
iii) to determine the level of inlet restriction to maintain free surface flow
     in pipes;
iv)  to conduct a pipe surcharge analysis.

     Whatever mode of operation is being used, the basic assumptions of the
model remain the same.  The model conducts an analysis of two interconnected
systems, the surface or major system and the pipe or minor system.  Since the
computations are done on  two levels, the surface and sewer network flows do
not necessarily have to be in the same direction.

     The major system is  formed by the street network and must follow  a
dendritic pattern converging to a downstream outlet.  The major  system should
be continuous, no water ponding is allowed except at storage locations. In
new subdivision, streets  can be designed so that low points are  avoided.
Existing development often have low points.  In recognizing this  the model
permits two types of inlets.  Normal inlets are those where flow partly enters
the minor system and is partly passed down the major system.  Storage  inlets
are located at low points, all of the water enters the minor system  at these
inlets.
                                      291

-------
     The sewer network  or  minor  system  can  follow either a dendritic or looped
pattern.  When a new system  is being  designed,  the pipes should follow a
dendritic pattern.  More sophisticated  analysis of looped or surcharged pipes
is possible with the EXTRAN  (Extended Transport) submodel.  The pipe slopes
control the flow direction in the minor system.  Water enters the minor system
through storm inlets which are connected  to manholes at sewer junctions.  At
each junction the flow  from  upstream  pipes  is added to the street inlet flow
giving the total flow to be  routed  to the next  segment.  Pipes are sized for
the peak flow at each junction.
MODEL OPERATION

     The model is  composed  of  four  main  submodels,  a surface runoff submodel,
inlet submodel, minor  system submodel  and  the major system submodel.

     The input data  consisting  of subareas,  street segments, pipe segments,
and storages are read  and connectivity matrices  set up.  The input data order
is shown in Figure C.i.  Computations  are  done in a number of steps.  First
the runoff for each  subarea is  computed  using the Runoff Block routine
borrowed from EPA-SWM  Model.  This  is  followed by the major system routing.
Starting at an upstream major system segment subarea runoff is routed down the
street segment with  any upsteam carryover  flow.  In conjunction with the major
system routing, inlet  flows are determined.  Any flow captured by the inlets
are stored for minor system analysis.  Excess flow not caputred by the inlets
form carryover flow  to be routed down  the  following major system segment.  The
street segment routing and  inlet capture are continued until all of the street
segments have been considered.

     The computations  for the minor system components are performed next.  The
user selected input  determines  how  the pipe  system is analyzed.  Design of a
new system starts with the  most upstream sewer segment and proceeds
downstream. The pipe flow is the total of  inlet  capture flow and the upstream
pipe flow.  Design or  resizing  of individual pipes is performed with the MINOR
submodel.  When the  pipes are sized, free  surface and a dendritic sewer system
is assumed. Pipes are  selected  based on  the  input slopes, pipe roughness and
computed peak flow,  using the Manning  equation.  The model selects the
smallest commercial  sewer size  that will maintain free surface flow.

     If the pipe surcharge  analysis option is selected the inlet flows are
saved in a separate  file, to be used by  EXTRAN subroutine.  Analysis with
EXTRAN can be conducted in  looped pipe systems and in systems where pipes have
insufficient capacity for free  surface flow.  Some small surcharging may be
desirable during major storm events, to  prevent  resizing of surcharged pipes.
Surcharge levels must be kept below foundations, to prevent basement flooding
damage.  The EXTRAN model must  be used to  determine if the surcharge levels
are acceptable.

     Storage can be  provided to control  either minor or major system runoff.
The minor system storage may be provided by  a superpipe, or other in system
detention facility.  For the major  system, street flow is typically directed
                                      292

-------
                               Subcatchments/Surface Runoff
                                                                    (9)
                          |   Storage  (Major System)
                                                                 (8)
                           Major  System Network & Inlets
                                                              (7)
                         Inlet Capture
                                                           (6)
                      Major System Characteristics
                                                         (5)
                    Storage  (Minor  System)
                                                      (4)
                  Minor System Network
                                                    (3)
           f
Rainfall
                                  (2)
             Run Control Parameters
                                               (1)
      Fig.  Cl   Main Card Groups  For Input Data to QTTSWMM
to a surface storage facility such as a depressed park or detention pond.
Minor system storage is provided to obtain economies in pipe sizing or to
ensure that peak flows do not exceed predevelopraents levels.  Major system
r^r?nr1%rCeSSary " ^^^ Street fiow.  «•* release it at a controlled
rate into the minor system or receiving stream.  The designer inputs the
stage-discharge curve to provide the desired control, the computer output
gives the required storage volumes.  !f a storage volume was input and it is
exceeded the overflow volume is given in the output.               a a in is

     The basic computer output of OTTSWMM provides the following information:

1.   a print-out of the input data as well as a summary statistics of the
     watershed:   number of subareas,  sewers,  storage units,  total drainaee
     area,  etc   density of inlets (number of inlets per unit area^avefLe
     distance between inlets, etc.;                           <"«a/, average

2.   required sizes of  sewers for free surface  flow conditions;

                                     293

-------
3.   inlet control requirements, that is locations of inlets which may need
     flow constricting devices, and limiting capacities if the latter is not
     specified;

4.   detailed simulation results for specified elements, in printed and
     plotted forms:

     a)   time history of surface runoff;
     b)   time histroy of major system flows and depths;
     c)   time history of sewer flows;

5.   a summary of simulation results including maximum flows and depths at
     various locations as requested.

6.   Storages:

     a)   required storage volume for major system flow;
     b)   required storage volume for minor system flow, or volume of overflow
          if a given storage volume is exceeded.

7.   with EXTRAN for sewer flow routing  the following information is obtained:

     a)   printout summary for water depth at junctions;
     b)   printout summary for conduits  showing design flows.
The work  described in this paper was not funded by the U.S.  Environmental
Protection  Agency  and therefore does not necessarily reflect the views of
the Agency  and  no  official endorsement whould be inferred.
                                     294

-------
          USE OF CONTINUOUS .SWMM FOR SELECTION OF HISTORIC RAINFALL
                         DESIGN EVENTS IN TALLAHASSEE

        by: Wayne C. Huber, Brett A. Cunningham and Kevin A. Cavender
               Department of Environmental Engineering Sciences
                            University of Florida
                         Gainesville, Florida  32611
                                   ABSTRACT

     Conventional design methods often utilize a "design storm" synthesized
from a total depth taken from an intensity-duration-frequency curve for an
arbitrary duration,, coupled with an assumed temporal distribution using any of
several available shapes for the hyetograph.   The  synthetic  design storm is
then used as input to a rainfall-runoff model.  Unfortunately,  the true return
period of runoff parameters such as peak flow or volume is unknown since
antecedent conditions must be arbitrarily assumed when using such a method.
In addition, there is seldom a firm basis  for the choice of the storm duration
or temporal distribution.   Nonetheless,  the use  of this method  is very common.

     As an alternative, a  model may be calibrated and verified for the drain-
age basin and  then used in  a continuous simulation for as  many  years as there
are available rainfall data.  A frequency analysis of the predicted runoff
events may then be used to select historic rainfall events for  use in design
based on desired return periods on the parameter of interest, such as peak
flow, runoff volume,  flood  stage,  pollutant load, or  pollutant concentration.
The historic rainfall events so identified may then be simulated in more
detail for design  purposes.

     The results of such an analysis for the Megginnis Arm Catchment in Talla-
hassee, Florida are described in  the paper.  SWMM  simulations with historic
and synthetic  design storms are compared with respect  to a frequency analysis
of peak flows.  Contrary to what is often assumed, it is found  that the
synthetic design storms are not  necessarily conservative  (i.e.,  they do not
necessarily produce higher peak flows) over the  range of return periods consi-
dered (up to 25 years).
                                     295

-------
                                  INTRODUCTION

 ANALYSIS  OF URBAN STORMWATER

      Urban drainage studies  may be conducted for several purposes,  including
 flood-control and water quality control.  "Flood control" may simply mean
 conventional drainage of urban streets  or be more comprehensive in  nature,
 e.g.,  for a whole basin.  The usual design criterion is peak flow,  although
 stage (or  hydraulic grade line) is likely more important.  Runoff volumes are
 also  of interest  when storage is used as a  control measure.  Quality control
 may be applied to limit pollutant  loads to downstream  receiving waters and may
 or  may not necessitate a detailed  simulation of  a pollutograph  (concentration
 versus  time) during a storm.   In most cases, receiving waters are sensitive
 only  to the total load and not detailed variations within a  storm (Drlscoll,
 1979;  Hydroscience, 1979); hence,  in  roost cases,  only  total  loads need to be
 predicted.

      For  both quantity and quality control,  design  conditions must  be speci-
 fied.   In  some instances, the design  engineer has  little  choice.  For in-
 stance,  in Florida, the Department of Environmental Regulation  (DER) specifies
 in  part that  stormwater quality criteria can b'e met by providing  for the
 retention  of the  runoff resulting  from  the  first  one inch or rainfall or of
 the first  half-inch of runoff,  for projects  of less  than  100 ac.  This leaves
 few design options to the engineer, although it  is easy  to administer.

      In the  area of drainage design,  another common guideline promulgated by
 agencies  is  to specify the use  of  a synthetic design storm.  This is usually
 constructed  by obtaining the depth for  a given frequency  and duration (e.g.,
 25-yr,  24-hr)  from an intensity-duration-frequency (IDF)  curve.  The depth Is
 then distributed In time by using an assumed temporal distribution  of the
 storm,  e.g., the Soil Conservation Service (SCS)  Type II distribution for the
 hyetograph.  What is usually  unknown  and  ignored  in such  decisions  is the true
 return  period  of  the  design.  What level of  protection has really been af-
 forded by  this design?   If the desired  level of  protection is,  say, 25 years,
 based on  rainfall  statistics,  are  the statistics  (frequencies) of peak flows
 or volumes consistent  with this value?  It is easy  to comprehend  that the
 return  period  of  a  rainfall event  is  unlikely to  be the same as the return
 period of  the  flow  peak or volume  caused by  the event because of variable
 antecedent  conditions  and the inherent nonlinearity of the catchment.  Is the
 resulting  design  conservative (over designed) and therefore uneconomic,  or the
 reverse?   It is difficult to  say  a priori, although comparisons have been made
 (Marsalek,  1978; Arnell, 1978,1982).   However,  most conventional wisdom
 assumes that  synthetic design storms  are  conservative, a  result that will be
 shown not  always  to be true.   For quality,  Geiger (1984)  presents convincing
 combined  sewer overflow data showing  that return  periods  of  event mean concen-
 trations can  be quite  different from  the  rainfall and  runoff that caused  them.

     The choice of  rainfall  input  for models has been  widely studied,  e.g., by
 Adams and  Howard  (1985),  Arnell (1978,  1982), Harreraoes  (1983), Huber et  al.
(1981),  James  and Robinson (1982), Marsalek  (1978),  McPherson (1978), Patry
and McPherson  (1979), Wenzel  and Voorhees (1981).   Adams  and Howard (1985)
make a particularly strong argument against  the use of design storms, synthe-


                                      296

-------
tic or historic,  and recommend continuous  simulation or derived frequency
distributions.

