EPA/600/9-83/015
              d States
              jnmental Protection
Environmental Research
Laboratory
Athens GA30613
EPA-600/9-83-015
September 1983
              arch and Development
             Proceedings of
             Stormwater and
             Water Quality Model
             Users Group Meeting
            January 27-28,  1983

-------
                                      EPA-600/9-83-015
                                      September  1983
             PROCEEDINGS
                  OF
  STORMWATER  AND  WATER QUALITY  MODEL
         USERS GROUP MEETING
         January  27-28, 1983
              Edited by

       Thomas 0.  Barnwell,  Jr.
  Center for Water Quality  Modeling
  Environmental  Research Laboratory
          Athens, Ga. 30613
  ENVIRONMENTAL RESEARCH LABORATORY
  OFFICE OF RESEARCH AND DEVELOPMENT
U. S. ENVIRONMENTAL PROTECTION AGENCY
           ATHENS, GA 30613

-------
                                  DISCLAIMER

  Mention of trade names or commercial products does not constitute endorse-
ment or recommendation for use by the U. S. Environmental Protection Agency.
Similarly, publication of studies reporting better results from one model vis-
a-vis others does not constitute endorsement.
                                     ii

-------
                                   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 exchange of
information and ideas between researchers and users, and between 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 interest to include models other than the
Stormwater Management Model and other aspects of modeling water quality in
urban and natural waters.  This report, a compendium of papers presented at
the January 1983 Users Group meeting, is published in the interest of dissemi-
nating to a wide audience the work of group members.

                                        William T. Donaldson
                                        Acting Director
                                        Environmental Research Laboratory
                                        Athens, Georgia
                                    iii

-------
                                   ABSTRACT

    This report includes 17 papers on topics related to the development and
application of computer based mathematical  models for water quantity and qua-
lity management presented at the semi-annual meeting of the Joint U.S.-Cana-
dian Stormwater and Water Quality Model  Users Group held on January 27-28,1983
in Gainesville, Florida.

    Topics covered included an applications of the QUAL-II  model  in the North-
eastern U.S. and Columbia, South America ;  applications of  STORM  in Singapore
and the Republic of China; and implementation of SWMM in Tunisia.  Modeling
pesticides in small ponds and mixing zones  in rivers was also discussed, as
well as sizing detention ponds to meet water quality objectives.   A program to
assess the impact of forest management practices on water quality was also
presented as was a mechanistic model for nonpoint source simulation.  Papers
related to precipitation discussed data management, estimation of missing
values, and areal intensity-duration-frequency curves.  The questions of hy-
drological regionalization, snowmelt, and porous pavements  were addressed.
Other papers discussed regional Stormwater  management facilities  and storm
sewer design optimization.
                                     iv

-------
                                   CONTENTS
                                                                          Page
FOREWORD	  i i i
ABSTRACT	   i v

Project Success - Sensitivity to Project Assumptions	    1
     Anthony Knepp and Grace Wood
     Dames a Moore

Application of the STORM Model to Design Problems in Singapore  and Kaoh-
  siung, Republic of China 	   22
     Brendan M. Harley
     Camp Dresser a McKee Inc.

An Attempt to Implement SWMM in Tunisia	   43
     Janusz Niemczynowicz
     Lund Institute of Technology

The Sediment-Water Interface in Modeling Pesticides in Sedimentation
  Ponds 	   53
     Raymond A. Ferrara and Peter Jaffe
     Princeton University

A Mixing Zone Model for Conservative Parameters 	   65
     Main R. Hutcheson
     Oklahoma Water Resources Board

Some Recent Adaptations and Applications of QUAL-II in the Northeast 	   87
     William W. Walker, Jr.
     Environmental Engineer

A Review of Model Use in Evaluating Nonpoint Source Loads from  Forest
  Management Activities 	  102
     George G. Ice and Raymond C. Whittemore
     National Council of the Paper Industry for Air and Stream  Improvement

Detention Pond Sizing to Achieve Water Quality Objectives 	  123
     Roger K Wells
     HMM Associates, Inc.

A Mechanistic Simulation for Transport of Nonpoint Source Pollutants 	  146
     Daryl B. Simons, Run-Ming Li and Kenneth G. Eggert
     Simons, Li a Associates

-------
Data Management for Continuous Hydrologic Simulation 	 161
     Jy S. Wu
     University of North Carolina at Charlotte

Estimation of Missing Values in Monthly Rainfall  Series 	 177
     Efi Foufoula
     University of Florida

Area! Intensity-Duration-Frequency Curves - A Possible Way of Improving
  the Rainfall Input 	 192
     Janusz Niemczynowicz
     Lund Institute of Technology

Hydrological Regionalization: A Question of Problem and Scale	202
     I. Simmers and E. Seyhan
     Free University, Amsterdam

Snowmelt Induced Urban Runoff in Northern Sweden  	 215
     Lars Bengtsson
     McMaster University

An Advancement in Hydraulic Modeling of Porous Pavement Facilities 	 237
     Gary Goforth
     Espey Huston a Associates Inc.

Planning and Implementation of Regional Stormwater Management Facilities
  i n Montgomery County, Maryland 	 255
     John M. Crouse(l), Vincent H. Berg(2)  and Linda J. D. Mitchell(l)
     vl)Greenhorne  a O'Mara and (2)Montgomery County, MD

Storm Sewer Design Optimization 	 272
     Dong Hoang
     City of Portland

ATTENDEES	 281
                                     vi

-------
                Project Success:    Sensitivity  to  Project Assumptions
                                         "The Medellin Case"
                                                    by
                                         Anthony  Knepp,  P.E.
                                                    and
                                             Grace E.  Wood
                                   Dames & Moo're,  Bethesda,  MD
                     Abstract
    The effect of major planning assumptions on the predicted water
quality of the R1o HedelUn 1s Investigated through application
of the  dual II model. The  R1o HedelUn  flows  through the center
of  MedelUn, the second largest dty In Colombia with a 1980 popu-
lation  of over 1.2 million people.  Currently, there Is no treatment
of municipal or Industrial waste.  The river receives raw  sewage
via  direct discharge and from numerous polluted tributaries 1n the
developed  area.  Nonpolnt  sources  of pollution  contribute large
quantities of sediment  to  the river. Practices such as 1n-stream
gravel  mining and direct dumping of solid waste add to the large
sediment and pollutant loads observed during wet weather periods.
All  of these factors result not  only  1n  very  poor biochemical
quality of the river, but frequent visual, color, and odor problems.
Public health 1s threatened  by the  poor  quality of the river, and
uses are severely limited.

    Preliminary  results of a major planning study (Greeley  and
Hansen, CCC, 1982) indicate that  a  configuration of four treat-
ment plants sited along the modeled  portion of the Rio MedelUn
would cost-effectively meet specified water quality criteria.

    This  paper presents the  results of a sensitivity analysis
using Qua! II which  demonstrates the degree predicted  water
quality 1s affected by initial project assumptions, such as:

    o  Expected ability to control  the nonpolnt sources
      of pollution, including solid waste and land runoff
    o  Availability of dilution water
    o  Likelihood of meeting wastewater collection goals.

The paper examines the Importance of meeting  basic assumptions  to
the  attainment of water quality goals. The  sensitivity  analysis
was accomplished  by establishing a "baseline" condition   and
examining  the effects on baseline quality of modifying a single
assumption or group of assumptions. Changes to the  baseline
condition  included  variations  1n flow,  percent   wastewater
collection, dilution water  Inflow,  nonpolnt  source  inputs, and
solid waste contributions.
                                                                 Study Area Location Map
 Introduction
      Every  water  resources  study team begins  a  project with the  expectation
 that  it  will  proceed  in  a  systematic manner  from problem  identification  and
 collection of  pertinent  information, through the development  of feasible
 solutions, to  the determination of  their  consequences.   Final results
 usually  include  the  identification  of  one or,  at most,  a  limited  number  of
 feasible solutions.    The decision  maker(s)  are then expected  to arrive at
 a final  decision  concerning  implementation  of  a  selected  alternative,  and

-------
then proceed to the next problem.  Experience suggests that in small pro-
jects such a systematic approach works quite well.  However, in many larger
water resources planning projects, such conceptual neatness is brought
closer to reality by the following observations of Loucks (1981):

     o   A final solution to a water resources problem rarely exists.

     o   Plans and projects are dynamic and evolve over time.

     o   The time to study a problem is never adequate to perform "state-
         of-the-art" mathematical work.  The converse of the above
         statement should also be recognized, i.e., if sufficient
         time does exist the objectives of the original study will
         have shifted by the time the study is completed.

     Recognizing that these difficulties exist, a major contribution to the
planning process is obtained when a sensitivity analysis is conducted for
the major project assumptions composing the "Model" used in the analysis of
feasible solutions.  "Model" is used here to mean the set of formulae
coefficients, parameters, etc., which reproduces both prototype behavior
and system response to future conditions.  In this way, the evolution of
project objectives is enhanced and the feasibility of project solutions is
recognized early in the planning stages.  Most importantly, the objectives,
feasibility of alternatives, and the interrelated aspects of the project
are brought into focus for the decision makers.  Since most water quality
planning studies are based on at least a 20 year planning horizon, a knowl-
edge of the impact of major assumptions to the overall success of a project
is invaluable to a decision maker as priorities and objectives change from
the time of project inception to the end of the planning horizon.

     This paper examines the sensitivity of meeting selected water quality
criteria to the assumed status of major project assumptions.  Emphasis is
placed on those project factors which are often classified as assumptions
in a study and which, as demonstrated here, can influence the meeting of
project objectives.

     The background for this  paper is a water quality management plan
developed for the metropolitan areas in the Aburra Valley surrounding
Medellin, Colombia.  As part of an overall study of the valley, a river
model was calibrated, verified, and used to evaluate potential  wastewater
treatment plant systems for the area.  Aspects of the study Included water
quality sampling, the estimation of nonpoint pollution, hydrology, sediment,
water diversions for dilution, and solid waste studies.  The planning
horizon for the sensitivity analysis is the year 2000.

     This paper is based on results of a prefeasibility study.   Simulation
results are used to illustrate the importance of carefully performing a
sensitivity analysis of major project assumptions with respect to the
information the analysis provides to decision makers for the direction of
the project.   Results and conclusions referenced in this paper may change

-------
considerably during subsequent  feasibility and design  phases  of the project.

Description of Project Area

     The Aburra Valley is  located  in  the Andean mountains,  7   north of the
equator at an elevation of 1,500 meters  (4,920 ft)  above  mean sea level
(MSL).  The City of Medellin  and nine other municipalities  comprise most of
the population of the valley.   The city  of Medellin, the  capital of the
Department (state) of Antioquia, is located at the  center of  the valley
(Figure 1) approximately 205  km (127  mi) from the Pacific Ocean and 295 km
(183 mi) from the Caribbean Sea.  It  is  the second  largest  city in Colombia
with a 1980 population of  over  1.2 million inhabitants.   Total  population
in the valley is approximately  2 million.
     The climate  is mild  with  monthly average temperatures  varying only
slightly over the year.   The mean temperature is 22.5°C (72.5°F), with
temperatures ranging  between 13  C (55°F)  and 33°C (91  F).   Average annual
rainfall totals 1,450 millimeters (57 inches).   Average annual  humidity is
70%.  Four seasons are distinguishable from changing rainfall  patterns.
Two relatively dry seasons  occur from December to March and July to August.
Two wet seasons occur from  April to June  and from September to November.
                                                           UNIDAD DE SANEAMIENTO HlDRlCO

                                                           HOYA HIDROGRAFICA DEL RIO M£DELUN
                       Figure 1: Aburra Valley - study area.

-------
     Most of the Aburra Valley is classified as humid subtropical and low
mountain humid forest.  The forest that remains is concentrated at the
upper and lower reaches of the valley.  In the Medellin area, urban zones
cover most of the available usable land, limited only by steep slopes.  The
metropolitan area is an important commercial and industrial center for
textile activity.  The non-urban areas are used for livestock grazing and
agriculture.

     The Rio Medellin flows north through the urban population centers.
The river is the natural drainage for the valley.  It receives flow from
numerous tributaries, the majority having steep slopes in the upstream
sections.  Tributary flows range from a few liters per second in the dry
seasons to several cubic meters per second during the wet seasons.  The Rio
Medellin extends for approximately 100 km (62 miles) from its headwater
near Primavera to its mouth at the confluence with the Rio Grande.  The
river receives 7.7 m /sec (180 MGD) of raw municipal and industrial waste.
Median flow at gage RM-12 (Figure 1) is 25 nr/sec (570 MGD).  Current uses
are severely limited and public health is threatened.

     The upper reaches of the Rio Medellin from its headwater to Itagui,
approximately 20 km (12 mi), flows through a narrow valley and falls
approximately 200 meters (656 ft) in 13 kilometers (8 mi).  The water
quality in this reach is relatively good.   It is downstream of this reach
that municipal discharges become significant.  Below Itagui the valley
widens, the river slope decreases, and the river passes through the heavily
populated and industrialized area of metropolitan Medellin.  The river is
channelized through much of this portion.   The water quality in this
stretch is severely degraded due to the large volumes of industrial,
municipal, and domestic effluents being discharged into the river as shown
by high levels of coliform bacteria, oil  and grease, detergents, and heavy
metals.  The five-day biochemical oxygen demand (BOD5) is very high, reach-
ing 120 mg/1 in the metropolitan area, and the dissolved oxygen (DO)
approaches zero, especially during the dry season.   Below the channelized
reach, the water quality continues to be highly degraded due to industrial
and domestic discharges from Medellin and Bello.   As a result, the DO
concentration remains near zero.   Because of the short travel  times and
high pollution load the organic matter from the Medellin area decomposes
only slightly in the river and water quality remains very poor.

     The characteristics of the river below gage RM-12 (kilometer 40, mile
26) changes radically.   The valley becomes very narrow,  the slope increases,
and water flow is very rapid passing through riffle areas and cascades.
This stretch represents a zone of high reaeration,  as rapids and cascades
oxygenate the water and sewage discharges  are minimized by the small  degree
of human development.   Thus, oxygen levels increase with distance down-
stream but BOD5, nutrients, toxic substances, and coliform bacteria
concentrations are still  elevated.

-------
     Near the mouth of the Rio Medellin, the valley widens and the slope
again decreases.  The Rio Grande and the Rio Santiago join the Medellin to
form the Rio Force and dilute the pollutants of the Rio Medellin.

     Figures 2 through 5 present water quality measurements taken between
March and August 1981 (Dames & Moore, 1981).  Flows were generally above
the average annual levels.  Sampling locations are indicated by the vertical
lines and ranges of measurements by short dashes.  The sampling points are
connected for clarity.

     Figures 2 and 3 indicate the extensive pollution load the river is
currently receiving.  Oxygen levels are recorded below 2 mg/1 over a 20-km
(12 mi) reach of the river.  BODS'concentrations are greater than secondary
effluent levels (30 mg/1) for nearly 40 kilometers (24 mi).  Ammonia
nitrogen and total phosphorus concentrations are well above potential
eutrophication causing levels.

Waste Management Practices in the Aburra Valley

     The city of Medellin and the surrounding municipalities are without
facilities necessary for the treatment of domestic, industrial, or
commercial wastewater.  Although the metropolitan area is served by
sanitary sewers, these discharge directly to the mainstem or to various
tributaries and open ditches draining the area.

     Existing management practices for solid waste disposal also contribute
to the poor quality of the Rio Medellin.  The indiscriminate disposal of
solid waste directly into and on the banks of the river and its tributaries
creates obvious visual and odor problems.  The organic load from floating
waste and the accumulation of debris creates a potential demand on in-
stream oxygen and elevates the concentration of BODS, thereby adding to
the degradation of the river.

Water Quality Criteria

     Quality criteria were established for a range of potential uses of
water from the Rio Medellin.  Criteria were developed for two categories
of pollutants: pollutants not significantly affected by the abatement
program, pollutants significantly affected by the abatement of organic
pollution.  The organic pollution criteria served a dual purpose.  First,
they obviously indicated a concentration of the given pollutant.  Second
(and more important) they served to indicate a level (albeit subjective)
of chemical pollution, originating from industrial and commercial
activities in the valley.  The implementation program is expected to
alleviate increasingly high percentages of pollution from the above two
sources as higher percentages of the population related pollution is
eliminated.

     Water quality parameters included as indicators of organic pollution
were dissolved oxygen, biochemical oxygen demand, ammonia, total phosphorus,
and fecal coliforms.

-------
   URBAN   CALDAS     ITAGUI  M6DELLIN BELLO COPACABANA  GIRARDOTA

   AREAS
        NOTE: Horizontal bars denote
            and maximum samples.
                                       40       50      60

                                         RIVER KILOMETERS
    Figure 2: Rio Medellin mean dissolved oxygen for March - August, 1981.
URBAN CALDAS ITAGl
AREAS [V3 F?\\


I
r
i
r
s
« 60-

20-







^^\







— —.






1
]f




I
/


Jl MEDELLIN
1 L\\\\\V\V





! 	