     This paper will illustrate the use of continuous simulation for selection
of historic design rainfall events for use in detailed design simulations.   In
this method, a model is first calibrated and verified for the catchment.  A
continuous simulation is then performed using as long a record as possible of
hourly or shorter increment rainfall data (about 25 years of  computerized
hourly values are typically available from  the  U.S. National Weather  Service).
A frequency analysis is then performed on the parameters of interest,'such as
peak flow, runoff volume, or quality loads and concentrations,  from  which
historic rainfall events that produce a runoff event of the desired  return
period are identified.   Finally, the storms  may  be  input  again  to the  model in
a more detailed design simulation.  By this means,  questions  of antecedent
conditions, storm duration, and storm shape are avoided,  and  a better  idea of
the (unknown) "true"  return period for the design is obtained.

     Why not use measured flows for this purpose?   Although this indeed  might
resolve the question of the true design frequency,  it  is  seldom possible in
urban areas because of lack of a gage altogether, a short or  incomplete
record, or changing land use during the time of the gaging.   Simulation  offers
a less exact option to bypass these difficulties.

OBJECTIVES

     Ultimately, it is desirable to determine the relative effect of using
various methods for drainage design on the design condition.   This paper will
partially address this question for peak flows using one catchment as  a  case
study.   Eventually it may be  even more  important to  determine the economic
impact of alternative design methodologies, both on the cost of the  project
and on the cost of the engineering design.  In other words,  if  the use of more
sophisticated and complex methods only results in a marginally different
answer from that obtained by conventional  means then it may be constructed for
almost the same  cost either way, but be much cheaper to design using the
conventional (simpler) methods.   This latter question of  costs is currently
under study at the University of Florida and will not be addressed in  this
paper.

     The immediate objectives of this paper are thus to:

     1. Determine design storms from historic rainfall series using  continuous
simulation, and

     2. Compare peak flow results obtained by simulation with historic design
storras, synthetic design storms and other methods.

METHODS

     The options will be compared  through a case study using the Megginnis
Arm Catchment in Tallahassee, Florida.  The steps are enumerated below.

     1. Calibrate and verify the SWMM model on the catchment.


                                     297

-------
      2. Perform a continuous simulation using the 22 year historic record of
 hourly rainfalls (1958-79).

      3. Perform a frequency analysis on predicted peak flows and runoff
 volumes.

      4. Select historic rainfall events that give peaks or  volumes of desired
 return periods.

      5. Run these historic storms through  the SWMM model again using a more
 detailed simulation and compare with the results using synthetic design storms
 and other procedures.

      6. Show the final comparisons on a plot of peak flow versus return
 period for the different methods.


                                SWMM SIMULATIONS

 WHY SWMM?

      The EPA Storm Water Management  Model (Huber et  al., 1981; Roesner et al.,
 1981)  was chosen because of its many applications to urban  areas (Huber et
 al.,  1985)  and  because it  may conveniently be used for both continuous and
 single event simulation.  Alternative choices  include STORM (Roesner et al.,
 1974;  HEC,  1977) and HSPF (Johanson et  al.,  1980).  However, STORM is  not well
 suited for  single event  simulation  and  generally  has  only simple hydrologic
 and water quality routines (which do not detract from its usefulness as a
 planning tool).   HSPF  might  be  a  viable  alternative,  but  it  is  less  well
 suited to urban areas than is SWMM.

 THE CATCHMENT

      The  2230 acre  Megginnis Arm^Catchment  in Tallahassee,  Florida (Figure  1)
 discharges  to Lake  Jackson  north  of the  city and  has  experienced both  quantity
 and quality problems in recent  years due to increased urbanization.  A runoff
 data base exists since 1973,  with rainfall-runoff data  available since 1979
 (both  collected  by  the USGS).  It is the site of a multi-million dollar exper-
 imental water quality control facility  consisting of  a  detention basin and
 artificial  marsh, and  has  been  included  by the  USGS in  th&ir urban runoff
 studies (Franklin and  Losey,  1984).  A summary  of reports available  for the
 catchment is given  by Esry and  Bowman (1984).  Preliminary  applications  of  the
 Runoff  Block of  SWMM have been reported by Huber et al. (1986)  for purposes
 of  predicting 5-year  peak  flows.  The more comprehensive results of  the con-
 tinuous and single event simulations will  be illustrated here .and compared
 with alternative engineering analyses of the basin for  prediction of peak
 flows.  The various methodologies will be  compared over a rango  of return
 periods, although there  will  be considerable uncertainty  in the peak flow
 estimates at  the end  of  the  22  year rainfall record.  The 5-yeair  results are
more reliable since  they fall in  the middle of  the range  of refturn periods and
will be singled  out for  special consideration of  the historic storms involved.
                                     298

-------
       Land to be
       Purchased
   1 mi le
1 mile
                            SCALE

Figure 1.  Megginnis Arm Catchment in Tallahassee,  Florida.
           Rainfall and flow measurements are made  at the
           entrance to Pond I.
                              299

-------
 SWMM CALIBRATION

      Calibration of SWMM first required estimates for the Runoff Block parame-
 ters listed in Table 1.   The  average  slope  was found  by selecting  eight  points
 at the edge of the catchment, calculating the path length of  each  point  to  the
 inlet,  dividing the path lengths by the change in elevation,  and taking  a
 weighted average of the eight slopes.  Percent imperviousness was  obtained
 from an earlier USGS modeling study by Franklin and Losey (1984).  Green-Ampt
 infiltration parameters were estimated by  identifying the soils in the catch-
 ment (primarily sandy)  from a county  soil survey  map,  finding the  hydraulic
 conductivity and capillary suction for each soil  from data published  by  Car-
 lisle et al.  (1981), and  then taking  a  weighted average over  the soil types.
 Manning's n values  were selected from charts based on average type of ground
 cover.   Average monthly pan evaporation data  were obtained from  National
 Weather Service (NWS)  values  published by Farnsworth  and Thompson  (1982), from
 which actual evapotranspiration  (ET) estimates  were calculated by multiplying
 by a pan coefficient of 0.7.   Final parameter estimates are shown  in  Table  1.
 The overall catchment  was schematized using only  one subcatchment  and no
 channel  routing in  order to maintain  a reasonable computation time for the
 continuous simulation.   Later, a five subcatchment  schematization  was used,
 but the  there was only a minimal improvement  in the predictions.   Hence, de-
 tailed  simulation results are reported  only for the single subcatchment  sche-
 matization.

       TABLE 1.   RUNOFF  BLOCK  PARAMETERS FOR THE MEGGINNIS  ARM CATCHMENT
                        Parameter                 Value
                   Area                       2230  ac   (903  ha)
                   Width                      6000  ft   (1830 m)
                   Percent  Imperviousness       28.3
                   Slope                         0.0216
                   Manning's  Roughness
                     Impervious                 0.015
                     Pervious                   0.35
                   Depression Storage
                     Impervious                 0.02 in  (0.5 mm)
                     Pervious                   0.50 in  (13 mm)
                   Green-Ampt Parameters
                     Suction                   18.13 in  (461 mm)
                     Hydraulic Conductivity     5.76 in/hr  (146 mm/hr)
                     Initial Moisture Deficit   0.15
     Rainfall-runoff data for calibration and verification were obtained from
available USGS records for the catchment.   Ten of the largest storms were
selected from  the period  1979-81 and  randomly divided  into two sets of five
storms each, one for calibration and  the other  for verification.   The larger
storms were  chosen  since  the ultimate use of  the modeling was to be drainage
                                     300

-------
design, for which calibration for  large  storms  is  preferable.  The model  was
calibrated for the five storms simultaneously, that is,  while maintaining the
same parameter values  for each (Maalel and Huber, 1984).   Calibration of
runoff volumes was sufficient using the  assumed value  of  catchment imper-
viousness (Figure 2);  thus, calibration  for peak flows was achieved  only  by
varying the subcatchment width (a  parameter  in  SWMM that is equivalent to
making changes in the slope or roughness).    The  results  for  peak flows are
shown  in Figure  3.

     Verification was accomplished by running five different  storms  using the
same parameters  that were used in the final  calibration runs.  Results of the
verification runs were comparable to the calibration runs and are also shown
in Figures 2 and  3.   Individual storms exhibit  varying goodness  of fit as shown
in Figures 4-7.   Figure 4 illustrates the best  of  the  ten fits and Figure 5
the worst, while Figures 6 and 7  illustrate  typical intermediate results.  It
should be emphasized that antecedent condtions  were not altered for  individual
storms which would have aided in  the fits.  (Continuous simulation eliminates
this problem.)  However,  a robust  calibration was  desired, which would produce
a good fit, on the average,  for many storms  (Maalel and Huber,  1984).   Hence,
the agreement between measured and predicted volumes and peaks shown in Fi-
gures 2 and 3 was considered adequate for this  study.
                              FREQUENCY ANALYSIS

     Upon completion of the calibration and verification exercises,  a contin-
uous run of 21.6 years  (259 months, June 1958 to December  1979) was  made using
hourly rainfall data from  the  Tallahassee  Airport  NWS rain gage,  located
approximately 7 miles  southwest of the study area.  Statistical analysis of
the predicted flows was performed using the Statistics Block of SWMM.  The
time series of hourly runoff values was separated into 1485 independent storm
events by varying  the minimum interevent time,  MIT, until the  coefficient of
variation of interevent times equalled 1.0,  yielding  a value of MIT  = 19
hours.   This method of delineating independent  events (Hydroscience,  1979;
Restrepo-Posada and Eagelson,  1980) is based on the fact that  the exponential
distribution is often fit  to interevent times,  and it has a coefficient of
variation (standard deviation divided by  the mean) equal  to 1.0.

     The SWMM  Statistics Block performs a frequency analysis on any  or all of
the following parameters:  runoff volume, average flow, peak flow, event dura-
tion, and interevent duration.  (If pollutants  were being simulated, frequency
analyses could also be performed on:  total load, average load, peak  load, flow
weighted average concentration, peak concentration.)   Storm events are sorted
and ranked by magnitude for each parameter of interest, and assigned an empir-
ical return period in months according the the Weibull  formula (T =  (n+l)/m,
where n Is the total number of months  and m is the rank of an event).  The
largest magnitude event for this  259-month simulation was thus assigned a
return period of 260 months.  Hence,  for this simulation, a five year event is
bracketed by return periods of 52 and 65 months (fifth and fourth largest
events, respectively),  both of which  were selected as "design events."

     On this basis,  the historic  storms producing the nine highest peak flows


                                     301

-------
     1  5
O


O
ui
o
Ui
ct
a
                0.2
                                                        1.2
                        0.6      0.8       1


                       MEASURED VOL.. IN

              OCALIBRAT1ON        • VERIFICATION


Figure 2.  Goodness  of  fit of  runoff volumes.
1.4
                  0.2
                              -i	1	r

                    0.4       O.6       0.8
                          (Thousands)
                   MEASURED PEAK FLOW, CFS

                                   I VERIFICATION
                      OC A LI BRAT I ON
       Figure 3.   Goodness  of fit  of runoff peaks.
                                302

-------
SVU. UUUUU
4OO. OOOOO
-
3OO. OOOOO
FLOW
IN
CF8
200. OOOOO

100. OOOOO
O. 0
8.

*
»*
#»
**
**
**+
«•» **•+
f\ 3 *»•»•+
*+ *
*+ *
*+ *
*+ »
»-»• •
•t *£
	 1 	 i 	 1 	 i 	 1 	 1 	 1
I
I


+
+
+•
+
+
j.
*+ * +
•+
»* »+
»» • *» •+
» «* +
-------
430. OOOOO



360. OOOOO
_



270. OOOOO

	 FLOW
IN
CFS
1BO. OOOOO





OJ
O 90. OOOOO
-^






Of*
. w
6

— 	 1 	 1 	 1 	 _- — i 	 1 	 1_ 	 _ 	 1 	 1 	
4> 4.
* 4- +-f 4-4.
+ 4- 4-4- 4- + 4-
4- 4- + +
4- +44.
~10m3/s / I I \
' *• % + t
4« 4-4- 4-
•4-« > + 4-
•+• + + 4-
•4- • 4-4- 4>
•4- • 4- 4> '+
• 4- * +4- 4-
•+ • ++ +
4- • * * 4-
4- • ** « +
+ « *• • 4. +4-f
*+ 4- »* » » 4>
4-4- 4- » » » * +
* * 4- » » • •*
+ + *•» » • * 4-
* 4- 4- * • »* 4-
4>4-4- •• » «• 4-
4- + •* •*• **• +
+ 4- * 4- •» «•* 4-
+ + « + » * • » 4-4
4- 4> *+ * * » 4-4-
4- 4>4> • •« « 4-M-
4-** * •••• • 4-4-f
• 4- » * * •«* 4-M-
•4- «»*•* «» !•>>» +
• 4- « •* 1 > 1 1 14-
*4- «•»« 4-4-4-t-4-4-+4-*-»+
*+ »***
4-4- «»**»«-»*»*«»«-»** I
O 7.*O 8."o 9.*O 10*0 ll.*0 12*0 13*0 14*0 19*0 16.
                                      TIME OF DAY.  IN HOURS
        STORM 4 MARCH  1O  198O
PREDICTED-*.  MEASURED-*
LOCATION   98
HYDROS*APH STATISTICS FOR  LOCATION  «78
VOLUME
CUBIC FEET
PREDICTED.
TOTAL TIME
MEASURED,
TOTAL TIME
PREDICTED,
OVERLAPPING
TIME
MEASURED.
OVERLAPPING
TIME
DIFFERENCES,
ABSOLUTE
X OF MEAS
0.

0.

0.


0.



O.
2B463E+07

3B339E+07

2B430E+07


9B339E4-O7



29889E+07
INCHES
0.

o.

o.


o.



0.
51.
392

721

391


721



369
233
PEAK
TIME. HR
8. 983

8. 667

8. 9S3


8. 667



0. OB3
FLOW
DURATION
FLOW. CFS
321.

414.

321.


414.



92.
22.
699

OOO

699


000



341
309
START, HR
6. 667

6. 667

6. 667


6. 667


Figure

END, HR
19. 833

19.790

1 9. 73O


19. 73O


LENOTH, HR
9. 167

9.083

9. OB3


9.083


5. Predicted
March 10,
NO.
POINTS
111

110

110


110


and meas
1980 (ve

-------
^9U. WUUU




2OO OOOOO








150. OOOOO

-
FLOW
IN
CFS


100. OOOOO



- -. 	




10
O 30. OOOOO
(Jl







Oft
, Q
2.
	 1 	 1 	 i 	 1 	 ;l 	 1 	 • 	 1 	 1 	 1 	 1
4- 4-
4- 4-
4- »
4 4-
4 4-
4- 4-
+ *
4- 4
4- 4-
+ 4-
4- 4-
4- +
4 4.
»4 +
»4> 4-
. *4-« •«•
- 4 «4-. 4-
3 / *+* +
«n /s ::% :
• 4 * 4>
• 4- * 4- 4-
• 4- » 4- 4-4-
4- » 4- 4- + 4-
4- #«4- +-*• 4-4-
+ * 4- 4- + •+ 4-
4- * 4- » 4- »4 4
4- »4- 4- 4- 4-« 4
4- * 4- +« +• *4-«* +
+ * 4- H-» 4- »4- » 4-
4. ^4. 4.4^ ^ 4.4-94. • 4.
• -»• » + «•**++» +
»+ «4 4- 4- » 4-4- * 4>
4- »4- » 4- 4-4- 4-4. » * * 4-
4.4. «+ **4-44- 4-4-4-4- «« » *•+
4-4- *+ •»+«» 4- » 4-4-
4-4-4- » » » 4-
• 4- 4- 4- » »* 4-4-
• 4- 44- »» 4-4.4-
4-*»»»4-4- »»
*+ • * **
*4- »»*
4-4-
• T T — T — — T . _^__ t _ -i«T « _^«_ f -i w i-. T ^ T ^_^
0 40 6. 0 80 10.0 UO 140 Ife. O 18. O 20. O 22.
i
TIME OF DAY, IN HOURS PREDICTED**. MEASURED-*
STORM2 MARCH 9 198O LOCATION 98
HYDROCftAPH STATISTICS FOR LOCATION 98
VOLUME PEAK FLOW DURATION NO.
CUBIC FEET INCHES TIME. HR FLOW, CFS START. HR END. HR LENGTH, HR POINTS
PREDICTED.   0 18009E+O7    0 222
TOTAL TIME

MEASURED.    O. 23986E+07    0. 296
TOTAL TIME

PREDICTED.   0. 17979E+07    0. 222
OVERLAPPING
TIME

MEASURED.    0 239B6E+O7    0 296
OVERLAP?INQ
TIME

DIFFERENCES.
  ABSOLUTE   O 6OO63E+O6    O 074
  V. OF MFAS                25 041
S. 500  158.767


5. 730  234 000


3 500  138. 767



3 750  234 000
0 230  75 233
       32 151
 3. 500   13. 300   10. OOO


 3. 300   13. 417    9. 917


 3.300   13.417    9.917
 3  300   13 417
                                      9 917
121


120


120


120
Figure  6.   Predicted  and measured hydrograph,
             March  9, 1980  (calibration run).

-------
190. OOOOO
—
120. OOOOO







9O. OOOOO


FLOW

IN

CFS


60. OOOOO







CO
g 30. OOOOO



Of\

-4
































11.2
STORM 1 SEPT 27

1 ! i _ !_ ! j
1.1 /S
+ *** •
+ * * *
+ +» *
+ •*• •
»+• » + *
* + + *
*+ + »
•••• + *
+ + * »
+ ++ • » «»
*+ + * * •

+ + • » •
* + + • » »
«•*+ + « * •
• •+ 4- *** *
* + + •«•»
» •» + ++*++•»• »
» * •»«*+*
» + ++ + * + » •*
• + •«•+•* •••* «
• + + «•»•»«
* + +++++* »• • •
* + +++ »• * * *
+ +4- • * »
+ •»+•«•+'»»«.»+ •«••»•*•»
+ * -M-t-l-
* •* •»•+
+ *
++ #*#
••*• »*»
*+
11.9 12.6 13.3 14.0 14.7 19.4 16.1 16.8 17.9 IB.
TIME OF DAY. IN HOURS PREDICTED-*, MEASURED-*
1979 LOCATION 9S
HYDROORAPH STATISTICS FOR LOCATION 98
VOLUME PEAK FLOW DURATION NO.
CUBIC FEET INCHES TIME.HR FLOW, CFS START, HR END. HR LENGTH, HR POINTS
PREDICTED. 0. 136S2E+07
TOTAL TIME
MEASURED. 0. 1197&E+07
TOTAL TIME
PREDICTED. O. 13636E+O7
OVERLAPPING
TIME
MEASURED. O. 11976E+07
OVERLAPPING
TIME
DIFFERENCES.
ABSOLUTE -0. 166O7E+06
X OF MEAS
0.169 13.000 122.836 11.730 17.730 6. OOO 73
0.148 12.667 119.000 11.730 17.667 3.917 72

0.168 13. OOO 122. B3B 11.730 17.667 3.917 72


0.148 12.667 119. OOO 11.730 17.667 3.917 72


-13 868 -0-333 -g- |3B Figure 7. Predicted and measured hydrograp
September 27, 1979 (Verification run)

-------
and.the nine highest runoff  volumes (total  flows)  are  indicated  in  Tables  2
and 3.   The  hyetographs of 52 and 65 month storms  for  each case  are shown  in
Figures 8 and 9 and their characteristics are listed in Table 4.
                     TABLE 2.   STORMS RANKED BY PEAK FLOW

Rank
1
2
3
4
5
6
7
8
9
Return Period
months
260
130
86.7
65
52
43.3
37.1
26
23.6
Date
9/ 8/68
7/21/70
7/16/64
7/21/69
9/ 3/65
9/20/69
12/ 3/64
11 9/65
6/30/64
Duration
hrs
6
34
73
58
11
88
61
81
48
Depth
in
6.52
8.21
10.16
6.11
4.35
13.79
9.92
6.01
4.67
Peak Flow
cfs
1907
1605
1505
1131
1104
1070
1028
807
805
            TABLE 3.   STORMS RANKED BY RUNOFF VOLUME  (TOTAL  FLOW)

Rank Return Period
months
1
2
3
4
5
6
7
8
9
260
130
86.7
65
52
43.3
37.1
26
23.6
Date
9/20/69
3/28/73
12/ 3/64
7/16/64
7/26/75
7/21/70
8/ 2/66
3/31/62
10/ 6/59
Duration
hrs
84
108
61
73
97
34
171
24
87
Depth Runoff Volume
in in
13.79
10.91
9.92
10.16
9.30
8.21
7.72
7.78
7.35
3.69
2.86
2.74
2.69
2.39
2.24
2.09
1.96
1.95
     Tables 2 and 3 show that the return periods  of  the  individual  storms are
quite different when ranked by another parameter.  This  is illustrated further
for the 5-year storms in Table 5.   Storm V-65 is  rare  by both flow measures;
in fact, it is the third largest storm of record on the basis of peak flow
(and the second largest rainfall volume event).  However,  when storm V-52 is
ranked by peak flow, and when storm  P-52 is  ranked by volume, they are seen to
be not especially rare, with return periods  of 7.4 and 6.8  months,  respec-
tively.   Return periods for rainfall  volumes shown in Table 5 were obtained
from the SYNOP program  for  rainfall frequency analysis  (EPA, 1976;  Hydrosci-
                                     307

-------
                                 7/21/69
OE
I
I/)
a
i
t
to

UJ

Z
-^ -r
32 -
3 -
28-
2 6 -
24 -
22-
2 -
1 8 -
1 6 -
1 4- -
1 2 -
1 -
08 -
O 6 -
O 4 -
02 -
o







-50
mm/hr






F'
r '
1 1 1














EP
rk.n
21 31 41 51 61
                               TIME. HOURS
        Figure 8a.  Hyetograph for  storm  P-65,  July  21,  1969.