"\







\







/



BELLO COPACABANA GIRARDOTA BARBOSA
rei r^i ^3 R^J




\
\







\
. \







^^
























~~~ 	 .







^^^
               10      20
        NOTE: Horiionul baridenote mrnimum
            and meximum umplet.
                                       40       &0      60      70

                                          RIVER KILOMETERS
                                                                      80       90
Figure 3:  Rio Medellin mean biochemical oxygen demand for March - August, 1981.

-------
URBAN   CALDAS     ITAGUI  MEDELLIN BELLO COPACABANA GIRARDOTA            BARBOSA
                      ixxxxxxxx^ 15^    k\xi    R\51                 Fx^l
0       10       20

 NOTE Horizonial bars denote mm
     and maximum samples.
                                     40       50      60

                                        RIVER KILOMETERS
 Figure 4: Rio Medellin mean ammonia nitrogen for March - August, 1981.
URBAN   CALPAS
AREAS    [X^
                  ITAGUI  MEDELLIN BELLO COPACABANA  GIRARDOTA
0       10       20

 NOTE: Horizontal b»n denote minimurr
     and maximum samples.
                                      40       50       60

                                        RIVER KILOMETERS
   Figure 5:  Rio Medellin mean total phosphorus for March - August, 1981.

-------
 Selected Abatement Strategy

      An extensive analysis was  performed  by  the  project team (Greeley and
 Hansen, and CCC, 1982) to determine  the cost-effective abatement strategy
 appropriate for the Rio Medellin.  Population, land use, solid waste, and
 river mining conditions were projected to the design year 2000 to estimate
 pollutant loads  The Qual II model was used  to simulate in-stream quality
 resulting from implementation of a given  point source treatment plant
 configuration and selected non-point source  controls.  Preliminary results
 of the study (Greeley and Hansen, and CCC, 1982)  indicate that a configu-
 ration of four treatment plants  (Figure 6) sited  along the modeled portion
 of the Rio Medellin would cost-effectively meet  selected water quality
 criteria.  The proposed configuration consists of two large secondary
 plants and two additional plants defined  as  providing preliminary treatment
 (screening, grit removal, and chlorination).  The interbasin transfer of
 high quality water for the purpose of dilution (20.5 m /sec) is also
 assumed in the treatment configuration.
                SECONDARY TREATMENT PLANT
TREATMENT
PLANT
LOCATIONS
                                      PRELIMINARY TREATMENT PLANT*

LUTION
ATER
PUTS
RIO
MEDELLIN

D = 2.5 cms D
\
0 10
o
= 5 cms
\


20
0




30 40
O
D = 18 cms
\


50 60
0




I I
70 80


R
M
I I
90 100
CITIES
ALONG THE
RIO MEDELLIN
GIRAROOTA
                               COPACABANA
•PRELIMINARY TREATMENT IS DEFINED AS

SCREENING, GRIT REMOVAL AND CHLORINATION
                  Figure 6: Point source selected abatement strategy.

-------
Sensitivity Analysis

     Adapting the calibrated model to year 2000 conditions required many
assumptions and projections to be made which describe the expected "state"
of the environment.  Foremost among these were the assumptions regarding:

     1.   Level of naturally occurring base flow used in the screening
          of alternatives.

     2.   Year 2000 status of the interceptor connection program to
          the major trunk lines.

     3.   Availability of an appreciable quantity of dilution water,
          particularly during drought periods when dilution water is
          most needed.

     4.   Level of solid waste control possible.

     5.   Effects of wet weather  (nonpoint) sources of pollution on
          meeting quality criteria.

     The implementation of a comprehensive plan requires a detailed
knowledge of Jiow the quality of the river responds to each of these
assumptions and which among them are impediments to meeting water quality
criteria.  The potential range of changes in regard to political and
social forces that are possible over a 20-year planning horizon make an
understanding of the sensitivity of the proposed solution to these
assumptions critical during the implementation phase of the comprehensive
plan.

     Sensitivities were developed by modifying either a single assumption
or group of assumptions and using the Qual II model to develop the
associated water quality response (Dames & Moore, 1982e).  A "baseline"
simulation was used as a standard to compare the relative sensitivities.
The "baseline" simulation represented a most likely set of assumptions
regarding the development of events influencing water quality.  For the
baseline simulation, a treatment plant configuration of four plants was
selected—two secondary plants located at Envigado and Bello, and two
plants providing preliminary treatment at Girardota and Barbosa.  The two
secondary plants were assumed to have 92% treatment efficiency for BOD
removal.

     The status of the five major assumptions in the Baseline simulation
are:

     1.   A flow likely to be exceeded 75 percent of the time was
          selected as the naturally occurring portion of the flow
          used in the baseline simulation.

-------
     2.   Seventy-five percent of the domestic, industrial, and
          commercial wastewater in the valley is collected
          and treated, with 95% collection and treatment of
          industrial waste from several  major industries in the
          valley.

     3.   High quality dilution water is available to the Rio Medellin.

     4.   Implementation of a solid waste management program
          eliminates the pollutant loadings attributed to unmitigated
          solid waste dumping.

     5.   Nonpoint sources of pollution are minimal  under baseline
          conditions.

    In the following discussion each project assumption is examined
separately.  The basis for each is outlined and the project details
surrounding the assumption are discussed, along with the sensitivity of
predicted water quality to a modification of the assumption.


     Assumption 1:  A flow that is likely to be exceeded 75 percent of
                    the time was selected as the naturally occurring
                    portion of the flow used in the baseline simulation.


     The selection of a design flow for use with the Qual II model in the
evaluation of the water quality impacts of abatement programs is a major
factor basic to the results of much of the analysis.  Currently, no local
institutional or legal constraints exist in Colombia to guide the selection
of this flow.

     An analysis of extreme events, the usual  procedure for determining  a
design low flow, was not possible due to the short length of records avail-
able.  The period of record was 2-10 years, with an average of 6 years at
each gage (Dames & Moore, 1982c).  Therefore,  for purposes of screening
abatement strategies a flow exceeded 75 percent of the time was established
for the analysis.  The exceedance frequency was selected in recognition  of
the severity of existing conditions and uncertainties in the data.

     To demonstrate the sensitivity of this assumption on predicted water
quality, the Qual II inputs were modified to reflect a more severe flow
condition, the average annual 14-consecutive-day low flow.  The effect of
modifying the design is shown in Figure 7.  The reduced flow corresponds to
an exceedance frequency of 90 percent.  The minimal  impact on oxygen
reserves of a lower flow is due to the large quantities of dilution water
and sanitary contributions to the total  flow in comparison to the smaller
"naturally" occurring base flow portion.  The  predicted maximum DO sag
level is less than 0.5 mg/1 more severe.  The  predicted in-stream BOD5
                                   10

-------
                                                        BASELINE SIMULATION (75%)*
                 20      30
                               40      50      60

                                 RIVER KILOMETERS
                                      LOW FLOW (90%)*
                                     BASELINE SIMULATION (75%)*
        MINIMUM LIMIT
                               40      50      60

                                 RIVER KILOMETERS
   •EXCEEDANCE FREQUENCY
Figure 7: Effect of changes in flow on biochemical oxygen demand and dissolved
           oxygen concentrations.
                                      n

-------
concentrations are only minimally affected.   The maximum and minimum limits
shown on Figure 7 are those limits established to minimize odor problems.

     With a knowledge of local hydrology, coupled with the low sensitivity
of water quality to changes in the base portion of the design flow, it be-
comes evident that further work in establishing a more restrictive or
reliable design flow should not be a high priority concern in the future
planning process.  Water quality is not particularly sensitive to changes
in this factor.
     Assumption 2:  Seventy-five percent of the domestic,  industrial
                    and commercial wastewater generated in the valley
                    is collected and treated.
     As indicated earlier, although there are no treatment facilities in
the Aburra Valley, portions of Medellin and the surrounding municipalities
are sewered.  In 1957 a plan was completed for construction of a sanitary
collection and treatment system for the City of Medellin (Greeley and
Hansen, 1957).  Presently, approximately 50 percent of the 152 kilometers
of the major interceptors (94 mi) originally planned are constructed.  A
major problem is that many of the trunks and smaller domestic lines are not
connected to the main interceptors and necessary manholes and cleanout
points are not built.

     A large-scale retrofit program for the system connections, manholes,
overflow outlets, etc., is necessary as part of the plan for 75 percent
wastewater collection in the valley.  A point of concern is whether 75 per-
cent collection is feasible from an engineering and cost standpoint.   The
topography of the region, as well as the status of existing housing
conditions, precludes collection of an uncertain percentage of waste flows.
Many of the dwellings in the valley, especially in the upland areas,  are
without plumbing or public water supply.  Collection of the wastewater
generated from these houses would be difficult to achieve.

     The results of changing the collection percentage from 75 to 50 per-
cent is shown in Figure 8.  The ability to meet minimum water quality
criteria is strongly influenced by the amount of sewage that can be
collected and treated before release.   The oxygen sag drops to a minimum
2.0 mg/1 while the BOD5 increases to a maximum of 48 mg/1.   The BOD5
maximum and DO minimum occur near the center of the most populated area.

     A major public works program to connect 75 percent of the wastewater
flows to the collection/treatment system must be recognized as a critical
factor in achieving water quality goals.  The indicated sensitivity of the
predicted water quality would lead a decision maker to recognize that the
program to connect interceptors to the main trunk lines is  a major factor
in the implementation of a comprehensive plan.
                                   12

-------
                                                   50% COLLECTION

                                                     BASELINE SIMULATION (75%)*
                                   40      50      60

                                    RIVER KILOMETERS
K
2
            MINIMUM LIMIT
                                           BASELINE SIMULATION (75%)

                                         50% COLLECTION
                                   40      50      60

                                    RIVER KILOMETERS
       •WASTEWATER COLLECTION
  Figure 8: Effect of wastewater collection levels on biochemical oxygen demand
             and dissolved oxygen concentrations.
                                       13

-------
    Assumption 3:  High quality water is available for dilution of
                   waste flows entering the Rio Medellin during
                   critical periods.


     Current planning calls for the diversion of water from the neighbor-
ing Rio  Grande watershed to the Aburra Valley for hydroelectric power,
water supply, and dilution purposes.   The proposed diversion includes the
building of gravity tunnels to the municipalities of Bello and Girardota.
The plan also includes construction of two hydroelectric power plants as
well as a water treatment plant.  An average of 35 m /sec (1236 ft3/sec)
would be diverted by the year 2000.  Although approximately 12 m /sec (424
ft3/sec) would-be used to meet increasing water demand, the remaining 23
m /sec (812 ft /sec) would be discharged into the river as dilution water
after power generation.  The baseline simulations accounted3for the above
by establishing inflows to the Rio Medellin of 5.0 and 18 m /sec (176.6 and
636 ft3/sec) at river kilometers 30 and 54 (miles 19 and 34),  respectively.
An additional 2.5 m /sec (88 ft /sec) will be available from another source
at kilometer 20 (mile 12).

     Since the availability of dilution water is dependent upon completion
of the rather complex proposed diversion plan, the ability to  meet the
initial project assumption is again important to the decision  makers.  To
test the sensitivity of water quality to the assumption of no  dilution
water, all dilution flows were eliminated from the simulation.  The results
of this modification are shown for BOD5 and DO in Figure 9, along with the
predicted baseline concentrations.  The BOD5 concentrations remain
elevated through the City of Medellin and only begin to decline downstream
as the organic load is diluted, river flows increase, and BOD5 inputs are
reduced.  Upstream of river kilometer 20 (mile 12), predicted  water quality
is the same for the baseline and modified assumptions simulations.   The
simulation of DO is shown to be very sensitive to the availability of
dilution water.  Without added dilution water a major zone of  anaerobic
conditions is likely to develop, even after construction of the assumed
collection and treatment facilities.

     The importance of dilution water to meeting project water quality
criteria is indicated to be very large, particularly near the
population center of Medellin, river kilometers 15 to 35 (miles 9 to 21).
Likewise the importance of the availability of water for dilution to the
success of the project is also large.


     Assumption 4:  A solid waste management plan is assumed to be
                    implemented thereby eliminating solid waste as
                    a cause of water quality degradation.
                                   14

-------
                                                   NO DILUTION WATER

                                                      BASELINE SIMULATION
            10      20
                                  40      50      60

                                   RIVER KILOMETERS
                   20      30
                                  40      SO      60

                                   RIVER KILOMETERS
Figure 9: Effect of dilution water on biochemical oxygen demand and dissolved
          oxygen concentrations.
                                      15

-------
     In 1980, the total quantity of solid waste produced in the Aburra
Valley was estimated to be approximately 460,000 metric tons (Grandjean,
1980).  The collection, treatment, recycling, and disposal of this solid
waste is a complex problem in Medellin and the surrounding municipalities.
Although a significant portion eventually reaches a controlled landfill or
the composting plant for ultimate disposal, in many areas of the valley the
location and methods of disposal are determined by convenience.  A common
practice is to dispose of solid waste on the banks of the Rio Medellin or
its tributaries.  Eventually, disposal sites adjacent to water courses
extend into the channel and waste is carried downstream during high flow
periods, causing widespread visual and odor problems and degrading in-
stream water quality.

     Implementation of a solid waste management program, independent of a
wastewater abatement program, is intended for Medellin.  For planning
purposes it was assumed that implementation of such a program would
eliminate solid waste as a cause of water quality degradation.   Therefore,
the estimated contribution of solid waste to the total pollution load to
the river from all sources (Dames & Moore, 1982b) was excluded as input to
the QUAL II simulation for the baseline condition.

     The above decision assumes a significant change in solid waste
disposal practices in the valley.  Solid waste recycling at all levels, at
the street receptacle, on the collection trucks, and at numerous disposal
sites, provides income to thousands of people throughout the valley.
Changing methods of disposal may have serious socioeconomic impacts on
many people, and therefore implementation of a program to completely
eliminate solid waste as a source of water quality degradation may be
quite difficult to achieve.

     The sensitivity of water quality to the assumptions regarding solid
waste is shown in Figure 10.  Solid waste is assumed to enter the river as
leachate from solid waste sites (a relatively minor source of pollution),
through direct dumping, and by entrainment with the runoff.   Figure 10 was
developed assuming a storm flow condition existed.  The naturally occurring
base flow of the river was increased to simulate an "average" storm.   It is
recognized that Qual II is a hydraulically steady state model and does not
simulate the time varying nature of runoff from a watershed.   However, the
model does provide a crude screening suitable for the sensitivity analysis
presented here.   Elevated BOD5 concentrations are at almost twice the base-
line concentration along the entire modeled portion of the river.   The
dissolved oxygen sag is approximately 1  mg/1  more severe at the point of
minimum DO concentration.   Recovery of oxygen to concentrations similar to
the baseline condition occurs downstream of kilometer 51,  (mile 32)  because
of the addition of 18 m3/sec (636 ft3/sec) of water for dilution.

     Meeting water quality criteria is shown to be very dependent on  an
effective solid waste management program.   The reliance of the  predicted
pollution abatement to the assumption regarding the solid  waste management
                                   16

-------
                                                     NONPOINT POLLUTION WITH SOLID WASTE
                                                        NONPOINT POLLUTION
                                                          BASELINE SIMULATION
             10      20
                                  40      50
                                    RIVER KILOMETERS
X
o
           MINIMUM LIMIT
      NONPOINT POLLUTION WITH SOLID WASTE
    BASELINE SIMULATION
. NONPOINT POLLUTION	
                    20      30
                                  40      50      60
                                    RIVER KILOMETERS
  Figure 10: Effect of nonpoint sources and solid waste on biochemical oxygen demand
              and dissolved oxygen concentrations.
                                           17

-------
program is clearly indicated, particularly with respect to organic
pollutant concentrations.  Future implementation of a comprehensive plan
should also rank a solid waste management plan as necessary to achieve
water quality improvements.


     Assumption 5:  The effect of nonpoint source pollution on water
                    quality.


     By definition here, nonpoint pollution occurs only during rainstorm
events when accumulated pollutants on urban and rural land surfaces are
carried to receiving streams.  Hence, this source of pollution is largely
dependent on rainfall patterns, topography, and land use in the region.
Since the selected design flow is a dry weather condition, nonpoint
pollutant loadings would be minimal and were, therefore, excluded as input
to Qual II in selection of an appropriate abatement program.

     An important concern to planners, however, is whether nonpoint
pollution in combination with point discharges would be significant enough
that water quality criteria would be violated during wet weather.

     To determine the nonpoint contribution to stream pollution, an
estimate was made of diffused, runoff-associated pollutant loads (Dames &
Moore, 1982d).   Figure 10 presents the results of the simulation.
Simulated BOD5 concentrations were slightly higher in wet weather relative
to the baseline condition.  Below river kilometer 40 (mile 24), concen-
trations were similar to baseline conditions because of the large
quantities of water available for dilution.  Dissolved oxygen levels,  how-
ever, were high throughout the stream.  The sag during wet weather was more
than 1 mg/1 less severe when compared to baseline conditions.

     Interestingly, wet weather conditions do not seem to affect seriously
the opportunity to achieve water quality criteria.   This is in part due to
the level of the criteria as indicated on Figure 10 and to the averaging
of flows and pollutant loadings required by the methodology followed.
First flush effects of urban areas are not simulated.  Results indicate
that for most water quality parameters, nonpoint pollution would have
little impact on achieving water quality criteria at the levels indicated.


      Assumption 6:  Cumulative Impacts


     In the above cases, changes to the baseline conditions included
variations in flow, percent collection, water for dilution, nonpoint
source pollutants, and solid waste pollutants.   In a subsequent step,  the
cumulative effect of the assumptions impacting water quality was also
evaluated by modifying several assumptions simultaneously.   These changes
                                   18

-------
                      PEAK-APPROX. 130BOD5 (MILLIGRAMS/LITER)
                                                 CUMULATIVE IMPACTS
                                       (LOW FLOW, NO DILUTION WATER, WITH SOLID WASTE)

                                          NONPOINT POLLUTION WITH SOLID WASTE

                                           50% COLLECTION

                                              NO DILUTION WATER
                                         40       50       SO

                                           RIVER KILOMETERS
           BASELINE SIMULATION
          AND NO DILUTION WATER
CC
2
                                      1 CUMULATIVE IMPACTS
                           -(LOW FLOW, NO DILUTION WATER, WITH SOLID WASTE)'
                                         40       50       60

                                           RIVER KILOMETERS
       Figure  11: Effects of cumulative impacts on biochemical oxygen demand and
                   dissolved oxygen concentrations.
                                                 19

-------
were chosen to reflect a potential "worst case" picture of resulting water
quality.  The simulation was based on the following assumptions, modified
from baseline conditions:


     Cumulative Impacts:  The simulated flow was changed to the lower flow
                          as described in Assumption I. All dilution
                          water inputs were eliminated.  No abatement to
                          the solid waste dumping was included.


Results of this simulation are shown in Figure 11 for DO and BODS.  The
results more clearly demonstrate the importance of both dilution water
availability and solid waste abatement.  A DO sag to zero occurs for
approximately 50 kilometers.

     BODS demands exceeding 30 mg/1 are predicted for approximately 88
kilometers of the 100-kilometer Rio Medellin.  Thirty milligrams per liter
of BODS was the maximum permissible concentration to meet minimum water
quality criteria.  This concentration was exceeded except in the most up-
stream reach.   A peak of 130 mg/1 resulted in the City of Medellin, where
organic loading contributions to the river are the greatest.   These results
again illustrate the importance of recognizing the relative importance of
the major assumptions to achieving water quality planning objectives.
Summary

     When used correctly, water quality models can enhance the likelihood
of.the success of comprehensive planning efforts.  The use of a model to
evaluate the effect on predicted water quality of major project assumptions
provides:

     1)  Understanding of the relative importance of each project
         assumption to the overall abatement program.

     2)  Basis for informed discussion on the direction the planning
         process should follow.

     An analysis of project assumptions for Medellin would indicate that in
addition to the extensive point source controls recommended, decision
makers should recognize that the implementation of a sewer connection pro-
gram, the abatement of solid waste, and the provision of dilution water are
very important to maintaining quality criteria during critical periods and
thus to the success of the project as currently planned.   Also, in this
case, the sensitivity of projected water quality to nonpoint sources and to
lower flows than used in the screening of abatement alternatives is small.
                                   20

-------
          Acknowledgements

               The  number of people involved in this study and their valuable contri-
          butions are  simply too numerous to mention here.  Their contributions to the
          analysis  presented in this paper are gratefully acknowledged.  Project
          directors from the organizations involved in the study were Messrs. Robert
          Zimmerman from Greeley and Hansen, Jaime Rodas from Compania Colombiana
          de Consultores, Richard C. Tucker from Dames & Moore, and Dr. Alonso
          Palacios  from Empresas Publicas de Medellin. The direction provided the
          project by their participation shaped the project and helped it successfully
          to meet its  objectives.
                                           REFERENCES


          Dames  & Moore, December 1981, "Existing Environmental Conditions,"  Pollution
              Control of the Medellin River and its Tributaries - Memorandum 34-1.

          Dames  & Moore, April 1982.  "Solid Uaste Pollution Assessment," Pollution
              Control of the Medellin River and its Tributaries - Memorandum 56-1.

          Dames  & Moore, May 1982.  "Hydro!ogic Studies," Pollution Control of the
              Medellin River and its Tributaries - Memorandum 59-1.

          Dames  & Moore, May 1982.  "Nonpoint Pollution Assessment," Pollution Control
              of the Medellin River and its Tributaries - Memorandum 25-1.

          Dames  & Moore, July 1982.  "Develop Initial Treatment Arrays," Pollution
              Control of the Medellin River and its Tributaries - Memorandum 22-1.

          EPA Environmental Research Laboratory, 1977.  User's Manual for Stream
          Quality Model (Qual II).

          EPA, Environmental Research Laboratory, 1977.  Computer Program
              Documentation for Stream Quality Model  (Qual II).

          Grandjean, 0, 1980, Estudio Sobre Recoleccion, Trataments y Dispocition
              Final de Desectos Solidos en al Area Metropolitano y Oriente Crecano,
              Ports 1 y 2.  Departmento Administraro de Planeacion y Direccion
              Desarollo Metropolitano, Governacion de Antioquia.

          Greeley and Hansen, 1957.  "Informe Sobre Recoleccion y Disposicion de las
              Aquas Negras de Medellin.  Medellin, EE. PP. M.

          Greeley and Hansen and Compania, Colombiana de Consultores, Ltda.,  July 1982.
              "Initial River Model and Treatment Model Analysis of Treatment
              Configurations and Selection of Treatment Configurations for Further
              Study," Pollution Control of the Medellin River and its Tributaries,
              Memorandum 28-1.

          Loucks, Daniel P. et al. Water Resources System Planning and Analysis,
              Prentice Hall, 1981.

          Orozco, A., 1978.  Solid Waste in Medellin and its Metropolitan Area.

          Revista Empresas Publicas de Medellin, Monografia Del Rio Medellin.  Vol. 3,
              Nos. 3 y 4, Julio/Deciembre 1981.  ISSN 0120-1239.



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

                                              21

-------
             Application of the STORM Model to Design Problems

             in Singapore and in Kaohsiung, Republic of China


                                     by


                             Brendan M. Harley

                         Camp Dresser & McKee Inc.

                        Boston, Massachusetts  02108
ABSTRACT

This paper presents two examples of the application of modified versions of
the STORM model to design situations overseas.   One of these, involved sim-
ulation of a raw water supply system where flow is captured via diversions
from the storm drainage system.  The second application used the model to
evaluate cost-effective CSO control measures in a situation where no separ-
ation of sanitary and storm sewers exist.


INTRODUCTION

Problems in managing urban runoff and combined  sewer overflow in rapidly ex-
panding cities in developing countries are generally quite major and, in
some cases, address urgent public health hazards with immediate needs for
solutions.  Use of simulation models to assist  in developing design criteria
is still quite rare, although a trend towards use of such models is develop-
ing.

Recently, COM has been involved in two overseas projects which presented the
opportunity to apply modified versions of the STORM model  to quite different
situations.  The first, in Singapore, involved  a study where urban stortnwater
runoff from "new town" areas is to be captured  and treated for use to augment
the island's water supply.  The STORM model  was used to evaluate the basin
runoff, to size the diversion/storage/pumping facilities which are spread
throughout the area, and to help determine the  effective "safe yield" from
the proposed project.  This project is now under construction.

The second application of the STORM model  was in the development of design
strategies to control heavily contaminated combined sewer overflows to the
Jen Ai River in the City of Kaohsiung, Republic of China.   This city of about
one million people  has essentially no sanitary sewerage system at all; all
sanitary wastewater together with stormwater flows via small open ditches and
channels to the river which is presently totally anoxic.  The design study
was for a series of major combined sewer overflow structures which, in con-


                                    22

-------
junction with a single trunk sewer, will reduce BOD loadings on the Jen Ai
River by over 90 percent.  Effective use was made of the ability of the
STORM model to simulate the effect of storage and local  treatment (simple
detention, swirl separators, etc.) at each of the control  facilities.


SELETAR PROJECT - SINGAPORE

BACKGROUND

The expansion of reliable water supply systems is a major requirement  in
areas of the world which experience very rapid population growth.  Some of
these developing countries lie in regions of the world having limited  rain-
fall, many lie in areas of abundant rainfall but possess very limited
capture/storage/transmission capability; in nearly all cases, large capital
investments have to be undertaken to meet the water supply needs.

The island nation of Singapore is handling the population growth demands
through the development of self-contained "new towns" containing both  resi-
dential and industrial areas; these "new towns" incorporate separate storm
and sewer systems designed to good modern practice levels.  Each "new  town"
is, in effect, a small city with a population of 200,000-400,000.  This
scheme utilizes runoff from such "new towns" areas to augment the water
supply requirements.  The scheme which is presently in the final design and
tender stages involves the following elements:

o  Diversions from main storm drains to a series of storage ponds located
   throughout the "new towns."  Diversion structures include both large
   drop-inlets and tightly controlled bascule-gate regulators.

o  Pumpage from these ponds to main reservoirs which also accept gravity
   drainage from sections of the "new towns."

o  Blending of this urban runoff with runoff from undeveloped watersheds.

The development of this scheme, which has a first stage capacity of 30 imgd
and an ultimate safe yield of 60 imgd, included the application of many
hydro!ogic analysis techniques including:

o  Generation of long term rainfall records, using disaggregation techniques,
   for use in yield studies.

o  Generations of design storm events, both for spillway design events and
   for the more frequent events where the diversion structures have to oper-
   ate effectively.

o- Use of a modified version of the STORM model to evaluate the runoff/diver-
   sion/pumping operations.

o  Application of the MITCAT model to simulate design event operation  of
   major components of the system.
                                     23

-------
o  Development of a system simulation model which emulated the operation of
   the overall collection/storage/treatment system and which was used to:

   -  evaluate overall system yield

   -  set pump station design rates

   -  develop the integrated operation rules for the multiple reservoirs in
      the system

o  Evaluation of expected water quality in the various components of the
   system.

o  Application of air-bubble systems to control  undersirable stratification
   in the reservoirs.

The overall scope of the project is to develop a water supply to augment the
present supply for municipal uses.  Treatment of this water to standards
suitable for municipal uses and its discharge at service pressure into the
distribution system are included in the project.

A shcematic layout of the water supply system is shown in Figure 1.  The
reservoirs will receive and hold the runoff to equalize the sometimes high
and very intermittent runoff rates to meet the steady demand on the system
for treated water.

The Seletar Reservoir will be developed by the construction of an earth em-
bankment dam with protecting sluice way across the mouth of the estuary.
The reservoir created by the dam will collect runoff which originates on its
tributary watershed and runoff from collecting systems located in adjacent
catchments.  These adjacent catchments are lands which are now being devel-
oped into "new towns" and nearby rural areas.  A large sand quarry pit will
form the Bedok Reservoir.  It will primarily receive runoff collected from
nearby urban land.  Runoff from these catchments will be collected in ponds
or lagoons and conveyed to the Reservoir by pumps and pipelines.   The water
collected in the Seletar Reservoir will be pumped either to Bedok Reservoir
or directly to the proposed treatment plant; the pipeline will be about 16km
long.


STORMWATER RUNOFF/CAPTURE ANALYSIS

One of the unique features of this scheme is the use of urban runoff as a
source of water supply.  The drainage system consists of a rather complex
group of catchment areas;  23 catchment basins are involved.   For twelve
of these, flow by gravity into the reservoirs is infeasible and a capture/
transfer scheme involving pumping is utilized.  The response of this system
to rainfall events was evaluated so that overall system yield could be esti-
mated, and the capture/transfer sites could be properly sized.
                                     24

-------
I	
                             U-l
                            oc
                            u.
                            o
                            UJ
                            a:
                     OAQ
             25

-------
STQRMWATER RUNOFF ANALYSIS

The model chosen to simulate the stormwater runoff characteristics of the
study area is STORM.  In this study, STORM was used to compute surface run-
off from precipitation, based on the hydrologic characteristics of the drain-
age basin.

In order to systematically and consistently apply the STORM Model  to each
basin, the standard procedure involving the following three steps  was employ-
ed:

1.  Identification of the most appropriate land use categories for the study
    area.

2.  Computation, for each drainage basin,  of the area in each land use cate-
    gory, and

3.  Estimation of the STORM Model parameters for each land use.

Within the study area, there are a variety of existing or planned  land uses.
To meet the requirements of this study, the usual six general land use cat-
egories were identified as follows:  major roads, industrial, institutional,
residential, open space, and natural.  Some variations in the development
levels and thus hydrologic features associated with each land use  category
exist.  However, it was established that the use of six standard land uses
adequately represents the hydrologic conditions.   Given the rainfall char-
acteristics of the study area and the STORM parameters, the average annual
percent runoff was established for each of six land uses and using the
standard model parameters for each land use, the STORM Model was systemati-
cally employed to simulate basin runoff conditions.

For a number of drainage basins within the study area, stormwater  runoff
cannot be transferred to system reservoirs by natural means.  Such cases re-
quire a pumping scheme to lift flows to a  reservoir.   Rainfall  in  the area
is characterized by relatively high-intensity short-duration storm events.
Due to the fairly large amount of impervious surface  and the short response
time of drainage basins in the study area, this rainfall is transformed into
runoff with no large attenuations.  This particular condition requires that
any runoff transfer scheme involving pumping to also  include a stormwater
storage facility to retain a portion of the runoff volume for pumping after
the storm has subsided.

A schematic of a typical storage and pumping facility is illustrated in
Figure 2.  In the case shown, stormwater flowing  along the main drainage
channel is diverted to the storage unit and from there is pumped to the
reservoir.  When flows exceed the pumping  rate, storage is used to contain
the excess.  When flows are less than the  pumping rate, the available pumping
rate is used to reduce the storage level.   If the storage capacity is ex-
ceeded, any excess flow is not diverted to the storage unit.

It is important to note that Figure 2 illustrates one possible storage
scheme, which is typically referred to as  "off-line"  storage.  An  alternative


                                     26

-------
Drainage
Channel
                 Diversion
                  Point
                                FIGURE   2

         SCHEMATIC  OF  STORMWATER STORAGE AND PUMPING FACILITIES
                                 27

-------
involves using the main drainage channel  as an impoundment for "in-line"
storage; pumping occurs directly from the channel.   Performance is based
only on the storage capacity (i.e., volume) irrespective of its location.

An additional  component of a stormwater transfer scheme involves the diver-
sion point itself.  The facility must include a means of selectively divert-
ing only a portion of the main channel  flow for the reasons as follows:
1) during low-flow periods a bypass may be desired  based on water quality
conditions, and 2) there may be a maximum hydraulic capacity where excess
flow cannot be diverted.

The stormwater transfer scheme described  above has  a number of basic design
parameters as follows:

1.  For the channel diversion, there is a low-flow  bypass and a high-flow
    capacity,

2.  The storage unit is parameterized by  its volume capacity, and

3.  The pumping unit has a hydraulic capacity.

Each of these features influences the ultimate capture achieved by the facil-
ity.  A primary objective of the study was to define the relationships be-
tween the design parameters (such as storage volumes) and capture rates.
Basic design questions addressed include  the following:   how much capture is
lost to low-flow bypass, what limits do the diversion capacity place on
capture, and what are the tradeoffs between pumping and storage?

To evaluate the above methods for each  basin in the study area, the DIVERT
module for STORM was developed.   DIVERT is intended to be a preliminary  de-
sign tool to assess the hydraulic performance of a  large number of process
configurations.

Storage-treatment was included in the model due to  the anticipated quality
of stormwater runoff and potential  need for treatment.   The DIVERT model was
developed primarily to evaluate capture achieved in schemes such as that in
this project;  however, it has a  number of very general  capabilities.

A schematic representation of the DIVERT  model  hydraulic components is shown
as Figure 3.  System computations are performed on  an hourly time step and
are typically carried out for a number  of years.  Basic  model  input is an
hourly time series of runoff generated  by STORM,  and model  output is an
hourly time series of system flows,  As shown in Figure  3,  there are six
basic system flows.

Hourly output time series are processed into relevant summaries (i.e., daily,
monthly, yearly, and simulation totals) for use in  assessing system perform-
ance.

The DIVERT model was used extensively in  the development of the sizing of
the various diversion/pumping alternatives considered.   A typical  output
from the simulation studies is the iso-capture plot shown in Figure 4 where

                                     28

-------
SELECTIVE
DIVERSION
 BLOCK
"TREATMENT/
  STORAGE
   BLOCK
    PUMP/
   STORAGE
    BLOCK
                   STORJKATER RUNOFF (from STORM)

                                     "RUNOr11
                 r
    Model Output
                                                                  "OVFLO"
                           FLOU TO RESERVOIR
                         FIGURE    3

          SCHEMATIC  DIAGRAM OF  DIVERSION  MODEL

                                  29

-------
         10      20
                10      15     20      25      30      35

                     PUMPING CAPACITY, P • Q/{A F) (m3/hour per hectare)
FIGURE  4.  CAPTURE NOMOGRAPH, ISO-CAPTURE AS  A FUNCTION OF PUMPING
AND  STORAGE CAPACITY PER  UNIT EFFECTIVE AREA
                                  30

-------
the capture effectiveness of a diversion site is expressed in terms of stor-
age pond volume and pumping rates.  Such a plot, combined with the relevant
cost data, can then be used to select both the optimum volume/pump capacity
at a given site, and also to allocate capture rates between the many sites
in use in this scheme.  A number of observations can be made from the data
shown in Figure 4.

1.  Overall, the spacing between iso-capture contours steadily increases,
    indicating an increasing marginal capacity required for additional
    capture.

2.  Beyond 50 to 60 percent capture, the marginal  capacity requirements be-
    come relatively large.

3.  Along the vertical axis (i.e., constant storage volume), there is an
    abrupt increase in the marginal pump requirements, and beyond about 3
    to 5 m3/hour per hectare very large additional pump capacity is required
    for further capture.

These figures are specific to hydrologic conditions for this scheme; similar
figures can be readily estimated for other sites.
SELECTIVE DIVERSION

Selective diversion potentially consists of two components - a low-flow by-
pass and high-flow diversion capacity.  It is assumed that the bypass must
be satisfied before diversion begins and further that a limited hydraulic
capacity at the diversion structure exists such that excessive flows cannot
be routed to the pump/storage unit.  Given this condition, the STORM/DIVERT
model was used to determine the relationship between the quantity of runoff
diverted and the diversion structure parameters.  The following general  ob-
servations can be made about the results obtained during this study:

1.  Overall, there is a steady decrease in the sensitivity of capture to in-
    creases in bypass and diversion capacity.

2.  For relatively low bypasses there is a vary rapid loss in capture ef-
    fectiveness.

3.  Beyond a certain maximum diversion capacity very little additional
    capture is achieved.

Bypass will be used primarily to exclude low-flows which may contain un-
acceptable water quality constituents.  From a capture point of view, re-
sults indicate that a relatively large amount of runoff can potentially be
lost via bypass flows, and thus only the minimum required to satisfy water
quality requirements should be excluded.  These are flows which are easily
captured from a hydraulics point of view and should not be bypassed unless
clearly necessary from a water quality perspective.
                                     31

-------
HATER QUALITY ASPECTS

Numerous water quality issues are important to the implementation of the
proposed scheme.  These issues are primarily stormwater runoff quality load-
ings and salinity and TDS levels in the reservoirs.   Stormwater runoff qual-
ity is of importance since the water supply system consists primarily of
this urban runoff; these flows are the main source of nutrients to the res-
ervoirs.

In general, storm runoff is characterized by relatively high pollutant load-
ings; these pollutants exist and are washed off from both developed and un-
developed land surfaces.  In this particular project, increased emphasis was
placed on runoff quality estimates because of the unusual nature of the
scheme as a whole.  A relatively straightforward approach was taken with a
somewhat modified version of the STORM model  being used in the simulation of
runoff quality loads.  Available local historic hourly rainfall records were
used to drive the STORM simulations, with a three year period of record be-
ing  found sufficient to provide reasonable estimates of long term averages;
generally a ten year hourly record was available for use.  An important set
of parameters in modeling runoff quality is the pollutant accumulation and
washoff rates for the specific area being studied.   Ideally, sampling data
from the area would be used to calibrate these parameters of the simulation
model.  Although specific data of this nature was obviously not available
from the yet-to-be-developed "new towns," some data  was available for nearby
similar developments.  This data, coupled with values available in published
literature was sufficient to enable the team to estimate the long term aver-
age loadings to Total Suspended Solids, Nitrates, Phosphates and Biological
Oxygen Demand resulting from the stormwater runoff.   Sensitivity studies
were undertaken to determine the impact of parameter uncertainty on the
loading estimates; the design of the overall  system  is sufficiently robust
to .adequately handle expected pollutant variations.

The modifications of the STORM model were primarily  to enable the model  to:

1.  Simulate the transfer of pollutants through the  diversion components in-
    cluding the actual diverters, settling of solids in the small  ponds, and
    the pump operations.

2.  Evaluate the operation of swirl concentrators which were considered for
    use at the diversion sites.  These were ultimately not recommended for
    use.

The runoff from the catchment areas is captured and  stored in the  reservoirs
prior to treatment.  The benthos deposits in the reservoirs are also a major
source of contaminant loadings in the system; in previous similar schemes
they have been a significant operational  nuisance.   It is expected that
phosphate, chloride and sulphates will be the contaminants of most impor-
tance in this scheme.
                                    32

-------
SUMMARY

This application required integrated use of many hydrologic modeling techni-
ques in the development of a highly complex system.  Their use in this pro-
ject permitted significant increases in the design "safe yield" of the sys-
tem and enabled the design team to evaluate the effectiveness of the many
quantity and quality control options which have to be considered during de-
sign development.  Problems do exist in the use of urban runoff for supply
systems; however, in situations where well-designed separate storm drain
systems are available, the runoff can be effectively used as a supply source.