                    65-month return period  based  on  peak flow.



                                 9/3/65
.•> 1 -
3 2 -
3 -
2 6 -
2 6 -
2 4 -
2 2 -
2 -
1 8 -
1 6 -
1 4 -
1 2 -
1 -
O 8 -
06-
04 -
02-






—50 mm/hr






// _
//

-------
                                7/16/64
cc
x
in
r
3.4 -i
3.2 -
3 -
2.8 -
2.6 -
2.4 -
2.2 -
2 -
1 .8 -
1 .6 -
1 .4 -
1 .2 -
1 -
O.8 -
0,6 -
0,4 -
0.2 -

— '






-50
mm/hr








1 ru r
i i






•

1

'
t
'
Ua
U n Lfi
M rlLJIrt! n ₯_
1 21 31 41 51 61 71
1 11 21 31 41 51 61 71

TIME. HOURS

Figure 9a. Hyetograph for storm V-65, July 16, 1964.
65-month return period based on runoff volume





(t
r
\
z
£
t/>
2
5




3 4 -,
3.2 -
3 -
2 8 -
26-
2.4 -
2 2 -
2 -
1 6 -
1 6 -
1 4 -
1 2 -
1 -
08 -
06 -
04 -
02 -
0 -
7/26/75






-Ml in 111 /lu
p



I
f n ?H(fl (l^Ui|imjr, Ik jWlH





L
1 11 21 31 41 51 61












1 « |.
I-,'.?M JS^.TI 1.1 1,,. i M'Hll llfl, ,WH
71 61 91












                               TIME, HOURS


        Figure 9b.  Hyetograph for storm V-52,  July  26,  1975.

                    52-month return period based  on  runoff volume,
                              309

-------
 ence,  1979).  Thus, although the return periods were computed slightly differ-
 ently,  Table  5  further  illustrates  that return periods of the same event
 ranked  by different  parameters are rarely the same.
           TABLE 4.   CHARACTERISTCS OF  5-YEAR HISTORIC DESIGN STORMS
   No.*    Date
 Runoff   Runoff  Rainfall  Peak   Time Since    Rainfall
Duration  Volume   Volume   Flow   Last Event   Last  Event
   hr       in       in     cfs        hr         in
V-65
V-52
P-65
P-52
7/16/64
7/26/75
7/21/69
9/ 3/65
73
97
58
11
2.69
2.39
1.61
1.19
10.16
9.30
6.11
4.35
1670
685
1131
1104
74
31
32
67
0.27
0.07
1.35
0.11
 * V means  ranked by volume, P means ranked by peak, and numbers are return
 periods  in months.
             TABLE  5.   RETURN  PERIODS  (MONTHS) OF  "5-YEAR"  STORMS
                       BY  VOLUME,  PEAK  FLOW ANT) RAINFALL
         No.   Date    Return  Period
                          by  Volume
                      Return  Period
                      by Peak Flow
 Return Period
by Rainfall Vol.
V-65
V-52
P-65
P-52
7/16/64
7/26/75
7/21/69
9/ 3/65
65
52
14
6.8
87
7.4
65
52
156
78
12
7.8
     Seven of the historic storms  defined  above  were then  run again through
the model (storms for  return  periods  of  37.1 and 43.3 months were omitted)
using hourly rainfall  inputs  but a 5-min time  step,  instead of the hourly time
step used in the continuous simulation.  In lieu of continuous simulation for
these detailed analyses, 4 to 6 days of prior rainfall were run through the
model prior to the  beginning  of each  design event.   This "pseudo-continuous
simulation" thus accounts for antecedent conditions  without the need for
running the entire rainfall time series at a short time step.   Results will be
described following a  discussion of the  generation of  synthetic design storms.


                           SYNTHETIC DESIGN STORMS

     Following the  techniques of Arnell  (1982), five synthetic design storms
                                      310

-------
were constructed for comparision with the historic storms in Tallahassee:  Soil
Conservation Service (SCS, 1964), Chicago (Keifer and Chu, 1957), Illinois
State Water Survey  (Huff, 1967), Sifalda (1973),  developed in Czechoslovakia,
and Flood Studies Report, FSR, (Natural Environment Research Council, 1975),
developed in Great Britain.   The choice of a 24-hour duration for the storms
was made  for two reasons: 1)  it is commonly used in engineering practice in
the'Tallahassee area, and 2)  approximate calculations of the time of concen-
tration of the basin using the kinematic wave equation  (Eagleson, 1970)
yielded estimates ranging from 13 to 60 hours,  depending on the choice of
rainfall excess.  Thus,  24 hours is  at least in  the range of possible times of
concentration.   But  the  very idea of having to arbitrarily select a  storm
duration in the first place  illustrates one of  the major difficulties in using
synthetic design storms.

     Five-year storms are considered as an example.  From the Tallahassee
region IDF curves (Weldon, 1985),  the 5-year, 24-hour average intensity is
0.305  in/hr,  giving  a 5-year, 24-hour depth of  7.32 in.   The five synthetic
design storms were then scaled to produce this  total storm depth.  Various 24-
hour depths frora the IDF curves are compared with  historic  storms in Table 6.
These were compiled from the SYNOP program (EPA,  1976) with  a minimum intere-
vent time of 5 hours and thus differ In magnitude slightly from the rainfall
volumes identified from the frequency analysis of runoff conducted using SWMM.
It may be seen that although  the IDF and historic depths are comparable, the
actual durations of the historic storms are certainly not 24 hours.
            TABLE 6.   COMPARISON OF 24-HOUR IDF DEPTHS WITH SYNOP
               FREQUENCY ANALYSIS OF HISTORICAL RAINFALL EVENTS

Storm
Date

9/20/69
IDF
7/17/64
IDF
12/13/64
7/28/75
7/21/70
IDF
8/30/50
6/18/72
Return
Period
yr
26.0
25
13.0
10
8.7
6.5
5.2
5
4.3
3.7
Duration

hr
54
24
33
24
36
53
20
24
36
46
Depth

in
13.41
10.08
9.76
8.64
9.73
8.84
8.18
7.44
7.34
7.17
     The final group of synthetic hyetographs is shown in Figures 10 and  11.
It is clear that they bear no resemblance to the actual historic storms shown
in Figures 8 and 9.
                                    311

-------
E
M
Z
     DESIGN STORMS  FOR  A 24 HR.  DURATION
                     AND A 5 YEAR RETURN PERIOD
    3.2
     1 -
    O.B
    0.6 -
    0.4 -
    0.2 -
 3 -
2.8 -
2.6 -
2.4 -
2.2 -
 2 --50mm/hr
1.8 -

1.4 -
          ILLINOIS
                           CHICAGO
                                 JTSR
                                     16
                                             2O
   O       4       6       12
   "                   TIME (hr.)
   Figure 10.  Chicago,  Flood Studies Report,  and
             Illinois  design storms.
                                                     24
                                                    24
01
z
     DESIGN STORMS  FOR  A 24 HR.  DURATION
                     AND A 5 YEAR KLTUKN PERIOD
3 -
2.B -
2.6 -
2.4 -
2-
l.B -
1.6 -
1.4 -
1.2-
1 -
O.B -
0.6 -
O.4 -
0.2 -




-5O mm/ hr







r— ^—
0 4
0












i —














SCS









^A
' i
^-^ -^=^=^= 	 _j
5 12 16 2O 2.
2'
TIME (hr.)
      Figure  11.  Soil Conservation Service and
                 Sifalda design storms.
                          312

-------
                           OTHER DESIGN TECHNIQUES

     The USGS has  performed a flood frequency analysis  on  15  basins  In the
Tallahassee area using the combination of continuous simulation and  regression
discussed earlier (Franklin and  Losey,  1984).  Their  predicted  5-yr  peak  for
Megginnis Arm is  1570 cfs.

     Finally, the venerable Rational Method could be applied.  The results  for
this method are also shown below, using a runoff coefficient of 0.27 (from  the
SWMM  modeling)  and  Weldon's  (1985) IDF data.
                                   RESULTS

     The main objective is a comparison of peak flows developed using alterna-
tive techniques.   For  this purpose,  seven  of  the historic storms listed  in
Table 2 (with antecedent conditions included) and five synthetic storms  were
run on a single event  basis using the calibrated version of SWMM described
earlier.  Seven historic storms from Table 3 (for total  flow)  were also  simu-
lated.   (Storms  for return periods of 43.3 and  37.1  months  were not run.)   For
all runs, a time step of 5 minutes was used,  but with hourly hyetograph
values, since historic rainfall data at finer time increments  were not avail-
able in time for this analysis.  (Copies of the microfilmed  weighing-bucket
charts must be ordered from the NWS National  Climatic Data  Center in  Ashe-
ville,  NC in order to obtain, say,  15 minute hyetograph increments.)   Peak
flows calculated by the model in  this way  are somewhat higher  than  for the
identical storms during the continuous simulation because SWMM calculates
average flows over the hourly time step  during  continuous  simulation  (in order
to avoid large continuity errors),  whereas instantaneous flows at the end of  a
time step are calculated during single event simulation.  Results for the 5-
year storms are given in Table 7 along with results for  the other methods that
have been applied to this basin.  Results  for all storms are shown  in Figure
12.  (The peculiar shape of the curve for  runoff volumes is because the
volumes are plotted at return periods corresponding  to a volume ranking, not  a
peak flow  ranking.)

     Before discussing the results, it is  interesting to note  that when  run on
a single event basis,  the  P-65 storm produces a lower peak than does  the P-52
storm,  in contrast to the results for the  same storms during the continuous
simulation (Tables 2 and 4).   This  is an artifact of  the numerical  methods
used in the SWMM  Runoff Block.  For continuous  simulation,  average  flows over
the time step (usually one hour) are computed in order to avoid continuity
errors when using long time  steps.  For  single  event simulation, instantaneous
flows at the end of the time step are computed. The peak instantaneous  values
are higher than the averages and respond more directly to peak rainfall  inten-
sities.  This factor,  coupled with the shorter  time step used  for the detailed
simulations were  enough  to alter the rankings of these two storms.   In gen-
eral, all the peak flows predicted during  the single event  simulations of  the
historic storms were  higher  by as  much as a few hundred cfs than those pre-
dicted during the continuous simulation, but the only change in rankings is
the one  discussed  above.
                                     313

-------
U)
v—>
-P.
             (ft
             LL
3.2
  3 -
2.8 -
2.6 -
2.4 -
2.2 -
  2 -
1.8 -
1.6 -
1 .4 -
1.2 -
   1 -
0.8 -
0.6 -
O.4 -
0.2
   0
                        -60m3/.s
                                                                                     HIST. PEAK
         OCHICAGO
                                                                               HlST.jrO'ML FLOW
       	1	1—
     0           4

           +    SCS
      OUSGS
                                                   T
            —I—
            12
T
                                              8          12         16
                                                 RETURN  PERIOD, YEARS
                20
24
o   ILLINIOS
 OPEAK  PLOW
       A    SIFALDA
     OTOTAL FLOW
        F.S.R
                         Figure 12.  Peak flows versus return period  predicted using historical storms,
                                    synthetic design storms and the  USGS method.  Historical storms
                                    based on total flow are plotted  at  return periods corresponding
                                    to runoff volume, not peak flow.

-------
      Considering a  5-year return period, for the Rational Method to predict a
 peak flow of  1748 cfs would require a duration on the IDF curve of approxi-
 mately 50 min,  much less than the actual (but  unknown!)  time of concentration.
 Hence,  it does  not  appear to be applicable.  Furthermore, the return period
 for  the method  is assigned on  the  basis  of  conditional frequency analysis of
 rainfall,  as  oppposed to the storm event analysis of runoff peaks or volumes
 of  the  other  methods.
        TABLE 7.  COMPARISON OF 5-YEAR PEAK FLOWS USING SEVERAL METHODS

Storm
V-65
V-52
P-65
P-52

SCS
Chicago
Illinois
Sifalda
F.S.R.
Peak Flow
cfs
1996
983
1384
1748

1926
1677
511
441
1235
Method Peak Flow
cfs
uses


Rational Method
tc = 24 hours
t_ = 13 hours
tc = 50 min



1570



187
289
1764



     Somewhat unexpectedly, the results for the synthetic design storms fall
on the low end of the estimates; only the SCS storm is consistently high.  For
example, a similar study by Marsalek (1978)  showed higher peaks using the
synthetic storms, but his basins were much  smaller (64 acre  and 321  acres).
The difference may be partly due to the choice of a hyetograph interval of 1
hour; synthetic hyetographs are sensitive to this choice since as the interval
becomes smaller, the peaks for shapes such as the Chicago and SCS storms tend
to become very high (and thus produce higher runoff peaks).   This is another
complication in the use of such methods.  On the other hand,  the historic
rainfall was also discretized at 1-hour intervals and might also be expected
to have higher peaks at shorter intervals (but the shorter interval data were
not available for this study).  Furthermore,  the  Megginnis Arm basin is large
with a long time of concentration;  in principle,  short time increment fluctua-
tions should be well damped by the basin response.

     Another reason why the peaks  are higher with the  historic storms lies in
the differences in storm volumes and durations.  For instance, the two "5-
year" historic  storms  chosen on the basis of volume are  both  larger than the
24-hr IDF value of 7.32  inches.    On  the  other hand, historic storm P-52 has a
duration of only 11 hours in which to concentrate  the rainfall  of  3.93  inches.
The anomaly is historic storm P-65, which has a lower rainfall volume and a
longer duration than do  the synthetic storms — but it has a  pocket of high
intensity rainfall (Figure 8a)  coupled with wet  antecedent conditions.
                                      315

-------
     The  main difference between the  results of this study and those of Marsa-
 lek  (1978) is  in the manner of choosing the historic storms.  Marsalek iso-
 lated  storms  on  the basis of rainfall criteria only, including a requirement
 that storms have a total depth greater than 4.9 in/hr and/or contain a 10-
 minute intensity greater than  5.9 In/hr.  His historic storms were further
 screened  by seeking maximum depths for durations of 10,  15,  30 and 60 minutes.
 In this study, the storms were screened solely on the basis of the- continuous
 simulation,  allowing  the catchment  to filter out the important elements of the
 rainfall-runoff  response (to the extent that it is properly modeled in SWMM).

     Arnell (1982) found that the Sifalda storm gave the best agreement with
 historic  storms,  in contrast to this example in which it gives a  low estimate
 for  peaks.  In general, this catchment produces higher peaks for  storms with
 late and  high peak rainfall intensities,  appearing after the infiltration
 capacity  of  the  basin is satisfied.

     The  USGS estimates (based on a log-Pearson Type 3 frequency analysis of
 simulated flows) do not differ widely from  those of the historic  SWMM runs at
 low  return periods, but all methods give  lower peak flows at high return
 periods than  that predicted by SWMM for the 260-month peak-flow storm.  As can
 be seen in Figure  12,  the SWMM simulations  of the historic storms produce
 lower  peaks than  those of the USGS and the Chicago and SCS design storms at
 low  return periods.  Hence, it cannot be  said unequivocally that synthetic
 design storms produce  conservative results.  On the basis of this single study
 on a single catchment,  the  SCS and Chicago  storms are conservative at low
 return periods but not  at  high  ones.

     However, there is  considerable uncertainty as to the true return period
 of the largest storm  runoff peak in the 22  year period.  Assuming on the basis
 of order  statistics that the empirical  frequencies assigned to the historic
 storms have a beta distribution  (Gumbel,  1958, Chapter 2), then +/- 25% confi-
 dence intervals  about the expected value  of the return period of the largest
 peak (23  years) are 11.1 and 47.5 years.   That  is,  the  probability is 50% that
 the true  return  period of the largest simulated peak in the 22 year record
 lies between  11.1 and  47.5  years.  These  confidence  intervals  narrow consider-
 ably at lower return periods.  Hence,  it can be  said  with more confidence that
 the SCS and Chicago storms are conservative at low return periods than that
 they are  not  conservative at high return  periods.   A much longer rainfall
 record would  resolve the question at  the  22 year return period, but there
 would always  be  doubt  at  the end point  of the record as to the proper fre-
 quency (even  if a frequency distribution,  such as the log-Pearson Type III,
 were fit  to the  data).  A main result  of  this study is  simply that there is no
 guarantee  that synthetic storms are conservative over the whole range of
 return periods.

     Ideally,   the measured runoff records should be analyzed to isolate  storm
 events and resolve the question  of which of the several curves shown on Figure
 12 is correct.  But the gage for the  Megginnis Arm Catchment has only been in
operation since 1973,  making it  difficult  to identify storms for return per-
iods greater  than about 10 years.  Another difficulty is the changing land use
within the catchment,  with rapid urbanization altering the nature of the
rainfall-runoff response.  This factor tends to direct  the analysis back to a
                                     316

-------
properly calibrated model.  However, for the record, the largest peak observed
during the intermittent sampling  since  1973 is 724  cfs on May 23, 1980 from a
storm of 1.43 inches.   It  is not suprising  that  the much higher rainfalls used
in the simulations  would  produce peaks  on  the order of 2000 cfs (Figure  12).
In general, maximum rainfalls after 1973 have been much less than before (see
Table 6).   Thus,  peak  flows simulated using 1958-79 rainfall data are quite
likely to be higher than those measured since 1973.


                       HISTORIC VERSUS SYNTHETIC STORMS

     The reader will probably not be surprised to learn that the authors favor
the historic storm  values for the following reasons:

1. They are based on a frequency analysis of peak flows (or runoff volumes),
not on a conditional frequency analysis  of  rainfall depths-   That is,  the
frequency analysis is  on the parameter of interest.

2. The frequency  analysis  of  the continuous  time  series  of  flows includes all
effects  of antecedent  conditions whereas the synthetic design storms do not.
Moreover, the detailed single-event  simulations  of historic storms may be
conducted in a "pseudo-continuous  manner," that is,  by  simulation of  several
days of  antecedent weather prior to  the  beginnning of the storm of interest.

3. The historic storms avoid the vexing  questions of storm  duration,  shape and
hyetograph discretization.   The  duration is  especially critical,  since peak
flows may arise  out of a storm  that lasts several days (e.g., storms  V-65,  V-
52, P-65) or out  of  a short,  intense one (e.g., storm P-52).  There is simply
no basis for establishing  a standard duration,  such as  24 hours,  for  all
design work, as seems to be  the unfortunate tendency in Florida.

4. Historic  storms can also  be used  for  analysis  of  volumes for design of
basins for detention or retention.  The volume of synthetic  storms is  arbi-
trarily  linked to the  assumed rainfall duration.   A given volume can  result
from an infinite number of combinations of  Intensity and duration,  with a
corresponding range  of return periods.

5. If  historic storms  are  used  for design,  the local citizenry  can be  confi-
dent that  the design will  withstand  a  real storm  that may be remembered for
its flooding by many,  as opposed to an "unreal"  synthetic storm.   Thus, .the
engineers  or agency  may make a statement such  as, "our  design will avoid  the
flooding that resulted from the  storm of September  3, 1965."
               PROBLEMS WITH USE OF HISTORIC STORMS  FOR DESIGN

     The analysis outlined in this  paper  is  neither  brief  nor  simple.   A
continuous simulation model must be calibrated and used,  first to identify
storm events and second to perform the more detailed design simulations.  This
means more work both for the design engineers and for  the  reviewing agencies
(leading to the  common arguments in favor of synthetic design  storms — they
are quick, easy,  and consistent from application to  application).   A possible


                                     317

-------
 compromise would be to compile an atlas of historic  storms and their antece-
 dent  conditions  for  use  in  design, based on generic runs of a continuous model
 for representative catchment sizes,  land  uses,  soil  types, climatic regions
 etc.  Research on this possibility is underway  at  the  University  of Florida.

      Another  consideration  is  the high  spatial  variability of real storms and
 the sensitivity  of runoff to storm direction  (James  and  Shtifter, 1981;
 Surkan, 1974).  Historic storms must still be derived from point rainfall
 data.  The catchment may be  more sensitive to the spatial dynamics of  storms
 than  to the relative rankings  within a  time series.  Of course, the same
 remarks apply to synthetic  storms.  However, the analysis of historic  storms
 may be carried further and  dynamic storms  constructed from historical data
 (James and Scheckenberger,  1984).   Spatial dynamics are more important as the
 size  of the catchment increases.

      Finally, the question of  economics versus  safety factors raised early in
 this  paper is unanswered.   Is  there  a savings in construction costs by using
 the type of analysis outlined  in this paper (with  the implication that the
 synthetic design storms commonly used in design are  too conservative)?  Or are
 the synthetic design storms not conservative  and therefore providing less than
 the level of protection assumed by design engineers?   What are the implica-
 tions on flood stage as opposed to peak flows?   Stage may only be affected in
 a minor way by differences of several hundred cfs in  flow for a wide flood
 plain.  In such  cases the design  may be insensitive to the design methodology
 used, and the quickest and fastest method  may suffice.  These questions are
 also  under investigation at  UF.


                                 CONCLUSIONS

      Given a calibrated and verified  model, selection of historic design storm
 events from a continuous simulation  offers the  advantages of a meaningful
 frequency analysis on the runoff  parameter of interest, such as peak flow or
 runoff volume; the frequency analysis need not be  tied to rainfall frequencies
 as with the use  of synthetic desfgn  storms.   For this study, historic  design
 storms were  selected from a frequency analysis  of  peak flows and runoff vol-
 umes  on the basis of a continuous SWMM simulation  using the 22 year period of
 available rainfall.  The model had previously been calibrated and verified
 using local rainfall-runoff data  for  the Megginnis  Arm Catchment in Tallahas-
 see.   The historic storms were then  simulated and  the results compared to
 those obtained using five different synthetic design storms, including the
 widely used SCS  Type II storm.  In general, the  SCS and Chicago storms pro-
 duced higher peaks at low return periods than produced by the historic storms,
 while the historic storms produced a higher peak at  the highest return period.
 Thus,  it cannot  be assumed that the synthetic storms are conservative  over the
entire range of  return periods.