JEN AI RIVER - KAOHSIUNG, REPUBLIC OF CHINA

BACKGROUND

Kaohsiung is a city of about one million people Icoated on the southwest of
the Island of Taiwan.

There are now essentially no sanitary sewers in Kaohsiung.  All  sanitary
wastewater, together with stormwater, flows via combined sewers  and open
channels to the Jen Ai River, a narrow, poorly-flushed tidal  estuary.   The
resulting very heavy loadings of BOD have rendered the Jen Ai  anoxic;  the
river is both odorous and an eyesore due to floatables.

The city government is presently undertaking a project to radically improve
the water quality in the Jen Ai.  This project has several major components
as follows:

1.  Build a trunk sewer of sufficient capacity to intercept all  the dry
    weather flow (DWF) and some of the wet weather flow at each  of 3
    principal tributary channels or sewers.

2.  Remove the majority of the anoxic sediments from the bottom  of the Jen
    Ai by dredging.

3.  Implement other means to maintain an improved water quality  in the river.

At this time, a trunk sewer has been constructed to intercept  combined storm
and sanitary sewage from 8 catchment areas, 6 on the east bank of the  Jen Ai
River, and 2 on the west bank (see Figure 5).  The area drained  by the 8
basins is about 5400 ha.  The total dry weather flow from the  8  areas  is 3
cms, and the total dry weather BOD load is 56000 kg/day.
                                     33

-------
<•
*\
     A      "1  I  Vx<-
     .»   s     V  •   / N •  """*> "^<.
  X   -•;..   'IVAV^

,%x:.  ^^.^
f'-^^XO';
                                  ~k
                                             c-'-cr
                                                        s
                                                        ac
                                                            CD
                                                            H-H
                                                            o:
                                                            Q-
                                                            UJ
                                                            O
                                                            Q-
                                                            
-------
The total design capacity of the trunk sewer and the pump station at its
downstream end is 6.5 cms, or about twice the combined total  dry weather
flow of the 8 areas.  Thus all dry weather flow and some storm flow will be
intercepted, but stormwater will still overflow to the river when flows ex-
ceed the trunk sewer capacity.

The analytic work described here was undertaken to evaluate the effective-
ness of the proposed scheme in terms of improvements in D.O.  in the Jan Ai
River, and to assist in the sizing of the diversion structure.  On the
larger tributaries, the provision of capacity to store significant quanti-
ties of stormwater, until the water can be drained to the sewer following
the storm, greatly reduces the frequency and magnitude of storm overflows,
and the mass of BOD discharged in overflows.

On the smaller tributaries with little or no storage capacity, where much of
the stormwater will have to overflow to the river, the use of swirl concen-
trators to remove TSS and BOD from overflow water before its discharge to
the river was evaluated.

As will be noted, the plan is to divert a significant fraction of the tota1
BOD loads from the overall watershed through use of quite a small intercept-
or, with essentially the whole drainage system diverted through it during
dry weather conditions.  The storm discharges from the basin are about 63
cms for a "typical" monthly storm (6mm/hour) and 540 cms for the 5-year
design event; these figures should be contrasted with the 6.5 cms peak
capacity of the interceptor.


APPLICATION OF STORM

The STORM model was used extensively in the following aspects of the study:

(1)  To determine runoff rates from the various watersheds.

(2)  To determine mass  loadings of BOD and TSS from the area.

(3)  To assist in sizing the various control facilities required to divert
     flows.

(4)  To help evaluate long term management strategies for the basin.

Hourly rainfall data was available for 12 years; typical simulations used
6 years of records which included a typical range of wet and dry years.  The
climate of Kaohsiung is characterized by an extremely dry winter from Novem-
ber through March followed by a very wet summer season.

Sol-ids buildup in the side drains and street-side ditches tends to be signi-
ficant during the winter months; during this period the flows in the system
are largely sewage and  grey water.  Little or no flushing of the estuary
occurs during this period.
                                     35

-------
The months of April and May are characterized by moderately intense rain-
falls; these first storms after the dry season flush much of the deposited
solids from the network and dump them in the river.  The assimilative capa-
city of the river during this period is overwhelmed by the total magnitude
of the BOD moads.

A modification was made to the STORM model  to enable us to simulate the
long term deposition of solids in the "sanitary" system since the system is
not self cleansing as in typical U.S. systems.  These depositional factors,
Fl and F2, are the mass fraction of the TSS and BOD in sanitary sewage which
in dry weather settles out of suspension and is deposited in the drains and
watercourses.  These deposits are then resuspended and flushed to the inter-
ceptor/river during storms.  It was not possible to perform field tests to
determine the actual deposit fraction for TSS and BOD during the course of
the project.  The values used for most STORM runs were 0.7 for TSS and 0.3
for BOD.  These values are probably conservatively high since channels at
low-flow sites cannot be assumed to be as efficient as primary clarifiers
would be.  However, since our concern was with the fraction resuspended dur-
ing the storm events, it was decided to adopt the above figures.

The overall system was evaluated by simulating each individual basin and its
interceptor junction characteristics; the allocation of interceptor capacity
against basins was as further indicated below.

The results of the simulation runs were largely as anticipated.   Figure 6
shows the impact of allocated interceptor capacity on the BOD overflowing
to the river from a typical basin; it will  be noted that the allocation of
flow capacity above the DWF results in only minor reductions in  overflows.
It proved possible to use disused canals as storage basins in several
instances; the impact of such storage on BOD overflows can be significant
and as shown in Figure 7 can reduce the fraction of BOD overflowing to the
river from about 30% to 5% of total BOD loads from the tributary basin.

The impact of using swirl concentrators for grit and solids removal at sev-
eral  of the smaller basins is shown in Figure 8.  Although effective on a
single site basis, they did not prove cost  effective in terms of the overall
Jen Ai control strategy.

Schematic layouts of a typical  interceptor  structure are shown in Figure
9-11.

Overall, it is estimated that the proposed  scheme will  reduce BOD loads  to
the Jen Ai River by about 90% from existing conditions.   Overflows will  still
occur about 20-40 times per year; many of the overflows  are quite small  and
the BOD loads minimal.

The computed reduction in BOD loading to the river is just about adequate to
enable the estuary to achieve a D.O.  level  of 2 ppm under average month
conditions.  Heavy loadings in  the spring months will  be sufficient to de-
oxygenate the river during these periods.   As a result,  it is proposed that
a series of air-bubblers be installed to ensure that the estuary remains
fully mixed and surface oxygenation maximized.   Simulations indicate that


                                    36

-------
  500
ec
  400
* 300
cc
LJ
IT
o
>-
o
z

o
CC
Id
  200
O  100
o
o
m
                                       SITE  D
                                     F = 0.3
                                     NO TREATMENT
                                     NO STORAGE
              .05       .10       .13       .tO       .25       JO

             ALLOCATED   TRUNK  SEWER   CAPACITY,  t(CMS)
         FIGURE 6.  INFLUENCE OF TRUNK CAPACITY,  t,
         ON BOD OVERFLOW RATE, y
                              37

-------
  o
  m
UJ

o
      0.6
       0.5
       0.4
       0.3
       O.I-
          )•
                             LETTERS DENOTE  SITES
                      H°
             .005      .010       JOI5

               STORAGE VOLUME
                                              .020
                                                  .025
               TRIBUTARY  AREA X RUNOFF FACTOR
                                                 m
         FIGURE  7.   INFLUENCE OF STORAGE VOLUME

         ON BOD  OVERFLOW RATIO
   o
   o
   UJ
   >
   O
    O.6



    0.5



    0.4-



S   O3




I   ....



    O.I -
                                           WASH:do    WAS":2;
           O.I   OlE   0.3  0:4  0:5   0&   0.7  aB

                BOD REMOVAL EFFICIENCY

            (IN SWIRL CONCENTRATOR FOR OVERFLOWS)
                                                     0.9
         FIGURE 8.  INFLUENCE OF" BOD  REMOVAL EFFICIENCY

         ON  BOD OVERFLOW RATIO
                                38

-------
6 VERTICAL- LIFT
FLOOD GATES
                                         ALTERNATE
                                         LOCATION  A
                 EXISTING
                 WORKING
                 SHAFT
                                                    ,  EXISTING
                                                    /  WORKING
                                                      SHAFT
ALTERNATE
LOCATION   B
   FIGURE 9.  SITE PLAN FOR  INTERCEPTOR STRUCTURE  AT "E"
   TWO ALTERNATE LOCATIONS ARE SHOWN.
                            39

-------
                                               UJ
                                               o:
                                               oo

                                               C£
                                               o
                                               I—
                                               D_
                                               LU
                                               O
                                               QL
                                               UJ
                                               U3
40

-------
                                                  01
                                                  ID
                                                  K-
                                                  O
                                                  p
                                                  Q.
                                                  C_J
                                                  a:
                                                  Di
                                                  O
                                                  O
                                                  C£
                                                  Qu
                                                  o;
                                                  ^)
                                                  CD
41

-------
 this will maintain  adequate  D.O.  levels in the river.


 SUMMARY

 In this  situation,  it  has  proved  possible to quickly restore adequate water
 quality  to an estuary  which  was grossly overloaded with uncontrolled sani-
 tary and stormwater flows.   The cost is quite low and well within the means
 of developing countries.   Of course, the problem is not yet fully solved -
 those areas  upstream of  the  interceptor still contain open sewers which will
 have to  be systematically  upgraded over the next 10-20 years.


CONCLUSIONS

Models such as STORM can be effectively used  in  design situations in devel-
oping countries.  In many instances,  the problems facing the designer are
much tougher than those encountered in  the U.S.  - fundamental  public health
improvements must be sought with moderate expenditure of funds for construc-
tion.  In the cases discussed in this paper,  STORM proved to be a very effec-
tive tool in assisting the designers  to achieve  the design goals.
 The work described in this  paper  was  not  funded  by  the  U.S.  Environmental
 Protection Agency.  The contents  do not necessarily  reflect  the views of  the
 Agency and no official  endorsement should be  inferred.


                                      42

-------
AN ATTEMPT TO IMPLEMENT SWMM IN TUNISIA
by Janusz Niemczynowicz
ABSTRACT

In 1980, the University of Lund started a research project in co.operation
with the University of Tunis. The aim of the project, was to implement the
SWMM in Tunis and to teach local research personel how to handle the model.
During the winter season, there are flooding problems in the cities of Tuni-
sia due to high-intensity convective rain storms and underdesigned sewage
systems. The greatest efforts in the city of Tunis until  now have been made
to solve the immediate problems as street flooding by constructing a huge
stormwater conduits. This will probably increase already severe pollution
problems in the receiving waters. It is obvious that some kind of system
thinking has to be applyed in this situation. Implementation of SWMM-model
was thought to be one of the masures possible to take in order to assure
meaningful! water management in th cities. In order to  obtain the input data
for calibration of the model, a catchment of 20 sq km was instrumented. Dif-
ferences between  Swedish and Tunisian urban areas, together with the climat
in Tunisia, couse significant differances in the input parameters in the
model. For example, roughnes in water courses can not be treated as a cons-
tant in Tunisia. During the dry period great amount of garbage is accumulated
in the water corses, during rainy season the garbage and lush vegetation is
succesively removed and the roughnes decreases gradualy. The paper describes
how the difficulties during field work, differances in working routines,
problems with understanding the"model ing philosophy", and problems with model
calibration influence the possibility of model implementation in Tunisia.

INSTRUCTION
The average, yearly precipitation in Tunis is about 450 mm. The major part of
the yearly rainfall occurs during October, November and December, usually as
short, intensive convective rainstorms. These rainstorms often take the form
of thunderstorms. The rest of the year, there is very little precipitation.
Tunis is surrounded by high hills which slope  steeply towards the central
part of the town causing severe problems with storm water management.
                                     43

-------
The storm water system presently existing in Tunis is not sufficiently design-
ed to handle storm water runoff resulting in street flooding several times a
year. During the past three years, great efforts have been made to reduce
flooding by constructing a huge storm water conduit which leads water from
the central part of the town to the nearby situated Lac de Tunis lagoon
(Coyne 1974). It is unlikely that even these efforts will significantly im-
prove the situation due to rapidly increasing urbanisation of the suburban
areas. Other problems will occur, i.e. pollution of the lagoon by storm water
discharge.

Most of the cities in Tunisia has the same problems. It is obvious that some
kind of system thinking has to be applyed in this situation in order to as-
sure resonable way to solve the design problems associated with storm water
systems. Implementation of SWMM-model was thought to be one of the measures
possible to take in order to assure meaningful 1 water management in the ci-
ties avoiding all the mistakes previously done in European countries in this
matter.

In 1980, a research project was started in co-operation with the University
of Tunis. The aim was to implement the Storm Water Management Model and to
teach local Tunisian research personnel how to handle this model.

In order to calibrate the model, a large amount of rainfall  and runoff data
had to be collected in an environment rather different from the Swedish one.
                           p
We decided to use the 20 km  Guereb-Roriche catchment in the northern part of
Tunis.

CATCHMENT AND INSTRUMENTATION
The Guereb-Roriche catchment consists of two areas:  the highly urbanised sec-
tion in the south and the rural  area in the north with a few single-family
dwellings.(See Figure 1).

The urbanised area of the catchment is densely populated and housing consists
of multi- and single-family buiIdings  andsome industries.  The character of

                                     44

-------
                                            GUEREB-RORICHE
                                            CATCHMENT
                             INSTRUMENTATION
                             O Raingouge           J/ /L°C de Tunis
                             * Run off measuring station
                             0 Quality sampler
                             *%S Densly populated aerea
                             ^ Water storage dams
Figure 1.       Instrumentation of the Guereb-Roriche  catchment.

the housing  is  quite different from European housing, which  has  significant
hydrological  consequences. For example, almost all  houses  have  flat roofs
with elevated edges, which causes roof ponding. Some  of  the  minor streets
are not  paved.  Roof drainage is rarely connected  to the  storm water system.
The percentage  of impermeable surfaces is very high - up to  60%  in some are-
as.

The southern part of the catchment is very flat and  varies  less  than 50 meters
in altitude.

The northern  part,on the other hand, is hilly with altitude differences up to
250 meters.  Agriculture dominates in this rural area  byt the uppermost region
is uninhabited  due to severe erosion of the steep hillsides.

The catchment is drained by two periodicly appearing  rivers,  the Guereb and
the Roriche, which join in the lower part of the  catchment and  then flow into
the Lac  de Tunis lagoon. In the upper part of the catchment,  there is no flow
in either river during the dry period. However, there is some baseflow
occuring in  the lower part due to the release of  waste water from the slum
                                      45

-------
areas. During the wet season, discharge from both revers can be as nigh as 15
m3/s.

The soil type within the catchment is laterit clay. Because of cracks
occuring in this clay during the dry season, it becomes very permeable.
After the first few rainfalls, permeability of the clay reduces drastically
(Infiltration tests made during the dry season have shown that the infiltra-
tion capacity exceeds 50 inches per hour during the next three hours). It
can be expected that infiltration tests made during the wet season will give
quite different results.

Urbanisation of the rural northern regions is proceeding rapidly and it is
expected that the entire catchment will be urbanised within a few years. The
development plans for the northern part include single-family homes, high
multi-family buildings and schools.

In order to take into account the nonhomogeneous character of the catchment,
and achieve rainfall area! characteristic, the runoff and rainfall measuring
stations were installed in 9 tield stations (Figure 1). (Niemczynoiwicz et al
1981).

As the distance between the stations is rather long, and there are no reli-
able telephone lines available, it was decided that all stations shouldrbe
equipped with separate clocks, data-loggers and automatic start-/stop-units.

The rainfall gauges are of the standard tipping-bucket type manufactured by
Rimco, Australia, with a bucket capacity of 0.2 mm. The funnel  diameter is
203 mm and the total accuracy is + 1% up to 380 mm/h.

The runoff gauge is an echosounder manufactured by Endress and Mauser, West
Germany. It permits a non-contact continuous measurement of the water level
over the weir.

In addition, there is also a mechanical water level recorder ensuring that
no data are lost during current-failure periods.

                                    46

-------
The datalogger, manufactured by A D Data Systems Inc., USA, Type ML-10, has
an internal quartz-controlled clock and a capability of monitoring up to 10
analogue signals and 32 bits of digital data. The data are recorded in a com-
puter-compatible format on a Philips-type data cassette. Data can be recorded
at pre-selected intervals which are controlled by the digital clock, by a
front-panel manual switch or by external input.

The recording interval is 5 minutes.

The automatic start/stop unit offers two possibilities of starting a recor-
ding sequence.

If the water rises to a certain pre-selected level, the recording starts; if
the raingauge counter reads two or more, the recording  is also started...An
additional timer keeps the recording for 2 hours after the last raingauge
tipping to make sure that the whole hydrograph is recorded.

The stations operate on a 220 V 50 Hz supply, but as we have had som bad
experiences with the power supply, especially during rainfalls, we have
changed the stations to battery power.
Figure 2 and 3 show two of the runoff measuring stations.
Figure 2       Runoff station

                                     47

-------
 Figure  3       Runoff station.

 RESULTS
 During  1981  and 1982,  great  efforts were made to collect and analyse avail-
 able  documents  and  maps  in order  to gather all data necessary for modeling.
 We  often discovered that important data were missing or could not be delive-
 red by  the officials  in  charge. There were also many maps of conduits that
 did not correspond  to  the actual  situation in the field. As a result, a lot
 of  field work had to  be  done for  verification of necessary data.

 The catchment was divided into forty-six subcatchments ranging in size from
      22
 0.06km   to 1.44  km   (see  Figure 4). The storm water system consists of seven-
 ty-seven sewer  pipes amd  two storages in the upstream region of the catch-
 ment  (see Figure 5).

 After two years of  data collection only about 8 rainfalls-runoff occurences
 were registratod on more  then one station.  Only service in  the field which
 had to  be done to assure  registration was changing the cassettes, paper
 charts ande the batteries in adequate periods of time.  Unfortunately,  most
of the high intensity rainfalls happen to occur while  batteries  were empty
or cassettes were full. During the third and  final  season of registrations

                                    48

-------
        Catchment segmentation
        Water divider
        Main water course
        Water divider between subcatchments
        Main conduits
        Water storage dams
Figure 4    Guereb-Roriche catchment in  Figure  5.   Guereb-Roriche catchment.
            Tunis                                     Diagram of storm water
                                                      system.
the schedule  for permanent  superintending by Swedish  personnel was estab-
lished. This  resulted in that  sufficient amount of  field data to make  a  reso-
nable calibration of.the model  was collected.
MODEL CALIBRATION
In co-operation with two students from the University  of Tunis, an attempt
was made  to  calibrate the model.  We discovered that our initial estemates of
the parameters were rather  inaccurate.

Fig 6 shows  an example of the  calibration runs.  The  first run was completely

                                       49

-------
different from observed hydrograph. We soon discovered  that  procentage  of
impermeable surfaces taken directly from the maps does  not represent the  real
situation. Only about 40% of all  impermeable surfaces is  connected  to storm
water system. The main problem  however is that this  figure changes  with time
during the rainy season. The reason is to find in erosion of soil between
paved surfaces, which gradually open small channels  connecting  new  areas  to
the system. One of two parameters depression storage or procentage  of imper-
meable surfaces must to be treated as a function of  time.
                M3/S  /\
                        TUNIS ELMENZ4H
                        EVENT 2Q.Ot.82
                         TUNIS 9CRUCT
                         EVENT laot.n
Figure 6  Examples of observed and SWMM-Simulated  hydrographs

Thewidthof the overland flow  (w-parameter), which  influences  the  shape  of
the hydrograph, calculated accordingly to the generally accepted rules,  ap-
peared to be far too high. This can  be explained by the fact  that  the  surfa-
ces are more undulated than we are used to, which makes the legth  of the
overland flow longer. Because of the high values of the infiltration para-
meters, used in the first simulations, no runoff was simulated from perme-
able surfaces. This is probably true for some of the first rainfalls in  the
beginning of the rain season. During the rain season, the infiltration capa-
city changes drastically and the infiltration parameters should be changed
                                        50

-------
accordingly.  During the first rainfalls, SWMM  overestimated  the peak flow
and runoff volume, but during the following veeks,  the simulations  were much
better (see Figure 7). This can probably be explained by the  variations of
the roughness in the main water channel. The river  Guereb and Roriche, which
drain the catchment, have rather flat sloping sides in the lower part of the
catchment. During the dry season, very lush vegetation covers the slopes of
the channel and most of the vegetation from the side slopes changing the
rouqhness of the channel drastically.
Fiqure 7
Examples of observed and SWMM-simulated  hydrographs
 In spite of those difficulties and differences in the magnitude of model
 parameters it is possible to reproduce the observed hydrograph quite well.
 The areal distribution of rainfull has to be taken into account because dif-
 ferances in rainfall intensity are very significant.
The one of the goals of the projekt was to implement SWMM on the French com-
puter                in the university of Tunis. The Guereb-Roriche catchment

                                      51

-------
was ment to be only a experimential field for demonstration of model applic-
tion, and for training Tunisien personnel in geting indata, running the model
and making design simulations. With other words, the project was thought to
be demonstration project  to teach  "modelling philosaphy". The main problems
which we faced in this context was the lack of hydrologists simultaneously
able to run the computer. On the other side there are no computer man who
understand hydrology.

Until now all the computer runs were done in Sweden.

REFERENCES

Coyne et Bellier,1974, Etude de 1'assainissement de Grand Tunis.  Etude de
facilite du plan directeur, ICN, Pays-Bas.

Dahlblom, P; Niemczynowicz, J; Hogland, W, 1982, "Significance of Water
Planning in Tunisia". Dept of Water Res. Eng., Lund University, 1982.

Dendrou, S A ; Delleur, J W, 1978, Planning Storm Drainage Systems for Urban
Growth, Journal of the Water Resources Planning and Management Division, Nov.

Hogland,  W; Niemczymowicz, J, 1980, Kvantitativ och kvalitativ vattenom-
sattningsbudget for Lunds centralort. Kompletterande matningar och metoder.
Institutionen for teknisk vattenresurslara, LTH/LU, Nr 3038, Lund.

Niemczynowicz, J; Grahn,  L, 1981, Hydrological  Instrumentation. Fifteenth
Anniversary Report, Department of Water Resources Engineering, University
of Lund, Report No 3053.
The work described in this paper was hot funde'd  by  the  U.S.  Environmental
Protection Agency.  The contents do not necessarily reflect  the  views  of  the
Agency and no official endorsement should be inferred.

                                    52

-------
THE SEDIMENT-WATER INTERFACE IN MODELING PESTICIDES IN SEDIMENTATION PONDS

                                    by

                  Peter R. Jaffe and Raymond A.  Ferrara

                     Department of Civil Engineering
                           Princeton University
                           Princeton, NJ  08544

Introduction

     Many pesticides adsorb onto soil particles upon application to fields.
Stormwater runoff then results in the transport of these soil  particles with
adsorbed pesticides to the receiving waters.  As a result,  peak pesticide
concentrations, sometimes reaching acute levels, can occur  in  receiving
waters during storm events that follow pesticide application.   Since parti-
culate material can be removed via sedimentation, the objective of this work
is to study what effect a sedimentation pond may have on the pesticide load
to receiving waters.

     The model presented in this study considers that the pesticide fraction
adsorbed onto the sediments will be removed during the sedimentation process,
and incorporated into the bottom sediments.  When after several storm events
the dissolved pesticide concentration in the overlying water column decreases,
the bottom sediments will release some of the pesticide contained in them.
The sedimentation pond will then behave like an equalization basin, capable
of buffering peak loads.  This buffer capacity is enhanced  by  the sedimen-
tation process, and the capacity of the bottom sediments to first retain and
then release again a portion of the applied load.  Furthermore, if decay
occurs, a pond could be designed  to provide sufficient detention time to
decrease the net total load to the receiving water.

Model Development

     The water column of the pond is represented as a completely mixed
system.  Pesticides are added to the pond during storm runoff in dissolved
and adsorbed phases.  These phases are at equilibrium as defined by their
partition coefficient  (i.e. a linear Freundlich adsorption isotherm).  The
dissolved phase can undergo any of a series of first order transformation
processes, such as chemical or biological degradation or evaporation, all
of which are additive.  Other sources and/or sinks of the dissolved phase
are the equilibrium interaction with the phase adsorbed onto sediments,and
diffusion into or out of the bottom sediments.  The adsorbed phase can also
undergo any of a series of first order transformation processes  in addition
to settling from the water column to the bottom sediments.   Consequently,
to model this system, equations are required to describe the changes in


                                    53

-------
volume, suspended sediments, adsorbed pesticide concentration, dissolved
pesticide concentration, and pesticide concentration in the bottom sediments.

a)   change in volume


                        - E)                                          (l)
               V -
                       2/3
if V £ VM, then QQ = 0
where:         A  -  surface area,
               E  =  evaporation,
               K  =  constant,
               L  =  length of the weir,
               P  =  precipitation,
              Q.  =  inflow,

              Q   =  outflow,

               R  =  change of volume due to the deposition of sediments,
               V  =  volume, and
              VM  =  holding capacity of the pond.

b)   suspended sediment

     Many chemical and physical properties of individual  sediment  particles
are dependent on their size distribution (e.g.  clay has  a higher adsorptive
capacity and a lower settling velocity than sand).   Consequently the size
distribution of sediments contained in storm water  runoff is important.
Furthermore irregularities in shape require the specification of a settling
velocity distribution for particles which are nominally  of the same size
(Ferrara and Salvage, 1982).  Then:


     3t  ' »l Cs1jk - «o Csjk - A \ Csjk
                                                             J. L
where:         Cs..   =  suspended sediment concentration of j   size
                 J      fraction with settling velocity  vk,

              Csi.,   =  Cs., in the inflow, and
                 j k       j k
                 vk  =  settling velocity.

then the term R in Equation  ( 1 ) is
                                    54

-------
          s        j k

where:         P   =  density of an average sediment particle,  and
                  =  connected porosity.

     The sediment-size distribution is divided into n size fractions, j = 1  -> n.
Each size fraction has a velocity distribution that is divided  into m velocities,
k = 1 ->• m, but m does not have to be the same for each j.   For  non-settling
solids (colloids), the velocity is zero.

c)  adsorbed pesticide concentration

     The change of concentration with respect to time of adsorbed pesticides
in a given size-velocity fraction of the sediments in the water column is:



     If (Cajk V> = Qi Caijk - QO Cajk - A vk Cajk -V kj Cajk -  Djk    (6>

The overall adsorbed concentration is then


     ^r (V E E  Ca,,) = Q. E E Cai .  - Qo z E Ca..  - A E E v, Ca ..
     01    j k    Jk     n j k    Jk      j k   Jk     j k  k   Jk

                      - V E E k'! Ca   - £ E D .                        (7)
                          j k  J   Jk   j k  Jk

where:         Ca.k  =  adsorbed concentration on j   sediment  fraction
                 J      with settling velocity v, ,
                                                K
              Cai.,   =  Ca..  in the inflow,
                 J K       JK
                 k".  =  sum of first order reaction rates  of the adsorbed
                  J     pesticide, and
                D-k  =  desorption term.


d)   dissolved concentration

     The equation describing the dissolved concentration in the water column
 is:
               . Q. -CC)  - k1 VC - AJ+ E I D ,                        (8)
               IT     o                .JI.JK
and
                                     55

-------
     j = _ A(on + Dr + DT) —      + d> v                              (9)
           VN B    S    I  3x        y  w                               '
                               x=0         x=0


where:         C  =  dissolved pesticide concentration,
              k1  =  sum  of all first order reaction rates,
               J  =  losses into the bottom sediments across the sediment-
                     water interface (advective plus diffusive flux),
              DR  =  bioturbation coefficient,

              D<.  =  molecular diffusion coefficient,

              DT  =  irrigation coefficient,

              v   =  velocity of water burial below the sediment water
                     interface, and
             D..  =  desorption term.


The equilibrium between the dissolved and adsorbed phases is


     Cajk - Kj C Csjk                                                 (10)


where          K.  =  partition coefficient.
                J
     Substituting equation (10) into equation (7)  to eliminate the adsorbed
concentration and then combining this equation with that for the dissolved
concentration, equation (8) eliminates the desorption terms  and yields the
final equation for the dissolved pesticide:

          Q  C.  Z (K. Csi.) - C[Q  +Vk;] Z  (K.  Cs.) - AC Z  v  K  Cs  +  C  Q
     f\r    1   1  -•   J    J       0     JjJJ       n-JJJ    11
     uu = 	J	sj	J	
     dt                             V[l  + Z  K. Cs.]
                                          T   J  J
                            (                 HV            dCsi
          CQn- k'VC- AJ- 
-------
     Ca - C Z £ K. Cs.,                                                (12)
            j k  J   Jk

and the total pesticide concentration (adsorbed and dissolved)  as:


     CT = C(l + z z K. Cs.. )                                          (13)
      1          j k  J   JK

Solution of this system of equations requires knowledge of the  dissolved
concentration gradient in the bottom sediments (equation (9)).   Therefore
solution of the concentrations in the bottom sediments  is  also  necessary.

e)   bottom sediments

     The development of the following equations is summarized from  the work
of Berner (1980).

     The mass balance equation in the sediments is:

     3C.     3F

     inrs -1*1 + z Ri                                               (14)

where:        C.  =  concentration of solid or liquid component i  in terms
               1     of mass per unit volume of total sediment,
              F.  =  flux of component i, and

              R.  =  reaction rates affecting i.

and
              3C.
     F. = - D -r-1 + v C.                                              (15)
      18X1

where:         D  =  diffusion coefficient, and
               v  =  velocity of flow relative to the sediment-water interface.

If no compaction or infiltration to the groundwater zone occurs, v  is the same
for the solid and liquid component.  Then v = v  = velocity of  burial of

solid particles and/or water below the sediment-water interface.

     The diffusion coefficient for the solid component  near the sediment-
water interface  (bioturbation and/or current mixing) is Dg.  For the dis-

solved phase, this same DR plus an irrigation coefficient  Dy and a  moleular

diffusion coefficient DM which is independent of  depth  apply.

     The concentration of a solid component Ts can be expressed in  terms  of
mass per unit mass of total  solids
                                     57

-------
     Cs = (1 -)p  Cs                                                 (16)


and for the dissolved component the concentration C is expressed as mass per
unit volume of pore water

     Cd = 4> C                                                         (17)

Then, combining each of equations (16) and (17) with equations (14) and  (15)
yields the general equations for the bottom sediments, which for a solid
component is
                                     r
     3['0 - *)p. CS]  8(D  -  -^ - }  3[(1 - *)v  p  CS]
                                                                 (1 _  j     R
           3t                  9X                    8X               v/ S

                                                                      (18)

where Rs is the reaction rate in mass per mass of solids per time, and for a
dissolved component

                                               v  C)
where Rd is the reaction rate in mass per volume of pore water per time.
For sorbing pesticides, equation (18) can be used for the adsorbed pesti-
cide C, and equation (19) for the dissolved pesticide.  The partition coef
ficient is used again to describe the equilibrium between the adsorbed and
dissolved phases

     C" - Kp C                                                         (20)


where K  is an average partition coefficient in the bottom sediments (equa-

tions (10) and (20) are equivalent).

Furthermore, a mass balance between the two phases must hold, so that the
sorotion rates R are related to each other by:


     Rs = ," * >   R.                                                (21)
          (1 - 4>JPS  d


 From the previous equations the final equation that describes the changes
 in concentration in the pore water of the bottom sediments is obtained:
                                      58

-------
                        9[(D
                                          v  £k + /_..'    \
                                          V       (      l
 it"    3X      1  + K1        3X           w 3X    1  + Kl

                                                                       (22)

where          K1 = (PS(! - 4>)  Kp).

This equation is solved numerically via finite difference methods with a
variable boundary condition, which is the dissolved concentration in the
water column.  The set of ordinary differential equations describing the
water column is solved using the Runge-Kutta method.   Because the two
systems are interacting with each other they are solved in sequence for
each time step.

Results and Discussion

     The results of several simulations are shown in the following figures.
Typical values for yearly storm runoff data, including runoff volume, sedi-
ment mass and pesticide load were used (Smith, et. al., 1977).   The sediment
size-velocity distributions of the inflow sediments were assumed constant
for all storms, and are listed in Table 1.  Using the relationships reported
by Karickhoff, et. al. (1978) the partition coefficients were estimated based
on the organic carbon content of the sediments and a range of typical octanol-
                                2   4
water partition coefficients (10 -10 ).  In the absence of reported data for
mixing coefficients in bottom sediments, reported values for mixing coeffi-
cients in near shore silts were used (Berner, 1980).   For the following simu-
lations, a detention capacity of the pond of 2000m  and a depth of 2 m was
selected.

     The bottom sediments were modeled over a depth of 0.2 m. Bioturbation
and irrigation was considered to occur only in the upper 10 cm while mole-
cular diffusion was modeled over the entire depth of the bottom sediments.
The molecular diffusion coefficient was estimated based on a molecular weight
of 200 and corrected for the tortuosity dividing it by the product of the
porosity and the formation factor.  A formation factor of 3 was assumed.

     In Figure 1 the inflow (runoff) pesticide concentration is shown over
time, as well as the pesticide concentration in the pond for two different
partition coefficients.  The effect of the pond can be seen clearly.  The
peak concentration in the pond is about 3% of the peak storm runoff concen-
tration without a pond.  As expected, if the partition coefficient is in-
creased the reduction of the peak concentration is even higher, because more
pesticide  is in the adsorbed phase, consequently increasing the pesticide
removal efficiency of the sedimentation process.  For longer times, the ef-
fect of the pond is to retain and slowly release pesticides.  This effect is
 increased for higher partition coefficients.

     In Figure 2 the total  cumulative pesticide load to the receiving waters
for consecutive  seasons  is shown.  As expected, if some decay does occur in

the water column  (k = 0.02 day"  ,  for the dissolved phase) the overall
pesticide removal rate is increased. It is interesting to notice that for the
                                     59

-------
case where decay occurs the system reaches its final  conditions sooner than
if decay does not occur.  For the later case, the final  conditions would be
such that the yearly cumulative loads with and without the pond are the same.
The reason for this is that the initial adsorptive capacity of the bottom
sediments is exhausted but a buffering capacity still exists.  In other
words, when the system reaches its final conditions,  there is no net gain in
pesticide mass in the bottom sediments over a season.   This is illustrated in
Figure 3, which represents the concentration profiles in the bottom sediments
at the end of each year for the cases of Figure 2.  The fact that the bottom
sediments continue to have a buffering capacity, even if their net yearly
gain of pesticides is negligible, is shown in Figure  4.   This figure shows
the pesticide profile in these sediments at the beginning of the second
season, after 50 days, and at the end, for a pesticide with decay.

     The effect on the cumulative load of different partition coefficients is
shown in Figure 5 in which a pesticide with decay is  modeled.  The increase
in pesticide removal efficiency for increased partition coefficients, during
season modeled, results because for higher adsorptivities the influence of
sedimentation increases, and also the bottom sediments have a higher capacity
to retain pesticides.

     For all of the previous simulations it was considered that 54% of the
sediments in the inflow are colloidal material or fine clays that will not
settle.  Theimproved pesticide removal efficiency when a coagulant is added
reducing the non-settling fraction by 75% is shown in Figure 6.   These simu-
lations do not include a decay rate.  Similar results were obtained for two
simulations using low (i.e. fine silt) as well as high (i.e. sand)  settling
velocities for the coagulated solids.

Summary and Conclusions

     The results of these simulations show that sedimentation ponds can be
used effectively to dampen out peak pesticide loads to receiving waters.   If
no decay occurs the total yearly load to the receiving waters will  be the
same with or without a pond once final conditions (i.e.  after a few seasons)
are obtained.  If some decay occurs, the total yearly load at final  condi-
tions will be less with a pond than without one.

     Bottom sediments and their interactions with the water column,  often ig-
nored for this type of analysis, have been shown to have an important effect
on the performance of the pond.  It would therefore be extremely valuable to
have a better understanding of the mixing conditions  in  the upper sediment
layers of ponds and lakes, as well as the reactions that occur for  pesticides
and other pollutants in adsorbed and dissolved phases  in these sediments.

References

Berner, R.A., Early Diagenesis, Princeton University  Press, Princeton, NJ,
     1980.

Ferrara, R.A., and Salvage, K.M., "Stormwater Pollutant  Settleability,"  in
     press.


                                     60

-------
Karickhoff.S.W., Brown, D.S., and Scott, T.A., "Sorption of Hydrophobia Pol-
     lutants on Natural Sediments," Water Research,  v.  13,  pp.  241-248, 1979.

Smith, C.N., et. al., "Transport of Agricultural  Chemicals  from Small  Upland
     Piedment Watersheds," USEPA, ORD, Env.  Res.  Lab.,  Athens,  Georgia, draft
     report, 1977.
                                  Table 1
             Distribution of the Solid  Particles  in  the  Runoff
                                 Particle Size (microns)

% organic
carbon
fraction of
the total
mass
settling
velocity
(m/day)
larger than 150
2.6%
o.ie
2070
0.03
6480
0.01
10800
between 1 and 150
5.3%
0.01
540
0.11
1620
0.04
2880
smaller than 1
4.9%
0.02
195
0.08
580
0.54
0
                                     61

-------
    0,06
•t
*
g   0.05

£   0.04
Z
z
I   0.03

£
g   0.02
<_>

£   0.01
M
                                   • INFLOW CONCENTRATION

                                   O CONCENTRATION IN THE POND FOR
                                     Ko.W " 10^
                                   ft CONCENTRATION IN THE POND FOR
                                     KO.W » 103
                 100
      200

TIME (DAYS)
                                                                   300
                                                                       1.5
                                                                       1.0
                                                                              s
                                                                             J
               FIGURE  1.   ATTENUATION  OP THE  CONCENTRATION.
                                              • WITHOUT THE POND
                                              • 2ND YEAR WITHOUT DECAY
                                              j, 1ST YEAR WITHOUT DECAY
                                              • 3RD YEAR WITH DECAY



i
I
w
>
§










0.22
0.20
0.18
0.16

0.14
0.12

0.10


0.08
0.06

0.04
0.02
0 1ST YEAR WITH DECAY
•

.

• •
.
X
Jt A *
I £ o °
• 4 o
• • *
• |l*
8
-
•
50 100 150
TIME (DAYS)
                  FIGURE 2.  CUMULATIVE LOAD FOR CONSECUTIVE SEASONS.
                                   62

-------

       0
       2.5
       5
   I  7.5
       10
       12.5
       15
       17.5
       20
                      X INITIAL CONDITIONS (2 YEAR)
                     • -50 DAYS LATER
                     A FINAL CONDITIONS (3 YEARS)
                       .   I   .    .    i    .    I
                               0.005               0.010
                                DISSOLVED CONCENTRATION (MS/*)
                                             0.015
                      FIGURE 3.   PERIODICAL VARIATION OF THE CONCENTRATION
                                 IN THE BOTTOM SEDIMENTS.
I
      0
      2.5
      5.0
      7.5
      10.0
      12.5
      15.
      17.5
      20.
A END OF THE FIRST SEASON/  WITHOUT DECAY
" END OF THE SECOND SEASON, WITHOUT DECAY
OEND OF THE FIRST SEASON,  WITH DECAY
»END OF THE SECOND SEASON, PITH DECAY
  .    I    i    i    .   .   I   .   .   i    .
                            0.005               0.010