                               ACKNOWLEDGEMENTS

     This research is supported by EPA Cooperative Agreement CR-811607 and by
USGS  project  USDI-14-08-0001-G1010.


                                     318

-------
                                  REFERENCES

  1.  Adams,  B.J. and C.D.D.  Howard.  Pathology of Design Storms.   Publica-
     tion 85-03, University of Toronto,  Dept.  of Civil  Engineering,  Toronto,
     Ontario,  1985.

  2.  Arnell, V. Analysis of rainfall data for use in design of storm sewer
     systems. Proc. First Int. Conf.  on Urban Storm Drainage, University  of
     Southampton,  Pentech  Press,  London,  71-86, April  1978.

  3.  Arnell, V. Rainfall Data for the Design of Sewer Pipe Systems.  Report
     Series A:8, Chalmers  University,  Dept. of Hydraulics, Goteborg, Sweden,
     1982.

  4.  Carlisle, V.W., Hallmark, C.T., Sodek, F., Ill, Caldwell, R.E., Hammond,
     L.C and V.E. Berkheiser.  Characterization  Data for Selected Florida
     joils.  Soil Science Research Report No. 81-1,  Soil Science Department,
     University of Florida, Gainesville, Florida,  June  1981.

  5   Driscoll, E.D.  in Benefit analysis for Combined Sewer Overflow  Control.
     Seminar Publication, EPA-625/4-79-013, Environmental Protection Agency,
     Cincinnati, Ohio,  April  1979.

  6.  Eagleson, P.S. Dynamic Hydrology.   McGraw-Hill, New  York,  1970.

  7.  Environmental Protection Agency. Areawide Assessment Procedures Manual.
     EPA-600/9-76-014,   Cincinnati, Ohio.   Il:E-l - E-115, 1976.

  8.  Esry, D.H.  and J.E. Bowman. Final  Construction Report,  Lake Jackson
     Clean Lakes Restoration Project.   Northwest Florida Water Management
     District, Havana,   Florida, 1984.

  9.  Farnsworth, R.K.  and  E.S. Thompson. Mean Monthly,  Seasonal, and Annual
     Pan Evaporation  for the United States. NOAA Technical Report NWS 34,
     Office  of Hydrology,  National Weather Service, Washington,  DC, December
     1982.

10.  Franklin, M.A. and G.T. Losey. Magnitude and Frequency of Floods from
     Urban Streams in Leon County^ Florida.  USGS Water Resources Investi-
     gations Report~54-4004, Tallahassee, Florida,  1984.

11.  Geiger, W.F. Characteristics  of  combined sewer runoff. Proc. Third Int.
     Conf. on Urban Storm Drainage,  Chalmers University, Goteborg, Sweden!
     2:851-860,  June  1984.

12.  Gumbel, E.J. Statistics of Extremes.  Columbia  University Press,  New York
     1958.

13.  Harremoes, P.  (Ed.). Rainfall as the Basis  for Urban  Runoff Design and
     Analysis.   Pergamon Press, New York,  1983.
                                     319

-------
14.  Huber, W.C.,  Cunningham, B,A.  and  K.A. Cavender. Continuous SWMM  modeling
     for selection of  design events. Urban Drainage Modeling,  Proc. Int. Symp.
     on Comparison of Urban Prainge Models with Real Catchment Data, Dubrov-
     nik, Yugoslavia, Pergamon Press, New York, 379-390, April 1986.

15.  Huber, W.C.,  Heaney,  J.P. and  B.A.  Cunningham. Storm Water Management
     Model (SWMM) Bibliography.  EPA/600/3-85/077 (NTIS PB86-136041/AS),  Envi-
     ronmental Protection Agency,  Athens, Georgia, September 1985.

16.  Huber, W.C., Heaney, J.P.,  Nix, S.J., Dickinson,  R.E.  and  D.J.  Polmann.
     Storm Water  Management Model User's Manual,  Version III.  EPA-600/2-84-
     109a (NTIS PB84-198423), Environmental Protection Agency, Cincinnati,
     Ohio, November  1981.

17.  Huff, F.A.  Time distribution of rainfall in heavy storms.  Water Resour-
     ces Research. 3(4); 1007-1019, 1967.

18.  Hydrologic Engineering Center. Storage,  Treatment,  Overflow,  Runoff Mo-
     del, STORM, User's Manual.   Generalized Computer Program 723-S8-L7520,
     U.S.  Army Corps of Engineers, Davis, California, 1977.

19.  Hydroscience, Inc. A  Statistical Method  for Assessment of  Urban Stormwa-
     ter Loads - Impacts - Controls.  EPA-440/3-79-023,  Environmental  Protec-
     tion Agency, Washington, DC,  1979.

20.  James, W. and M.A. Robinson. Continuous models essential for  detention
     design.  Proc.,  Engineering  Foundation  Conference  on Stormwater Detention
     Facilities Planning Design Operation and Maintenance,  Henniker, New Hamp-
     shire, American Society of  Civil Engineers, New York,  163-175, August
     1982.

21.  James, W. and R. Scheckenberger.  RAINPAC - a program  for analysis of
     rainfall inputs in computing storm dynamics.  Proc. Stormwater and Water
     Quality Meeting, Detroit, Michigan, EPA-600/9-85-003 (NTIS PB85-
     168003/AS),  Environmental Protection Agency, Athens,  Georgia,  81-100,
     April 1984.

22.  James, W. and Z. Shtlfter.  Implications of storm dynamics on  design storm
     inputs. Proc. Stormwater and Water Quality Managment Modeling and  SWMM
     Users^ Group Meeting, Niagara  Falls, Ontario, Report CHI-81,  Dept. of
     Civil Engineering, McMaster University, Hamilton, Ontario, 55-78, Sep-
     tember 1981.

23.  Johanson, R.C., Imhoff, J. C.  and  H.H. Davis. User's .Manual for Hydrolo-
     gical Simulation Program ^ Fortran (HSPF).   EPA-600/9-80-015, Environmen-
     tal Protection Agency,  Athens, Georgia,  1980.

24.  Keifer,  C.J.  and H.H. Chu.  Synthetic  storm pattern for drainage  design.
     J. Hyd. Div., Proc. ASCE, jtt:(HY4),  July 1957.

25.  Maalel,  K. and  W.C. Huber. SWMM calibration  using continuous  and  multiple
     event simulation.  Proc. Third Int. Conf. on Urban Storm Drainage, Chal-


                                    320

-------
     mers  University,  Goteborg,  Sweden.  2i595-604,  June  1984.

26.  Marsalek, J. Research on the Design Storm  Concept.  ASCE Urban  Water
     Resources Research Program  Tech. Memorandum No. 33, (NTIS PB-291936),
     American  Society  of  Civil Engineers, New York, 1978.

27.  McPherson,  M.B. Urban Runoff Control Planning.  EPA-600/9-78-035,  Environ-
     mental  Protection Agency, Washington,  DC,  October 1978.

28.  Natural Environment  Research Council.  Flood  Studies Report.   Five Vol-
     umes. Institute of Hydrology,  Wallingford, UK, 1975.

29.  Patry,  G. and  M.B. McPherson (Eds.).  The Design Storm  Concept.  EP80-R-8,
     GREMU-79/02, Civil Engineering Dept.,  Ecole  Polytechnique  de  Montreal,
     Montreal, Quebec, December  1979.

30.  Restrepo-Posada,  P.J. and P.S.  Eagleson.  Identification of independent
     rainstorms.  J^ Hydrology. ^5j309-3l9,  1982.

31.  Roesner,  L.A.,  Nichandros,  H.M., Shubinski, R.P., Feldman, A.D., Abbott,
     J.W.  and A.O. Friedland.   A_ Model for Evaluating Runoff-Quality J£ Metro-
     politan Master Planning. ASCE Urban Water Resources Research Program
     Technical Memorandum No. 23 (NTIS PB-234312), American Society  of Civil
     Engineers, New York, April  1974.

32.  Roesner,  L.A-,  Shubinski,  R.P.  and  J.A. Aldrich. Storm  Water  Management
     Model User's Manual,  Version III:  Addendum I±  EXTRAN.  EPA-600/2-84-109b
     (NTIS PB84-198431),  Environmental Protection Agency, Cincinnati,  Ohio,
     November  1981.

33.  Sifalda, V.  Entwicklung  eines  berechunungsregens fur die beraessung von
     kanalnetzen.  Gwf -•  Wasser/Abwasser.  114(H9),  1973.

34.  Soil Conservation Service. National Engineering Handbook,  Section 4,
     Hydrology.  U.S. Dept. of  Agriculture,  Washington,  DC,  1964.

35.  Surkan,  A.J. Simulation  of storm velocity effects on flow from  distri-
     buted channel networks.  Water  Resources  Research. 10(6):1149-1160, Decem-
     ber 1974.               	

36.  Weldon,  K.E.  F.D.O.T. rainfall  intensity-duration-frequency  curve genera-
     tion.   In M.P.  Wanielista and  Y.A.  Yousef (Eds.), Stormwater  Management,
     _An Update, Publication 85-1, University of Central Florida, Environmental
     Systems Engineering  Institute,  Orlando,  Florida.  11-31, July 1985.

37.  Wenzel,  H.G., Jr.  and M.L. Voorhees. An Evaluation of  the Urban Design
     Storm Concept.  Research  Report  164, Water Resources~~Center, University^ of
     Illinois,  Urbana-Champaign, Illinois,  August  1981.

The work described in  this paper was partially  funded by the U.S.  Environmental
Protection Agency and  has been subject to the Agency's peer and administrative
review.  It has been approved for publication as  an EPA  document.


                                     321

-------
 An Expert System Prototype for RECEIV-II Using TURBO PASCAL
                                     by

                             Robert E.  Dickinson
              Department of Environmental Engineering  Sciences
                            University  of Florida
                         Gainesville, Florida  32611

                                Ivan B.  Chou
                    Applied Technology  & Management, Inc.
                         Gainesville, Florida  32602

                               Fred V.  Ramsey
                    Applied Technology  & Management, Inc.
                         Gainesville, Florida  32602
                                  Abstract
     A  pre-processor written  In TURBO  PASCAL was created  for the RECEIV-II
model.  The graphics based  pre-processor enables trained  engineers and
scientists familiar with  estaurine hydrodynamics to easily run the
receiving water model.  The pre-processor acts similarly  to programs such
as LOTUS-123  in enabling  a  higher percentage of people to become competent
modelers.
                               Introduction
    The use of the personal computer by the engineer and scientist Is a
landmark advance that greatly increases their analytical tools.
Spreadsheets, Word Processing programs, and Data Base Management programs
increase productivity by (1) compressing the time between the initial
conceptualization and the final product;  and (2) eliminating some of the
mechanical impediments to original creation.  The PC aids both in the
creation and production of engineering and scientific products.