                               DISSOLVED CONCENTRATION
                                           0.015
              FIGURE 4.
 EFFECT OF A DECAY RATE ON THE CONCENTRATION  IN
 THE BOTTOM SEDIMENTS.
                                          63

-------
                0.22

                0.20

                0.18

                0.16

              i 0.14

              | 0.12

              > 0.10

              3 0.08
              CJ
                0.06

                0.04

                0.02
A
•
                                 WITHOUT THE POND
                                 K*S COMESPONDIN6 TO A Ko.M - 102
                                    S CORRESPONDING TO A Kb.tt. - 10*
                                   'S CORRESPONDING TO A KO.N. - 10"
                                        0  •
                                      SO
                                  100
                                TINE (DAYS)
                                                  150
                     FIGURE 5. EFFECT OF DIFFERENT PARTITION COEFFICIENTS ON THE CUMULATIVE LOAD.
               0.22

               0.20

               0.18

               0.16

             ~ 0.14
             3
             g 0.12

             u 0.10
             >
             § 0.08

             * 0.06
               0.04

               0.02
  • o*
                               • MITHOUT THE FOND
                               O WITH THE P0»
                               A KITH THE POND AND COAGULANT ADDITION
                                     50
                                                    150
                             100
                       TINE (DAYS)

FIGURE 6. EFFECT OF INCREASED SEDIMENT REMOVAL ON THE CUMULATIVE LOAD-
The work  described  in  this paper was not funded by the  U.S.  Environmental
Protection Agency.   The contents do  not necessarily  reflect  the views  of  the
Agency and no official  endorsement should  be  inferred,.
                                                64

-------
               A MIXING ZONE MODEL FOR CONSERVATIVE  PARAMENERS

                                     by

                              Main R.  Hutcheson
                        Oklahoma Water Resources  Board
                                  SECTION 1
                                INTRODUCTION

There are  two  basic ways in which  point  source  discharges  are regulated.
One way  is  through the application of  technology  based  permit limits for
dischargers.   However,  the current  administration strongly supports the
water quality  based approach to regulating  pollutants (Eidsness, 1982).   In
this approach a wasteload allocation is performed to determine permit limits
which will  protect the water quality standards  assigned  to  the receiving
stream.

The water  quality  standards apply everywhere outside the mixing zone.   In
the past,  the  thrust has been to ignore mixing zones and perform wasteload
allocations for  entire planning  segments.   However, this approach has not
proven feasible  (Hutcheson, 1979).   More states are  now  including  mixing
zones.   In  1971  twenty-two states mentioned mixing zones, while at present
all but  three states  include  mixing zones  in their  regulations  (Neely,
1982).   Therefore,  a mixing zone model applicable  to single discharges  or
dischargers with overlapping mixing zones can have general applicability.

Mixing zones  may be defined in many different ways.  Several  states use an
arbitrary definition involving  flow volume, stream width, or distance from
the source.   Oklahoma defines  a mixing zone  as  being thirteen times the
stream width,  with three  quarters of  the  flow volume protected  as a  zone of
passage,  which effectively  limits  the mixing zone to about a fourth of the
stream width (Figure 1).

In order to assign discharge permit limits (to an industrial  point source,
for example)  which will protect water quality standards outside the mixing
zone, it is necessary to employ a  dispersion model in the wasteload allo-
cation process.  The dispersion  model predicts pollutant concentrations in
the receiving  stream.   The most basic  and  widely  used  model  is the  mass
balance  approach,  wherein  a steady  state  system exists  and pollutant is
mixed uniformly throughout  the receiving stream.   There are several  problems
with  this  approach, the most important being its  inability to  protect the
zone of  passage  (depicted in Figure 1).  Another major problem with the mass

                                     65

-------
 balance approach  is  verification.   Usually the pollutant  becomes uniformly
 distributed within the  stream so far downstream  that the signal  to noise
 level  is too  low  to  relate the  effects  of the discharge  to the instream
 concentration.  Sometimes  resuspension,  nonpoint sources, partitioning and
 sedimentation contribute  significantly to  the concentration of conservative
 parameters.   If the  mixing zone is arbitrarily  defined, as is the case in
 many states,  then  the assumption of uniform mixing through the  stream at the
 end  of the mixing zone  may be invalid.   In Oklahoma  complete mixing at the
 end  of the defined mixing zone  occurs only  on very  small  streams, so in
                         Mixing lone
                                   Portion of the nixing zone
                                  which is also a zone of passage
                 Figure 1
               Portion of receiving stream in which water quality
               standards are not valid for an isolated discharge
general,  the concentration  outside  the mixing zone  is much greater  than
predicted by mass balance  (Hutcheson and Gopal,  1981a).
Since dispersion  across a  stream  must be taken into
zone, applicable  models  must include  this  effect.
methods  for  accounting for cross stream dispersion.
that the concentration due to a
half-normally distributed  across
occurs.
for with
concentration  distribution  is
(Hutcheson and  Gopal, 1981b).
                                             account in the mixing
                                             There are  two basic
                                             The easier one assumes
                         steady  state point source on the  banks  is
                         a  stream until  reflection  from the far bank
Due  to  the  principle of superposition,  reflection may be  accounted
a virtual point source.  The assumption of  a  half-normal  (Gaussian)
                       generally reasonable  in Oklahoma  streams
                       There are many situations where a wasteload
allocation methodology based on a Gaussian dispersion model will  adequately
determine  discharge wasteloads  (from  which  permit limits  may  be derived)
which  will  allow  instream concentrations specified  by  the water  quality
standards  to  be met.   There are, however,  situations  where the  observed
distributions  in  Oklahoma do not  approximate normal  ones.   Furthermore,  it
has  been  shown that  the  dispersion mechanism is  dependent both  upon  the
turbulence of  the stream and the configuration of the pollution  (Hutcheson,
                                     66

-------
1979b).   In other words, turbulent eddies disperse different sized plumes at
differing  rates.   Therefore,  the Gaussian  model  may not predict concen-
tration  distributions  as well as a  numerical  solution to the dispersion
equation which  computes new  dispersion  coefficients  at  various  distances
from the source  does.   Furthermore,  processes affecting the concentration
distribution which  depend  on  the  hydrology  of  the stream are better handled
with a  numerical  solution.   An example  is  partitioning and sedimentation.
In general, conservative  elements settle out  of the water column when the
flow speed  is  slow, not uniformly as would be assumed in a Gaussian model.
Therefore,   it  is  deemed worthwhile  to pursue  a  numerical  solution  to the
dispersion  equation.   This  type  of  model is more amenable to incorporation
of other mechanisms affecting pollutant concentration besides dispersion.

The objective  of  this  report is to describe the development of a wasteload
allocation process which incorporates a  numerical solution to the dispersion
equation,  and  to  demonstrate  the utility of  the process.  The development
consists of the  following  steps.  First, the  dispersion equation  is simpli-
fied through the  use of assumptions  so  that  it may  be solved numerically.
Then a  relationship between the standard deviation  of a  Gaussian distri-
bution  and the  dispersion coefficient is derived.   Dye studies are used to
determine  the  standard deviations.    A  relationship  for  the standard
deviation,  hence  the dispersion  coefficient is developed based on hydrology
and plume  dimensions.   The dispersion equation  is  solved  numerically for
pollutant  concentration at grid  points  throughout  the mixing zone.  The
maximum  concentration  at the end of  the mixing  zone or on the boundary  of
the zone of passage, the related source strength,  the background concen-
tration  and the  water  quality standard  are used to determine the wasteload
allocation  which  will  protect the water quality standard.   This process is
applicable  to multiple  discharges as well as isolated ones, using the super-
position principle.

                                  SECTION 2
               DERIVATION OF THE  APPROPRIATE DISPERSION EQUATION

 Dye  studies performed in Oklahoma showed that the Natural  coordinate  system
 is appropriate for dispersion modeling.   The Natural Coordinate system,
 advocated  by Yotsukura and Sayre (1976),  features  curvilinear horizontal
 coordinates.   The x direction is along  the thalweg of the stream, the hori-
 zontal  coordinate perpendicular to this is the cumulative discharge,  q,  and
 the  vertical  component, y, is  stream depth.   The cumulative discharge  is
 defined as the discharge  from  the  injection  bank to a point in the stream.
 The maximum q  is  the total  stream discharge.

 Hutcheson  and  Gopal (1981b)  showed that, in general, dye concentrations  are
 more  like  normal distributions when  the Natural  coordinate system is used
 rather  than a Cartesian coordinate system.   Yoksukura and Sayre  (1976)  also
 observed this  to be the  case.   Therefore,  the Natural system yields  better
 estimates  of the  dispersion coefficient. Another important reason for using
 the  Natural system  in  Oklahoma  is  the  definition of the zone of passage.
 One  quarter of the  flow volume  occurs  at  the point  at which the ratio  of
 cumulative to  total discharge is one quarter.  Therefore,  in  a  coordinate
 system  using  cumulative discharge, the  physical  dimensions  of the zone  of
 passage may be easily  determined.

                                    67

-------
The standard  diffusion  equation is expressed in the  Cartesian  system.   A
coordinate transformation  is  required  to convert to the Natural  system,  and
many  assumptions  may be  made to  simplify  the  equation  to facilitate a
numerical solution.

In Cartesian  coordinates  the  dispersion equation may be written (Yotsukura
and Sayre, 1976)

            afl   3(6wv)    8(6vO
where 6  is  instantaneous concentration,  and w , w  and w  are flow speeds
in the  x (downstream  along the bank), y  (ve&iccrl)  and zz (transverse)
directions.   Steady  state  conditions  are  assumed to exist.   Therefore, (1)
may be reduced to a three dimensional  equation by time averaging (Hutcheson,
1980).  The gradient transfer hypothesis  (concentration flux proportional  to
concentration  gradient)  is  assumed valid  because the  Oklahoma Water
Resources Board  (OWRB) dye studies showed that  the concentration  distri-
butions  were  more  or less  Gaussian.   Further  simplification  of  the
dispersion  equation  requires  the  assumption of  uniform mixing in  the
vertical.   It  has been shown that  vertical mixing is complete at a distance
downstream  from  the  source  equal  to  50  to  100 times the depth of  the
discharge point.   In  shallow Oklahoma streams  the  assumption of uniform
vertical mixing  is valid virtually throughout the mixing zone.   It is also
assumed  that the mean velocity is along  the thalweg (x direction)  and that
longitudinal dispersion  has negligible influence on  steady  state  mixing.
While longitudinal dispersion  is  the  primary dispersion mechanism in  non-
steady state situations,  the longitudinal  gradients for the concentration
distribution which evolves  from a steady state  source renders  the  longi-
tudinal dispersion mechanism ineffective.

Using the above  assumptions, dispersion in the q direction and advection in
the x direction are the only mechanisms which must be  accounted for in order
to predict  concentration distributions of  conservative  parameters  in a
mixing zone.   Therefore, employing the  appropriate coordinate transfor-
mations, (1) becomes (Hutcheson, 1980)


                           If • f^x"2"^''                      (2)
where c  is  the  time and vertically averaged plume concentrations,  h is
stream depth,  u  is  stream velocity in the x direction, K is  the dispersion
coefficient and m  is  a factor introduced to correct for  differences  between
distances along curved coordinate lines and those measured along the  x axis.
Along the  thalweg m   =  o, and at  other  locations  the approximation  of
Yotsukura and Sayre ft.976) may be used:

                                ALv
                           \ - -s
where AL  is the distance between grid points along a constant q.   Eq (2)
may be solved if an expression for K can  be obtained.


                                    68

-------
                                  SECTION 3
       DETERMINATION OF  AN EXPRESSION  FOR THE DISPERSION  COEFFICIENT K

Dispersion of pollution  from  a steady state point source is dependent upon
the relationshp  between the  characteristics  of the  prevalent turbulent
eddies and  the  size of  the pollutant plume.   In order  to  determine the
characteristics  of  the eddies, perturbation (as opposed  to mean) quantities
must be measured  and  assimilated in a statistical  fashion.   Because of the
difficulties in accomplishing this,  it was decided to use the relationship
between the variance of  the concentration distribution  and K  to determine
the dispersion coefficient  indirectly.   This  approach may be  used because
the Oklahoma dye concentration usually approximated a normal  distribution in
the Natural  coordinate system.

For the conditions  under which a normal  distribution is valid, h, u and K
are independent  of q,  and m  = 1.  Under these conditions (2)  may be written


                         H - Kh2u w = "•
The solution to  (4) is the Gaussian plume, expressed by

                           c = -^_ exp (-3L)                     (5)
where a2  is  the  variance of the  concentration  distribution  and S is the
source strength.   Substitution of (5) into (4) yields
This is the  diffusivity expression required to satisfy (4) for a Gaussian
plume.   It will be assumed that it also holds when the concentration distri-
bution is approximately Gaussian.

In order  to  develop  an expression for diffusivity  using  (6),  dye studies
were performed on  selected Oklahoma streams (Hutcheson and Gopal , 1981b).
Ten field experiments  were conducted on six different stream reaches.   The
streams chosen were typical of the larger streams  in central  Oklahoma.   They
were broad,  flat, shallow streams with smooth sand or mud  bottoms containing
few rocks  or other obstructions.  Flow  velocities  were generally on the
order  of  .5 to 3  feet per second.  Dye was injected  into the receiving
streams at  a  rate  of  .2 to  .3  ml /second.   Samples  for dye concentration
analysis and  flow  measurements  were taken at frequent enough  intervals to
adequately  define  the  stream hydrology  and  plume dimensions to thirteen
stream widths from the source.

To  determine the variance of the observed dye concentrations symmetry  about
the  injection bank must be  assumed  since a bank release produces a half-
normal  distribution.   The concentrations observed  in  the river were placed
at  the same distance  from  the  injection  banks on the  inland side to obtain
the entire  normal  distribution.   The  variance of  these concentration distri-
butions  were computed  using  a  method employed by Meyer (1977), for equally
spaced data

                                    69

-------
                              I[c.(qq)2]
where  c.  is the normalized  steady  state  concentration at the cumulative
discharge q. and
Equally spaced  concentrations  were  extracted from normalized distributions
derived from  the analyzed  samples  collected during the dye experiments.

The dye samples were collected on cross sections at various distances from
the source.  Therefore, the variances computed using (7) are valid for these
distances.  These computed variances at the cross sections (solid lines) are
plotted for seven of the  field experiments in Figures 2, 3, 4, 5, 6, 7, and
8 as a  function of distance from the  source.  Data from three experiments
could not  be  used,  due to problems with the dye injector on two occasions,
and because an  injection  into the middle of the stream, rather than a bank
injection, was  attempted once (Hutcheson and Gopal ,  1981b).

An examination  of Figures 2, 4, 5, 7, and 8 show a tendency for the computed
variance  to  eventually become quasi-constant  at some  distance  from the
source.   However, since the  dye must eventually  spread  uniformly across the
river,  the computed  variance must continue to increase with distance from
the source.   Several  theories  have  been advanced  for the behavior displayed
in these  figures,  but  no definitive  answer  has been  found.   The  most
important  behavior  displayed in  Figures  2-8 is  the  rapid  increase in
variance with distance from the source.

Various mechanisms  by which the  variance  increases have  been  considered.
Since only gross  features of the flow were measured,  it is not possible to
relate the dispersion coefficient to turbulence at  this time.   As shown in
Figures 2-8,  the variance increases with distance from the source.   In order
to predict variance using distance, it was assumed that

                                   a2 = ax ,

where a and b are arbitrary  constants.  The constant values may be obtained
using a least squares technique with the computed variances up to the point
they become quasi-constant, so


                              a2  = .0002 x 2'3.                     (8)

This equation is  used to  predict variance, and the  results are depicted in
Figures 2-8 as  the dashed line, so that predicted and  computed  variances may
be compared.   The  dashed  lines are identical in each  figure,  but appear to
shift due to the changing scales.
                                   70

-------
                           o  »^
71

-------
72

-------
73

-------
                  o2(xlOO ft2)
                 240 -
                 2)0 -
                 180 -
                 150 -
                                    3000    4000
                                       Distance (ft.)
                                                5000
                    Figure 8 Observed and Predicted Variance versus Distance from Source
                         U.S. Highway 177B
The correlation between predicted and observed  concentration variance is  .7.
This indicates  that (8) does not predict the variance  very accurately.  Even
so, it  is tempting  to use (8)  due  to the simple  concentration prediction
scheme  which would  result.   The expression  for a may  be  substituted into
(5), and,  when the mean flow is known,  the concentration may be computed  as
a function of position.  This is much simpler than trying  to predict concen-
trations by  solving (2) numerically.

It  may  be shown that,  close to  the source, the  dispersion coefficient
changes  with  distance  downstream.    A  constant  coefficient implies  a2
increases linearly  with x, from (6).  However,  (8) shows the relationship is
not linear.   Therefore, K is  not constant and, therefore,  published dis-
persion  coefficients are  of  little  use in predicting  concentration distri-
butions  near the  source.
A  relationship between
during the dye  studies.
volume of  discharge  and dispersion was  observed
 In order to account for upstream  dispersion as well
                                     74

-------
as this relationship, the  integral of the cumulative discharge to .3 of the
maximum concentration with respect to distance from the source was used as a
predictor.   The integrated value is used because the variance of the concen-
tration distribution depends  upon  the dispersion which has occurred at all
points downstream  from  the source.   The relationship may  be  expressed as

                              a2 = a+b/P dx,

where CD is  cumulative  discharge from the  injection banks to the location
where the normalized concentration equals .3 and z* is  the distance from the
injection bank  to  this  location.  The a  and b are  arbitrary constants  found
by the least squares technique, so that
                                        rn
                         a2 = 351.9+.02j^f dx.                     (9)

Again, the  constants were  evaluated using  only  variance  and cumulative
discharge data  up  to the point where the increase in variance essentially
stopped.

The correlation between the predicted and observed concentration variance is
.94.  This  indicates that (9) is more  capable  of  estimating  concentration
variance than  (8)  is.   Less than 50% of the variance of a2 is explained by
(8), while  (9)  explains nearly 90%.   Predictions  using (q) are depicted as
the dash-dot  lines in  Figures 2-8.  Since the value of the constant in (9)
is immaterial, because the expression will be differentiated to obtain K, a2
is set equal to 0  at x = 0, and = 20 when JCD/z*dx<200.

Although  (9)  is apparently a good predictor for a2, the physical  mechanisms
governing  dispersion remain undefined.   Therefore,  it  is  anticipated that
much  work  will  be  required  to  adequately  predict variance.   However,  (9)
will  be  used  to  show the  utility of  the  dispersion equation.

Note  that  in Figures 2 and 3 there is an initial  slow  increase  in variance,
then  a rapid increase.   In  both stream  reaches the  flow was  intitally  near
the  bank opposite  the dye  injection,  and  it shifted over to the  injection
bank  at the  point where  the  concentration variance  started its  rapid
increase.

In  Figures  4 and 5 the rapid increase in variance starts much closer to the
source.   In these  reaches the thalwegs were near the injection  banks at the
injection points.   Predictions  using (9) are able  to anticipate the  point  at
which  the  increase in  variance  starts, while those using (8)  cannot.

The  same reach is  depicted  in  Figures  7 and 8.   Flow  conditions are the
same,  and  the only difference  is  that  in Figure  7 the wind  is blowing  very
strongly away from the  injection bank, while in Figure 8 the  wind is blowing
strongly towards  the injection  bank.  This  may  cause a  large  enough  increase
in  turbulence near the  dye  injector to cause the  much  greater  variances
observed  in Figure 8.

If  (9) or  a similar  expression  is   used  to predict variance, then an
analytical  expression  such as (5) is not appropriate for predicting concen-

                                     75

-------
tration.  The  variance must be known in  order  to use (5) to  predict  the
concentration distribution, but the concentration distribution must be known
in order  to  determine CD and  z*.   To resolve this  dilemma,  the  dispersion
equation  must  be  solved using numerical methods.  This requires the use of
an expression  for  dispersion,  rather  than one  for variance.   Since the
concentration  distributions observed in  the  dye studies  are reasonably
normal  in the  cross-stream direction,  (6) should yield a  good estimate  of
the dispersion coefficient.  Substitution of (9) into (6) yields

                               K - -01 CD2                         ™
                               K "   ** '                        U0)
If h and u are constant, then CD = uhz* and (10) reduces to

                              K = .Oluz*.                         (11)

Since  z*  increases  as  the plume spread across a  river,  K must always
increase, according  to  (11).   However, K probably becomes constant at some
distance from  the source,  and at this  point  (10) will  no  longer be  valid.
However, the  distribution  has  flattened out so  much  that,  as  seen from  (4),
c changes relatively slowly anyway.   Therefore, (10) may  be  assumed to  be
valid  for thirteen stream widths downstream from the source without unduly
degrading the concentration distribution prediction.


                                  SECTION 4
                NUMERICAL  SOLUTION TO THE DISPERSION EQUATION

Once all of the independent variables have been determined, the equation may
be solved for c.   Expanding the right hand side (r.h.s.) of (2) and making
the appropriate assumptions about K and m  yields
                                         /\
where k = m h2uK.

Eq  (12)  implies  that K and m   change  slowly  enough in the q direction so
that  their  cross stream gradients  are  negligible  in comparison with the
gradients of u and h2.  Certainly the m  gradient is negligible so long as a
stream reach  is  relatively straight, and the manner in which  K is derived
requires that it remain constant in the q direction.

In order to solve (12) numerically, it must first be finite differenced.   To
avoid stability  problems,  the  Crank-Nicholson finite difference scheme is
employed.  In the x direction, forward differencing is  used:

                      9c _ c(m,n+l)-c(m,n)
                      9x ~       Ax       '

where Ax is an incremental distance along the  thalweg.
                                    76

-------
The  m's  indicate locations on  the  cumulative  discharge coordinate (across
the  stream)  and the  n's  indicate distance downstream  along  the  thalweg.

In the cross-stream direction,

            §c _ c(m+l,n+l)-c(m-
            8q                    4Aq

and  similarly  for  In h2 and  In u.   The Aq is  the incremental  discharge in
the  cross-stream direction.  Using the  same  type of finite differencing:
                                                                  (15)

Since  In  h2  and In u  are  known at the grid points,  their gradients  may be
determined a priori.   For this  reason

                      Y
Substituting (13),  (14),  (15)  and  (16)  into  (12), defining \ = kAx/2Aq, and
separating the concentration terms  involving the n+1 grid points from those
involving n yields
                                            -l,n)                (17)

All the quantities  in this equation are known except the concentrations at
the grid points  involving n+1.   Therefore, a  system  of  M equations in M
unknowns (the  concentrations at  the n+1  grid points) may  be developed.

To solve the  system of equations  initial  and  boundary  conditions  must be
specified.   The initial condition  is required to obtain the concentration at
the grid  points where n  = 1.   Based upon the dye  studies  a half- normal
concentration  distribution located  near  the source  is  the  appropriate
initial condition.   A half-normal  distribution may  be expressed by  (5) with
S  replaced  by  2S,  to double the  concentration (absolute  reflection).  The
concentration variance must be  known in order to use (5).  Because concen-
trations were  measured in the dye studies and the  industry example, the
initial variance  is computed from the observed  concentration.  However, a
prediction  for  a2  very near the source will be required when concentration
measurements are  not available.   When the concentration at n = 1 is deter-
mined,  (17) may be  used  in an iterative procedure,  so the concentrations at
succeeding distances downstream may be determined.

The boundary  conditions  may be obtained from the anticipated concentration
distribution.   The  half-normal distribution  implies very  small gradients in
concentration at  the banks near  the source.  Far from the source, a uniform
concentration distribution again implies  very small gradients.  Therefore,

                                     77

-------
the appropriate boundary  condition  is  9c/3q
this is the case if:
                                             = 0 at  the  banks.   From (14)
          c(o,n) = c(2,n), c(o,n+l) =
                                                c(M-l,n) =
Expanding (17) into M equations and employing the boundary conditions yields
a system of  simultaneous  equations which may be written  in  the form of a
matrix equation
where t is the column vector
                                c(M,n+l)
and F  is the column vector
                                                                   (18)
and A is the tridiagonal matrix
                                         •0   0
                                                  2\
                                                  Aq
                                     78

-------
Due to the tridiagonal  nature  of  A,  it  is  possible to  invert the matrix and
solve for C directly, or an iterative approach may be used.

The mathematical formalism  to  determine  the concentration distribution from
a continuous point  source  located on the bank is now complete.   To reiter-
ate, the concentration  is  first  initiated by  assuming  a  Gaussian  distri-
bution.   Computation  of the dispersion coefficient is accomplished in step
two using  (10).  Averaged values  of  h and  u are  used,  and K is incorporated
in  the  A's in  the  tridiagonal matrix.   Concentration at the next grid
distance downstream  is  determined using the last computed concentrations,
the newly  computed dispersion  coefficient, and  the  required  hydrological
parameters which are  obtained  from the  observed  flow  and  stream morphology.
Step 2  is repeated  using  the  new concentration data,  establishing  an
iterative procedure  which  is repeated until the  concentration at  the grid
points farthest from  the source is determined.

If  the  input  data  is totally consistent, then the procedure described will
yield a  concentration distribution which conserves mass.  This, however,  is
often not  the case,  and can cause errors in the  concentration computations.
In  order to ensure conservation of mass, a variational formalism is employed
(Hutcheson, 1981c).

A numerical solution to the diffusion  equation  is not new, and a  computer
program could have been readily obtained from  several sources.   However, the
concept  of  allowing  the model   to determine the  dispersion  coefficient at
each grid  distance  downstream  based on  the dimensions of the plume and the
hydrological  characteristics of  the portion of  the  stream  affecting dis-
persion  is  unique.   Therefore, it  is  deemed  more efficient to build the
model  from basic principles  rather than  trying to   modify an  existing
computer program.

                                   SECTION  5
                             MODEL VERIFICATION

Data obtained during  the dye studies  is  used to  verify the dispersion model.
The dye  concentrations  were  measured  on  cross  sections extending to thirteen
stream  widths  downstream from  the  source  (Hutcheson  and Gopal, 1981b).   A
comparison of the actual dye concentrations with  those predicted is given  in
Figures  9-12.   Only the best  comparisons  are  shown here.  Eighteen others,
some where the  comparisons  are  not nearly  so good, are displayed in the Task
111 Final  Report  (Hutcheson,   1981c).   The x's  represent the actual dye
concentrations  and  the solid lines represent  the predictions.  The vertical
dashed  lines  represent the point at  which the cumulative discharge equals
one-fourth  the  total  discharge,   i.e.,  the edge  of  the zone  of passage.
Initialization  occurred on  cross  section A, using the  observed concentration
variance.   The  cross sections  are  essentially equidistant from each other,
with  A  being  this distance  from  the  source.   The cross sections in Figures
9-12  are 2,770, 2,029, 1,255,  and 854 feet from the source,  respectively.
                                     79

-------
                                                                 t
l NOU*V1N33H<
                                             §   g
                                                      E  »
                                                      |  f
                                  80

-------
-s
•s
 81

-------
The  model  adequately reproduces  the  concentration distributions for the
cases presented.  This does not mean that in every instance the measured and
computed concentrations  are  identical.   In some cases major discrepancies
between actual  and  predicted  concentrations exist.  More refinement of the
model is required  before all  concentration distributions  can be predicted
accurately.

When no concentration data is available, initialization must occur  close to
the  source, where the concentration variance is known to be very small.  If
the  plume  occupies  only a small portion of the discharge, then small grid
intervals  are  required  in order to have enough points within the plume to
achieve  adequate  concentration  computations.   However,  smaller   grid
intervals  require more  grid  points,  which require added manual  analysis of
the data and more entries to the computer.   Therefore,  it is not feasible to
initialize the  model  on an operational basis unless an objective analysis
scheme is developed to alleviate the manual analysis and data entry problem.
Since the  initial  variances  measured  in the two stream reaches examined on
the  Washita River are very small,  a finer  grid mesh than is employed by the
dispersion model is  required.   For this reason, these cases were not run.

It is very difficult to make the input data compatible with the dispersion
model using a manual analysis.   The amount of  time required to manually
adjust the  flow velocity, depth,  and total discharge on a  grid dense enough
for  use in the dispersion model  is prohibitive, especially  since  m  must
also be  accounted  for.    Therefore,  an objective  analysis scheme  which
ensures compatibility between the  model and input  data must be developed to
ensure conservation of mass.

In order  to make the model valid for  all  situations,  a better dispersion
coefficient is required.  The  K  currently used in the model is not related
closely enough  to the turbulence  of the flow to be useful  in all situations.
For  example,  the effect of bottom roughness on  turbulence  is  not taken  into
account.   Therefore, the  current dispersion  coefficient  cannot be valid
except for the  smooth sand or mud  bottoms  on which it was  developed.

                                   SECTION  6
                      WASTELOAD ALLOCATION METHODOLOGY

The  wasteload  allocation technique uses the  assimilation  capacity of the
receiving  stream as a  basis  for  permitting.  This is  accomplished through
the  use  of a  dispersion model.   It is  recognized,  however, that technology
based  permit  limits and  socio-economic considerations are also vital parts
of the wasteload allocation process.   Even when the assimiTation capacity  is
virtually  unlimited, dischargers  should be required  to meet some  minimal
level  of  treatment, such as  best practicable treatment.   In some cases the
assimilation  capacity is so small  that  it  is impossible to achieve  stringent
water  quality standards, so  socio-economic factors come  into play.  These
factors are not considered here,  however.
                                     82

-------
To obtain wasteload allocations,  concentrations generated by the dispersion
model must  be used.   Because the principle  of superposition  applies to
conservative  parameters,  it  is  easy  to extend  the  dispersion model,
developed to  predict  concentrations  for a  single source, to  multiple
discharges.   Therefore, the  model may be run  for  each discharger, and the
results added to determine total  concentration at the grid points.

The wasteload allocation  process  is designed to  use  the  model  results to
determine permit  limits which restrict discharges  so  that concentrations on
the boundary  do  not exceed standards.   First  the  background concentration
must be  estimated.  This  may  be done  in  any  one of several ways (Hutcheson,
1982).    Next, concentrations  on the boundary formed by the mixing zone and
the zone  of  passage,  as shown  in Figure 1  are determined using the  dis-
persion  model.   For wasteload allocation purposes,  any  reasonable source
strength, S,,  may be  used.   However,  for multiple discharges the source
strengths must be related to  each other  so that

                              S.  = a.S1  2
-------
                                  SECTION 7
             APPLICATION OF THE WASTELOAD ALLOCATION METHODOLOGY
                        TO AN INDUSTRIAL POINT SOURCE

Farmland  Industries  discharges  into Skeleton Creek.  Velocity,  depth,  and
water  samples  were  taken  at appropriate  locations  in the  mixing zone
(Hutcheson and  Gopal,  1981a).   The water samples were  analyzed for zinc,
chrome,  and  total dissolved  solids (TDS).   Farmland Industries does  not
discharge continuously.   However,  its period of discharge is assumed  long
enough  for  a steady state situation  to  be  established within  the  mixing
zone.   Therefore,  the wasteload  allocation  methodology may be applied.

Even though  wasteload  allocations  for three different parameters were per-
formed,  it was  necessary to run the dispersion model only once to obtain C
for  a  given S-^  (Hutcheson,  1982).   The  background concentrations were
obtained  from analyzed samples  and C1 for each parameter from the Oklahoma
Water Quality Standards.  These standards have since been revised.   Eq.  (20)
was used to obtain the source strength (proposed permit limit) for the three
parameters.

The wasteload allocations  for Farmland Industries can  be made  without  the
use  of  the  dispersion  model,  since  the  maximum concentrations on the
boundary  are known  from an analysis of the  samples  collected.   Therefore,
the observed concentrations  and  observed source strengths may  be  used  in
(20) to determine allowable  wasteloads.   A comparison  of  the  proposed
allowable loads is shown in Table 1.

Table 1.  Comparison of wasteload allocations.
 C-B         S,
(mg/L)    (mg/Sec.)
Wasteload
Using Observed
Concentration
  (mg/sec.)
Wasteload Using
Dispersion Model
    (mg/sec.)
                                  Chromium
.2435      28.664


.2745       27.56


 1619     524,872
     4.12

 Zinc
    499.4

  TDS
  116,415
       6.83


      962.1


   69,456.5
 Wasteloads using the water  samples  are  shown in column three, while those
 obtained using the dispersion model  are  shown in  column  four.   The  first  two
 columns show the difference  between  the observed maximum concentration on
 the boundary  and  the  background,  and  the  observed source  strength,
                                      84

-------
respectively.   For  chrome  and zinc, the allowable wasteload produced using
observed concentrations  is smaller than that  using  the  dispersion model.
However, for  IDS  the allowable wasteload produced using  observed concen-
trations is larger.

While there are discrepancies between the wasteload allocations,  they are
relatively insignificant.  A more important comparison is between the waste-
loads and  column  2, which represents the actual wasteload being discharged
at the  time of observation.   The  allowable wasteload (represented  by either
column 3 or 4) for chromium and IDS is much lower than the actual wasteload.
Therefore,  regardless of which  method  of computing maximum allowable waste-
load is  used,  a  great reduction  in the  observed wasteload is required to
meet the water quality standards which were applied.   However, the allowable
wasteload  for  zinc  is much larger than the observed wasteload.   Therefore,
zinc loading  could  be increased without violating water quality standards.

                                  SECTION 8
                                  CONCLUSIONS

A wasteload allocation  methodology has been  developed.   It does  not  rely on
the  assumption of a uniform  concentration distribution, and can be applied
to either  isolated  or multiple outfalls.  The methodology provides a means
for  determining  the  wasteload  allocation  required to meet water  quality
standards, thereby  protecting the beneficial  uses assigned to a receiving
stream.  The  central  component  of the  methodology is a dispersion  model for
the  mixing zone for conservative  parameter discharges.

There are  several  improvements  which  should  be made  to the dispersion model
in order for  it  to be  used  on  an operational  basis  in the wasteload allo-
cation  procedure.   Input to the  dispersion  model  is  much too tedious for
operational use,  so  an  objective analysis  must be developed to  minimize
manipulation  of data to make it  more compatible with the dispersion model.
The  initialization  procedure is  too  dependent upon  the judgement of the
person  operating  the model;   it must be  made more automatic.  The  determi-
nation  of  the dispersion coefficient is not based on physical  principles,  so
more data  should be gathered  and  a  coefficient which explains the dispersion
mechanism  more fully should  be developed.   Other mechanisms  which affect
concentration, such as  settling,  must also be incorporated into the model.

There are  many situations  in  which  the wasteload allocation described is not
applicable.   Because  it is based upon a steady  state mixing zone model for
conservative  parameters, it  cannot be  used  if the discharge interval is too
short to allow steady state conditions to develop in the mixing zone, if the
pollutant  is  not "conservative," or  if  a  classical  mixing zone  is not
appropriate for the receiving water body.  This  occurs when there is no flow
velocity,  as  in a lake, or when the discharge forms  a majority of  the flow.

However, even in  its present form, the utility  of the wasteload allocation
methodology has been shown.   It  has been demonstrated that the  dispersion
model can  predict concentration distributions well enough to  develop permit
limits  which  will  allow water quality  standards  in a receiving stream to be
met.

                                     85

-------
                              REFERENCES

Eidsness,  Frederic  A.   1982.   Prepared  Text on Water Quality  Standards.
     Presented at the Governor's Water Conference, Oklahoma City.

Hutcheson,  Main  R.    1979a.   Select and Verify  Hydrologic Model(s)  for
     Conservative Parameters.   Final Report,  208 Work  Element No. 613,
     Oklahoma Water Resources Board.

Hutcheson, Main R.  1979b.  Analysis of Ensemble Averaged Concentrations and
     Fluxes in a  Tracer Puff.   EPA-600/4-79-002, National Technical Infor-
     mation Service, Springfield, VA.

Hutcheson,  Main   R.    1980.  Refine and Update  Model  for  Conservative
     Parameters.   Working Paper, 208 Work Element  111.   Oklahoma Water
     Resources Board.

Hutcheson,  Main  R.  and B. K. Gopal.  1981a.   Mixing Zone  Survey.   Final
     Report,  208  Work  Element  No.  245, Oklahoma  Water  Resources Board.

Hutcheson,  Main   R.   and  B.  K.   Gopal.   1981b.   Model Calibration and
     Verification Studies  for Conservative Parameters.   Final  Report,  208
     Work Element 232.  Oklahoma Water Resources Board.

Hutcheson,  Main   R.    1981c.   Refine and  Update Model for Conservative
     Parameters.   Final Report, 208 Work Element  111.   Oklahoma  Water
     Resources Board.

Hutcheson, Main  R.   1982.   Develop a Wasteload  Allocation  Methodology  and
     Determine Maximum  Allowable  Loads  for Conservative Parameters.  Final
     Report, 208 Work Element 121.   Oklahoma Water Resources Board.

Meyer, W.   1977.  Transverse Mixing in the Mobile River, Alabama.  Journal
     Research, U.S.  Geological  Survey, Volume 5, No. 1.

Neely, W.  Brock.  1982.  The Definition and Use of Mixing Zones.  Environ.
     Sci. Techno!. Vol. 16, No.  9, pp.  518-521.

Oklahoma Water Resources  Board.   Oklahoma Water Quality Standards  (1982).

Yotsukura,  Nobuhiro  and William W.  Sayre.  1976.   Transverse  Mixing in
          Natural Channels.   Water  Resources  Research,  Vol.  12,  No.  4,
          695-704.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency.  The contents do not necessarily reflect the views of the
Agency and no official endorsement should be inferred.


                                    86

-------
    SOME RECENT ADAPTATIONS AND APPLICATIONS OF QUAL-II IN THE NORTHEAST

                              presented at the

           Stormwater and Water Quality Model Users Group Meeting
           University of Florida, Gainesville, January 27-28, 1983

                                     by

               William W. Walker, Jr., Environmental Engineer
               1127 Lowell Road, Concord, Massachusetts 01742

                                Introduction

     The  QUAL-II  model  has an extensive history which can be traced to the
original work of Streeter and Phelps (1925)  and  has  been  widely  used  in
wasteload  allocations  and  other  aspects  of  river  basin  water  quality
management.  Several versions of  the  program  have  appeared  (Texas  Water
Development Board, 1970, Water Resources Engineers, 1972, Meta Systems, 1979,
Roesner  et  al.,  1981).   Generally,  all versions provide a capability for
simulating  longitudinal  transport  and  transformation  of  water   quality
components  in  one-dimensional,  vertically-mixed  systems with steady-state
hydraulic conditions.  This paper describes certain modifications which  have
been  made  in the model structure to improve simulations of dissolved oxygen
conditions in rivers heavily impacted by photosynthesis.  The code  has  also
been  adapted  for use on microcomputers.  The development of this version is
traced to wasteload  allocation  studies  in  Vermont  (Meta  Systems,  1979,
Vermont  Department  of Water Resources, 1982) and to recent studies in Maine
(Walker, 1982) and Massachusetts (Walker, 1983).


                             Model Modifications

     The simulation of nutrient cycles and algal growth kinetics tend  to  be
more  important  in  rivers and shallow impoundments with low velocities.  In
these situations, photosynthesis by suspended phytoplankton, aquatic  plants,
and/or  periphyton  may represent important components of the oxygen balance.
Several alterations have been  made  in  the  QUAL-II  structure  to  improve
simulations  of  phytoplankton, nutrients, and oxygen under these conditions.
These include:

    (1) addition  of  detrital  organic  phosphorus  and   organic   nitrogen
        compartments;

    (2) provision  for  algal  uptake of ammonia and/or nitrate nitrogen (vs.
        nitrate only in previous versions);

    (3) provision  for  self-shading  by  phytoplankton;   (computing   light
        extinction  coefficients  as  a function of chlorophyll introduces an
        important feedback control on peak biomass  levels  in  nutrient-rich
        environments);


                                      87

-------
    (4) specification of alternative (vs. multiplicative) nutrient limitation
        by nitrogen or phosphorus;

Control  pathways are depicted in Figure 1.  The addition of detrital organic
nutrient compartments essentially closes  the  nutrient  cycles  and  permits
model  calibration  and  testing  against  observed  total nitrogen and total
phosphorus  data,  as  well  as  individual  nutrient  species.   The   other
modifications are designed to reflect the kinetic formulations used in state-
of-the-art  phytoplankton models, as applied to lakes and estuaries (Di Torro
et al., 1977).  Details on the  equations  and  functional  forms  are  given
elsewhere  (Vermont  Department  of  Water Resources, 1982, VanBenschoten and
Walker, 1982).
     Other additions and structural modifications used  in  the  applications
discussed below include:

    (1) calculation   "Apparent   BOD-5"  concentrations  as  a  function  of
        carbonaceous BOD and estimated 5-day algal respiration at  20  deg  C
        (permits calibration and testing against observed BOD-5 data);

    (2) calculation  of  Secchi  depths  from chlorophyll and non-algal light
        extinction  coefficients  (also  useful  in  calibration  and  impact
        assessments) ;

    (3) output   of  several  diagnostic  variables  useful  for  identifying
        controlling processes, including  breakdowns  of  oxygen  sources and
        sinks  (g/m3-day)  and  algal  growth  factors  in each computational
        element;

    (4) provision for specifying benthic sources and/or sinks for  any  state
        variable  by  reach  in  g/m2-day, including plant photosynthesis and
        respiration;

    (5) provision for including dam reaeration at the downstream end  of  any
        reach;

    (6) provision  for  simulating  eddy-diffusive  exchanges with downstream
        water bodies by specifying  far-field  concentrations  and  effective
        dispersion rate (useful for simulating fresh-water estuaries or other
        backwater  situations  where  the  downstream  water quality boundary
        condition is fixed);

    (7) provision for estimating longitudinal dispersion coefficients using a
        function  developed  by  Fischer  et  al.  (1979),  reportedly   more
        realistic than the Elder (1959) equation;

    (8) provision  for  simulating  diel  oxygen  fluctuations  attributed to
        photosynthesis  and  respiration  using  an  approximate  formulation
        described by DiTorro (1968);

The   last   modification   permits   calculation  of  daily  minimum  oxygen
concentrations without applying the model in a non-steady-state  mode.   Diel
flucutations are calculated around the steady-state solution as a function of


                                     88

-------
                 Figure 1

Control Pathways in the Modified QUAL-II Model
              ATMOSPHERIC
              REAERATION
r+ORGANIC N
  AMMONIA N
          \
BENTHIC DEMAND


C-BOD 	
  NITRITE N
  NITRATE
ORGANIC P
    i
DISSOLVED P
            CHLOROPHYLL-A
                (ALGAE)
           BOTTOM  SEDIMENT
                   89

-------
reaeration rate, photosynthesis rate, and day length.
     In  the  Vermont  application,  most  of  the  above  changes  have been
implemented by modifying certain subroutines in the EPA code and  adding  new
subroutines  for  simulating  detrital  organic  nutrients  and  diel  oxygen
variations.  A SAS (SAS Institute, 1979) interface has  also  been  developed
for   manipulating   model   output   and  plotting  observed  and  predicted
concentration profiles (Walker, 1980).   The  interface  has  been  extremely
useful in streamlining calibration, testing, and report writing.
     In  subsequent  applications,  the  program  code  has  been  completely
rewritten and adapted for use on  a  microcomputer.   The  micro  version  is
written   in   standard   FORTRAN  and  uses  overlays  to  overcome  storage
limitations.  It has also been run on mainframes.  Experimentation  indicated
that  a  version  written in intrepreted BASIC would be too slow and storage-
limited to  be  practical  in  most  applications.   Application  limits  are
determined  primarily  by  the  number  of  computational  elements, computer
memory, and the memory overhead consumed by the FORTRAN compiler, linker, and
input/output operations.  On a system with  approximately  40K  of  available
storage  (64K  less  overhead),  the  current  version  can  handle up to 190
computational  elements.   Provisions  for  branching   or   non-steady-state
simulations  are  not included, primarily because they have not been required
in applications to date.  These restrictions could be lifted with  additional
code modifications and overlaying.  The micro version has been tested against
the  modified  EPA  code using input files developed for the Vermont Winooski
River application.  A separate program has also  been  written  for  plotting
observed  and  predicted  profiles.   Execution  times for a typical problem,
including printing and plotting, are on the order of ten minutes on an 8-bit,
4 mhz microcomputer equipped with a special  arithmetic  chip  available  for
most systems.
     Upgrading  of  the  algal  growth  kinetics (especially the inclusion of
self-shading) reduces the linearity of the equations has been found  to  pose
problems  for  the  steady-state  solution algorithm, particularly in systems
with long hydraulic residence times.  In the current  EPA  code  (Roesner  et
al.,  1981),  convergence is guided and tested based upon algal growth rates.
In the modified code, convergence is tested based upon stability of the state.
variables from one iteration to the next.  Certain modifications of the  tri-
diagonal   matrix   formulation   have  also  been  successful  in  improving
convergence properties, especially in systems with long  residence  times  or
significant algal growth limitation by self-shading or nutrients.
                     Calibration and Testing Procedures

     Model  inputs  may be broadly classified as (1) "boundary conditions" or
(2) "system parameters".  Boundary conditions include such factors  as  river
flows,  waste  inputs,  morphometry,  and  climate;    these  can  be directly
measured  or  independently  estimated.   System   parameters   include   the
fundamental  rate  and stoichiometric coefficients which are used in process-
level simulations and which are usually difficult to  measure  directly.   In
simulating  a  given river system, the boundary conditions vary from one time
period to another, depending upon various driving forces,  while  the  system
parameters should be relatively constant.
                                     90

-------
     The  development  of  input estimates within each category is based upon
combinations of the following:

    (1) direct monitoring data
    (2) empirical functions (e.g. reaeration rate formulae)
    (3) literature values and guidance manuals (e.g., Zison et al., 1978,
        NACASI, 1980)
    (4) empirical adjustment ("tuning") to observed water quality profiles
    (5) field reconnaissance

The calibration process would be relatively straight-forward  if  all  inputs
could  be  directly measured.  This is generally infeasible, however, because
of the complexity of the model, implicit nature  of  some  coefficients,  and
limitations in monitoring resources and technology.  Calibration of this type
of  model  requires  subjective judgments because the feasible ranges of most
coefficients are wide and more than one set  of  coefficients  can  often  be
selected  to  fit  a  given set of field data.  Because of these limitations,
user experience  and  field  reconnaissance  are  valuable  assets  in  model
applications.   Sensitivity  analysis  should  be  employed  to determine the
importance of parametric and structural assumptions.
     One test of model generality is based upon  stability  of  the  process-
level parameters from one time period to another in a given river.  This test
involves  simulating  observed  water  quality  conditions under at least two
different sets of boundary conditions using a fixed  set  of  parameters.   A
more  severe  test  of  generality would require parameter stability from one
river to another.  A model with this property, while  difficult  to  achieve,
would  be  extremely useful because it would ease the calibration and testing
requirements for each application.
     Because of the desirability of achieving generality  and  learning  more
about  the  strengths  and  weaknesses  of  the model structure, a relatively
conservative approach involving minimal adjustment of system  parameters  has
been  taken  in  calibrating  the model to the systems described below.  Most
parameter adjustments  have  been  limited  to  characteristics  which  could
logically  vary  from one river to another, such as benthic photosynthesis or
benthic oxygen demand; these are essentially forcing functions, as opposed to
system   parameters.    Stoichiometric   coefficients   (e.g.,    respiration
equivalents,  algae  chlorophyll-a  and  nutrient  contents)  and  many  rate
coefficients have been held  constant  at  "reasonable"  values,  based  upon
literature ranges and accumulated experience with the model.
     Many  parameters have been estimated with the aid of published empirical
relationships (e.g. reaeration rate), rather than adjusted to fit  individual
profiles.   Direct field measurements of reaeration rate would be preferable,
but  are  usually  infeasible  because  of  time  or  economic   constraints.
Variations  in  certain  process-level parameters, such as algal growth rate,
respiration rate," and settling velocity, from one reach to another in a given
river would introduce too many degrees of freedom in the calibration  process
and  are  less  defensible on a scientific basis; accordingly, reach-to-reach
adjustments in these parameters are not part of the calibration procedure.
                                      91

-------
                                Applications

     The model has been applied in studies  of  three  river  basins  in  the
Northeast:   (1)  Lower  Winooski  River,  Vermont (VanBenschoten and Walker,
1982); (2) Upper East Branch of the Sebasticook River, Maine (Walker,  1982);
and  (3)  Sudbury/Concord  Rivers, Massachusetts (Walker, 1983).  While there
are several unique features associated with each of these  applications,  all
include  low-velocity reaches in shallow impoundments or backwater areas with
high concentrations of algae (generally greater then 30 mg  chlorophyll-a/m3)
during  critical periods.  Key parameter estimates are summarized in Table 1.
Model generality is reflected by  the  stabilities  of  some  parameters  and
instabilities  of  others  from  one  system to another.  Each application is
described briefly below.
     The Winooski River originates in the  Green  Mountains  of  Vermont  and
empties  into  Lake  Champlain.   In  its last 32 kilometers, the river flows
through the metropolitan Burlington area, where there are two hydropower dams
and several industrial and municipal point  sources.   Violations  of  the  6
mg/liter  dissolved  oxygen  standard  have  generally been observed near the
mouth of the river, where the elevation  gradient  is  relatively  small  and
where maximum summer algal populations of about 30 mg/m3 are generally found.
The oxygen violations are usually associated with large diel fluctuations (up
to  8  mg/liter),  attributed to combined effects of (1) algal photosynthesis
and respiration and (2) daily  fluctuations  in  river  flow  resulting  from
hydropower  operations  (typically  50  to  1200  cfs).  Sustained periods of
oxygen  violations  have  also  been  observed  at  various  locations  under
conditions  of  maximum  temperature  (>  30  degrees C) and die-off of algal
blooms.
     The Vermont Agency of  Environmental  Conservation  (VAEC,  1980,  1982)
conducted   a   wasteload  allocation  study  in  order  to  assess  possible
requirements for advanced waste treatment.  The unsteady flow regimes induced
by hydropower peaking operations posed several  potential  problems  for  the
modeling effort.  Because of the complexities and extensive data requirements
involved  in  application  of . a non-steady-state hydraulic and water quality
model, the Agency elected to use a steady-state  model  and  to  conduct  the
allocation  runs  under steady 7Q10 conditions, under the assumption that the
utilities would be required to pass a minimum of 7Q10  during  non-generating
hours.  The model described above was calibrated and tested against data from
week-long  intensive  surveys conducted during two different years.  With the
cooperation of the power company,  stable  flow  conditions  were  maintained
during both surveys to provide suitable data sets for calibration and testing
of  a  steady-state model.  Observed and predicted concentration profiles for
the second survey are shown in Figure 2.
     The Sebasticook study (Walker, 1982) examined the impacts of a  combined
municipal/industrial  discharge  on  a  small stream which discharges into an
inlet of Lake Sebasticook, a eutrophic lake in central  Maine.   The  shallow
inlet  is  characterized  by  high  algal  and  aquatic  plant  densities and
hydraulic  exchanges  with  open  lake  waters  are  probably  important   in
determining  concentrations  at the lower end.  Preliminary model simulations
indicate that the inlet is functioning as an  oxidation  pond,  since  oxygen
turnover  rates  in  the  water  column consist largely of photosynthesis and
respiration by algae and aquatic plants.  A detailed evaluation  is  hindered
by  lack  of  diurnal  sampling  for dissolved oxygen and limited spatial and


                                     92

-------
                                  Table 1
Parameter Estimates Used in Winooski, Sebasticook, and Sudbury Applications
 Parameters
 Value/Comments
 Reaeration Rate
O'Connor & Dobbins (1958) equation
constrained to K2 > 1 / mean depth (Su)
 Longitudinal Dispersion    Elder(1959) (Wi), Fischer (1979) (Se,Su)
 Decay/Oxidation Rates (I/day)
     BOD-U
     Ammonia N
     Nitrite N
     Organic N
     Organic P

 Algal Parameters
     Maximum Growth Rate
     Respiration Rate
     Settling Velocity
     Chlorophyll Content
     P Content
     N Content
     Light Extinction
     Ammonia Pref. Factor
     Photo. Oxygen Equiv.
     Resp. Oxygen Equiv.
.2 + bed activity (Zison et al.,  1978)
.2-2 (Wi), .3-1 (Se), .6 (Su)
2.0 (Wi.Se), 3.0 (Su)
.1
.1
2.3 I/day (Wi.Su), 2.5 I/day (Se)
.12 I/day
.75 m/day (Wi.Su), .60 m/day (Se)
.010 mg Chl-a / mg Algae
.011 mg P / mg Algae
.080  mg N / mg Algae
43.2 m2/g Chl-a
.9
1.6 mg 02 / mg algae
2.0 mg 02 / mg algae
 Half-Saturation Constants
     Algal P Uptake          .005 g/m3
     Algal N Uptake          .03 g/m3
     Algal Growth vs. Light  1.5 calories/cm2-hr

 Benthic Oxygen Fluxes (g/m2-day)
     Plant Photosynthesis    0-2 (Wi), 0-2 (Se), 1.5
     Plant Respiration       0     (Wi), 0-2 (Se), 1.5
     Other Benthic Demand    .5    (Wi), .5    (Se),   1
                              10 (Su)
                              10 (Su)
                               3 (Su)*
 Other Benthic Fluxes (g/m2-day)
     Dissolved P Source **  0 -  .005 (Se,Su)
     Ammonia N Source   **  0 -
     Ammonia N Sink         .05
     Nitrate N Sink         .10
     025 (Su, channel)
         (Su, overbank)
         (Su, overbank)
   Wi = Winooski, Se = Sebasticook, Su = Sudbury
   parameters are zero in other applications if some initials are given
   parameters are used in all applications if no initials are given
   ranges refer to reach-specific values, all rates at 20 degrees C
*  plus wetland benthic loading component
** benthic sources of phosphorus and ammonia in impounded reaches
                                     93

-------
                                  Figure  2

            Observed and Predicted Water  Quality  Profiles
                        Lower  Winooski  River 1979
   Dissolved Oxygen (mg/liter)   1979

 12-

 10-

  8

  6

  4-

  2

  0-i
    0     4     8     12    16    20
               River  Mile

   5-Day BOD (mg/liter)         1979
     	C-BOD
      —C-BOD + Algal Respiration
3.5

3.0

2.5-
2.0

1 .5-

1 .0-

0.5
0.0
          4,8     12    16    20
               River  Mile
                               1979
Chloropbyll-a (ug/liter)
                                         Nitrogen (mg/liter)
                                            	Total N
                                            --Total Kjeldahl N
                                                                        1979
                                                            12    16
                                                     River  Mile
                                                                           20
                                          Nitrogen (mg/liter)          1979
                                            	 Nitrate-N + Nitrite-N
                                            	Ammonia-N
                                          .10

                                          .08

                                          .06

                                          .04

                                          .02

                                          .00
Phosphorus (mg/liter)
  	Total P
  	Dissolved P
                                                                        1979
                                 20
                                                 4     8     12    16     20
                                                     River  Mile
                                     94

-------
temporal sampling frequencies for  all  water  quality  variables.   In  this
application,  reasonable  simulation  of  observed  chlorophyll and afternoon
oxygen (between predicted daily mean and daily maximum) profiles was achieved
with minor adjustments in the parameters used in the Winooski simulations.
     The model has been recently applied to  the  Sudbury/Concord  Rivers  in
Massachusetts  (Walker,  1983),  as  part  of  an assessment of the potential
environmental  constraints  involved  in  diverting  waters  from  the  upper
watershed  for  water  supply  purposes.   The  study  extends  over 51 river
kilometers, 44 of which  consist  of  a  shallow  impoundment  surrounded  by
wetlands,  including  the  Great  Meadows National Wildlife Refuge.  Both the
model  and  monitoring  data  indicate  that  water  quality  conditions  are
controlled  largely  by  the  hydraulic  geometry  and  by  interactions with
tributary wetlands and that point sources are of minor importance.  Hydraulic
simulations using HEC-II and detailed channel cross-section  measurements  at
278  locations have been used to define hydraulic geometries in water quality
model.
     As a result of the channel  and  floodplain  morphometry  and  backwater
effects,  the river undergoes relatively large changes in width as a function
of flow in certain reaches.  Bottom sediments in  the  channel  and  overbank
areas  are  highly  organic  in  nature and reflect the export, settling, and
decay  of  organic  materials  from  adjacent  wetlands.   Figure  3  depicts
variations  in dissolved oxygen and dissolved oxygen deficit as a function of
river flow measured during summer months at the lower end  of  the  impounded
area  most  stongly influenced by adjacent wetlands.  A unique aspect of this
system is that dissolved oxygen  levels  tend  to  be  lower  (and  deficits,
higher)  during high-flow periods.  The relationships in Figure 3 reflect the
combined influences of (1) higher organic loadings during high-flow  periods;
(2)  higher  benthic demands in overbank areas; (3) less algal growth because
of the increased flushing rate during  high-flow  periods;  (4)  die-off  and
decay  of  wetland  vegetation during infrequent summer flooding events.  The
last condition results in the most severe water quality conditions, including
depression of oxygen levels below 2 mg/liter over  most  the  impoundent.   A
significant fishkill was reported in 1938 following the largest recorded July
flood.
     Preliminary  calibration of the net benthic oxygen demands in each model
reach  under  high-flow  and  low-flow  conditions  indicated  that   demands
increased  with the ratio of tributary wetland area to river surface area and
increased with average basin runoff.   Accordingly,  a  simple  mass  balance
model  linking  the  apparent  benthic  demand to tributary wetland areas and
runoff rate has been used to represent wetland interactions.  The  calibrated
wetland  organic  export  rate of .17 grams of oxygen demand per square meter
per day represents only about 1.3 to 2.1 percent of the literature range  for
net  primary productivity of freshwater emergent macrophytes on fertile sites
in temperature regions (Wetzel, 1975).  The remainder of the  organic  matter
produced  in  the" wetlands  is  apparently  decomposed  in  place or flushed
downstream  with  little   decomposition.     Another   wetland   interaction
considered  in  the  model  is  benthic  uptake  of  nitrate (.1 g/m2-day) in
overbank  areas  and  is  attributed  to  plant  uptake  and  denitrification
supported by organic substrates, as observed in other wetland systems (Kadlec
and Kadlec, 1978).
     The  model  has  been  calibrated  and  tested  against  data from three
intensive surveys (1) August 1973, summer low-flow;  (2)  July  1973,  summer

                                    95

-------
                     Figure 3

Summer Dissolved Oxygen Concentration and Deficit
  at Route  117  vs.  Concord River  Flow at Lowell
    14.0


    12.0


    10.0


     8.0'


     6.0


     4.0-


     2.0


     0.0-i
               -range of diel measurements
June - Sept.
o « Concord DNR
x - Mass. DEQE
      1.7      2.1       2.5      2.9

                LOG(FLOW,CFS)
                                 3.3
   10.


 	I
 \
 CD  6.

 ^  4.
 I—
 U  2-

 £  o.
 (=3
  . -2.


 a" -4-

   -6.


0
o
'.
0
X
0 0
X
o
o '
.7 2.1
0 $
i
o
0
0
o
0
o
o
j — -range of diel measurements
June - Sept.
1 o = Concord DNR
x = Mass. DEQE
2.5 2.9 3.3
               LOG(FLOW,CFS)
                       96

-------
flood;  and  (3)  June 1979, late-spring flood.  While the average flows were
similar in the  July  1973  and  June  1979  surveys,  the  major  hydrologic
difference  is  that the former occurred a week after a summer storm event of
about seven-year freqency, whereas flows were  decreasing  seasonally  during
the  latter,  when  only  flood-tolerant  vegetation  would have developed in
adjacent wetlands.  Over most of the impoundment,  diel  oxygen  fluctuations
were  less  than 1 g/m3 during the July 1973 survey, as compared with a range
of 2 - 8 g/m3 for the other surveys; this reflects  a  suppression  of  plant
photosynthesis associated with summer flooding.
     Observed  and  predicted  daily mean oxygen profiles for each survey are
shown in Figure 4.  The severe conditions  during  July  of  1973  have  been
simulated by setting benthic photosynthesis rates to zero, as compared with a
1.5  -  10  g/m2-day range calibrated to various reaches for the other survey
periods. All other model parameters are fixed for the three simulations.   An
oxygen  sag attributed to high benthic demands and low velocities is apparent
below river kilometer 48, the approximate upper end of the backwater  effects
created by the impoundment at river kilometer 7.
     Observed  and  predicted  chlorophyll-a  profiles are shown in Figure 5.
Using the same kinetic  and  stoichiometric  parameters  used  in  the  Lower
Winooski  algal  simulations,  the  model  simulates the peak algal densities
observed  during  the  low-flow  and  spring-flood  surveys  with  reasonable
accuracy,  especially  considering  that  the  observed points are based upon
single grab samples.  For the low-flow survey,  the  chlorophyll  profile  is
over-predicted  below the Assabet River (RKM 25).  The time-of-travel through
the impoundment during that survey was about 20 days and it is possible  that
the lower observed chlorophyll levels reflect higher flow conditions previous
to  the  survey.   Other possible explanations include effects of zooplankton
predation (more likely to be  important  at  long  residence  times  but  not
simulated  by  the  model),  or  shading by floating duckweed which have been
observed in this portion of the  river.   Despite  this  problem,  the  model
adequately  simulates the observed chlorophyll-a profiles in the upper end of
the impoundment, where the oxygen sag is located.

                                 Conclusions

     This paper has described adaptations  and  applications  of  QUAL-II  to
three   New   England   river  basins.   Incorporation  of  organic  nutrient
compartments and updating of algal growth kinetics increase the  realism  and
generality  of the model.  Interfacing the model output with SAS and plotting
routines  facilitates  calibration,  testing,  and  statistical  analysis  of
observed  and predicted water quality profiles.  The revised code can be used
on microcomputers or mainframes with FORTRAN capability.
     Reasonable generality is indicated  by  model  calibration  to  observed
water  quality  profiles  with  minimal adjustment in key parameter estimates
from one  application  to  another,  especially  in  those  parameters  which
determine  nutrient  and algal profiles.  Relatively large variations in some
parameters, including nitrification rate, benthic oxygen demand, and  benthic
photosynthesis,  reflect  inherent  limitations  in  this  type  of model and
dictate needs for calibration and testing in each application.  Modifications
of the code to permit systematic sensitivity analysis (Walker,  1982a)  would
improve  user  perspectives of key processes and assumptions in a given model
application.


                                     97

-------
                                 Figure  4


Observed  and  Predicted  Mean Dissolved Oxygen Concentrations
                        Sudbury/Concord Rivers
   11.000
   10.421
   9.842
   9.263
   8.684
   8.105
   7.526
   6.947
   6.368
   5.789
   5.211
   4.632
   4.053
   3.474
   2.895
   2.316
   1.737
   1.158
     .579
     .000

      60
August  1973: Mean Dissolved Oxygen  (g/m3)

   SUMMER,  LOW-FLOW
"oo   53.37   4&!?5   40.12   33.50    26ts7
                       River Kilometer
                                          2ot25   13.62    7.00
   11.000
   10.421
    9.842
    9.263
    8.684
    8.105
    7.526
    6.947
    6.368
    5.789
    5.211
    4.632
    4.053
    3.474
    2.895
    2.316
    1.737
    1.158
     .579
     .000

       60
July 1973:  Mean Dissolved Oxygen (g/m3)

   SUMMER,  FLOOD
                                           MEAN D . 0
.00   53.37   46.75   4o!l2   Sslso   2&!s7   2ot25   13.62    7^00
                       River Kilometer
   11.000
   10.421
    9.842
    9.263
    8.684
    8.105
    7.526
    6.947
    6.368
    5.789
    5.211
    4.632
    4.053
    3.474
    2.895
    2.316
    1.737
    1.158
     .579
     .000
 June 1979: Mean Dissolved Oxygen (g/m3)

   LATE  SPRING,FLOOD
                                   •   »•
                                •  >!• 'IT»
                                »T
              53.37
             46.75   40.12   33.50   26.J
                       River Kilometer
                                                   2ot25   13.62    ?toO
                                    98

-------
                            Figure  5


 Observed  and Predicted  Chlorophyll-a  Concentrations
                   Sudbury/Concord Rivers
.037
.035
.033
.031
.029
.027
.025
.023
.021
.019
.018
.016
.014
.012
.010
.008
.006
.004
.002
.000

  60"
    August  1973: Chlorophyll-a (g/m3)

     SUMMER,  LOW-FLOW
                                      CHLOROPHYLL-A
    00   53.37  46.75   40.12   33.50  26.87   20.25   13.62    7.00
                           River Kilometer
.037
.035
.033
.031
.029
.027
.025
.023
.021
.019
.018
.016
.014
.012
.010
.008
.006
.004
.002
.000
    July 1973: Chlorophyll-a (g/m3)

       SUMMER,FLOOD
  60.00   53.37   46.75
                        4ol2   33sO   26
                           River Kilometer
                                              20.25   13.62
                                                             7.00
     June 1979: Chlorophyll-a (g/m3)

       LATE   SPRING, FLOOD
.037
.035
.033
.031
.029
.027
.025
.023
.021
.019
.018
.016
.014
.012
.010
.008
.006
.004
.002
.000
  60.00   53.37   46.75
                        40.12   33.50   26.87
                           River Kilometer
                                              20.25   13.62
                                                              7.00
                                 99

-------
                                 References

DiTorro,  D.M.,   "Algae  and  Dissolved  Oxygen", in "Basic Models of Natural
Systems", prepared for U.S.  Environmental  Protection  Agency  by  Manhattan
College, New York, 1968.

DiTorro,  D.M.,   Thomann, R.V., O'Connor, D.J., and Mancini, J.L., "Estuarine
Phytoplankton  Biomass  Models  -  Verification  Analyses   and   Preliminary
Applications", in The Sea, John Wiley and Sons, Inc., New York, 1977.

Elder,  J.W.,  "The  Dispersion  of  Marked  Fluid  in Turbulent Shear Flow",
Journal of Fluid  Mechanics. Vol. 5, pp. 544-560, 1959.

Fischer, H.B., E.J. List, R.C.Y. Koh, J. Imberger, and N.H. Brooks, Mixing in
Inland and Coastal Waters. Academic Press, New York, 1979.

Kadlec, R.H. and  J.A.  Kadlec,  "Wetlands  and  Water  Quality",  in  Wetland
Functions  and  Values:  The  State  of  Our  Understanding ,  American Water
Resources Association, November 1978.

Meta Systems, Inc., "Documentation for the Meta Systems Version of the  QUAL-
II   Water   Quality  Simulation  Model",  prepared  for  U.S.  Environmental
Protection Agency, Water Planning Division, 1979.

National Council  for Air and Stream  Improvement,  Inc.,  "A  Review  of  the
Mathematical  Water  Quality  Model QUAL-II and Guidance for Its Use", Stream
Improvement Technical Bulletin No. 335, 1980.

O'Connor,  D.J.   and  W.E.  Dobbins,  "Mechanism  of  Reaeration  in  Natural
Streams", Trans. ASCE. Vol. 123, 1958.

Roesner,  L.A.,   et  al.,  "Computer Program Documentation for Stream Quality
Model (QUAL-II)", U.S. Environmental  Protection  Agency,  Center  for  Water
Quality  Modeling,  Athens  Environmental  Research Laboratory, Georgia, EPA-
600/9-81-014, 1981.

SAS Institute Inc., "SAS Users Guide", Raleigh, NC,  1979.

Streeter, H.W. and E.B.  Phelps,  "A  Study  of  the  Pollution  and  Natural
Purification  of  the  Ohio  River",  U.S. Public Health Service Bulletin 146
(reprinted 1958), 1925.

Texas Water Development Board, "Simulation of Water  Quality  in  Streams  and
Canals", Program Documentation and User's Manual, 1970.

VanBenschoten, J.B. and W.  W. Walker, "Calibration and Application of QUAL-II
to  the  Lower Winooski River", draft manuscript submitted to Water Resources
Bulletin. American Water Resources Association, August 1982.

Vermont Agency of Environmental Conservation, "Lower Winooski River Wasteload
Allocation Study - Part A:  Report of Data",  Department of Water Resources and
Environmental Engineering,  Montpelier, December 1980.

                                     100

-------
Vermont Agency of Environmental Conservation, "Lower Winooski River Wasteload
Allocation Study - Part B: Mathematical Modeling Report", Department of Water
Resources and Environmental Engineering, Montpelier, January 1982.

Walker, W. W., "A SAS Interface for QUAL-II", prepared for U.S. Environmental
Protection Agency and Vermont Agency of Environmental Conservation, 1980.

Walker, W. W., "Calibration and Application of  QUAL-II  to  the  Upper  East
Branch  of  the  Sebasticook  River between Corinna and Coburn", prepared for
Kleinschmidt and Dutting, Inc., Pittsfield, Maine, 1982.

Walker,  W.W.,  "A  Sensitivity  and  Error  Analysis  Procedure   for   Lake
Eutrophication  Modeling", Water Resources Bulletin. American Water Resources
Association, Vol. 18, No. 1, pp. 53-61, February 1982a.

Walker, W. W., "Downstream Water Quality Impacts of Diversions  from  Sudbury
Reservoir  -  Model  Calibration and Testing", prepared for Interdisciplinary
Environmental Planning, Inc., Parsons Brinkerhoff Quade & Douglas, Inc.,  and
Metropolitan  District  Commission,  Commonwealth  of  Massachusetts, January
1983.

Water Resources Engineers, Inc., "Progress Report on Contract No. 68-01-0713,
Upper Mississippi River Basin Model Project", prepared for U.S. Environmental
Protection Agency, September 1972.

Wetzel, R.G., Limnology. W.B. Saunders Company, Philadelphia, 1975.

Zison, S.W., W.B.  Mills,  D.  Deimer,  C.W.  Chen,  "Rates,  Constants,  and
Kinetics  in  Surface  Water Quality Modeling", U.S. Environmental Protection
Agency, Athens Environmental Research Laboratory, Georgia,  EPA-600/3-78-105,
December 1978.
 The work described  in  this  paper  was  not  funded  by the U.S. Environmental
 Protection  Agency.   The  contents  do  not necessarily  reflect the views of the
 Agency and  no official endorsement should be  inferred.
                                     101

-------
                A REVIEW OF MODEL USE IN EVALUATING
      NONPOINT SOURCE LOADS FROM FOREST MANAGEMENT ACTIVITIES

                        1                          2
           George G.  Ice   and Raymond C. Whittemore
                            INTRODUCTION
     The water quality of runoff from forest land is a topic of
continuing interest; particularly regarding the influence of
management activities.  Models and predictive techniques are now
being used to compare alternative management strategies.  This
paper will discuss:  (a) the unique regulatory requirements that
mandate predictions of nonpoint source (NFS) contributions for
forest lands; (b) procedures which are being developed or used to
predict sediment loading to streams; (c) areas of continuing need
for modelling development; and (d) forest industry research to
improve our understanding of forest-water quality relationships
and modelling.


 II  NEED  FOR  MODELLING  YIELDS  FROM  SILVICULTURAL  NONPOINT SOURCES
     There  is  currently  a  urgent  need  to validate  procedures  to
predict water  quality, streambed  and stream  biota  responses to
forest practices.   In part this need results  from  the  Clean Water
Act and the diverse silvicultural nonpoint source  (NPS)  control
programs developed  under Section  208.  However  even  more pressing
are the recent questions that  have been raised  about modelling in
regard to state  forest practice acts,  state  water  quality stand-
ards, and legislation covering planning on National  Forests.  In
addition there is a continuing evolution in  the perspective of
forest planners  regarding  NPS  loading  which  is  further compli-
cating both modelling approaches  and solutions.

     Like other  NPS activities, the Clean Water Act  required
that, where necessary, control programs for  silvicultural activi-
ties be developed by the states.  These programs have  emphasized
the use of  Best  Management Practices (BMP's).   The development of
BMP's has required  an understanding of the mechanisms  influencing
NPS loading.   For example, in  Florida  recommended  practices under
the voluntary  silvicultural NPS control program are  based on  a
site-sensitivity classification which  is a function  of steepness-
of-slope, soil erodibility (K-factor from the Universal  Soil  Loss
Equation),  and proximity to open  water (1).
1  Research  Forester, NCASI, Oregon State University, Corvallis,
   Oregon
2  Research  Engineer, NCASI, Tufts University, Medford, MA

                               102

-------
      A recent survey of state silvicultural NFS control programs
 found that, in general, these programs have been well implemented
 and are effective in reducing NPS loading  (2).  However, as part
 of an ongoing process, changes in BMP's continue to be considered.
 Using Florida again as an example, some changes in guidelines
 have been proposed based on an understanding of "source area
 differences" between physiographic provinces (3).  In New Mexico,
 an extensive BMP development program is ongoing.  Robert Brozka,
 Water Quality Project Forester for the New Mexico Forestry
 Division writes that,

      ".  .  . soil  movement was quantified by means of the
      Modified  Universal Soil Loss Equation using procedures
      described in the EPA handbook titled "An Approach to
      Water Resource Evaluation of Non-Point Silvicultural
      Sources'.  Soil movement on roads was Quantified sep-
      arately utilizing a mathematical sediment yield model
      developed specifically for forest roads.  This same
      model is  currently being adapted specifically to the
      State by  the New Mexico Water Resources Research Insti-
      tute.  Data  is being collected from around the State on
      runoff and sediment production rates with a rainfall
      simulator.  Ultimately, different road designs and
      erosion control practices will be evaluated with the
      model so  that the best alternative can be chosen." (4)
     In some states, predictive procedures are being used or are
proposed to be used to indicate appropriate forest management prac-
tices and compare alternatives on a site-specific basis.  In
Northern California, where the silvicultural NPS control program
is enforced through the State Forest Practice Act, an Erosion
Hazard Rating  (EHR) has been used to identify areas of high erosion
potential.  In the past the EHR has combined both surface and mass
wasting (landslide) erosion hazard.  Two recent studies (one by
California Division of Forestry and the other by Humbolt State
University) found that the EHR was not a good indicator of observed
erosion rates  (5,6).  In response to this, the California Board of
Forest has recently adopted a new EHR for surface erosion and is in
the process of developing a new EHR for mass wasting.  See Table _!_
for factors considered by the California surface soil EHR.  Rules
have recently  been adopted in California which require EHRs down
to 10 acre parcels in some cases.

     In Idaho,  USDA Forest Service operations in one watershed
were recently halted because a USDA Forest Service model pro-
jected that harvesting and reading would result in sediment
loading which would cause a 20 percent reduction in fishery
potential.  This projected 20 percent reduction in fishery poten-
tial was interpreted by the Idaho Department of Health and
Welfare as violating  state water quality standards that protect
beneficial use.
                                103

-------
      In Oregon,  a petition has  been submitted to the Board  of
Forestry which requests that an erosion model developed by  the
Siuslaw National Forest for forest-wide planning be  adopted as
part of the Oregon Forest Practice  Act to  be used  in evaluating
Forest Operations along the Oregon  Coast.   The petition requests
that operations  be limited to those that,  according  to the
Siuslaw model,  restrict sediment loading increases to less  than
200  percent of the natural rate.  A further discussion of the
model is presented later.
           TABLE 1  SURFACE SOIL EROSION HAZARD RATING
                   	FORM FOR STATE OF CALIFORNIA
            STATE OF CALIFORNIA
            DEPARTMENT OF FORESTRY
            ESTIMATED SURFACE SOIL
            EROSION HAZARD
            FORM I (12/81)
            L SOIL FACTORS
A. SOIL TEXTURE FINE MEDIUM COARSE
1. DETACHABILITY
RATING
2. PERMEABILITY
RATING
LOW
1-J
SLOW
5-4
MODERATE
10-18
MODERATE
3-2
HIGH
11-30
RAPID
1
FACTOR
RATING


Illlip
            B. DEPTH TO RESTRICTIVE LAYER OR BEDROCK
RATING
SHALLOW
f.io"
15-9
MODERATE
20-31"
DEEP
>370%
26-35

            HI.PROTECTIVE VEGETATIVE COVER REMAINING AFTER DISTURBANCE
RATING
LOW
0-40%
15-6
MODERATE I HIGH
41-80%
7-4
81-100%
3-1

            IV. TWO-YEAR, ONE HOUR RAINFALL INTENSITY (HundrtJtb Inch)
See attached map
RATING
LOW
4.40
1-3
MODERATE
40-59
4-7
HIGH
60-69
8-11
EXTREME
>«^
12-15

                                     TOTAL SUM OF F ACTORS |j
                         EROSION HAZARD RATING

<50
LOW
I 50-65
'MODERATE
66-75
HIGH
>75
EXTREME
i* ,,*v *'
THE DETERMINATION IS |
                                  104

-------
      The Forest Service has been the developer and user of
 numerous sediment loading models recently because the National
 Forests are subject to special planning requirements under the
 National Environmental Policy  Act  of 1969 and National Forest
 Management Act  (NFMA).  Under  NFMA the  USDA  Forest Service  is
 required to make projections about resource  outputs  resulting
 from management options.  The  results of these decisions  will
 influence the availability  of  timber, particularly for the
 Western United States.  However  a  recent evaluations about
 available predictive  techiques for forest resources  found that
 11.  . . it is impossible to  accurately predict  [  ] on-site and
 offsite effects  ..." and  the land manager  is currently
 ".  . . unable to define limitations to  use,  if and how manage-
 ment practices might  be applied, and possible trade-offs  in  the
 various uses and values,  even  though he is required  to do so for
 environmental analysis reports"  (7).

      On top of these  regulatory  and legislative  requirements for
 the use of models in  forest management  planning,  the forestry
 community (particularly in  the West) has recently begun to
 consider the consequences of multiple forest operations on  water
 quality.  A California Board of  Forestry task force  has raised
 the concern that ".  .  . effects  may appear to be  insignificant
 when a harvesting project is viewed alone, but may become cumula-
 tive and cause demonstrable damage when the  effects  of other
 harvesting projects are considered" (8).  Some of the potential
 cumulative effects that have been  propose include:   (a) peak flow
 increases from snow accumulation and melt rate changes in cleared
 areas, (b) peak flow  increases due to soil compaction and altered
 drainage networks, (c) channel changes  due to accelerated inputs
 of  large sediment and organic  debris, and (d) channel changes due
 to  removal of large wood  debris  and streamside vegetation (9).

      in  addition to  the  consideration  of multiple operations,
 forest  planners  are  now  being  required  to  project not only changes
 in sediment  concentration  (the  traditional NFS measure  for forest
 practice  impacts on  water  quality)  but  also  changes  in  channel
 conditions  and  ultimately  the  beneficial uses of  the  water.


     From these examples we  can summarize that there continues to
be increasing sophistication demanded in projecting the water
quality consequences of forest  activities at both the state and
federal  level.   Models are being used both in long-term planning
on federal lands and by states  as a  tool to identify general
management practices which are  appropriate for use as BMP's.   In
order for planning  decisions to be  the best possible it is
necessary that the models being used be  tested and their limita-
tion considered.
                                105

-------
          Ill    PROCEDURES  TO  EVALUATE WATER  PUALITY
                RESPONSE  TO SILVICULTURAL OPERATIONS
     Because sediment  is considered to be the most  important NPS
contribution from forest activities we will concentrate our dis-
cussion on this parameter.  It is recognized, however, that in
other more specific situations nutrients, biochemical oxygen de-
mand, temperature, and herbicides can result in localized non-
point source pollution problems.
A.   WRENSS

     In 1980, EPA published "An Approach to Water Resources
Evaluation of Nonpoint Silvicultural Sources" (WRENSS) (10).
WRENSS was written by forest hydrologists with the USDA Forest
Service and is now being used (in various modified forms) by sev-
eral states and National Forests Management groups.  Because WRENSS
involves many of the modelling elements currently used we will
provide a brief description of the sediment prediction components.

(1)  Overview - WRENSS models changes resulting  from  forest
activities that cause increases in water available for runoff,
and also increases  in both  surface erosion and landslides.  All
these modifications  are  then considered using sediment-rating
curves to evaluate  potential channel geometry changes.

(2)  Water Quantity  - Water available for runoff  is modeled using
a water-balance approach where the change in available water is
a function of changes in interception and evapotranspiration (ET)
resulting from vegetation removal.  Modifications in  interception
and ET are a function of the change in leaf area  index (LAI) or
forest cover density (CD).  Response of discharge is  given in
terms of either a seasonal  hydrograph for snow dominated  regions
or a flow duration  curve for rain dominated regions.

(3)  Surface Erosion - Sediment loading to streams is a  function
of three processes;  surface erosion, sediment from landslides
(mass wasting), and  channel scour.  Surface erosion is calculated
with a version of the Universal Soil Loss Equation which  has been
modified for application to forest sites.

     The modified soil loss model (MSLE) is:

                A =  R K  L S VM
                               106

-------
          where:    A =  The estimated  average annual soil  loss
                       per unit area  in tons/acre.
                   R =  The rainfall factor, usually expressed
                       in units of rainfall-erosivity index,
                       El, and evaluated from an iso erodent
                       map.
                   K =  The soil-credibility factor, is usually
                       expressed in tons/acre/EI units for a
                       specific soil  in cultivated, continuous
                       fallow, tilled up and down the slope.
                   L =  The slope length factor is the ratio of
                       soil loss from the actual field slope
                       length to that from a 72.6-foot (22.1 m)
                       length plot.
                   S =  The slope gradient factor, is the  ratio
                       of soil loss from a given field gradient
                       to that from a 9-percent slope.
                   VM = The vegetation-management factor,  is the
                       ratio of soil  loss from land managed
                       under specified conditions to that from
                       the fallow condition (10).
      A delivery  ratio is  calculated from  the site  conditions to
predict  the amount of soil  loss  (A) actually reaching the  stream
system.   This portion of  V7RENSS  might be  used to predict surface
erosion  frorn site-preparation activities  as are shown in Figure I.
                                                     FIGURE 1

                                                  BEDDING  AS A  SITE
                                                PREPARATION TECHNIQUE
                                                      IN FLORIDA
(4)   Landslides - In  the West,  landslides can  be major  contributor
to the sediment .loading of  forest streams.  See Figure  2.  Of  the
two  general  types of  landslides (slump-earth flows which are deep,
slow movements in plastic soils, and  debris avalanche-debris flows
which are rapid, shallow failures)  it is the debris avalanche-
debris flow  type which is most influenced by forest activities.
                                   107

-------
In order to predict landslide rates using WRENSS, a  similar  nearby
site which has previously experienced comparable management
activities must be inventoried to determine landslides under
undisturbed and man-induced conditions.  Site and practice hazard
similarity can be judged using a hazard index based  on site  condi-
tions.  From this first inventory, an accelerator factor  (F)  is
developed where:
                   F =
man-induced failures
                         natural failures
                                                  FIGURE  2

                                             I'APS VTASTING IN  THE
                                               OPECON CASCADES
     Another  inventory of  the  site being modeled, provides  the
natural delivery rate which  is multiplied by  the accelerator
factor to yield the delivery of man-induced mass failure  material.

(5)  Total Potential Sediment  and Channel Modification  -  System
modifications are combined to  predict changes  in total  sediment.
The following steps are  involved.

(a)  From the pre and post management flow-duration  curves  and a
sediment rating curve developed by stream measurements, pre and
post management suspended  sediment loads are developed.   [Stream
stability may be used in post-management sediment rating  curve
adjustments] .  This step accounts for channel.-derived sediment
load increases.  See Figure  3.

(b)  Bedload sediment rating curves are developed as above  and the
same type of analysis is used  to generate pre  and post  management
bedload rates.