    The primary usage of the PC, however, is restricted to programs equally
applicable to any profession, trade, or occupation.  The scientific and


                                   322

-------
engineering community has only begun to realize the potential of the PC as
design aids and expert systems.

    The purpose of this paper is to introduce a prototype expert system for
the RECEIV-II model (Raytheon, 1974) constructed using the Borland
International program TURBO PASCAL (Borland,  1985).


                              Expert Systems
     An expert system incorporates the human expert's knowledge into a
computer program so that others can solve the same type of problem (Van
Horn, 1986).  Some of the characteristics of "real" expert systems as
described by Van Horn(1986) are: (1) the program contains the heuristic
knowledge of an expert; (2) the program is able to interface with the user
and developers in natural language; (3) the program asks questions to
obtain needed data; (4) the program is easily refined and upgraded without
extensive reprogramming;  (5) the program can explain its conclusions; (6)
the program can accept uncertain input and assign it a certainty factor;
(7) the program gives answers with a certain level of confidence; and (8)
the program learns from its own performance.

    A combination pre-processor and water quality model on the PC addresses
most of the requirements of an expert system using a liberal interpretation
of the definition of an expert system.  The model itself functions as the
rule base of the expert system.  The inference engine is built into the
structure of the program,  and the pre-processor acts as a quasi natural
language interface.

     The key gain in using a pre-processor and model in the PC environment
is the increase in modeling competence, performance, and productivity.
The model can now be used by any engineer or scientist with a knowledge of
coastal hydrodynamics and limited computer experience.
                               Pre-Processor
     This section will explain the actual construction of a menu driven
pre-processor in TURBO PASCAL.  The tools used to construct the pre-
processor include TURBO SCREEN (Pascom Computing, 1985), TURBO PASCAL
(Borland International, 1985), and TURBO Toolbox (Borland International,
1984).

     TURBO SCREEN is a program used to construct graphical menus for data
input.  The menus are made using a menu driven text editor with WORDSTAR
commands.   The program generates TURBO PASCAL code that will mimic in a
compiled TURBO PASCAL program the menus created by the programmer. The
menus, which consist of colored or shaded foreground text on colored or
shaded backgrounds, are restricted to 24 row by 80 columns. The text
explains the data input requirements for each data field and function as
                                    323

-------
built in help screens.  The pre-processor data fields, programmer defined,
are either real, integer, string or character.

     A graphical menu approach to the pre-processor was chosen over a
simple question and answer approach for aesthetics, program modularity, and
the ability to use window management.  Each menu is defined as a picture or
window and is manipulated as a record in PASCAL.  PASCAL records are
similar to arrays in FORTRAN except records can have mixed types, e.g.
integer, real, and string components in the same record.  Each menu is
easily manipulated by the programmer by simply specifying the picture
number.  The beauty of PASCAL over BASIC and FORTRAN is this ability to use
records.

     The pre-processor either creates a new input file or edits an existing
file.  The user of the RECEIV-II model does not have to learn the FORTRAN
field codes for data input; the pre-processor handles the actual ASCII file
manipulation.   The pre-processor allows movement between fields in the
screen by reading the keyboard buffer.  The data input stream does not have
to be sequential.  It can be entered vertically or horizontally in the
menu.  As an example entering channel data can be done channel by channel
or field parameter by field parameter.

     The creation of the pre-processor (1950 PASCAL statements)  took
approximately three days from start to working product using the advanced
programming tools available from Borland International.  The program
creation was greatly aided by the sequential nature of the menus built
using TURBO SCREEN and the editing environment of TURBO PASCAL.
                                Discussion
     Compilers constructed specifically for the PC such as TURBO PASCAL and
TURBO PROLOG (Borland, 1986) offer an enhanced editing environment, faster
compilation, and the advantages of a higher level language.  The
programmer's productivity is increased because: (1) low level language
errors such as undefined variables are eliminated; (2) there is better
error checking during compilation; and (3) faster compilation speeds (less
than 1 minute for a 2000 line program).  The structure of PASCAL is similar
to FORTRAN.  A FORTRAN programmer can easily learn the ancillary techniques
to program in PASCAL.

     The most expensive component of software development is the
programmer's time in writing and debugging source code (Brooks, 1975).
Languages and compilation systems that enhance the creation and
implementation of programs will: (1) increase programming productivity; and
(2) enable a higher percentage of the population to program.  As
experienced FORTRAN programmer's we know the frustration of late night
sessions hunting for a bug in the program.  A language that eliminates most
bugs in the compilation stage has a tremendous evolutionary advantage over
lower level languages.
                                     324

-------
        Over the last 40 years as computer languages have advanced from
 machine code to assembly language to BASIC and FORTRAN to PASCAL to MODULA-
 2 the ease of programming has increased and the percentage of the
 population capable of programming has Increased.  The popularity of
 programs such as LOTUS-123 and DBASE-III Is partly due to the need of
 everyone to appear as competent programmers and computer "literate".  These
 programs allows a novice programmer the almost Instantaneous ability to
 generate programs.

      A pre-processor linked to a model on the PC allows not just the
 experienced programmer but the trained engineer and scientist to be a
 competent modeler.  The "natural language" interface is similar to LOTUS-
 123 in permitting a higher percentage of the population of engineers and
 scientists to become competent RECEIV modelers.
                                 References

  Borland  International, TURBO PASCAL, Borland  International, Scotts  Valley,
  Ca.,  1985.

  Borland  International, TURBO TOOLBOX, Borland International, Scotts  Valley,
  Ca.,  1984.

  Brooks,  F.P, The mythical man-month, Addison-Wesley, Reading Massachusetts,
  1975.

  Van Horn, M., Understanding Expert Systems, Bantam Books, Toronto,  1986.

  Pascotn Computing, TURBO SCREEN Programmers Manual, Pascom Computing,
  Cleveland, Ohio, 1985.

  Raytheon Company, Part 1 - RECEIV-II Water Quantity and Quality Model,  U.S.
  Environmental Protection Agency, Washington, D.C., 1974.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore does not necessarily reflect the views of
the Agency and no official endorsement should be inferred.
                                    325

-------
                      MODELING FLOOD HYDROLOGY USING HYMO
                          by:  James E. Scholl, P.E.
                               CH2M HILL
                               Gainesville, Florida 32602
                                   ABSTRACT

     HYMO is a computer program well suited to model flood hydrology for a
wide range of watershed conditions.  Any design storm duration and
distribution can be input along with site-specific unit hydrograph shape
parameters.  Stream or channel routing is accomplished using the variable
storage coefficient method; the storage-indication method is used for
reservoir routing.  The original program was developed for mainframe
equipment, but can easily be adapted for use on microcomputers.

     After a brief summary of HYMO commands, formatting, and storage, the
paper describes hydrograph analyses performed in the U.S. Virgin Islands.
Results of the analyses were used to develop a basis for assigning unit
hydrograph parameters to model flpod hydrology.
                                HYMO DESCRIPTION
      HYMO,  derived from the words hydrologic model,  is a computer  language
developed by  the  Agricultural  Research Service,  U.S.  Department  of Agricul-
ture,  in cooperation  with the  Texas  Agricultural Experiment  Station, Texas
A&M University  (1).   The original program was developed for  mainframe equip-
ment,  but can easily  be adapted for  use on microcomputers.   HYMO commands
are simple  to use and offer a  great  deal of flexibility.  For example, any
design storm  duration and distribution can be input  along with site-specific
unit hydrograph shape parameters.  Flood hydrographs  are  developed using
unit hydrograph theory and the SCS rainfall-runoff relationship.

     The variable storage coefficient  {VSC)  flood-routing method is used for
stream routing.   The  VSC method accounts for changes  in channel  velocity or
reach  travel  time with flood stage.  Although an iterative solution is used,

                                     326

-------
 the VSC method requires  minimal  computer  time  and  is  free  of  convergence
 problems.   Channel  section  rating  curves  used  for  VSC flood-routing  calcula-
 tions  can  be  input  directly or computed internally by HYMO for  uniform flow
 using  Manning's Equation.   Channel section data  for internal  calculations
 can be segmented  into  flood plain  and main channel components with appro-
 priate slope  and  n  values for each segment.

     The storage-indication method as documented in the SCS National
 Engineering Handbook,  Section 4, Hydrology (2) is  used by  HYMO  for reservoir
 routing computations.

                                 HYMO COMMANDS
     The  HYMO  language includes  17 commands  for performing flood hydrology
 and  sediment yield calculations.  Commands are expressed in the first  20
 columns of  the data card, with columns 21 through  79 used for numeric  data
 and  keywords.  Column 80  is reserved for a page change code (an asterisk  in
 column 80 causes  the printer to  advance to a new page).  Continuation  cards
 are  allowed when  59 characters are insufficient to present the data.

     The  data  can be written in  any format,  but at least one blank space
 must be left between data items.  A decimal  is required for numbers
 containing  fractions, but not for whole numbers.  Keywords can be written
 with the  data  to  describe individual data items.  Comment cards may be used
 at any point in a HYMO program by punching an asterisk in column 1 and the
 comment in  columns 2 through 79.

     Six  hydrographs can be stored in a HYMO program at a time, identified
 by storage  location numbers 1 through 6.  The storage  location numbers must
 be.repeated for watersheds requiring more than 6 hydrographs.  However, no
 more than six  hydrographs are ever needed at one time because HYMO programs
 begin at  the head of a watershed and work downstream through one reach at a
 time.  Because the first hydrograph is lost  when a storage location number
 is used to  store  or compute another hydrograph, the user should be sure that
 the  hydrograph will not be referred to again before using the storage
 location  number for another command.

     Details of rules for HYMO commands are presented in the Users Manual
 (3) along with an example problem.

                              HYDROGRAPH ANALYSIS


     To establish a site-specific basis for using HYMO in the  U.S.  Virgin
Islands,  published and unpublished streamflow data were obtained from the
U.S.  Geological Survey (4)(5).   Because long unbroken periods  of record were
not available,  observed data did not provide an adequate basis to develop an
annual maximum flood series.   The data did,  however,  contain several
isolated single-event runoff hydrographs which could be analyzed to
determine site-specific flood hydrograph parameters.   These included the
time to peak,  the total runoff volume,  the peak flow rate,  and the SCS unit
hydrograph shape factor.


                                    327

-------
     Streamflow data from four gaged watersheds on St. Croix were used to
establish a basis for assigning unit hydrograph parameters for ungaged
watershed areas.  Streamflow records considered in this study are summarized
in Table 1.
TABLE 1.  USGS STREAMFLOW DATA AVAILABLE FOR ST. CROIX, U.S. VIRGIN ISLANDS

Station Name
River Gut at River
River Gut at Golden Grove
Jolly Hill Gut at Jolly Hill
Station
Number
3320
3330
3450
Drainage
Area (mi2)
1.42
5.12
2.10
Period of
Record Examined
1963
1963
1963
- 1967 (5 years)
- 1971 (9 years)
- 1968 (6 years)
Creque Gut above Mount
Washington Reservoir
3470
0.50
1965 - 1967 (3 years)
     The first step in the hydrograph analysis was to obtain the original
stage hydrographs  {strip charts) from the USGS files in Puerto Rico (6).
Strip charts were obtained for selected days for all four stream gaging
stations listed in Table 1.  These strip charts were then screened for
single-peak events which resulted from short duration rainstorms.  Since
synchronized rainfall records were not available at these gages, this
selection process was largely a matter of judgment.