(c)  Coarse and fine sediment  contributions are separated from
material delivered to the  stream from mass wasting.
                               108

-------
(d)  Total  suspended sediment  load  is determined as the  cumulative

yield from  rating curve increases,  surface erosion increases,  and

the wash  load  from mass failures.   All surface erosion material is

assumed to  be  suspendable.


(e)  Total  sediment available  for  transport is determined  by

including bedload and coarse mass-wasted material.


(f)  The  potential for channel  modification are assessed by com-

paring maximum bedload transport to introduced coarse  material.
E

Z
Z
o
H
EH

s

z
w
          EH
          z
          U
          £
          H
          a
          w
          co

          a
          w
          a
          z
          w
          a.
          tn
          CO
          O
          o
          u
           I
            1,000,000
              100,000 .
              10,000 .:;
    1,000 '
                100 ••'::
                 10 .
                          1       10      100     1000


                            INSTANTANEODS DISCHARGE IN CSM
                                            10,000
                              FIGURE 3



            SEDIMENT RATING CURVE WITH  STREAM STABILITY

           RATING  (SSR)  ADJUSTMENTS - DEVELOPED FOR FOX

       PLANNING  UNIT,  SIX RIVERS NATIONAL  FOREST,  CALIFORNIA

       (POOR SSR  115; FAIR SSR  115 AND   75; GOOD SSR  75)
                                 109

-------
B.   NCASI Activity Involving WRENSS

     Because WRENSS is being applied to an increasing number of
assessments, this procedure and its derivatives will be the
subject of numerous performance evaluations.

     Recently, a portion of WRENSS was used to predict baseline
sediment yields from undisturbed basins.  As previously mentioned,
the NFMA has necessitated that sediment yields along with other
factors, be considered for planning purposes on National Forests.
For this reason methods which predict changes in sediment yield
are being developed and applied for specific conditions.  In 1981,
the USDA Forest Service Northern and Intermountain Regions pre-
pared a working- draft of procedures to predict sediment yields.
However, one problem was the lack of baseline data for most sites
to show sediment yields prior to harvesting (12).
     A method was proposed in this document which would estimate
natural sediment yields.  Data from undisturbed watersheds (repre-
senting low to high sediment producing landscapes) were reported
to show a range of 10 to  100 tons/mi /year.  For  undisturbed
forest watersheds "...  the source of natural sediment is pri-
marily stream channel erosion processes"  (12).  The authors
concluded that "since natural surface erosion is  considered
insignificant, the variation in natural sediment  is assumed
attributable to differences in mass erosion hazards and delivery
differences" (12J-  Therefore, using the boundary values of 10
to 100 tons/ mi /year a relationship between average natural
sediment yield and the WRENSS mass erosion hazard rating was
developed.

     In order to rapidly  evaluate this natural sediment yield
estimating procedure, a USDA Forest Service Pacific Northwest
Region document, "Erosion and Sedimentation Data  Catalog of the
Pacific Northwest" was utilized (13).  The catalog summarizes the
results and site characteristics for 19 small watershed studies
as well as other erosion  plot and large-basin studies.  Several of
the summarized studies include data for control (undisturbed)
basins.  Using site characteristics listed and knowledge of these
watersheds, hazard ratings were developed and plotted against
measured yields.  See Figure 4. The results showed that the
approach could not be applied throughout the Northwest.  Further,
the hazard index used could not descriminate between wide ranges
of average outputs even for nearby watersheds.

     Although there are many reasons why short periods of sediment
monitoring might not agree with the hazard index, the need to
explore the proposed relationship is apparent.
                               110

-------
iiUUU -




N
1
I
Sediment 1
n c
D C
g yw
Z
E
0)
>


4 -
3
9

























^











.
! ^
.s*^\
^^ •
•

•

•






...!#
^^
•









X
^
^







y


Function as originally plotted
through single data point.





•
                      10     20     30      40     50      60
                              Mass Erosion Hazard Rating


                              FIGURE 4

                COMPARISON  OF PREDICTED TO ACTUAL
               SEDIMENT  YIELDS FOR NORTHWEST STUDIES  (20)
     NCASI is now collecting information from landslide surveys
which can be used to  evaluate predictive methods.   This informa-
tion is now being organized  into a catalog similar to the "Erosion
and Sedimentation Data  Catalog of the Pacific Northwest" which was
refered to earlier.   One  example of the arrangement of information
is shown below for  a  survey  of landslides in the Mapleton Panger
District.  See Table  2.

     Until recently WRENSS has involved a laborious, hand calcula-
tion procedure.  The  USDA Watershed Systems Development Group in
Fort Collins, Colorado  has  just released a draft copy for review
of computerized WRENSS  solutions, WETT.WET and WET.SED.  NCASI,
                                111

-------
along with others, is now reviewing these programs.  (Although the
hydrograph  portion of WPENSS is simulated by WETT.WET., the
sediment portion of WRENSS is only partially reproduced.)

C.  Other Procedures to Predict Sediment Yields from Silvicultural
    Yields
     While this discussion has centered on WRENSS there are
numerous other procedures being touted to assist in forest plan-
ning.  A recent publication by the USDA Forest Service Pacific
Northwest Region summarized more than 10 different techniques
being used to quantify surface erosion, mass movements (slides),
and sediment delivery (14).  See Table 3 for a short description
of these models.


      The  Siuslaw National  Forest provides another  example  of the
number of models being  proposed for use  in  forest  planning.
Passing over  the erosion  prediction options  provided  in  Table 3,
hydrologist,  soil  scientists and fisheries  biologists  on the
Siuslaw National Forest have developed 5 separate  procedures to
predict sediment  loading  to  streams and  the  response  of  anadrom-
ous  fish  as a  result  of forest reading and  harvesting.   These
procedures included:   (a)  a  forest-wide  planning procedure;  (b) a
district-wide  procedure to estimate cumulative effects;  (c)  a
district  procedure which  estimates landslide-only  sediment load-
ing;  (d)  a site-specific  procedure which requires  on-site  esti-
mates of  the  effectiveness of mitigative measures, and  (e)  a
steam-threshold procedure  to estimate when unacceptable  sediment
loading occurs  to  the stream system.

      As previously discussed, the forest-wide procedure  for  the
Siuslaw National Forest has been proposed as a predictor for
defining  appropriate  practices under the Oregon Forest Practice
Act.  The Siuslaw  Model benefits greatly from numerous on-site
studies and landslide inventories and this empirical model  is
based in  part  on these  relationships.  Components  predicted
include natural erosion rates, in-unit failures and volumes,  road
related failures and  volumes, surface erosion rate (dry  ravel
from  broadcast burning) and delivery, and effectiveness  of pro-
tective measures (leave  areas).  The procedures weaknesses  include
lack  of validation, lack of discrimination between a wide  range
of practices, and  the need for gross generalizations about  site
considerations.

      NCASI has developed computer programs to solve the  Siuslaw
National Forest procedures as well as many of those presented, by
the Pacific Northwest Region in order to compare predicted
sediment  loading rates with existing data sets.
                               112

-------

-------

-------
       TABLE  2     EXAMPLE  OF  MASS  WASTING CATALOG
               Inventory of Mass Erosion in  the Mapleton Ranger District
  Siuslaw National Forest and Pacific Northwest Forest and Range Experiment Station

   Location:  Selected units in the Mapleton Ranger District, Siuslaw National Forest,
   Oregon were evaluated.  See Mapleton District - Ketcheson and Froehlich for the location
   of the of the Mapleton District.

   Objective:  With improved road building,  in-unit slides are- suspected as possibly
   becoming more important (proportional) to the overall sediment budget for steep terrain
   sites.  This is especially true because road right-of-ways occupy only a fraction
   (approx. 8%) of the overall area being harvested.  Therefore, an inventory of slides
   was made to determine the number of failures in forested, in-unit (clearcut) and road
   right-of-way sites.  The focus of the study was on the most slide-prone soil types in
   the Mapleton District.

   Site Characteristics;  See Mapleton Ranger District - Ketcheson and Froehlich for a
   complete description of soil types, topography, and precipitation.  This inventory focus-
   ed on SRI soil type 47 which is the most  slide prone in the Mapleton District (hazard
   rating of 5) and has shallow, coarse-textured soils with steep, deeply-incised slopes.

   Procedures;  Color air photos (scale - 1:15,840) made in 1972 were used to determine
   road and in-unit slide frequencies.  Field measurements were used to establish the size
   distribution of these slides along with dating by dendrochronological methods.  In-forest
   erosion rates were determined exclusively using field observations.  Slide volumes less
   than 10 yds  were excluded.  The inventory covered time periods of 10, 15, and 20 years,
   respectively, for clearcut (in-unit), road right-of-way, and natural failures.  No
   accounting was made of debris torrents following slide entrance into the stream channel.

   Results:  Comparison of in-field and air  photo methods showed good agreement for clearcut,
   and road-in-clearcut sites, but poor agreement for road-in-forest sites (due to shadows)
   so in-field techniques were used for the  forest and road-in-forest calculations.  Clear-
   cuts adjacent to roads did not significantly increase the-slide rate of road right-of-
   ways.  Photo inventory methods were found to detect 40 yd  or larger slides, while in-
   field measurements detected 10 yd  or larger failures  C< of in-unit slides 40 yd  or less).
   The frequency of events for in-forest sites was found to be 20x the rate for other sites
   studied in the Northwest, but because the average volume was 4% the volume found in
   other studies, the in-forest slide erosion rate was similar.  A decrease in importance
   of roads between this study and the initial Mapleton landslide inventory was noted (change
   in the ratio of clearcut area to road failures from 7.3 to 31).

        In an Appendix report, debris torrents were surveyed for Cedar Creek which is a
   tributary to Sweet Creek and the Siuslaw  River, 5.5 miles SW of Mapleton.   The analysis
   used cumulative acre-years in order to account for changing land-uses ef the period of
   records.  The authors concluded that a similar adjustment to the Mapleton data could
   increase the man-caused rate of sliding 2 to 4 times.

                LANDSLIDES FOR FORESTED,
               CLEARCUT, AND ROADED SITES
         Unit Associated Road Associated

i Slope Class
K
   0-58
   58-70
   70-84
   14-100
  100-115
   1U-
       ict
       I
      St
       i
      SKI
     »jfic«tion
                 3
                18
              11
              10
              14
              1}
              13
              12
              II
               9
 3
20
55
19
 3
13.
20.
12.
 5.
12.
 9.2
K.I
                                       Natural
                                      For
 4
15
33
31
 8
 1.2
 4.1
14.3
12.2
16.3
14.3
12.2
11.4
                                                                    DEBRIS TORRENTS ON
                                                                   CEDAR CREEK DRAINAGE
FQrevt
ClMrcut
CuMulAtive
34,544
10,751
Debris , Frequency
Torrents ( Event! /Acre- Yr )
5 0.00014
8 0.00074
frequency to
Debris Torrents
For Forest Areas
j.
S.3X
                                                                          8.0056
o « *
      47
     others
               o.so
               0.20
               0.33
11.5
 5.1
 1.4
 0.15
 0.28
 0.17
                                                113

-------
TABLE  3     AVAILABLE  MODELS  AND  TECHNIQUES  FOR  COMPUTING  ON-SITE
                  EROSION,  MASS  MOVEMENT,   SEDIMENT  DELIVERY  INDICES,
                     AND  EROSION HAZARD  RATINGS  AS  LISTED  BY THE
                  USDA  FOREST  SERVICE  PACIFIC  NORTHWEST  REGION  (14)
              Technical
               Name
              EROSON
              ONEROS 3
              OSLE
              MSLE
              Hegahan
              Time
              Trend
              Erosion
              Model

              Region 5
              Erosion
              Hazard
              Rating
              System

              Region 1
              and 5
              Guidelines
              for
              Predicting
              Sediment
              Yields

              VIRENSS
              Mass
              Erosion
              Hazard
              Rating

              WRENSS
              Sediment
              Delivery
              Model

              Region 5
              Sediment
              Delivery
              Model
Description

Program  computes maximum erosion hazard
rating in inches of  soil lost  per year
based on slope, precipitation, and
erodibility coefficients, and  a basic
erosion  rate.  Cover conditions are then
used to  estimate actual soil loss.

Modified Musgrave method which is adapted
to forest applications and used to predict
sheet-erosion in tons/acre. Cully erosion
and delivery of on-site erosion to streams
can be included in calculation.

Calculates sheet erosion based on rain-
fall, soil erodibility, slope-lengths,
cover and erosion control factors in tons/
acre (or desired units).

Modified DSLE which  uses precipitation,
soil erodibility, topographic, and vegeta-
tion management factors to predict on-site
erosion  in tons/acre.  (Modified for forest
use).

Forest level planning model which distri-
butes erosion over time in tons/acre based on
long-term normal erosion (no new disturbance),
material available to erode after disturb-
ance, and potential  recovery rate.

Point total based on site characteristics
defines  hazard rating.
Erosion predicted  as a function of land
type and management activity.  Guidelines
include functions  to deliver eroded
material to streams and route  it to criti-
cal areas.
Two erosion hazard  ratings (for debris
avalanche and slump-earthflow types fail-
ures)  based on site conditions and a point
scale  for component hazard


Sediment delivery index based on available
water,  texture of eroded material, ground
cover,  slope gradient, slope shape, sur-
face roughness, and delivery distance.

Delivery of sediment is a function of soil
erosion hazard rating, drainage dissection,
slope  position, vegetation cover, and
entrapment coefficient.
Computer
Solutions
Available
                                                114

-------
     IV   AREAS OF CONTINUING NEED FOR MODELLING DEVELOPMENT


     A general understanding of many of the basic mechanisms con-
trolling watershed-sediment response to forest-management activi-
ties has been developed and this understanding is incorporated
into some of the predictive techniques discussed.  For example,
WRENSS has identified  key components including channel scour, mass
wasting, and surface erosion which can contribute to sediment load
changes.  However, because of the complexity of forest systems
and the difficulties of quantifying those systems (particularly
the hydrologic cycle), no single model has yet proven to be uni-
versally acceptable.   In fact some models are being used prema-
turely without thorough testing simply because immediate answers
and management decisions are required for planning.

     The further development of models to predict NFS response to
forest practices will  require five steps.  First our understanding
of some of the mechanisms currently modelled as cause-and-effect
need to be refined.  For example, several models use sediment-
rating curves to predict sediment load response to forest prac-
tices by predicting hydroyraph changes following harvesting.  Yet,
it is known from sediment concentration-flow patterns for season,
storm sequence, and position in the storm hydrograph, that flow
is only one of the independent variables influencing sediment
concentration.  When discharge is increased, as a result of
reduce evapotranspiration for example, the sediment-rating curve
relationship could overestimate sediment loading because the
rainfall-intensity factors generating material available for trans-
port are overestimated by the flow parameters.  Conversely, the
sediment-rating curve  relationship could underestimate sediment
loading because practices could make more material available for
transport.  Further examples can be drawn from the bulk of re-
search dealing with infiltration and interflow (22).

     Second, key mechanisms need to be modelled which have been
avoided to this point.  An example is the lack of groundwater
modelling for mass wasting analysis.  A special soils task force
reporting to the Oregon State Forester concluded that "nearly all
slope failures in the  forest environment occur in response to a
higher than average pore water pressure condition..."  and there-
fore the development of engineering groundwater models is nec-
essary to improve road location and design (15).

     Hewlett (21), in  a recent paper on modelling philosophy,
points out that hydrological modellers are currently "plagued with
an embarrassment of riches "for journals contain numerous attempts
to explain and/or model individual hydrological processes without
providing a clear-cut  choice for managers to implement.  Hewlett


                               115

-------
 further  suggests  that  decision  making  should  be  the  "heart  of the
 research enterprise".    Models  which simulate  nature accurately
 may  become  inefficient in  application  to decision  making  because
 the  cost is  prohibitive for  collecting  input meteorological and/or
 hydrological data,  and measurement  of model parameters.

      Difficulties may  arise  when  users  attempt to  interpret cause
 and  effect  relationships from a model which was  not  designed to
 deterministically simulate the  effect of interest  or concern.   A
 surge in concentration of  an aerially applied  pesticide  in  storm
 flow,  for example,  may be  attributed to basin-wide flushing if  an
 overland flow model is used  to  fit  the  hydrograph.   The  same
 effect,  however,  could be  attributed to stream channel factors  if
 a  subsurface storm  water flow model is  used.   Both models may fit
 the  hydrograph  for  predictive purposes, but neither  may  imitate
 the  actual  runoff process.

     Third,  complete models  providing the linkage between forest
activity, site conditions, and  the water quality factors  upon which
decisions are based need to  be  provided.  in reviewing research
priorities,  Brown noted  that the next step for research is
"...integrated, process-level studies of system function.   For
example,  instead  of measuring sediment concentration at the  outlet
of a small watershed after clearcutting, sediment production will
be defined as a function of  soil disturbance,  soil characteristics,
terrain,  and  climatic  variables.  The transport and deposition  of
sediment  in  the stream  channel  will be defined as a function of
the hydraulic factors  involved.   These conditions  then will  be
related  to the variables that influence aquatic organisms"  (16).
Models based  on fundamental  factors and directed  at the key
response  parameters will be more transferable  and useful.  This
system's  engineering   approach  is clearly not  a major focus  of  the
current modelling activity being discussed.

     Fourth,  and perhaps most importantly,  there  is a need  to
develop workable  techniques  for collecting  the key on-site param-
eters which  influence  site response.  For example, for sites sub-
ject to mass  failures,  a small  change in the depth of soil,
presence  or  absence of  a hardpan,  groundwater  drainage, or orien-
tation of bedrock bedding planes can make the  difference between a
site failing  or remaining  intact.   It needs  to be acknowledged
that there may be limits or  uncertainty to  the information we can
collect and  use operationally.   Where site-specific information
is not available  for key parameters, defacto "best estimate"
data is often used and  the modelling problems  of  this approach
need to be recognized  and  incorporated  in the  management decision.
     Finally, models need to be thoroughly tested and documented
before operational use to evaluate their accuracy and precision.
This need is similar to that expressed by NCASI concerning point


                               116

-------
source water  quality models  in Technical  Bulletin  No.  367  (23) .
This technical  bulletin  addressed the practical  concerns of  model
selection,  calibration,  and  verification.   The findings of this
study show  the  need to define  goodness of fit criteria for eval-
uating models,  and to develop  a broad enough data  base to uncover
and subsequently  represent mathematically major  physical, chemical,
and biological  mechanisms  in the basin before making  long-range
forecasts.


           V   RESEARCH TO IMPROVE OUR UNDERSTANDING
                 OF FOREST WATER QUALITY MODELING


     Each year since 1979, about $9,000,000 has been spent on re-
search to address NFS questions  (17).  This research includes
efforts to provide the basic data necessary to model NFS mechanisms
and also efforts to improve or validate models.   An excellent
example of industry,  university, and USDA Forest Service research
cooperation is centered in Gainesville, Florida under IMPAC
(Intensive Management Practices Assessment Center)  program.   This
program is filling a real information need about the effects  of
harvesting and site preparation  in flatwoods conditions on water
quality.  Perhaps some of the most interesting results from this
study have concerned the water quality coming from undisturbed
flatwoods sites.

     Many other research programs are scattered throughout the U.S.
and information on these programs can be obtained in an annual
NCASI Technical Bulletin.  An example of the type of basic data
gathering research underway  is seen in the South Central U.S.
Until recently there has been little or no research on controlling
sediment losses from forest roads in the South Central U.S.   Four
road projects are now being conducted by the forest industry, states,
and universities in Oklahoma, Arkansas (2), and Mississippi.   See
Figure 5.

     Similarly, in the West, studies have been conducted to assess
sediment losses from road and skid trails in Idaho, California,
Oregon,  and Wasington.  A forest  industry conducted study in Idaho
has demonstrated the effectiveness of scattering slash over skid
trails as a BMP to minimize erosion (18).  This study has also
been used to compare observed sediment losses from skid trails to
those predicted by commonly used sediment loss equations.   See
Figure 6.

     A review of non-point source models  applicable to the assess-
ment of the effect of silvicultural practices on water quality is
also underway at the Northeast  Regional Center of NCASI.  The over-
land flow and catchment models  described  in Tables 4 through 6^
are being examined for their ultimate  utility to the pulp and
paper industry.  The information in the Tables 4 through 6^ was
assembled in a  recent review by  Ambrose  (24).
                                117

-------
                                             FIGURE 5
                                            FOREST INDUSTRY STUDY
                                            OF  BEST MANAGEMENT
                                            PRACTICES FOR FOREST
                                            ROADS  IN THE SOUTH -
                                            A PROPORTIONAL SAMPLER
                                            COLLECTS RUNOFF FROM
                                            THE ROAD
                A             FIGURE 6             °

   FOREST  INDUSTRY STUDY OF EROSION  FROM  SKID TRAILS IN  IDAHO
SHOWS THE  EFFECTIVENESS OF SCATTERED DEBRIS IN REDUCING  EROSION
   A;  SKID  TRAIL WITH SCATTERED  DEBRIS    B;   BARE SKID  TRAIL
                               118

-------
      The models can be  further  segmented  for forested  areas
according to  sophistication by  the use of  a  classification scheme
proposed by Ambrose (24).   The  segmentation  and classification
are provided  in Table 7 .   The models in Table 7 are  the focus  of
current NCASI  interest.
Physical Properties Spactial Properties
Iirper- Agri- Wet Channel
Model vious cultural Forested lands Single Multiple Routing


TABLE 4



PRE-SCREENED
OVERLAND MODELS-




CATCHMENT PROPERTIES















Model
ACTMO
ACRUN
Amberger
et al.
ANSWERS
ARM-II
CNS
CREAMS
CSU
EPA RRB (URO)
Haith and Tubbs
HSPF
Hydroscience
MRI
MUNP
tfS
OKI -RRB
PTR
80S
SCRAM
SEMSTORM
SOGREAH-
CAREDAS
Simplified
SWMM
STORM
SKMM-Level I
SWMM-II
Texas A 4 M
URS
UTM-TEHM
WEST
HQAM
WRENS















Continuous Event Annual Monthly
X
X

X
X
X
X X
X
X
X X
X
X
X X
X X
X
X
X X
X
X
X
X X

X

X
X • X
X X
X
X
X X
X

X X
X X















Dally
X
X

X
X
X

X
X


X


X
X
X
X
X


X

X


X


X



ACTMO X X
AGRUN X X
Amberger
et al. X X
ANSWERS X X
ARM-II X X
CNS X X
CREAMS X X
CSU XX X
EPARRB CURD X X X X X
HSPF X X X X X
Haith and
Tubbs X XX
Hydroscience X X
MRI XXX X
MUNP X
NPS X X X X X
OKI-RRB X X
PTR XXX
QQS X X
SCRAM X X
SEMSTORM X X X X X
SOGREAH-
CAREOAS X X
Simplified
SWMM X X
STORM X X X X X
SWMM-Level IX X
SWMM-II X X X X X
Texas MM x X
URS X X
UTM-TEHM XX X
WEST
WQftM X X X X X
WRENS X X
















TABLE 5 PRE-SCREENED
OVERLAND MODELS-
TIME DOMAIN*














X
X
X
X











X
X



X
X
































 •This reflects the effective predictive capability of the model and is
 considerably larger than the computational time step.
                                  119

-------
TABLE  6    PRE-SCREENED
          OVERLAND  MODELS-
       CONSTITUENT  SYSTEMS

Model
ACTMO
AGRUN
Amberger
et al.
ANSWERS
ARM-II
OB
CREAMS
CSU
EPA RRB (URI)
Haith and
Tubbs
HSPF
Hydroscience
MR I
HJtf
tes
OKI-RRS
PTR
DOS
SCRAM
SEMSTORM
SOGREAH-
CAREOAS
SimplifiK)
SWrt"t
STORM
SMMM Level I
SMMM-II
Texas A4H
URS
UTM-TEHM
WEST
NQAM
KRENS
Hydro-
logy
X
X

X
X
X
X
X
X


X
X



X
X
X
X
X
X

X

X
X
X
X
X
X
X
X


Sed-
iment
X
X

X
X
X
X
X
X
X

X
X
X
X

X
X
X
X
X
X

X

X
X
X
X
X
X
X
X
X
X
Arb- Pnos- Carbon
itrarv Nitrogen phorus (organics)
1 X X
J

X X
X
XX X X
X X
XXX
XXX X
XXX X

X
XXX X
X
X

X
X
X
X
X
10

X

X
X
X
X
XXX X
X
X

X
X
         TABLE  7    NFS  MODELS  AND  SOPHISTICATION LEVEL

                MODEL              SOPHISTICATION  LEVEL
                   CSU
                   HSPF
                   SWMM-II
                   NPS
                   SEMSTORM
                   STORM
                   WQAM
                   WRENS
 4
 4  or  3
 4  or  3
 3
 2
 2
 1
-1
            LEVEL  1 - Totally empirical approach to watershed hydrology and
            sediment generation.  Useful for long term (annual)
            predictions.  Handbook approach.

            LEVEL  2 - Simplified empirical  hydrology simulation via
            computer program.  No attempt is made to account for moisture
            between storm events.  The Rational Formula and/or SCS Curve
            Number approaches to runoff from impervious and pervious
            surfaces are used.  Effective temporal resolution is 1 month.

            LEVEL  3 - Watershed hydrology is calculated explicitly via a
            process oriented approach.  Pollutant generation is handled
            empirically by link to sediments or the use of washoff
            functions.  Effective temporal  resolution is of the order of
            one day or less.

            LEVEL  4 - These models treat important hydrologic, nutrientr
            and sediment processes in detail.  Nutrient interactions can
            occur  in both particulate and dissolved form.  Detail structure
            is built into the model to account for various management
            strategies.  The time resolution is of the order of one day or
            less.
                                        120

-------
        VI     LITERATURE  REFERENCES
 (It   "Summary  of  the  Current  Status  of  Silvicultural  208 Programs-
      1980,"  Special Report No.  80-12 (December  1980).

 (2)   "A  Summary of Silvicultural  Nonpoint  Source Control Programs
      for the United States-1982," NCASI Special Report  83-01
      (January  1983)

 (3)   Olszewski, R.J.,  "Streamside Management  Zones  in Florida and
      the Southeast,"Presented in  Technical Bulletin No. 389
      (November 1982).

 (4)   Brozka, R.J., Water Quality  Project Forester,  Forestry
      Division, Natural Resources  Department,  Santa  Fe,  New Mexico,
      Personal  Communication (August  4,  1982).

 (5)   Hauge,  C.J., Furniss, M.J.  and  Euphrat,  F.D.,  "Soil Erosion
      in  California's  Coastal  Forest  District,"  Calif. Geo. 32  (6)
      120 (June 1979).

 (6)   Datzman,  P.A.,  "The Erosion  Hazard Rating  System of the
      Coast Forest District: How Valid is it as  a Predictor of
      Erosion and  Can  a Better Prediction Equation be  Developed?"
      M.S. Thesis, Humbolt State Univ.,  Arcata,  California  (June
      1978).

 (7)   Nobel,  D., "Silver Creek - Field Laboratory in the Idaho
      Bathalith,"  Forest Resources West  4 (November  1980).

 (8)   O'Leary,  S.,  "Silvicultural  208 Activities in  California,"
      Presented in Technical Bulletin No. 389  (November  1982).

 (9)   Swanson,  F.,  "Cumulative Stream Impacts,"  Paper  presented at
      Workshop  on  Stream Management, Oregon State University,
      Corvallis, Oregon (May 1982).

(10)   "An Approach to  Water Resources Evaluation of  Nonpoint
      Silvicultural Sources; A Procedural Handbook," EPA-600/8-
      80-012  (August  1980).

(11)   Laven,  R.D.,  and Lehre', A.K.,  "The Effects of Timber Har-
      vest and  Roads  on Sediment Yield and  Channel Morphology in
      the Fox Planning Unit,"  USDA Forest Service, Six Rivers
      National  Forest  (July 1977).

(12)   Cline,  R. ,  et  al. "Procedure for  Predicting Sediment Yiels-
      A Working Draft," USDA For.  Serv., North Region  and Inter-
      mountain  Region  (No Date).

(13)   Lassen, K.R., and Sidle, R.C.,  "Erosion  and Sedimentation
      Data Catalog of  the Pacific Northwest,:  USDA For.  Serv.,
      Pacific Northwest Region R6-WM-050-1981  (September 1981).


 (14)  Howes,  S., and Hughes, D. ,  "Descriptions of Available Models
      and Techniques for Computing On-Site  Erosion Losses and Mass
      Movement, Sediment Delivery  Indices,  and Erosion Hazard
      Ratings," Forest  Soil  Horizons  7,  USDA For. Serv., Pacific
      Northwest Region,  Portland,  Oregon (May 1981).

 (15)  Spiesschaert, D. ,  Carleson,  D.,  Carter, G., Duncan, S.,
      Madison,  R., Manson,  R., and Pyles, M., "Minimizing Debris
      Avalanches on Forest  Land,"  A Report  to the State Forester,
      Oregon  State Department  of Forestry (December  1982).

 (16)  "Pollution Control  in the  Forest Products  Industry,"  EPA
      625/3-79-010 Envir.  Res.  Info.  Center, Cincinnati, Ohio
      (August 1979) .

 (17)  "Annual Survey of  Ongoing  Research on the  Impact of Forest
      Management Practices  on  Water Quality and  Utility - 1982,"
      NCASI Technical Bulletin No.  373 (June 1982).
                              121

-------
               (18) McGreer, D.J.,  "A Study of Erosion  from Skid Trails in
                   Northern Idaho," in Measuring and Assessing the Effective-
                   ness of Alternative Forest Management Practices on Water
                   Puality, NCASI  Technical Bulletin No. 353 (August 1981).

               (19) "Factors Affecting Changes in the Percent of Fine Sediment
                   in Gravel Bedded Channels," NCASI Technical Bulletin No.  354
                   (August 1981).

               (20) McGreer, D.,  Forest Hydrologist, Potlatch Corp., Lewiston
                   Idaho, Personal Communication (December 28, 1981).

               (21)  Hewlett, J., "Models in Land Use Hydrology",  Technical
                    Report No. TR  113, Department of Water Affairs,
                    Forestry, and  Environmental Conservation, Pretoria,
                    So. Africa (May, 1980).

               (22)  Novotny, V.  and Chesters, G.,  Handbook of Non-Point
                    Source Pollution, Van Nostrand Reinhold, New York (1981).

               (23)  "A Study of  the Selection, Calibration, and Verification
                    of Mathematical Water Quality Models", NCASI  Technical
                    Bulletin No. 367, New York, NY (March, 1982).

               (24)  Ambrose, R. , et al., "Models for Analyzing Euthrophication
                    in Watersheds-A Selection Methodology", Chesapeake
                    Bay Program, US EPA, Annapolis, MD (1981).
The work described  in this paper was not  funded  by the U.S.  Environmental
Protection Agency.   The  contents do  not  necessarily  reflect  the views of the
Agency  and no official  endorsement  should be  inferred.

                                            122

-------
          DETENTION POND SIZING TO ACHIEVE WATER QUALITY OBJECTIVES

                            Roger K. Wells, P.E.

                            HMM Associates, Inc.


                                Intr odu ct ion
     HMM Associates has recently performed a modeling study of the Black
Warrior River in Tuscaloosa County, Alabama.  The study area consisted of
the reach downstream from the Oliver Lock and Dam to the Warrior Lock and
Dam--approximately 75 river miles.  The goals of the study were twofold:
(1) to perform the necessary water quality modeling to support an applica-
tion for discharge permit under NPDES and  (2) since the reach is water
quality-limited during a portion of the year, to demonstrate the adequacy of
a proposed discharge detention system.  As part of (2), the size of the pond
was to be determined.

                               Overall Method

     As part of the effort, several mathematical modeling techniques were
employed.  The water quality of the reach was modeled using QUAL-II.  Runs
were made with and without the proposed discharge.  QUAL-II was used in the
flow augmentation mode to generate data to provide a correlation between
river conditions and allowable discharge.  Once this relationship was es-
tablished, a synthetic streamflow model was developed to simulate a realis-
tic sequence of daily flows, and finally a detention pond model was devel-
oped to determine the annual maximum storage requirements for a 50-year
period of simulation.  Figure 1 shows an example of the desired frequency
distribution.  Extreme value distributions such as this are difficult to
construct solely on the basis of existing records because of the distribu-
tion's skew.  Skew is a measure of the third moment, which is greatly in-
fluenced by rare outlying events.  Rarely are streamflow records long enough
for reliable estimates of the third moment.  Using a reliable estimate of
this relationship, designers can pick the required pond volume with a knowl-
edge of the return period of conditions which may affect facility capacity.
The following sections discuss in turn the analytical steps to develop this
curve.

                      Model Set-Up and Input Parameters

     The first step in setting up a QUAL-II simulation is to schematize the
river so that important waste loads, tributaries, and geometrical charac-
teristics may be accounted for.  For this study, a total of 154 1/2-mile
computational elements for the reach between the Oliver Lock and Dam and
the Warrior  Lock and Dam downstream were used.  These computational
elements were grouped into reaches, each with its own reaction rates, tem-
perature, and geometry.  Fifteen reaches with 11 different hydraulic charac-
teristics were incorporated into the model.  Each reach also receives dis-
tributed runoff flow.  Eleven computational elements were identified as
point source discharges or withdrawals.  Small  tributary flows were con-
sidered point sources in this study.  Figure 2  shows the river schematic
developed for this study.
                                    123

-------
    ~|	 -

                                                            33
                                                             s
                                                       ^_^	I  _
a *
                    2u
                                       So
                      Figure  1
                          124

-------
         REACHES
                                 PT
2.0
2.1

3.0
4.o
5.o

Go
7.0
&>
10.Q
 ll.O
  \.l
           OLIVE?
         CL£ME\iTS
         BJTECSTATC
         EA6LE
         SUCfcLS
             SAklCX CX.
          MOUkJO
          STATt
ELLIOT6 CR

WV4A.TLCY 6R
          CUTOFP
         O41LDS
         BROW us  ec
         WlSHWAY-14.
            PT
                         3U
                         sfo
                         500
                         296
                         292
                         236
                          274
                         210
                                        OIL
                                       GOODRJCW
                         TUSC^LOOSA  STR
                              IklTMCE
                              DISCHARGE.
                         LAWTELK.- CHEMICAL
                                    Noerw
• D^iOTS CR. (vMOUMDVIUL SDUTU LWOOI4)
                          Figure 2

                  Black Warrior  River Model
                             125

-------
Hydraulic Data
     Hydraulic data were developed as a result of field work which was un-
dertaken in support of this study.  The field study is discussed later in
this paper.  These data were used to develop trapezoidal approximations of
each section for model verification.  Since different flows were necessary
for subsequent model runs, velocity, flow, and depth relationships were de-
rived for later runs.  These relationships are of the following form:
      V = aQb

      D = a(f

where V = stream velocity, in ft/sec;

      Q = stream flow, in ft^/sec;

      D = stream depth, in ft; and

      a, b, a, and 3 are coefficients determined as part of this study.

Table 1 summarizes the coefficients so determined.

Reaction Coefficients

     The reaction coefficients which must be specified include the carbon-
aceous BOD reaction rate (Kj), the NH3 oxidation rate (CKNH3), the N02 oxi-
dation rate (CKN02), and the atmospheric reaeration rate (K2).

     K_j.  The BOD reaction rate is a parameter which expresses the rate (in
units of days"1) of oxidation of carbonaceous BOD.  During the field mea-
surement program, a sample was taken near the proposed discharge location
for a 30-day BOD analysis.  These data were analyzed for reaction rate using
a non-linear least squares regression technique known as the Gauss-Newton
method.  Figure 3 summarizes the results of this analysis.  It can be seen
that the oxygen demand is still increasing substantially up to 30 days,
which indicates a low reaction rate.  It was also felt that the point at
five days might be invalid because of its low value, so regressions were
performed both with and without this value.

     In selecting a reaction coefficient (Kj) for modeling, the values from
Figure 3 were used as limits.  The rate was varied until a balance was
struck between measured BODs and measured DOs within this range.  The final
rate selected for modeling was 0.02 day"1.  This value is considerably below
the values traditionally used for modeling, i.e., 0.1-0.2 day"1.  Since this
test is the only known long-term BOD test from this reach of the Black
Warrior River and reasonable agreement was obtained between field measure-
ments and calculations, it was decided to use the value of 0.02 day"1 for
modeling.

     CJCNHj.  A value of 0.05 day"1 was selected for the ammonia oxidation
rate.  This is based on discussions with the regulatory authorities and re-
ports of their work in the Black Warrior River, which indicated that nitri-
fication is not an important oxygen-consuming process in-stream.

     CKNO^.  The nitrite oxidation rate chosen was 2.0 day"1.  This corres-
ponds to the idea that nitrite is merely a short-lived intermediate step in
the process of oxidizing NH, to N03.  Without detailed nitrogen measurements
to support the use of another value, the upper end of the range suggested in
the QUAL-II user's manual was chosen.

                                    126

-------
         Table 1
Stream Flow Coefficients
Reach
Number
1
1
2
2
3
4
5
6
7
8
9
10
11
11
11
.0
.1
.0
.1
.0
.0
.0
.0
.0
.0
.0
.0
.0
.1
.2
3
3
1
1
3
3
3
2
3
1
2
1
1
1
1
a
.408E-4
.408E-4
.943E-4
.943E-4
.840E-4
.155E-4
.417E-4
.538E-4
.203E-4
.953E-4
.072E-4
.472E-4
.314E-4
.314E-4
.314E-4

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
b
.929
.929
.956
.956
.904
.934
.931
.949
.933
.949
.946
.952
.968
.968
.968

11
11
20
20
10
11
8
16
9
21
19
27
29
29
29
a
.085
.085
.407
.407
.054
.368
.756
.817
.496
.187
.240
.066
.237
.237
.237

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3
.0523
.0523
.0329
.0329
.0561
.0514
.0617
.0383
.0584
.0320
.0345
.0261
.0245
.0245
.0245
           127

-------
                            30-Day BOD
Time
(days)
5
10
15
20
25
30
BOD
(fflg/1)
0.05
0.65
1.45
1.55
1.75
2.05
 2.0
 1.5
1.0
0.5
0.0
                      5 pts
                                     6 pts
0
5
10
15
Days
20
25
30
Data Base
6-point
5-point
Reaction Rate
0.0118
0.0260
Corr. Coeff.
0.91517
0.90644
                        Figure 3
                            128

-------
Reaeration Rate

     Atmospheric reaeration is the fundamental source to replenish in-
stream DO.  Unfortunately, a wide range of estimates of the coefficient
is available.  The method of Tsivoglou and Wallace was used for the
calibration/verification, a method recommended by EPA in the QUAL-II docu-
mentation.  The form of this estimate is:

      K2 = 3600 K Se U

where K2 = reaeration rate, in hour  ;
      K  = a constant - 0.05-0.1 ft"1;

      Se = energy gradient slope, in ft/ft; and

      U  = stream velocity, in ft/sec.

It can be seen that K2 depends on the energy gradient slope.   At low flows,
the Warrior Pool becomes more lake-like, with a near-zero energy gradient,
and a method more appropriate for a lake environment is used.   Therefore, it
was decided that a method outlined by Brown* would be used.  This method re-
lates the surface transfer coefficient K^ to the wind speed:

      KL = 0.05 w2

where K,  = surface transfer coefficient, in m/day; and
      w  = wind speed, in m/sec.

For the low flow portion of this study, K^ was estimated to be 0.25 m/day
(0.82 ft/day), based on a 5 mph wind speed.  Table 2 summarizes the K2
selected for each reach for the low flow runs.
Initial Conditions
     For a steady-state simulation, the only parameter which needs to be
specified is the temperature since it affects the adjusted reaction coeffi-
cient and the saturation value of DO.  A value of 85.1°F (29.5°C)  was chosen.
This value is the mean August temperature based on daily measurements at
Northport, Alabama, during the most recently published water resources data
period (USGS, as of June 1982).

Headwater Sources
     The only headwater considered in this study is the Black Warrior River
at the Oliver Lock and Dam.  The USGS in Tuscaloosa was contacted to deter-
mine the most recent estimate of the 7Q^Q.  This parameter is significantly
affected by the operating policy of the upstream dams.   The current operat-
ing mode was instituted in 1964.  The values of 7Q}Q for this period were
obtained from the USGS.  The value 605.7 cfs is used as the starting point
for all low flow modeling.  As mentioned before, in-stream temperature is
taken to be 85.1°F, and the headwater DO level is taken to be the saturation
value at 85.1°F, or 7.4 mg/1.  The value of BOD^ at the headwaters is taken
to be 0.45 mg/1.  The source of this estimate consists  of measurements taken
during the field study conducted by HMM as part of this study and measure-
*Brown, Russ T.  Modeling the Effects of Wind Reaeration.   Report No.  WR28-
 2-520-117.  Tennessee Valley Authority.  March 1981.


                                     129

-------
                             Table  2
Reaeration Values (K2) for Each Reach, Based on 5 mph Wind Speed
Reach Number
1,0, 1.1
2.0, 2.1
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0, 11.1, 11.2
Depth (ft)
15.5
25.2
14.4
15.8
13.0
21.5
13.8
26.0
24.0
32.0
34.2
K2 (day"1)
0.053
0.033
0.057
0.052
0.063
0.038
0.059
0.032
0.034
0.026
0.024
                               130

-------
ments provided by AWIC.  Table 3 summarizes these data, as well as measure-
ments of NH3, N03, and P04 (1980 only) included in 1980 and 1981 studies by
AWIC.  The mean value of BOD5, N03, and TSS was included in the headwater
concentration levels for these constituents.  The concentration of NH3 was
entered as zero, and PC^ was not modeled since all measurements were below
the detection limit.

     The QUAL-II model provides for a complete description of the relation-
ship among algae, nutrients, and oxygen.  In the field measurement programs
conducted by AWIC in September 1980 and in June-July 1981, numerous oxygen
samples were taken that indicated that the water was supersaturated with
oxygen.  This also occurred in the HMM measurements taken in August 1982.
These findings indicate that a net oxygen credit, due to photosynthesis by
algae, is occurring, since this is assumed to be the only process producing
oxygen in-stream.  In August 1981 AWIC indicated, in their analysis con-
cerning the Hunt Oil Company expansion, that a net contribution of 0.15 mg/1/
day of oxygen was being made by algal processes.  This rate was incorporated
into the HMM modeling effort.  The QUAL-II model couples algal growth to the
availability of nutrients and light.  It also allows for creating NH^-N
from algae through a coefficient which represents the fraction of respired
algal biomass which is resolubilized as NH3-N by bacterial action.  Unless
the reaction rate for NH3-N to N02-N is very low, a net oxygen deficit will
result from these coupled processes.  Thus, the supersaturated levels of
oxygen also support the idea that little or no nitrification is occurring
in-stream, an opinion expressed by AWIC when evaluating the Hunt Oil expan-
sion.  In QUAL-II, the presence of algae is considered to be represented by
the measurement of chlorophyll A.  This was input to the model at 20 ug/1,
which yields an oxygen production rate of 0.15 mg/I/day at the headwaters.
This value rises slightly until about river mile 325, when increasing river
depths limit the availability of light for algal growth, and the downstream
chlorophyll concentration drops to about 5.25 yg/1 at the Warrior Dam,
yielding an oxygen production rate of 0.04 mg/l/day.

Existing Point Source Loads

     Point source loadings in QUAL-II include unmodeled tributaries as well
as industrial and municipal discharges.  Sources of data on existing dis-
charges include the Tuscaloosa County 208 Report and AWIC.  Table 4 summar-
izes the characteristics of all modeled discharges.
New Discharge

     The new discharge is subject to new source performance standards.  The
following parameters summarize the discharge's characteristics.

                            QD = 5.028 cfs;
                           TSS = 201 mg/1;

                          BOD5 =84.7 mg/1; and

                          BOD /BODC = 2.62
                             u    5

                          Field Measurement Program

     On August 3-4, 1982, HMM Associates undertook a field measurement pro-
gram on the Black Warrior River to obtain data necessary to characterize the

                                    131

-------
                  Table 3
BOD5, N03, NH3, and P04 Near the  Headwaters
Date
8/82

6-7/81


9/80





Mean =
Std.
Dev. =
TSS
8.4
3.6
4.4
-
-
-
6
5
6
3
4
11
5.71

2.56
B
0
0
0
1
0

0
0
0
0
0
0
0

0
\J LJ f-
.05
.15
.25
.3
.3
-
.4
.5
.4
.5
.4
.7
.45

.33
River
Mile
311
320
329
337
337
337
337
337
337
337
337
337



Source
NC
3 NH, PO,
Field Study -
Field Study -
Field Study
AWIC
AWIC
AWIC
AWIC
AWIC
AWIC
AWIC
AWIC
AWIC



-
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.

0.
-
92 < 1
74 < 1
54 < 1
43 1 -
45
44
42 - < .2
41 - < .2
37 < .2
524

185
                    132

-------
 0)
 o
 f-l

 o
oo
 c
•H
 O
a,

4-1
 o

 X
 fn
 n)
 3
CO
;7
ffi~b6
s
^^
s


cfi I~l
O bO


r-l
Q bO
~

O
Q^ N^ V
H


5^
O j^j
Du ^
v-'

















cu
o
rl 0)
|3 B
O to
COS
r-l CM
OO O O
VO CM
r-l co
CO 0 0
CT* O O
r-» co
t—l CM CM

CO OO
co i — co
rH "*
CM CM CM
^ ^ ^
LO LO LO
co co
^ ^ rH
LO LO LO
o^ o^ oo

CM CM CM
^ ^ ^
o>
LO CO O
O r-l 
i3
C
cO
CO

bO
•rl
PQ



r-H
^j-
CO
rH
i-H

O
co
CM
rH

O
LO

i— 1
LO
OO



OO
r-H
o







rH
CO
O
•H
B
CU
r;
O

^_l
cu
4-J
Jg
CO




rH
LO
CO
rH
rH

f^
LO
VO

^
LO

rH
LO
OO

rH
VO
OO
rH
O



0
O
bO
cd

- —
•
j^

cu
r-H
r-H
•rH
|>
13
C

O
s



cr\
LO
r-l


vO

OO
r-^

CU
rH
i-H
•rH
-O
C
3
0
S

A;
CU
CU
M
o

co
4_)
O
•rl
r-H
i— 1
W



O
CO

^

o
0


LO
^

r-l
LO
OO

LO


l—l
0








x:
u
q

j_i
PQ

t>-i
CU
rH
4J
cO
rj
^



CO
r^!
CO
^
00
LO
r-H


LO
r--

r-H
LO
00

LO


00
CM







^
cu
CU
}_l
U

CU
i-H
•H
^

CU
$>
•rf
PM

Es
o
, — 1
CU
&
co
CU
o
q
cu
cu
LH
CU
M

CU
4-) 4J
«J -H
z e
•H j_|
•0 cu
M pi ,
•rl
o

i— 1
1
CO
CO
4-i bO
O  q
cO 13 4->
O CO 0 CU
B O 4-) rH
CU O CO rH
S rH 0
CO 13 PQ
U 0 i-H
M CO CU 
-------
hydraulic characteristics of the river, to measure depths in detail in the
vicinity of the intake  and discharge areas, and to obtain DO, BOD, TSS, and
color samples throughout the modeled reach of the river.  Samples were also
taken from Big Sandy Creek, Elliot's Creek (upstream of the Moundville South
Lagoon), Whatley Branch, and Five Mile Creek.

     The hydraulic characteristics were determined by bathymetric measure-
ments taken in the river with a Raytheon DE719B recording fathometer.  The
daily discharges at the Northport Gage for the two days of the study were
obtained from the USGS.  The flow for August 3 was 6530 cfs, and the flow
for August 4 was 5790 cfs.  In modeling, the mean flow of 6160 cfs was used,
since the measurements of DO and BOD spanned the two days.  The water qual-
ity parameters were sampled at each bathymetric measurement point.  _Tn situ
DO and temperature measurements were taken with a Yellow Springs ModeT 54B
DO meter.  Table 5 summarizes the samples taken during the field study.

                              Modeling Results

Calibration/Verification

     The QUAL-II model was set up as described with the measured hydraulic
characteristics as input.  Using the data measured on August 3-4 as a ref-
erence, the model was run to verify that reasonable results were produced.
Minor adjustments were made in reaction rates and the method of treating the
hydraulics until reasonable agreement was reached between calculated and
measured DO and BOD.  The results of this process are shown in Figure 4.   It
should be noted that the City of Tuscaloosa Sewage Treatment Plant (STP)  was
not operating properly on the days of measurement.   The level of BOD5 in the
STP effluent was at an average level of 211 mg/1.   The level of NH3 assumed
in the STP discharge was the permit level of NH? scaled by the ratio of
actual BOD5 to permit BOD5:

                    NH3 discharge = 20 x 211/30 = 141 mg/1
It can be seen that reasonable agreement was  obtained,  particularly in the
case of DO.  The large amount of scatter in the measured BODs can be
accounted for by incomplete mixing of the STP discharge in the portion of
the river where the greatest changes in BOD levels  would be expected.

No Mill

     The model was next run with the river flow at  the 7Q}Q level of 605.7
cfs and all present dischargers operating at  their  full permit limits.   The
minimum calculated DO was 5.39 mg/1, occurring at the downstream end of the
reach.
With New Discharge

     When the new discharge was input at the  levels described above and the
river flow was 605.7 cfs, the flow augmentation option was used.   In this
case, a total augmentation of 60.9 cfs was required to maintain 5.0 mg/1  of
DO.  Several runs were made to provide information  necessary for input  to
the retention plan analysis.   Since the maximum daily discharge permitted
under NSPS is twice the 30-day average, this  level  is considered to be  the
upper limit for the discharge.   Computer runs were  made at 25%,  50%,  75%,
150%, and 200% of the 30-day average discharge levels to determine the  cri-
tical flows for these discharge levels.  The  results  are plotted in Figure 5,

                                   134

-------
    X
    rH
    aj

    S

    LO
LO


(U
    CD
    •H
, c
co ,2 2*
o ^ ^
•" S!£-
CO
4-1
•oS
•H H
XI Z

3
H
4J
C
r«d CD
ox ^
rH PU CO
0 < p.
0 _tx
rH
C/> 	
gjf
y^
inrH
O CO
10 •£.

IK?
vVO

rH
8^,
00
•H
EH
OJ
CO
a






c
0
•rt
JJ
CJ
P-)


Z Z



O ON

o in
rH




0 0
NO O
CN

O CO

ao r~
CO
m in
0 rH
o o

in -> • IS •
O Orf Otf

aw £c&

s_, ^, — 2
2 p-^b^^>4>H!Xb^l?>l^>4?>Hi^



in ooocNoincMCNvoo-i CTNCM

o\ cvj i"1-* vo r**- r^* LO vo vo in i/> VD
co




o ioir>u~ioou*ioou"1 in in
in CSi-HrHCNi— IrHCMCMi— ( 1 i 1 i— 1 t— 1
t—t

0 0»O^ • •

P*. C l-^CMCMOOO^O^O^OOr^^^OO
a>- cu-rocococneocMcMOJCNCMCMcoco
i-l O- CO Pi
« en bow SSSSrSSrSrSSSSSS:
3 03-* CQ G.' Oi O^O^Cdt^OtJodP^C^Pi Qd D^ Pd

2




^O

vO





o
CM



vO
rH

t

ON
Ol

CM
ao
ro
o
m
^

CO






rH
i-H
CO
S
cd
                                              135

-------
 DO Calibration
BOD Calibration
   Figure 4
      136

-------
            Mill Discharge Fraction vs. River Flow
 River  Flow (cfs)
Discharge Volume (70)
       465.0
       623.2
       666.6.
       786.0
       877.0
          0
         75
       100
       150
       200
Note:   100% discharge is 5.028 cfs and 84.7 mg/1  BOD,- or
        2300 Ibs/day.                                  D
                                                        4-
           -H-
                             t\ '• I !
                          \0ki OF FULL
              Regression Analysis of Critical River Flows
                            Figure 5
                               137

-------
and a regression line was calculated for use in subsequent analyses.  The
equation of the regression line is given by:

     X = 2.0                          881 < Q
     X = -2.229 + 0.004798 x Q        465 < Q < 881

     X = 0                            otherwise

where X = fraction of maximum new discharge; and

      Q = river flow.

The correlation coefficient is 0.9982.

                         Persistence of Daily Flows

     It is evident from the results reported in the previous sections that
the allowable discharge level is highly dependent on daily river flow val-
ues.  Since the current river operating system went into effect in the early
1960's , the daily flow fluctuations have been greatly reduced, and prolonged
periods of low flow are rare.  This is evident in the fact that for the
period of record of the Northport Gage (1895-present), the 10-year 7-day low
flow (7Qi0) has risen from approximately 96 cfs to 605.7 cfs.  Since the
volume of retention is greatly dependent on day-to-day flow variations, an
analysis has been performed to study the actual persistence of day-to-day
flows.  As an example of the issue at hand, the 7Q}Q is 605.7.  This could
be taken to mean that for at least 7 days, no discharge could occur (if the
limiting flow happened to be at or above this value), resulting in 7 days of
required storage.  However, this 7-day average could also be the result of
the following sequence of flows (in cfs):  248, 248, 1500, 248, 248, 1500,
248--resulting in 2 days of required storage (if the limiting flow was less
than 1500).  Thus, modeling the day-to-day variation is fundamental to
studying this problem.

     The USGS has supplied statistics on the daily flow records from the
Northport Gage from 1964 to 1981.  The distribution of daily flow values is
shown in Figure 6, along with a calculated log normal distribution which
shows that the log normal distribution is a reasonable model of the daily
river flow.  The parameters of this distribution are:
     logq = 8.31  (mean)

       VQ = 1.30  (variance)

        N = 6575  (number of samples)

                      Synthetic Stream Flow Generation

     This section presents a method for generating daily stream flows by
computer.  These stream flows will correspond to the actual stream flows on
three measures of performance — the mean and the variance of the logarithm of
the daily flows, and the mean annual 7-day low flow.  The method used has
been described by Loucks et^ jil_. *  A simple autogressive model is used.  This
type of model is a first-order Markov process which recognizes that the flow
*Loucks, D. P., J. R. Stedinger, and D. A. Haith.  Water Resource Systems
 Planning and Analysis.  Prentice-Hall.  1981.  Chapter 6.

                                    138

-------
> J
Q
88
W

                          (Wx^T-iaapH+T
focffLriMflfrt!l3'?>li »m-i| >t*i|: frfrft
ii Tl i i  i T I ill Ti i! TriT : !.ii.!;  : in
                                                 1O     11
                                                               11
            -5-1
                                   Figure  6

                      Distribution  of Daily Flow Values

                                       139

-------
today is to some  degree  dependent on the flow yesterday.  The generation  of
synthetic stream  flows requires  the marginal distribution of flow on  one
day given the  flow  on  the previous day.  Thus, if there  is  a low flow today,
a low flow is  more  likely than a high flow tomorrow.  The preferred method
of generating  such  flows is  to transform the flows to a  normally distributed
random variable.  In this case,  this corresponds to dealing with the  logar-
ithms of the flows.  A first-order Markov model for such a  variable is given
by:
      Xt+1 = V +  Px(Xt
where X = log  Q;
      u = mean of X;
     p  = lag  1 autocorrelation  coefficient;
      J\.
      t = time, in  -days;
     a  = standard  deviation of  X; and
      A.
     V  = unit normal random variate.
In this case,  y = 8.31;
             a = /~O"  = 1.14;  and
               J\.
               p is  varied until  the mean annual 7-day low flow is reproduced.
Table 6 summarizes  a number of computer runs for various values of p.  For
this study, a  value of 0.45 is used.
                            Retention Pond Model
     The basic mathematical expression of a detention basin is given by the
following ordinary  differential  equation:
     ^| = i(t) - o(to                                                   (i)
where s = storage volume;
      I = inflow;
      0 = outflow;  and
      t = time.
In this case,  I is  a constant 3.25 mgd.   The problem becomes one of deter-
mining 0(t).   The following procedure is used.   The flow generation model
described above is  used to generate a daily flow value.   The relationship
determined previously is used to determine the  maximum permissible discharge:
     f = 0                            QR < 465
     f = 0.00478Q - 2.229             465 < QR  < 881
     f = 2.0                          QR > 881
where f = fraction  of daily average discharge (3.25 mgd); and
     QR = river flow, in cfs.
                                    140

-------
                             Table 6
Dependence  of 7-Day Low Flows on the  Lag 1 Autocorrelation Coefficient
p
0
0
0
0
0
0
0
0
.3
.4
.5
.6
.7
.8
.9
.45
Mean 7 -Day Low Flow
1364
1146
959
806
695
631
600
1050.3
USGS value  is 1058.11
                               141

-------
     Equation (1), in discrete form and rearranged, is used to compute the
daily change in required storage:

     AS = [I - 0]*At

If storage goes negative, it is set equal to zero.  Another flow is gener-
ated, and the process repeats.  A flow chart representing this procedure is
shown in Figure 7.
     An algorithm was written to implement this procedure on a small com-
puter.  A run modeling performance for 50 years was made and the resulting
distribution of annual extreme maximum storage volumes plotted.  Table 7
contains the results of this run.  A Pearson Type III extreme value prob-
ability distribution was fitted to these observations.  Using this distri-
bution, a graph was constructed to show the required storage volumes for
different return periods (Figure 8).

                            Proposed Control Plan

     Since development of this plan has been based on the distribution of
low river flows, it is proposed that the level of discharge be controlled
in coordination with present flow levels.  In order to do this, river flow
must be measured or estimated locally.  Using the measured flow level, the
discharge may be regulated according to the following relationship:

     QD = 0                           QR < 465

     QD = 24.12E-3 x QR - 11.21       465 < QR < 881

     QD = 10.06                       881 < QR

where Qn = discharge flow, in cfs; and

      QR = river flow, in cfs.
     The purpose of the discharge control plan is to maintain Black Warrior
water quality while at the same time allowing the facility to operate at
design capacity.  To this end, retention storage has been incorporated into
the overall facility design.
                                 Conclusion

     A method of generating a time series for the modeling of extreme
values, in this case annual maximum storage volumes, has been presented.
The dependence of these processes on higher moments of the underlying dis-
tribution makes use of existing records unreliable, since rarely are hydro-
logical time sequences long enough for the estimation of higher-order
moments.  In the future, models using higher-order autoregressive and/or
moving average parameters may be fruitfully explored.
                                    142

-------
Flow Chart for Retention Pond Performance Computer Program
                                 143

-------
.a
 ro
                   o
                  p*
                   ff
                  Tl

                   01





                  I
                                                                                                                          CM  CO       r-l

                                                                                                                          •J-  r-  T-l  CM

                                                                                                                          CO  CJN  IO  l/"t
                                                                                                        CM  O  i-l   ON

                                                                                                        CO  ^  r-l   CO
                                                                                                        CM  tn  CM



                                                                                                         n    n    n
 03

 0)
O
LA
 (U

 CD

 <0
  111

  01
 fs<
  I

  01
  M
                                                                                                         01
                                                                                                         no
                                                                                                         eg


                                                                                                         S   S  I
                                                                                                         00   C    .
                                                                                                                      &
                                                                                                                      H
                                                                                                                                   4J   O


                                                                                                                                   01   01
                                                                                                                                   6   M
 H
 o
 E
 X
 (0
 C.

 C
                          ,-<   co  m  r-*  ON
                          o   o  o  o  o
                                   CO   to  f^  O">   r~I  CO  I/I

                                   r-l   ?-l  i-l  i-l   CM  CM  CM
r~-

 o;
 o
I—I
PL.
                 T3

                  01
                                                             -*  CO
                                                                          •-!  eg  oo  r-i
                          O'lONONONr-IU'iOCOOin
                                                                                                         ON  ON  ON  ON  ON  ON  ON  ON   ON  ON  ON  ON
                                                                                                                                                o>  c^   tn
                                                                                                                                                CM  i-l   CM
                                                                        144

-------
Storage Requirements for Various Return Periods
Storage Volume Days at Return Period
(acre-feet) 3.25 mgd (years)
27.8 2.79
32.2 3.23
41.5 4.16
44 . 8 4 . 49
-|=-=|:|:||fi
i
i




(3 2 	 -r-^-H-J 	 (-»-rH--
° ] * ' J x T^' i " •[ n ~ -^~
™ J,t_- 	 h 4-pI 	 1 	
*"• sW^^^^'W^^yr
i! |^t|P|l
" ' ' i ' a I
141 yj. 1 2 "^ iHn


^IJ4V^:I|f 4^
5 i i ' /'
* ! ! ' [ ! ! 1 ! i
i ^jg^s*±^

' ' ! | i
1 ' 1 i 1 i j
« nuiitit IT iii
a o |0
5
10 'I^fl^E

i ' ! i ^
50 £^i:^&g

a!
;

	 h ~i~t^~1 — | — M T"^ — H I ' — ] t \jf J 't • ' — *~H 	 j~^t — ^
"BBffiSffiS^'Stt^
i I i r 1

" " , p /i
" ~ " " 1 + > ' - - i

igS::-HllfflT^^iriS:^ia

-r;Hq n j 1 4~~*^ -U--- 4--^ 1 t-i | — H-J — l-|-j~ ' ' 1 j -j-7
„ | 1 i 	 4_|_ ^4-4- J-l 	 h-H- 1 t 1 1 I i 1 1 I 1 I 1 llj -|-
Hl 1 ; ! i j
"IM ---^r- M 4L l


:::::^:::T:::T|::j;r::;g+:^;^:T
| |
i it t "•

TJTT ±:^j:+r:j: jiq::H^ tn4: |m +Hj- ^ S 5
| 1 i |
I | ' ' "i ' ' : ' \
\ [ [ 111'
a
^^Za^III85
/T : ! : ! 1
• ' i ii

"^""uHUtlJ- *
1 4 , ' 1 »
j !; S
3wmrpnj^

i ;
! j


1 ! |

^-^-p- |N •!' -8
- ij-i-i i i-l-i Hil.J-i.iJg

::::lt:l|S"




-M 1 I- -f-j 1 1 -1- 4 ^
: ! i CM
, 	 r^ l^^^f °

- i i 1 1 -1 — i 	 L-rM- «*
	 r.l j .. _j__| 	 r4- 1 ) O


, : ; i ! : : i S
Z<> 3o ^ 50 feo 70 *
                                 Figure 8

The work described in this paper was not funded by the U.S.  Environmental
Protection Agency.  The contents do not necessarily reflect  the views of the
Agency and no official endorsement should be inferred.
                                    145

-------
             A MECHANISTIC SIMULATION FOR TRANSPORT OF NONPOINT
                             SOURCE POLLUTANTS

                                     BY

                         Daryl B. Simons, President
                   Ruh-Ming Li, Executive Vice President
             Kenneth G. Eggert, Senior Water Resources Engineer
                       Simons, Li & Associates, Inc.
                           Fort Collins, Colorado
                                  ABSTRACT

     Nonpoint source pollution represents an integration of the behavioral
response of the ecosystem to the impacts of land use activities on indivi-
dual components of the ecosystem.  By analyzing basic ecosystem processes
and impacts of land use activities on specific processes governing produc-
tion and transport of nonpoint source pollution, it is possible to predict
cause-effect relationships between these activities and water quality.  In
order to accomplish this goal, a simulation model for evaluation of alter-
native management practices as a function of both environmental and manage-
ment parameters is presented.
     The simulation may be used to predict watershed response to land use
including both planned management practices and unplanned activities asso-
ciated with catastrophic events.  Nonpoint source pollution from these acti-
vities and events may include loading of streams by:  1) sediment from sur-
face erosion, mass wasting, and channel bottom and bank erosion, 2) thermal
energy, 3) biological contaminants, 4) organic debris, 5) nutrients and
dissolved solids, 6) pesticides, and 7) other waste materials either in
solution or adsorbed to sediments.  The simulation is composed of the
following components:  1) a soil-plant-atomspheric water simulation for
adjusting soil moisture as a function of evaporation, evapotranspiration,
soil water hydraulics, and snowmelt, 2) a kinematic wave surface water
routing component, 3) a hydraulically-based sediment yield model, 4) a sub-
surface flow component, 5) a temperature and dissolved oxygen component, 6)
pollutant routing, and 7) streambank erosion and forest litter routing
routines.

                                INTRODUCTION

     The theoretical foundation of the watershed simulation component of the
generalized planning model for evaluation of alternative management prac-
tices is described in this paper.  The model is a cause-effect simulation
based on mathematical descriptions of watershed processes related to


                                    146

-------
nonpoint source pollution.  Nonpoint source pollution represents an integra-
tion of the behavioral response of the ecosystem to the impacts of activi-
ties on individual components of the ecosystem.  Management activities
include both planned management practices, such as road construction,
watershed management, stream improvement, low head impoundments, application
of pesticides and fertilizers, foliage production and grazing, mechanical
site preparation, and disposal of wastes by land treatment, and unplanned
activities associated with catastrophic events such as large runoff events.
The model as originally formulated was addressed to management activities
involving timber harvest, timber planting and replacement, grazing, mechan-
cial site preparation and prescribed fire.  However, the mechanistic basis
of the model makes it applicable to nearly any watershed activity.  Specific
applications to agricultural activities as well as waste land treatment and
disposal are well within the scope of the model formulation.
     The watershed process model is structured in modular form.  The modules
contain related physical process simulations and are described individually
later in this paper.  The basic structure of the model is shown in Figures 1
through 3.  The model structure was designed to allow a high level of repre-
sentation of physical processes but still remain within the range of memory
constraints normally encountered at most computing centers.  The modular
nature of the model structure also facilitates substitution of new or
improved simulation components as they are developed without extensive revi-
sion of the code.  The approach has both advantages and disadvantages.  The
basic advantages are the modular nature as described above, and the capabi-
lity of using simulation routines that are not overly limited by computer
storage.  The disadvantage of the structure is that input/output sequences
and file management are machine specific and must be tailored to the host
computer.  However, the basic file management functions used are generally
available at most computing centers.  Presently, the model has been
installed on CDC Cyber 172, HP 1000 and IBM 370 systems.  Therefore, the
advantage of a more precise representation of physical processes is felt to
outweigh any disadvantages due to machine specific job control language.
     As shown in Figures  1 through 3, the basic model components are as
follows:

     1.   Water balance
     2.   Water and sediment routing
     3.   Pollutant routing
     4.   Heat and dissolved oxygen routing
     5.   Streambank erosion
     6.   Vegetative litter routing
     7.   Sequential GOAL Programming

                               WATER BALANCE

     Since infiltration exerts a fundamental control on the storm water
runoff hydrograph, any long-term hydrologic simulation must have a component
for calculating the changes in soil moisture content as a function of time.
The primary processes affecting the amount of soil moisture are infiltra-
tion, percolation, evaporation, evapotranspiration and drainage.  These
                                    147

-------
                                                            4
                                                            Cn


                                                            £
                                                            C
                                                            a

                                                            -d
                                                            o>
                                                            N
                                                            •H
                                                            rH
                                                            
-------
CD
II r
111
UJ ^ «

CD
u_ ^ o
c 5

II*

CO
z E

LL. -
S

CONxxxx (File)
Wotershed
Configuration (C)
1
1
1
1
... ... ,
1
1
1
CD
« -a
J!
- -
SORT /MERGE
System Sort
Temporal-Spatial
                                                                          r'
            ^\~\
                  is I
                  UJ tf> =


                  11?

                  Sis

                  S«8
              r
                                                           '^ S
r
i	
     •~i
  z •

  "
o
"If
111
•2






o
I||
111
z °-
          55
          LL. -o
                            O.T)

                          "* 2 Si


                          LL o1 S
a> . —
3 CD
7*?
|l(
•?«
0






0
SSI
8*Jf







s£
if!
ill
>•£
Q1"
ffm

^1?
~|s
ri.
^F?
h s a








"B
u] « 6
S C|S
si|







w m

• 1?
-II
Sff
' ^1
                        	zr

IS
                                    I
                                                   HI
                                                                          i	
                                                           5~ «

                                                           9s
                                                           ~ E o
                                      in


                                      •O


                                      (fl
                                      8.
                                      T3
                                 I*1" |  <"



                                 ||t I  |


                                  ?tn CT I
                                   .r i
                                      Oi

                                      •H
                                 1*1

                                 It!
                                 Z~f
                                 v at *^
                                    149

-------
r



, i
MClxxxx (File) TAPEI5 (File)
Hydrogroph Timing — --*r* ~ Number of Events
Parameters (C ) ' with Minimum Runoff (C)
! I



i
1 	 TAPE3 (File)
Plane Water
F09xxxx(File) [ ' Discharge (C)
Computational (-»• |



FIOKXXX (File) 1
Channel Data -»— i — 	 	 	 	 	 *


t
QUPxxxx (File) QPKxxxx (File)
Upstream -*• Peak
Inflow (C) Discharge (B)

r



»
ALIxxxx (File)
Alternative One
Hydrogrophs (B)
(






MULT (Prgm)
Multiple Watershed
Program
	 1 	
±
. - .j ... ..._
TAPE7 (File)
Plane Sediment a
Water Discharge (C
i
1
USFSORT [Prgm}
Spliting and Sorting
Of Two Fi les
1
1
TAPE4 (File)
Plane Sediment
Yields (C)
J

1
MCHNL (Prgm)
Multiple Channel
Mode!

1
y-t- _-
SSMxxxx (File)
Total Sediment
Yield (WS) (C)


t
ALTPLOT (Prgm)
Merger of
Three Files
(
t
AL2«xxx (File)
Alternative Two
Hydrographs (B)
j
i
PREPLOT (Prgm)
Merger of up to
Four Files
1
ALSxxxx (File)
Hydrographs for A] 1
Alternatives (C)


ALOxxxx (File)
Alternative •
Parameters (8)
	 ... t
	 P
J J
TAPES (File) TAPE9I (File) 1
Watershed Sediment Julian Dates r~
8 Water Discharge (C) of Storms (B)|
|



| |
TAPEI2 (File) TAPEI3 (File)
Watershed Water Watershed Sediment
Discharge (C) Yields (C)
J_ _J






-n " "" "~T ' "' "1
1 HYGxxxx (File) TAPE92 (File)
| Storm Storm *
1 Hydrographs (C) Hydrographs (8) ,
_|




I 1
AL3xxxx (File) AL4xxxx (File)
Alternative Three Alternative Four
Hydrographs (B) Hydrographs (B)
j 1



Plotting Package to
	 • TEKTRONIX
Plotrtr '
l
Figure 3.  Water and Sediment Routing
                  150

-------
interrelated processes involve hydrologic, biologic, atmospheric and soil-
specific aspects.  Therefore, a physically-based water balance model must
simulate all of these aspects and properly account for their interrelation.
After a literature search for an existing water balance model, a decision
was made to modify and implement a simulation presented by Goldstein and
Mankin (1972).  This program, known as PROSPER, has been widely tested in a
variety of locations with generally good results (Swift et al. , 1975;
Luxmoore et al., 1977).  The success of these applications is probably based
on the flexibility built into the model.  It is possible to adapt model
subroutines to a particular watershed environment by modifying the methods
used to calculate the resistances to water flux through the soil and plant
components.
     PROSPER is a plant-atmosphere-soil water flux simulation which imple-
ments an energy balance and aerodynamic calculation of evapotranspiration
using the Penman method with a multilayer Darcian soil model.  The model
simulates the fluxes of water through the soil and plants in response to
atmospheric and solar conditions.  The simulation uses a time increment of
one day.  All hydrologic plant and atmospheric processes are averaged on a
daily basis.  The model as implemented uses an electrical circuit analogue
for the soil and plant system.  The current in the circuit loop represents
the water flux through the respective soil, plant, or atmospheric component.
     Since PROSPER was written for predicting daily variations in water
flux, its formulation was found to be inadequate for the prediction of
infiltration and interception on a storm-by-storm basis.  Further, since the
time history of infiltration during a storm is particularly critical when
determining water and sediment runoff, the effects of watershed management
on infiltration parameters could not be properly simulated without modifica-
tion to PROSPER.  Therefore, the water balance component was modified to
include modified Green-Ampt infiltration component for layered soils for
more precise determination of storm water runoff and a storm water intercep-
tion routine.  Both infiltration and interception components provide greater
sensitivity to management activities in the watershed than the original
PROSPER, and the infiltration routine provides for the interfacing of the
water balance component with the water and sediment routing and water
quality subprograms.
     Figure 4, which is modified from Goldstein and Mankin (1972), shows a
schematic representation of the water balance system.  Resistances in the
current analogue are formulated in terms of physically defined watershed
moisture parameters.  These parameters include soil layer hydraulic conduc-
tivities, root resistance, and empirically measured species specific stoma-
tal resistances.  By varying the values of these parameters to fit the phy-
sical setting of the watershed, it is possible to achieve a more precise
representation of physical processes than provided by previous simulation of
similar scope.  •
                                    151

-------
                 Infiltration to Layer I
Layer  I
                            RSBLI .>
                                                  RR2 + RRI/2
                 Infiltration to Layer 2
                            RSBL2
                 Infiltration to Layer 3-
                            RSBL3
                 Infiltration to Layer4-
                            RSBL4
                 Infiltration to Layer 5-
Layer 2
Layer 3
Layer 4
Layer 5
                                To Drainage

                    Figure 4.  Soil-plant  flux schematic.

     In many  watersheds the most significant  contribution to runoff  is  from
snowmelt.   Often the most severe erosion  and  sedimentation events  occur as
combinations  of high intensity spring rainfall and snowmelt runoff.   This
aspect of  the hydrologic cycle is particularly important in high altitude
watersheds of the western United States and in northern latitudes.   After
examining  available snowmelt models, it was felt that a modified version of
the model  developed by Leaf and Brink (1973)  would provide the best  approach
for incorporation into the generalized planning model.   This model is a
simulation originally developed for snowmelt  prediction in Colorado
subalpine  watersheds.  It is a mechanistic approach with data requirements
consistent with previously described planning model components, in par-
ticular, evapotranspiration and water balance components.  The snowmelt
model as formulated is designed to provide daily runoff water yields  and
therefore,  it requires modification to interface with the other water
balance routines.
     The snowmelt model is a combination  of the fundamental laws of  conser-
vation of  mass (water balance) and conservation of energy.  The thermal
                                     152

-------
state of the snowpack  is  described by the calorie deficit  of  the  pack on any
given day.  The  calorie deficit is defined to be the  number of  calories  (per
unit area) required  to bring the pack to an isothermal  state  at 0°C.   It is
assumed that no  melt can  occur until the pack has reached  this  state.  The
general scheme of  computing snowpack accumulation,  temperature  change, and
melt begins with an  assessment of the effects of the  day's precipitation, if
any, on the pack.  This includes the effect of the snow or rain introduced
to the pack at the average  air temperature, as well as  the addition  of water
to the pack as snow  or as rainfall.   After the effect of rainfall has been
assessed, an energy  balance calculation is begun with a determination of the
net radiation input  (or loss)  to the snow pack.   Depending on the thickness
of the snowpack  and  the amount of free water existing in the  pack, a thermal
diffusion model  is used instead of a radiation balance.  The  new  pack tem-
perature is calculated using either the thermal diffusion  model or the
radiation balance  directly.   Based on the new temperature, a  new  calorie
deficit or melt  is calculated, and the simulation proceeds to the next day's
precipitation and  radiation data.  Water balance relations are  adjusted
daily based on the above  outcome.
     The original  Leaf and  Brink (1973) snowmelt model  was designed  to pro-
duce daily runoff  water yields.  As in the previously discussed case,  this
type of runoff description  is  inadequate for prediction of the  water
hydrographs required for  simulation of nonpoint source  pollution  transport.
A means for transforming  the melt water volume into a snowmelt  hyetograph is
required.  Since the melt occurs in response to the input  of  solar energy,
it seems reasonable  to approximate the snowmelt hyetograph by distributing
the melt over the  sunshine  period using the hourly insolation divided by the
total daily insolation as a weighting function.   Since  hourly measurements
of insolation at the snowpack are not usually available, it is  assumed that
the corresponding  ratio of  hourly extraterrestrial to total extraterrestrial
radiation will suffice.   This  insolation weighted meltwater runoff function
is subjected to  infiltration at a rate equal to the saturated hydraulic  con-
ductivity, and the resulting excess runoff is treated by the  routing models
in the same manner as  the excess rainfall described above.

                          WATER AND SEDIMENT ROUTING

     The water and sediment routing model is designed to route  storm water
and sediment runoff  from  watersheds of complex geometry.   In  order to
accomplish this  task,  the complex watershed geometry  must  be  simplified  into
a representation suitable for computer simulation.  The geometric approxima-
tion used in this  component is an arbitrary number of two  plane,  one-channel
"open book" subwatersheds and planes linked together  by channels.  Figure 5
illustrates a sample representation of such a system  of planes  and channels.
     For simplicity, a numerical solution to the kinematic wave problem
could have been  used for  both the subwatershed units  and the  linking chan-
nels.  However,  an analytical solution such as the method  of  characteristics
approach allows  more efficient use of computer storage  and usually more
rapid calculation  of the  runoff hydrograph.  Therefore, whenever  possible,
analytical solutions are  employed.  To be consistent, in the  portions of the
watershed where  the  analytical methods is used to route the water, the sedi-
ment yield is also computed by an analytical method.  Likewise, in the
                                     153

-------
CH-5
PL -9 PL- 10
|

^\
WS-7 J
V^/
CH-4
PL-7 PL-8

,.
CH-6
PL-II PL- IE

CH-7
PL- 13 PL- 14
1
CH-8
PL-15 PL- 16
i
CH-9
PL-17 PL- 18
J
t
CH-IO
PL-19 PL-20
1
CH-II
PL- 21
         Figure 5.  Schematic diagram of a typical routing network.

portions where water is routed numerically, sediment yield and transport are
computed by a numerical routing scheme.

Overland Flow and Primary Channels

     The two tasks performed by this program are the determination of sedi-
ment yield by size fractions and the routing of water for all plane and
upstream watershed units that form the entire watershed.  The water routing
is performed analytically by applying the method of characteristics to the
kinematic wave approximation to the momentum and continuity equations.  The
sediment yield is calculated by comparing the supply due to detachment by
both rainfall and runoff, and the potential transporting capacity.  The
transport capacity is determined by size fraction.  The suspended sediment
transport capacity is calculated using Einstein's (1950) suspended sediment
equation and bed load is calculated using the Meyer-Peter, Muller (USSR,
1960) equation.  These methods are used in both the overland flow and pri-
mary channel flow routines.

Main Channel Water Routing

     The main channel routing program uses a numerical scheme to route the
water in the downstream main channels.  It uses the discharges calculated in
                                    154

-------
the overland flow and primary  channel  routines  as upstream and  lateral
inflows into the channel units.   In  addition, an infiltration routine calcu-
lates the amount of water which  infiltrates  in  the channel units and
subtracts it from the lateral  inflow.

Sediment Routing

     The sediment routing in the main  channel routine uses a similar prin-
ciple to the sediment yield calculations  in  the overland flow and primary
channel model—the process of  balancing supply  and capacity for each sedi-
ment size.  In addition, the main channel includes the effect of armoring on
the sediment transport  rate.   The process of armoring occurs because of the
difference in sediment  transport capacity between the different sizes,
resulting in a layer of large  size fraction  formed on the surface.  If the
erosion processes continue, this layer of larger size fraction will protect
the smaller one from detaching or dislodging.
     The sediment calculations in the  main channel are also different in
that the numerical method used allows  for the sediment to be routed through
the channels at each time increment  and then integrated over the time incre-
ments to arrive at a total yield for each size  fraction.  The method used in
overland flow and primary channel model can  only provide a total yield for
each size fraction, but cannot truly route the  sediment through the channel.
The main channel model  is able to route the  sediment because the balancing
of transporting capacity and supply  can be compared at each time and space
increment along the channel due  to the use of the numerical scheme.  The
method used in the overland flow and primary channel does not allow for
this, since only average conditions  are determined.
     Since the main channel model uses the numerical method, which requires
sediment transport rates at each time  increment, and needs to use the yields
calculated in the overland flow  and primary  channel model as upstream and
lateral inflows, the yields in the latter must  be transformed into sediment
hydrographs.  This is accomplished by  distributing the yields from overland
flow and primary channel model in proportion to the water discharge at each
time increment.  A more detailed discussion  of  the water and sediment
routing component may be found in Li et al (1979).
     This water and sediment routing method  has been successfully employed
on a variety of watersheds.  Results have shown it to provide very accurate
characterization of watershed  hydraulics. A considerable degree of emphasis
has been placed on the  previously described  water balance and the water and
sediment routing components.   Pollutants  are, of course, chiefly transported
by surface and subsurface  flow components.  Therefore, the characterization
of the hydraulics of infiltration, overland  flow, channel flow, and the
mechanics of sediment yield and  transport are critical and of central impor-
tance to accurately estimating pollutant  migration.  The described hydraulic
components provide the  level of  accuracy  necessary to represent these pro-
cesses and through the  model structure remain practical for implementation
on most computing hardware.

                               POLLUTANT ROUTING

     As originally formulated, the pollutant routing module addressed the
routing of nutrients.   Nutrient  compounds are a source of nonpoint

                                     155

-------
pollutants affecting water quality.  Nutrients, such as nitrogen and
phosphorus, are of  concern because of their role in eutrophication pro-
cesses.  A physical process simulation model was developed for predicting
nutrient losses from forest and agricultural watersheds associated with  sur-
face runoff and sediment transport.  Mass balance and loading function con-
cepts were the basic principles utilized in formulating this model.  The
model was developed to predict loadings of organic nutrients, nitrate, ammo-
nium, and inorganic ortho-phosphorus to streams and rivers.
     Natural nutrient input to the ecosystem comes mainly from precipita-
tion, litter fall,  and geologic weathering.  Precipitation and litter fall
were considered the primary external inputs of nutrients from the
atmosphere.  These  average inputs were routed into the litter layer where
microbial degradation occurred.  The products of degradation were then
routed to the stream and into the soil layer.  Within the soil layer, these
products were again evaluated along with plant uptake and soil adsorption.
The products of these processes occurring within the soil were then routed
to the stream.  Generally, nutrient constituents cannot move unless
transported by sediment and water and therefore, water and sediment are  the
major carriers of nutrients through the ecosystem.  Evaluation of these
carrier amounts is necessary for predicting nutrient losses from the
watershed.
     The nutrient simulator proposed here is basically a nutrient budget
model.  All of the processes mentioned above except the immobilization pro-
cess were taken into account when simulating average nutrient concentrations
in the soil.  The quantities of nutrient losses to streams during storms
were predicted by the incorporation of the loading function concept.
     As stated earlier, the pollutant model was originally developed, as
described, for nutrient routing.   However, it is currently under modifica-
tion to allow the addition of pesticide routing processes.  Again, an
existing accepted pesticide model was selected.  The pesticide model chosen
for this study is a simple mechanistically based model developed by the
Agricultural Research Service (Knisel, 1980).  It is directly coupled to the
water and sediment yield model through the use of water and sediment runoff
volumes as inputs.  The model also requires an estimate of the initial
moisture deficit in the soil, which is obtained directly from the water
balance routines.
     The pesticide component (as  reported by Knisel, 1980) was developed on
simplified concepts of processes  and designed to be responsive to different
management options.  Foliar- and soil-applied pesticides are separately
described so that different decay rates can be used for each source of the
same chemical if necessary.  Usually pesticide residing on foliage dissi-
pates more rapidly than that from soil.  Also decay rates can be made site
specific if information is available.  Movement of pesticides from the soil
surface as a result of infiltrating water is estimated using differences of
rainfall and runoff for the storm and pesticide mobility parameters.  Pesti-
cide in runoff is then partitioned between the solution in water and that
adsorbed to sediment.  This aspect is particularly important when examining
management options that may limit sediment yield.   Further,  the partition
coefficient system used is particularly useful since the coefficients are
lab measurable.
                                   156

-------
                   TEMPERATURE  AND  DISSOLVED  OXYGEN ROUTING

     Thermal energy  content,  dissolved oxygen (DO), and biological oxygen
demand (BOD) of runoff water  can  directly  or  indirectly affect the tem-
perature and oxygen  content in  the  stream.  Based  on mass and energy
balance,  the temperature  and  dissolved oxygen model is included  in this
simulation.  This model is useful to  evaluate the  thermal and dissolved  oxy-
gen loading to the stream through surface  runoff.
     Overland flows  transmit  thermal  and DO loading from land surfaces to
the stream.  Temperature  and  dissolved oxygen loading of the stream result
from high temperature, high bio-chemical oxygen  demand, and low  dissolved
oxygen in runoff water.   Temperature  and DO effects of subsurface flow are
not included in this model.
     The three mechanisms of  heat transfer, radiation, conduction, and con-
vection,  are included in  this model.   Each mechanism plays a role in the
heat transfer process.  Conduction  is the  only predominant mechanism for
heat transfer between soil layers and heat transmission between  soil and
surface flow.  Convective heat  transfer occurs because of relative motion
betwen various parts of the heated  body or fluid.  Convection plays an
important role in heat transfer from  water surfaces, particularly in eva-
porative processes.  In this  model, these  heat transfer mechanisms are used
to formulate equations for (1)  atmospheric processes, (2) canopy-ground
cover processes, and (3)  surface  runoff processes.
     The oxygen concentration in  the  stream water  at any given time is
determined by the solubility  of oxygen in  the water, the rate at which this
oxygen is consumed by various biological processes (represented  by biologi-
cal oxygen demand),  and the rate  at which  this depletion is replenished.
Deoxygenation of the water due  to the bacterial  decomposition of car-
bonaceous organic material and  reaeration  caused by the oxygen deficit and
turbulence are the most fundamental processed occurring in natural water.
The basic theory used to  describe the deoxygenation and reaeration processes
was proposed by Streeter  and  Phelps (1925).   The rate at which the BOD is
exerted was presumed to be identical  to that  observed while using the
laboratory BOD test.  A proportionality is assumed to exist between the
reaeration rate and  certain hydraulic parameters of flow.  The DO effects
include concentration reductions  due  to purging  action of gases  rising from
the benthal layer, plant  aspiration,  diffusion into the benthic  layer, and
DO addition photosynthesis.

                              STREAMBANK EP.OSION

     Much of the sediment production  of watersheds and channel systems ari-