     Once the stage hydrographs were selected for analysis, they were
converted to discharge hydrographs by application of the appropriate rating
table, and the following hydrograph parameters were measured:

     1.   Time to peak {T ), in hours, defined as the time from the
          beginning of rile to the peak of the hydrograph

      2.   Peak flow  rate  (q  ) of the  runoff  hydrograph,  in cfs

      3.   Total volume of  runoff  (Q),  in  inches

      Using these measured  parameters,  the hydrograph  shape factor, B, was
computed for each  event as follows:
                          B =
                                                                        (1)
where:
           B  =  Hydrograph shape  factor
           A  =  Watershed drainage area,  in square  miles
               {all  other terms  are as previously  defined)

These parameters  are  reported in Table  2 for each event analyzed.
                                     328

-------
TABLE 2.  HYDROGRAPH ANALYSIS RESULTS, BASED ON USGS DATA

Location
River Gut at River,
No. 3320

Mean
River Gut at Golden
Grove, No. 3330




Mean
Jolly Hill at Jolly
Hill, No. 3450


Mean
Creque Gut above
Mt. Washington
Reservoir, No. 3470

Mean
T
Event Date p
4/06/65
5/30/65
11/09/65

8/01/63
11/02/63
11/21/63
10/24/69
10/30/69
12/02/69

1/03/63
1/04/63
12/12/65
12/12/65

10/13/65
11/12/65
11/09/65
5/17/65

(hours)
1.7
3.0
1.0
1.9
4.00
1.30
1.25
1.50
5.50
2.75
2.72
0.60
0.50
0.60
0.70
0.60
1.0
1.0
0.5
0.5
0.75
Q (inches)
0.033
0.064
0.023

0.240
0.100
0.021
0.112
0.062
0.070

0.024
0.097
0.009
0.023

0.297
0.250
0.174


S (cfs)
11.44
11.78
18.91

144.0
123.0
28.1
169.0
26.5
41.5

39.5
177.0
17.5
39.5

91.1
79.8
83.7


B
415
389
579
461
467
311
325
440
457
317
386
470
434
556
572
508
613
638
481
462
549
     Equation 1 is an algebraic transformation of the hydrograph peak rate
equation employed by the Soil Conservation Service (SCS)  (7).   The standard
value for B used by the SCS in the majority of hydrologic design applica-
tions is 484.  However, Mockus (1972)  reported that the hydrograph shape
factor has been known to vary from about 600 in steep terrain to 300 in very
flat swampy country.  From equation 1, it can be seen that the larger the
hydrograph shape factor, B, the larger the peak rate of runoff generated by
a given volume of runoff, Q.  Thus, the hydrograph shape, which is generally
related to topography, is also an important factor influencing flood
potential.

     The hydrograph shape factor is a watershed characteristic and should be
a constant for each watershed analyzed.  However, considerable variation is
reported in Table 2 for individual event B values at each gaging station.
These variations are primarily due to the fact that all errors present in
each measured hydrograph parameter (i.e., T , Q, and q )  are combined in the
computation of B.  The time to peak, T  , is^a particularly difficult para-
meter to measure accurately due to the time scale of the original strip
charts  (1 inch = 10 hours) and the short times of observation.  Therefore,
                                     329

-------
the mean of the event B values is probably the best estimate of the hydro-
graph shape factor for each watershed analyzed.

     Plotting the results from Table 2 as shown in Figure 1, the relation-
ship between the unit hydrograph shape factor, B, and drainage area provides
a basis for modeling ungaged watersheds.  A total of 13 watershed areas in
the U.S. Virgin Islands ranging in size from 365 acres to 3,396 acres have
been modeled using this procedure.  Nine of these watersheds were on St.
Croix and four on St. Thomas.  The results of this modeling work are
contained in three technical reports by CH2M HILL (8){9}(10).

                                  CONCLUSIONS
     HYMO is a practical tool for evaluating flood hydrology for a wide
range of conditions and project requirements.  The analysis of observed
streamflow hydrographs, as demonstrated using data for the U.S. Virgin
Islands, can provide a site-specific basis for assigning unit hydrograph
parameters for ungaged watersheds.

     The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official endorsement
should be inferred.

                                  REFERENCES
1.   Williams, J.R. and Hann, R.W.  HYMO:  Problem-Oriented Computer
     Language for Hydrologic Modeling—Users Manual.  ARS-S-9, Southern
     Region Agricultural Research Service, U.S. Department of Agriculture,
     1973.

2.   Mockus, V.  Section 4, Hydrology.   In:  SCS National Engineering
     Handbook  (NEH-4).  U.S. Department  of Agriculture, Soil Conservation
     Service, Washington, D.C., 1972.

3.   Williams, J.R. and Hann, R.W.  HYMO:  Problem-Oriented Computer
     Language for Hydrologic Modeling—Users Manual.  ARS-S-9, Southern
     Region Agricultural Research Service, U.S. Department of Agriculture,
     1973.

4.   Robison, T. M. et al.  Water Records of the U.S. Virgin Islands,
     1962-69.  U.S. Geological Survey, San Juan, Puerto Rico, 1972.

5.   McCoy, J.  Personal Communications.  U.S. Geological Survey, San Juan,
     Puerto Rico, 1979.

6.   McCoy, J.  Personal Communications.  U.S. Geological Survey, San Juan,
     Puerto Rico, 1979.
                                    330

-------
Hydrograph Shape Factor— B
8 8 § 1 § 1
Crequc
	
B = 4



•h-
> Gut Above ML W
1
1 	 J
M Slandar



dSCS



-L.
ashinglon
Reservoir

telue








**-,
^^M



•—




^
^M



'-





• ^*"*-
River Gut at River
Rh


• Jolly Hil

er Gut at (


Gul at
-^
iolden <


Jolly 1
•^
5rove


Hill
«_
• "*




•**




*


          0,2      0.3    0.4   0.5  0.6    0.6   1.0           2.0

                             Drainage Area (square miles)
3.0
     4.0   5.0  6.0
                 8.0
          FIGURE 1. Hydrograph Analysis Results for the U.S. Virgin Islands.
7.   Mockus, V.  Section 4, Hydrology.   In:   SCS National Engineering
     Handbook  (NEH-4).  U.S. Department  of Agriculture, Soil Conservation
     Service, Washington, D.C.,  1972.

8.   CH2M HILL.  A Flood Damage  Mitigation Plan for the U.S. Virgin Islands.
     Prepared for the Disaster Programs  Office, Office of the Governor,
     Government of the U.S. Virgin  Islands,  1979.

9.   CH2M HILL.  Planned Drainage Basin  Studies for the Protection of Roads
     from Flood Damage in the U.S.  Virgin Islands,  Volume 1—St. Thomas.
     Prepared for the Public Works  Department,  Government of the U.S. Virgin
     Islands, 1982.

10.  CH2M HILL.  Planned Drainage Basin  Studies for the Protection of Roads
     from Flood Damage in the U.S.  Virgin Islands,  volume 2—St. Croix.
     Prepared for the Public Works  Department,  Government of the U.S. Virgin
     Islands, 1982.
 The  work described in this paper was not funded by  the  U.S.  Environmental
 Protection Agency and therefore does not necessarily  reflect the views of
 the  Agency and no official endorsement should be  inferred.
                                     331

-------
                                   ATTENDEES
Name

John Aldrich

Richard Baker

Tom Barnwell

Christina Barrett

Vern Bonner

John M. Grouse

Brett Cunningham

Geoffrey Dendy

Roy R. Detweiler

Robert E. Dickinson

Forrest Dierberg

Jon Dobson

Eugene Driscoll

J. D. Edgman

Jeff Einhouse

Andrew C. Eversull

Raymond Ferrara

Huqh Fraser

David R. Gaboury

Victor Gaghorado

Richard Gietz


Lou Grant

Thomas T. Griffin

Judy Grim
Representing

COM

Metcalf & Eddy

EPA

Greiner Engineering, Inc.

Hydrologic Engineering Center

Greenhorne & O'Mara, Inc.

University of Florida

Greinger Engineering, Inc.

Chados Ford Enterprises, inc.

University of Florida

Florida Institute of Technology

Gainesville, FL

Oakland, NJ

Dallas Water Utilities

Miller, Miller, Sellon, Einhouse

Louisiana State University

Lafayette College

Cumming-Cockburn Associates,  Ltd,

Woodward-Clyde Consultants

Poland, FL

Regional Municipality of Ottawa-
Carleton

Bio Environmental Services

Najarian & Associates, Inc.

Briley, Wild & Associates
                                 332

-------
                                   ATTENDEES
Name




Ji Han




Michael C. Hancock




James P. Kearney




Jeffrey D. Holler




Wayne Huber




William James




Howard A. Jongedyk




Roger T. Kilgore




Stephen Kintner




Tinny H. Lee




Agustin E. Maristany




Robert McCarthy




Joseph M. McGinn




S. Wayne Miles




Michael Morrison




Ananta K. Nath




J-Rene' Noiseux




J. Robert Owen




Richard J. Pfevfrer




Larry A. Roesner




Mark Robinson




Robert Roussel
Representing




Post, Buckley, Schuh, & Jernigan  Inc.




University of Florida




University of Florida




South Florida Water Management District




University of Florida




McMaster University




Federal Highway Administration




GKY 6 Associates




Brevard County Water Resources Dept.




Briley, Wild & Associates, Inc.




N.W. Florida Water Management District




Dallas Water Utilities




CB MacGuire Inc.




Raleigh, NC




Hayes, Seay, Mattern & Mattern




Nebraska Natural Resources Commission




Les Consultants Dessan, inc.




Colorado Dept. of Health




South Florida Management Division




Camp, Dresser, & McKee




Computational Hydraulic Group




Les Consultants Dessan, Inc.
                                   333

-------
                                    ATTENDEES
Name




A. Charles Rowney




Marty Sanders




James E. Scholl




Linda Seigle




Phil Shelly




Himat Solanki




Wing H. Tang




Wilson Timmens,  Jr.




Ben Urbonas




Michael G. Waldon




Raymond E. Wiles




Dr. P. Wisner




Yousef A. Yousef




Claudia L. Zahorcak
Representing




Queen's University




Lindahl, Browning, Ferrari & Hellstrom




CH2ZM-Hill




Florida Dept. of Transportation




Washington Analytical Sers. Center Inc.




Smally, Wellford & Nalven, Inc.




Univ. of Puerto Rico




Brevard County Water Resources




Urban Drainage & Flood Control District




LSU - Louisiana State Univ.




GeoScience, Inc.




Univ. of Ottawa




Univ. of Central Florida




Brown and Caldwell
 *U.S. GOVERNMENT PRINTING OFFICE: 1986-646-116-40655
                                      334

-------