ses from streambank  erosion.  In  the  United States it has been estimated
that approximately 500 million  tons/year enter the drainage systems of the
country (Barnes, 1963).   Therefore, a mathematical model of the  process  of
streambank erosion by channel widening is  included in this simulation.   The
predictive capability of  the  model  is enhanced by  its pheriomenological
structure, although  empirical data  are needed in the stream morphology com-
ponent.  The model estimates  the  total amount of streambank erosion and  the
fraction of it that  goes  into suspension.  Threshold channel conditions,
                                    157

-------
bank characteristics, and the hydrologic events are input to the model.
     As formulated, the streambank erosion model estimates the total  amount
of erosion that is likely to occur in the transition from a condition of
geomorphic equilibrium to another condition of equilibrium.  As such,  it
does not provide information on the rate of streambank erosion; rather, it
gives a total value assuming the new equilibrium condition is eventually
reached.  In practice, however, the rate of streambank erosion is a function
of the time history of hydrologic events which are not explicitly considered
in the present model.  Therefore, the calculated values are to be regarded
as estimates of the total amount of streambank erosion that is associated
with a certain level of hydrologic excitation.  Further refinements will
need to be implemented if the model is to provide information on the  rate of
streambank erosion.

                                FOREST LITTER

     A first approximation to the rill formation and the loading of forest
litter is included in this simulation.  The model is based on the assumption
that the amount of forest litter loading is directly proportional to  the
areal extent of rilling.  This is a reasonable assumption in view of  the
demonstrated effectiveness of concentrated flow in transporting sediment and
debris through upland watershed drainage networks.  This approach allows the
conversion of the forest litter loading problem into that of determining the
areal extent of rilling (rilling density), given a set of topographic,
hydrologic, and morphologic conditions.
     The quantity of forest litter delivered to a stream is a direct  func-
tion of the areal extent of rilling and the amount of forest litter produc-
tion.  The areal extent of rilling will, in general, be determined by large
events.  Smaller subsequent events will not entirely fill the established
rill network.  Therefore, the litter washed out of the rill network will be
detached from the area defined by the top width of the flowing water.  This
top width may be obtained from an "at a station" relationship provided by
Li, Simons and Stevens (1976).

                         SEQUENTIAL GOAL PROGRAMMING

     The problem of management of a basin depends very much upon the  selec-
tion of the best land-use strategy to optimize specified socioeconomic
objectives under certain constraints on water and sediment transport  con-
ditions in the stream channels and on water quality standards in the  basin.
The quality of a basin management plan depends on the quality and availabi-
lity of data.  However, the efficiency of planning depends on the optimiza-
tion tool to be selected and the accuracy of simulation models used to pro-
ject future basin-system response.  A review and evaluation of multi-
objective programming techniques used to solve basin planning problems has
been conducted by Loucks (1975) and later by Cohon and Marks (1975).   One
technique of increasing popularity often encountered in water resources
literature is goal programming.  It is designed to evaluate (possibly
conflicting) goals as well as goals of differing priorities.
     The planning model is designed as multi-level.  The lower level  is used
to select optimal land-use strategies based on different alternative


                                    158

-------
management practices and subject  to  some  specified  land-use constraints.   In
this level, the elements of the resource  response matrix are calculated by
various land-use process models which  use the same  type of spatial and tem-
poral information  (i.e., same  soil-vegetation units and time-step).  After
selecting a set of optimal strategies  for land-use  management, the upper
level is then used to select the  best  management strategy for the entire
basin system based on different land-use  strategies selected from the first
level and the outcomes of hydrologic and  water quality component processes,
subject to some specified socioeconomic constraints.  In this level, complex
models of water and sediment yields  and water quality are used to calculate
the elements of the process response matrix which then serves as the nece-
sary input for the selection of the  best  basin management strategy.

                            POTENTIAL  APPLICATIONS

     Potential applications of the simulation are initially intended to aid
in the evaluation of watershed management alternatives.  Management activi-
ties that could be considered  initially were vegetation growth (over story
and under story), timber harvest, foliage utilization (grazing), site pre-
paration, waste disposal and prescribed nutrient and pesticide applications.
The linkage of the management  activity models, the  process models and the
multiple objective programming model form the preliminary generalized
planning model.  This preliminary planning model would be useful in eva-
luating selected alternatives  as  a function of environmental goals.
Environmental goals relate to  control  of  sediment,  nutrients, pesticide,
hazardous waste, thermal, and  dissolved oxygen pollution.  Also, resource
management goals can be identified that maximize income and/or implement the
best management practices.  The changes caused by management activities are
reflected in changes in input  model  parameters.  The effects of these
changes may be simulated by adjusting  soil and vegetative and chemical para-
meters.  Chronological simulation allows  those parameters to be varied
either instantaneously or as functions of time.  By interfacing with the
planning model, these parameter values may be adjusted in order to select
the best management alternatives  and scheduling.

                               ACKNOWLEDGEMENT

     The writers wish to acknowledge Mr.  W. T. Fullerton, Mr. J. N-H. Ho,
Dr.  N. Duong, and Dr. V. M. Ponce for their valuable contributions to the
formulation of this simulation.   Financial support  for this study was pro-
vided by the Environmental Protection  Agency, Environmental Research
Laboratory, Athens, Georgia; and  by  the USDA Forest Service, Rocky Mountain
Forest and Range Experiment Station, Flagstaff, Arizona.

                                  REFERENCES

Barnes, R. C., Jr.  1968.  Streambank  erosion.  Soil Conservation
33(6):126-128.

Cohon, J. L., and  D. H. Marks.   1975.  A  Review and Evaluation of
Multiobjective Programming Techniques. Water Resource Research
11(2):208-220.

                                     159

-------
Einstein, A. H.   1950.  The Bed Load Function for Sediment Transportation in
Open Channel Flows.  U.S. Department of Agriculture Technical Bulletin
No. 1026.

Goldstein, R. A., and J. B. Mankin.  1972.  PROSPER:  A Model of Atmosphere-
Soil -Plant Water Flow.  Proceedings, Summer Simulation Conference, Los
Angeles, pp. 1176-1181.

Knisel, W. G.   1980.  "A Field Scale Model for Chemicals, Runoff, and
Erosion from Agricultural Management Systems," for U.S. Department of
Agriculture, Conservation Research Report #26.

Leaf, C. F. and G. E. Brink.  1973.  Computer Simulation of Snowmelt Within
a Colorado Subalpine Watershed, USDA Forest Service Research Paper RM-99,
Rocky Mountain Forest and Range Experiment Station, Fort Collins, Colorado.

Li, R. M., D. B. Simons, W. T. Fullerton, K. G. Eggert and B. E. Spronk.
1979.  "Simulation of Water Runoff and Sediment Yield from a System of
Multipole Watersheds," presented at the XVIII Congress of the International
Association for Hydraulic Research, Cagliari, Italy.

Li, R. M., D. B. Simons, and M. A. Stevens.  1976.  Morphology of Cobble
Streams in Small Watersheds.  J. Hydraul. Div., ASCE, 12(HY8):1101-1112.

Loucks, D. P.   1975.  Conflict and Choice:  Planning for Multiple
Objectives.  In:  Economy Wide Models and Development Planning, edited by C.
Blitzer, P.  Clark, and L. Taylor, Oxford University Press,  New York.

Luxmoore, R. J., D. J. Van Roagen, F. D. Hale, J.  B. Mankin,  and R.  A.
Goldstein.  1977.  Field Water Balance and Simulated Water Relations of
Prairie and Oak-Hickory Vegetation on Deciduous Forest Soils.  Soil  Sci.
123(2):77.

Streeter, H. W., and E. B. Phelps.  1925.  A Study of the Pollution  and
Natural Purification of the Ohio River.   Public Health Bulletin 146, U.S.
Public Health Service, Washington, D.C.

Swift, L. W., W. T. Swank, J. B. Mankin, R. J.  Luxmoore,  and R. A.
Goldstein.  1975.  Simulation of Evapotranspiration and Drainage from Mature
and Clear-Cut Deciduous Forests and Young Pine Plantations.   Water Resources
Research 11 (5):667.

U.S.  Bureau of Reclamation.  1960.  Investigation  of Meyer-Peter,  Muller
Bedload Formulas.  Sedimentation Section, Hydrology Branch,  Div.  of  Project
Investigation,  U.S. Department of the Interior.
 The work described  in  this  paper was not funded by the U.S. Environmental
 Protection  Agency.   The  contents do not necessarily reflect the views of the
 Agency  and  no  official endorsement should be inferred.

                                    160

-------
                        DATA MANAGEMENT FOR CONTINUOUS

                           HYDROLOGIC SIMULATION

                             Jy S. Wu, Ph.D.1


                              INTRODUCTION

     The use of models to simulate rainfall-runoff events is receiving
considerable attention by water resources planners and engineers.  One of
the major reasons for this concern is the limited capability of field
sampling programs in obtaining a continuous data-base which reflects the
watershed responses to changes in hydrologic conditions and management
practices over an extended period of time.  The conduct of a continual
field monitoring and sampling program is often quite expensive; even a
well managed sampling program can still be handicapped by the failure of
sampling devices and unexpected difficulties during the sampling period.
Mathematical modeling thus becomes a useful tool for analyzing existing
data and predicting future conditions as a result of implementing watershed
management practices.  In some cases, the collected data must be utilized
to statistically generate missing records for calibrating a continuous
hydrologic simulation model.  This paper  describes the application of the
U.S. EPA's Agricultural Runoff Model (ARM) on one of the agricultural
watersheds of the Chowan River Basin in North Carolina.  The technique of
regression analysis was employed to generate missing runoff data as
required for calibrating the ARM model.

                          WATERSHED DESCRIPTION

     The Cutawhiskie Lateral monitoring site is located in Hertford County
of northeastern North Carolina with a total drainage area of 667 acres.
Runoff from this monitoring site drains into the Cutawhiskie Creek which,
in turn, drains into the Chowan River near the North Carolina-Virginia
border.  The site is relatively flat with an average overland slope of 0.4%.
Land use is about 50% woodland and 50% cropland of peanut, soybean, cotton
and corn.

     The North Carolina State University, in cooperation with the Division
of Environmental Management, has been conducting a field sampling and
monitoring program for selected agricultural watersheds including the
Cutawhiskie site, to assess nonpoint source effects on streams.  The program
also calls for the survey of producers within the monitored watersheds to
assess land use, management practices and opinions on 208 planning and
  Assistant Professor, Department of Urban and Environmental  Engineering,
  University of North Carolina at Charlotte, Charlotte, N.C.  28223

                                     161

-------
implementation options.
have been collected for
 A field smapling data-base
a period of three years.

     RESEARCH RATIONALE
and survey information
     The selected period for hydrologic calibration of the ARM model  was
chosen from October, 1980, through September, 1981.  The hourly rainfall and
daily evaporation records were incomplete during this particular period.  It
was necessary to explore alternate procedures for generating missing  records.
Available data from nearby meteorological stations were utilized to construct
isohytes of rainfall and evaporation.  Estimations were then based on inter-
polating adjacent isohytes of the study area.

     The recorded runoff information was also incomplete.   A total of eleven
runoff events having well defined hydrographs were identified.  The technique
of multi-regression analysis was employed to generate missing runoff  records.
Runoff volume of each event was correlated with rainfall volume of each event,
duration of rainfall, and number of dry days prior to rainfall.  A regression
equation was developed and used to calculate the single event, monthly and
annual runoff volumes.  This allows calibrating the ARM model based on annual
and monthly water balances and individual storm comparison.  A detailed
description of data preparation is given in subsequent sections.

                         ARM MODEL  DESCRIPTION

     The ARM model is a continuous hydrologic and water quality simulation
model (Donigian & Harley, 1978).  It simulates runoff, sediment, pesticides
and nutrients contribution to stream channels from both surface and sub-
surface sources.  Major data requirements for hydrologic simulation include:

A.  Rainfall records on 5 minutes, 15 minutes, or hourly intervals.

B.  Daily evaporation data, and the following hydrologic parameters:

     Nominal upper zone soil moisture storage, UZSN
     Initial upper zone soil moisture storage, UZS
     Nominal  lower zone  soil moisture storage, LZSN
     Initial  lower zone  soil moisture storage, LZN
     Overland  flow length,  L
     Average overland  flow  slope,  SS
     Fraction  of  impervious  area,  A
     Maximum interception storage,  EPXM
     Index  to  actual monthly evaporation,  K3
     Potential evapotranspiration  correction  factor, PETMUL
     Mean infiltration rate, INFIL
     Interflow parameter,  INTER
     Interflow recession  rate,  IRC
     Fraction  of  groundwater recharge percolating to deep groundwater, K24L
     Groundwater  recession  rate,  KK24
     Fraction  of  watershed  area where groundwater is within reach of
           vegetation, K24EL
                                     162

-------
     Initial groundwater storage, SWG
     Parameter for variable recession rate of groundwater
          recharge, KV
     Initial interception storage, ICS
     Initial overland flow storage, OFS
     Initial interflow storage, IPS

Of all  the above parameters; LZSN, INFIL, UZSN, INTER, and K3 are the
major parameters subject to calibration.  Hydrologic calibration involves
the comparison of simulated and recorded runoff volumes on annual, monthly
and single storm event basis.

                              INPUT DATA

Rainfall

     Rainfall records were obtained from the U. S. Geological Survey
(U.S.G.S.) meteorological station no. 02053175 at Cutawhiskie Creek
tributary near Menola, N.C.  The recorded rainfall amounts were found to
fluctuate due to the expansion and contraction effect of the rainwater
collecting devices.  It was necessary to compare rainfall records with the
nearby Murfreesboro and Lewiston stations to ensure the probable occurence
of rainfall.  If no rain was noted from these two reference stations, the
recorded fluctuations of rainfall amounts were ignored for the Cutawhiskie
site.  The U.S.G.S. data provides rainfall records from October through
December of 1980; and from January, April, May, June and July of 1981.
The missing rainfall records for February, March, August and September of
1981 were estimated by constructing rainfall isohytes from recorded data
of the nearby stations.  Once the monthly rainfall for those missing months
were estimated, the hourly rainfall records from the Murfreesboro station
were adjusted proportionally and used as the input hourly rainfall for the
Cutawhiskie site.  The monthly rainfall data are given in Table 1.

Evapotranspiration

     Potential evapotranspiration is assumed to be equal  to lake
evaporation estimated from Weather Bureau Class A pan records.  The ARM
model allows the use of a single variable to adjust pan evaporation data.
The actual evapotranspiration is computed as a function of the potential
evapotranspiration and soil moisture conditions.  Chang (1968) reported
average ratios between potential evapotranspiration and pan evaporation.
These ratios for cotton, grass and corn are summarized in Table 2.
                                     163

-------
                             Table 1.

          Rainfall Data for Cutawhiskie Site, 1980-1981

           	Rainfall,  in	
Month

October
November
December
January
February
March
April
May
June
July
August
September
U.S.G.S.*   Cutawhiskie
  6.23
  3.42
  4.59
  4.64
  5.00
  4.88
  4.24
  4.84
             Murfreesboro
 4.78
   66
   13
   02
  02
  02
  13
0.94
               2.55**
   20**
   00
   55
   21
   81
 7.00**
 1.20**
36.11
 .65
 .42
 ,29
 .16
 .35
 .29
                               7.10
                               1.20
                              37.57
Lewiston

  3.96
   .13
   .71
   .19
                               2.25
    03
    58
    11
    36
    98
              6.86
    Values include effects of equipment expansion and contraction
    and may be high.
    Estimated from isohytes.
                            Table 2.

    Ratio of Potential Evapotranspiration to Pan Evaporation
                       (After Chang,  1968)
                                        Ratio
              Types of Crop

              Cotton
              Corn
              Grass
                          Mature     Ripening
                    0.2  0.85-1.00
                    0.27    0.90
                    0.84 (average)
                        0.1-0.4
                          0.4
                               164

-------
There are four meteorological stations recording pan evaporation data near
the study area, namely John H. Kerr Dam in Virginia; Chapel Hill, Aurora
and Hoffman Forest (formerly Maysville) in North Carolina.  The following
guidelines were adapted to estimate the evaporation records for the study
area.

1.  John H. Kerr Dam is the closest station; therefore, the evaporation
    data was taken directly from this station except the missing period
    for November and December of 1980 and January through March of 1981.

2.  During the missing period, evaporation data from Chapel Hill or
    Hoffman Forest was used as the substitution.  Preference is given
    to Chapel Hill since it is closer to the study area.

Table 3 summarizes the montly evaporation data estimated for the Cutawhiskie
site.

                                  Table 3.

                    Monthly Evaporation Data, 1980-1981
                                Monthly Evaporation,in
     Month

     October
     November
     December
     January
     February
     March
     April
     May
     June
     July
     August
     September
Cutawhiskie
(Estimated)
John H.
    Dam
                                        Kerr
      39
      89
      54(1)
     3.39
Chapel
 Hill

 3.27
 2.89
    1.86(2)
      .24(3)
      ,98
      ,16
      ,98
      .23
      .60
      .45
       ,16
       ,98
       .23
       ,60
       .45
  .86
  .69
  .58
  .51
Hoffman
 Forest

  3.37
  1.97
  3.98
  6.21
    4.80
   50.12
     4.80
 6.08
 4.48
  7.14
  7.22
  6.79
  4.76
  4.86
     1)  Based on  Hoffman  Forest  data  from  December 7-20 and
         average value  from  December 7-20 for  the remaining days,
     2)  Based on  Hoffman  Forest  data  of 1980.
     3)  Based on  the partial  available data from Hoffman Forest
         and  average value for the  remaining days.
                                     165

-------
Parameter Evaluation

The following summarizes calibrated values of the ARM model parameters for
hydrologic simulation (see previous section for parameter definitions).
     UZSN:  0.93 inch
UZS:  0.72 inch
LZSN:
L:
A:
PETMUL:
INTER:
K24L:
K3:
6.60 inch
800.0 ft.
0.00%
0.50
2.00
1.00
0.5,0.5,0
                                          LZS:
                                           SS:
                                         EPXM:
                                        INFIL:
                                          IRC:
      6.52 inch
      0.40%
      0.15 inch
      0.05 in/hr
      0.50
              5,0.5,0.5,0.6,0.67,0.7,0.75,0.8,0.8,0.8,0.65,0.5
   KK24,K24E1,SWG,KV,ICC,OFS,IPS:  0.00

     The nominal capacity of soil was estimated using the information
provided by the ARM model user's manual and was subsequently fine-tuned
in the calibration process.  The pan evaporation data was adjusted by a
factor of 0.5 in accordance with Chang's work (Chang, 1968) as described
in the previous section.  It was found during trial  runs that values of
greater than 0.5 would cause a large loss of water due to evapotranspiration,
making annual water balance undesirable.  The infiltration capacity, INFIL,
was estimated from soil characteristics and was fine-tuned during calibra-
tion.  The INTER parameter which controls runoff from interflow and peaking
of runoff hydrographs was initially estimated from guidelines given in the
user's manual and was subject to calibration.  The K3 parameter was based
on 50% watershed area with deep vegetation and was varied monthly to re-
flect seasonal  cropping periods.  Other parameters were estimated following
the guidelines  given in the user's manual.

                               RUNOFF DATA

     Calibration involves the comparison of simulated and recorded runoff
volumes.  The existence of a good data-base of runoff records is essential
for model  calibration.   Discharges from the Cutawhiskie site were recorded
by U.S.G.S.  The flow records were incomplete for five months and totally
missing for three months.  Therefore, it was necessary to generate flow
records using the technique of multi-regression analysis.  A similar
analysis was used by the Principal Investigator (Wu  and Ahlert, 1978) in a
study to predict total  runoff and sediment loads during selected runoff
events.

     In this project, runoff volume of each event was correlated with
rainfall volume of each event, number of dry days prior to rainfall, and
duration of rainfall.  A total of fifteen runoff hydrographs were obtained;
however, only eleven of them were used in the regression analysis.   This is
because the other four hydrographs were recorded during the period in which
rainfall data was estimated from nearby stations.  There is no assurance as
                                    166

-------
to the consistence of the timing of rainfall-runoff events.   The following
procedures were used for generating runoff data:

1.  Hydrographs were plotted from U.S.G.S. data.

2.  Area under the hydrograph was computed to derive the volume of direct
    runoff for each event.

3.  Number of dry days prior to rainfall and duration of rainfall  were
    determined from the continuous hourly rainfall  records.   A minimum
    of 3 hours was chosen to define an individual  rainfall  event.

4.  Regression analysis were performed using the Minitab Computer Program
    (Ryan et al., 1976) available at the computer center of the University
    of North Carolina at Charlotte.  Table 4 summarizes runoff data used
    for regression analysis.  The standard residuals of prediction were also
    included in Table 4.  Standard residuals normally range from -2 to
    +2 for good correlation.  Figs. 1-4 illustrate relationships among
    recorded runoff, rainfall, dry days, rainfall  duration and results of
    regression analysis.
                               Table 4.
                 Rainfall-Runoff Events for Regression Analysis
     Date
Recorded
 Runoff
 (R).in
Rainfall
(P), in
 Dry Days
(D), days
10/19/80
10/25/80
10/30/80
11/21/80
11/24/80
12/10/80
1/21/81
4/05/81
4/19/81
4/24/81
7/03/81
0.043
0.298
0.031
0.017
0.151
0.055
0.045
0.225
0.029
0.010
0.056
                            32
                            93
                          0.52
                          0.54
                          1.06
                          0.99
                          0.82
                          0.74
                          0.66
                          0.30
                          0.73
                         7.
                         5,
                         1,
                         2,
                         2.
               .21
               .04
               .33
               .83
               .08
             12.42
             13.83
              6.21
             10.63
              3.63
              0.79
 Rainfall
 Duration
(RD),  hrs

   13
   14
   17
    8
    6
   14
   16
    5
    4
    2
    3
   Standard
Residuals From
  Regression
                             -1
                              1
      .64
      .01
     0.56
    -0.45
     0,07
    -0.21
     0.39
     2.13
    -0.55
    -0.29
    -0.79
    A regression equation was developed:

               R = 0.1676(P) - 0.0031(D) - 0.0043(RD)

     This equation was then applied to each rainfall event to calculate the
per storm runoff volume for periods in which missing records were noted.
The-monthly and annual runoff volumes were computed as the summation of
individual storm runoff volume.
                                     167

-------
   0.3
   0.2
"S o.i
o
o
a;
c£
   0.0
               0.5         1.0          1.5

                      Recorded  Rainfall,  in


          Fig.l  Runoff versus  Rainfall for  Storm

                   Events Shown in Table  4
                                                       2.0
   0.3
                i  •
•-  0.2
o
c
3
o:

-o
O)
•o

o
o
OJ
o:
0.1
   0.0
                • »i
                            8

                      Dry Days, days
                                            12
              Fig. 2 Runoff versus Dry Days for Storm

                      Events Shown in Table 4
16
                         168

-------
   0.3
•^  0.2
o
c
CtL
O)

"g  0.1
o
o
Ol
   0.0
                   5           10            15          20
                        Rainfall Duration,  hrs

             Fig.3 Runoff  versus  Rainfall  Duration  for
                     Storm Events Shown  in Table  4
    0.0
                   0.1          0.2          0.3
                          Predicted  Runoff, in

             Fig.4  Recorded  versus Predicted  Runoff
                       from Regression Analysis
                          169

-------
                            RESULTS AND DISCUSSION
Statistical Analysis
     For regression analysis, a good size of sample population is needed
to obtain statistically reliable results.  A total  of eleven runoff events
were employed in the regression analysis; a majority of them occurred in
the Fall and Spring with only one runoff event recorded in July.   Rainfall-
runoff relationships usually vary seasonally; therefore, the regression
results of runoff volume for the Summer season are  unlikely to be accurate.
However, due to the fact that a good runoff data-base is unavailable, the
regression technique may serve as a guide for generating missing  records to
calibrate the ARM model.

Results of Calibration

     A comparison of runoff volumes between the field data and the ARM
model output is given in Table 5.  It can be seen that the annual and
monthly runoff volumes compare reasonably well.  Figure 5 presents the
model output of monthly rainfall, runoff, evapotranspiration, soil moisture
changes, and groundwater recharge.  The per storm comparisons are given in
Figures 6-11 and in Table 6;they also agree reasonably with recorded
hydrographs.  Thus, the ARM model has been successfully calibrated for
the Cutawhiskie site.  The calibration of water quality parameters could
not be undertaken since there is only one set of water quality data avail-
able for the study period.
                                Table 5.

            Predicted Runoff for the Catawhiskie Site, 1980-1981

ARM
Month Output
October 0.379
November 0.210
December 0.166
January 0.053
February 0.377
March 0.108
April 0.066
May 0.090
June 0.599
July 0.126
August 1.123
September 0.049
3.346
* Partial Records
** Missing Records
Runoff* in
U.S.G.S. Data
+ Regression
0.377
0.153
0.076
0.035
0.552
0.146
0.160
0.299
0.414
0.473
0.903
0.107
3.695



U.S.G.S.
Records
0.372
0.153
0.035*
0.034*
0.552
0.146*
0.160
**
**
0.050*
0.059*
**



                                    170

-------
CD
                Monthly Rainfall, inches
                (1980-1981)
     Oct    Nov    Dec    Jan   Feb   Mar   Apr   May   Jim    Jul    Aug   Sep
OJ
-S'
                Monthly  Runoff,  inches
                (1980-1981)
     Oct   Nov   Dec   Jan   Feb   Mar   Apr   May   Jim   Jul    Aug   Sep
OJ
•5*
                Monthly Actual  Evapotranspiration,  inches
                (1980-1981)
     Oct  Nov    Dec   Jan   Feb   Mar   Apr   May   Jun   Jul   Aug   Sep
           Fig.5  Summary of Simulation Results from ARM Model
                                     171

-------
OJ
o
c
                                  I     I
                Monthly Soil  Moisture Changes,  inches

                (1980-1981)
                           -I      L
 -2
 -4
          I	I
1	I	I	I
I	I	i
     Oct   Nov   Dec   Jan   Feb   Mar   Apr   May   Jun   Jul    Aug   Sep
6
oi 4
-C
(J
c
1 — 1
2
0
	 , 	 1 	 1 	 1 	 1 	 r — • 	 1 	 1 	 1 i '
Monthly Groundwater Recharge, inches
(1980-1981)
.

_

	 |~ ' 1 	 1
     Oct   Nov   Dec   Jan   Feb   Mar   Apr   May   Jun   Jul    Aug   Sep


            Fig.5   Summary of Simulation Results from ARM Model

                    (Cont'd)
                                    172

-------
CFS


 60


 40


 20


  0
                 	  Simulated by ARM
                 	  Recorded
   0
CFS

1.2


0.8


0.4


  0
 8       12       16       20      24      4       8      12
           Time, hrs
 Fig.6   Hydrograph  Comparison for Storm 10/25/80
                  T
        T
T
T
                         Simulated by ARM
                         Recorded
   0
CFS


1.2


0.8


0.4


  0
       12      16      20
            Time, hrs
               24
                   Fig.7  Hydrograph Comparison for Storm 10/30/80
                         Simulated  by ARM
                         Recorded
                              12
   0
8      12      16      20      24      4        8
            Time, hrs
 Fig.8  Hydrograph Comparison for Storm 11/21/80
                                      12
                                    173

-------
CFS
  3
                                   	  Simulated  by ARM
                                   	  Recorded
          16
20
16
       24      4       8
            Time,  hrs

Fig.9  Hydrograph  Comparison  for  Storm 11/24/80
20
24
                                         Simulated  by ARM
                                         Recorded
           8
12      16      20     24       4       8      12
             Time, hrs

 Fig.10  Hydrograph Comparison for Storm 12/10/80
               16
                                          	  Simulated by ARM
                                                Recorded
                  16      20      24      4       8      12      16
                               Time, hrs

                   Fig.11  Hydrograph Comparison for Storm 1/21/81


                                     174
                                                      20

-------
                                  Table 6.
     Date

   10/25/80
   10/30/80
   11/21/80
   11/24/80
   12/10/80
    1/21/81
                   Comparison of Single Runoff Event

                                  from

                    Recorded Data and ARM Model Output
                      Peak Time
Recorded Computi
6:30
13:30
13:00
19:20
15:30
5:00
12:00
10:30
18:20
15:00
                                        Peak Discharge. CFS
             22:00
                            19:00
Recorded
53.0
1.0
0.9
20.8
2.8
1.6
Computed
54.5
0.8
0.5
14.2
1.9
0.9
1.
2.
3.
4.
                           CONCLUSIONS

Use of continuous hydrologic simulation model  can be a useful  tool
for evaluating the effectiveness of watershed management practices
over  an extended period of time.  However, it requires the
existence of a good data-base for calibration and verification.

The regression technique can be used to generate missing records
for calibration; however, great care is needed in interpreting
the regression results.

A well planned sampling program is critical so that the collected
data can be utilized and served for its intended purpose of the
sampling and analysis program.
                                        stations in the Chowan
                                        determine the
    As there is a few evaporation measuring
    Basin,  a study is recommended to better
    evapotranspiration of the Chowan Basin.

                            ACKNOWLEDGEMENT

     This study was supported by the North Carolina  Uater  Resources
Res-earch Institute.  Appreciation is due Ms.  Beverly Young for
providing field data and other pertinent information,  and  Mr. Robert Davis
for preparing input data and performing computer work.
                                    175

-------
                              REFERENCES

     Chang, Jen-Hu, "Climate and Agriculture-An  Ecological  Survey",
          Aldine Publishing Company,  1968.

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

     Ryan Jr., T.  A.,  et al. "Minitab Student  Handbook",  Belmont,
          California,  1976.

     Wu,  Jy S. and Robert C. Ahlert,  "Assessment of Methods  for
          Computing Storm Runoff Loads",  Water Resources  Bulletin,
          Vol. 4,  No.  2, pp. 429-439, 1978.
The work described in this paper was not funded  by  the  U.S.  Environmental
Protection Agency.  The contents do not necessarily reflect  the  views of  the
Agency and no official endorsement should be inferred.


                                    176

-------
                        ESTIMATION OF MISSING VALUES
                         IN MONTHLY RAINFALL SERIES

                                     by

                           Efi Foufoula-Georgiou

ABSTRACT

     Infilling of missing values is often necessary prior to the practical
use of hydrological time series.  In this paper, three different types of
infilling methods are considered reflecting the following basic ideas:
     (1) the use of regional-statistical information in four simple techniques:
         - mean value method (MV),
         - reciprocal distance method (RD),
         - normal ratio method (NR),
         - modified weighted average method (MWA);
     (2) the use of a univariate stochastic (ARMA)  model which describes the
         time correlation of the series;
     (3) the use of a multivariate stochastic (ARMA) model which describes the
         time and space correlation of the series.

     An algorithm for the recursive estimation of the missing values by a
parallel updating of the univariate or multivariate ARMA model is proposed
and demonstrated.  All methods are illustrated in a case study using 55 years
of monthly rainfall data from four south Florida stations.

INTRODUCTION

     Many different kinds of statistical analyses may be performed on a given
data set, e.g., determination of elementary statistical parameters, auto-
and cross-correlation analysis, spectral analysis,  frequency analysis, fitting
time series models.  For routine statistics (e.g.,  calculation of mean,
variance and skewness) missing values are seldom a problem.   But for
techniques as common as autocorrelation and spectral analysis missing values
can cause difficulties.  In multivariate analysis missing values result in
"wasted information" when only the overlapping period of the time series is
used in the analysis, and in numerical inconsistencies (Valencia and Schaake,
1973; Fiering, 1968; Slack, 1973) when the incomplete series are used.  The
evaluation of the estimation methods analyzed has utilized monthly rainfall
records from the South Florida Water Management District (SFWMD), and has
been based upon:  a) the statistical comparison of the methods to each other
at a fixed level of percent of missing values, and b) the performance of each
individual method at different levels of percent of missing values.  Gaps
(missing values) have been artificially created in the complete record of the
interpolation station (station whose missing values are to be estimated) with
1 Graduate student, Department of Environmental Engineering Sciences,
University of Florida, Gainesville.

                                     177

-------
the following procedure:  First, the lengths of the_gaps have been generated
from a discrete exponential distribution with mean k months.  Then, for a
given percent of missing values, m, the mean interevent length (missing
values between two successive gaps), T~, has been calculated as T = k (100-m)/m
and the interevent lengths have been generated ^randomly from an exponential
distribution with mean T.  The values used for k and m are based on a
frequency analysis of missing values__in SFWMD monthly rainfall records
(Foufoula-Georgiou, 1982) and are:  k = 2.4 months and m = 2, 5, 10, 15 and
20%.  Overlapping and concurrent periods of 55 years of monthly rainfall data
of the four SFWMD stations shown in Fig. 1 have been used in the analysis.

TRADITIONAL ESTIMATION TECHNIQUES

Mathematical Representation

     In all the following equations y will be used for the interpolation
station and x. for index station j, j = 1, 2, ..., n.  An estimated value at
time t is y'.J

Mean Value Method (MV) —

The simplest method simply replaces the missing values with the sample mean:

                           y^ = y                                        (i)

This method results in a reduced variance and a spurious correlation
coefficient especially at a high percent of missing values.

Reciprocal Distance Method (RD)

     A missing value v  is estimated as:


                          y=?aX                                   (2)
The weighting coefficients a^ are calculated from:


                   _  	 , i , j \ ^  /  V   / 1 / J \r
                   a •
                                  1=1
where d. is the distance between index station j and the interpolation
station, and n is the number of index stations used for the estimation.  It
has been concluded (Shearman and Salter, 1975; Wei and McGuinness, 1975;
Dean and Snyder, 1977) that P=2 better approximates the isohyetal map drawn
by conventional methods.

Normal Ratio Method (NR) —

     A missing value y  is estimated as:
                                     178

-------
                                      T Aa^JJCdtr..... 1
Fig. 1.  The four south Florida rainfall stations used  in  the
         analysis.
            " A: 6038, Moore Haven Lock 1
             1: 6013, Avon Park
             2: 6093, Fort Myers WSO AP.
             3: 6042, Canal Point USDA
                              179

-------
where R and R.  are  the  normal  annual  or monthly  rainfall at  the  interpolation
and index station j  respectively.  This method is recommended  (Paulhus  and
Kohler, 1952) when  the  normal  rainfall at any of the  index stations  differs
from that of  the interpolation station by more than 10 percent.

Modified Weighted Average Method  (MWA) —

     The RD and NR  methods may be both written in the general  form of a
weighted average scheme:
                            y; - A xt                                   (5)

where A is_a  row vector (1 x n) and Xt is a column vector (n x 1).   To
preserve the  mean, ~y, and variance, sy, estimated from the available data, a
modified scheme may be  used:
                                                                         (6)
                        yj = B  Xt + (y - B X)
where
                            B  = A^                                     (7)
                                  V
and s , is obtained  from:
     Y    2              T     n    n
         s ,  =  A cov[X] A  =   E    I  a. a. c..                          (8)
          7                  i=l  j=l 1  J  1J

where c-jj is  the covariance between elements x^  and xj of the  rainfall
series ox the index stations i and j  (Kottegoda  and Elgy, 1977;  Foufoula-
Georgiou, 1982).

Comparison of the Methods

     Evaluation of  the  methods is based on the statistical comparison of the
estimated series (mixture of existing and estimated values)  to the incomplete
series (what  is really  available in practice) and to  the actual  series
(unknown in practice but known in this artificial case).

     The following  notation is introduced:

y  .  s ,  r   =  mean, standard  deviation and serial correlation coefficient of
 e    e    e
_               the  estimated series;
y.,  s.,  r.  =  same  as  above but for  the incomplete series, where
 1    x    x    i=l,2,3,4  and 5 for the five different percentages  of
_               missing  values;
ya,  s ,  r   =  parameters of the actual series;
      3.    3-
yj.            =  mean  of  the residuals  (estimated - actual values);

s^,  s        =  variance of the residuals over the whole series and over only
 r    r'e       the  estimated values respectively.
                                      180

-------
     The criteria used for the comparison of the methods are:
(1) the b±as jun the mean_as measured by
           ~    and
(2) the bias in the standard deviation as measured by
    (a) se/si and (b) s /s ;

(3) the bias in the serial correlation coefficient as measured by re - r ;
(4) _the bias of the estimation model as given by the mean of the residuals,
    yr (this is also a way to detect a consistent over- or under-estimation
    by a method) ;
(5) the accuracy as determined by the variance of the residuals s^ and s^  ;
(6) the significance of the biases in the mean, standard deviation and   'e
    correlation coefficient as determined by the appropriate test statistic
    for each.

     Regarding comparison of the means the following can be concluded from
Table 1:
(1) the bias in the mean in all cases is not significant at the 5%
    significance level as shown by the appropriate t-test;
(2) the bias in the mean of the incomplete series is relatively small but
    becomes larger the higher the percent of missing values;
(3) at high percents of missing values the NR method gives the less biased
    mean ;
(4) except for the RD method which consistently overestimates the mean (the
    bias being larger the higher the percent of missing values) , the other
    methods do not show a consistent over or underestimation.

     Regarding comparison of the variances the following can be concluded
from Table 2:
(1) although slight, the bias in the standard deviation is always significant,
    but this is so because the ratio of variances would have to equal 1.0
    exactly to satisfy the F-test (i.e., be unbiased) with as large a number
    of degrees of freedom as in this study;
(2) the MV method always gives a reduced variance as compared to the variance
    of the incomplete series and of the actual series, the bias being larger
    the higher the percent of missing values;
(3) the bias in the standard deviation of the incomplete series is small;
(4) there is no consistent over or under-estimation of the variance by any
    of the methods (except the MV method) ;
(5) the MWA method does not give less biased variance even at the higher
    percent of missing values tested, as compared to the RD and NR methods.

     Regarding comparison of the correlation coefficient the following can be
concluded from Table 3:
(1) the bias in the correlation coefficient is in all cases not significant
    at the 5% significance level as shown by the appropriate z-test;
(2) the MV method gives the largest bias in the correlation coefficients, the
    bias increasing the higher the percent of missing values;
(3) all methods (except the MWA method) consistently overestimate the serial
    correlation coefficient of the incomplete series but not the serial
    correlation of the actual series, and therefore this is not considered

                                      181

-------
Table 1.  Bias In the Mean.


2%
5%
10%
15%
20%
INC

0.
0.
0.
0.
0.
MV

0.009
-0.012
-0.010
-0.089
0.042
RD
 ya
2%
5%
10%
15%
20%
-0.010
-0.013
0.018
0.009
-0.044
-0.001
-0.025
0.008
-0.020
-0.002
-0.002
0.001
0.024
0.051
0.105
-0.012
-0.021
-0.006
0.009
-0.001
-0.013
-0.010
0.001
0.008
0.042
4.126




Table 2.  Bias in the Standard Deviation.


2%
5%
10%
15%
20%


2%
5%
10%
15%
20%
INC

1.
1.
1.
1.
1.


1.002
0.999
1.009
0.999
1.008
MV

0.995
0.983
0.972
0.957
0.944


0.997
0.983
0.981
0.956
0.952
RD
se/si
0.998
1.007
0.996
0.988
1.006
s IB
e a
1.000
1.006
1.004
0.988
1.014
NR

0.996
1.001
0.986
0.978
0.973


0.998
1.000
0.994
0.978
0.980
MWA

0.998
1.013
1.005
0.994
1.011


1.000
1.013
1.014
0.994
1.019

si
3.680
3.671
3.705
3.671
3.701
s
a
3.673




                                     182

-------
Table 3.  Bias in the serial correlation coefficient.
INC
MV
RD
NR
MWA
(rl,e - rl,a) rl,a
2%
5%
10%
15%
20%
0.005
0.006
0.013
0.033
0.042
0.001
0.003
0.014
0.006
0.004
0.002
0.001
0.011
0.013
0.011
-0.003 0.366
-0.002
0.010
-0.009
-0.012
Table 4.  Accuracy—Mean and Variance of the Residuals,
          N = total number of values = 660
          N  = number of missing values
INC


2%
5%
10%
15%
20%


2%
5%
10%
15%
20%

2%
5%
10%
15%
20%
MV


-0.043
-0.440
0.007
-0.175
0.037


5.037
8.610
7.892
7.620
5.224

0.084
0.406
0.720
1.112
1.016
RD
= (y
r J e
-0.061
0.034
0.156
0.338
0.502
5 = (yQ
r ,e e
2.874
3.656
4.239
4.630
4.891
sr = (ye -
0.048
0.172
0.387
0.675
0.951
NR
- y )/N
a o
-0.570
-0.380
-0.113
0.074
0.038
- y )2/(N -2)
a o
3.149
3.411
3.484
3.958
3.681
v )2/(N-2)
3.
0.053
0.161
0.318
0.577
0.716
MWA


-0.589
-0.176
-0.046
0.105
0.200


4.585
5.340
5.187
5.816
4.898

0.077
0.252
0.473
0.849
0.953

N
o
13
33
62
98
130













                                     183

-------
    a problem;
(4) the RD method seems to give the less biased correlation coefficient  even
    at the higher percentage of missing values .

     Regarding accuracy of the methods the following can be concluded  from
Table 4:
(1) no method seems to consistently over or underestimate  the missing  values
    at all percent levels, but at high percent levels the  missing values are
    overestimated by all methods;
(2) the NR method is the most accurate method especially at high percents of
    missing values (i.e., it gives the smallest mean and variance of the
    residuals) .

ESTIMATION BY A UNIVARIATE STOCHASTIC MODEL

Introduction

     The observed monthly rainfall series, y£, is normalized using the square
root transformation (Roesner and Yevjevich, 1966; Stidd, 1970;  Delleur and
Kavvas , 1978) and the periodicity is removed by subtracting the monthly
means and dividing by the standard deviations (Kavvas and  Delleur, 1975) .
The reduced series, zt, approximately normal and stationary is then modeled
by an ARMA(1,D model:

                        Zt = * Zt-l ~ 9 at-l + at                        (9)

where , 6 are the autoregressive and moving average parameters respectively,
and at is a sequence of independent random variables from  a normal
distribution with zero mean and unit variance (white noise) .

     For an ARMA(1,1)  model the minimum mean square error forecasts z'(£) of
Z  ,, where £ is the lead time are:
                              =  zt - 6 £t              , £=1           (10)

                              =  z^U-1)                , £=2, ..., k

as developed by Box and Jenkins (1976).

Proposed Estimation Algorithm

     The estimation of the missing values in the series is performed
recursively by the following procedure:
Step 1:  The incomplete series S  is filled-in with any initial estimates of
         the missing values giving the complete series, _£]_.
Step 2;  An ARMA(1,1)  model is fitted to the series _S, and the maximum
         likelihood estimates (MLE) of the parameters  and 6 are found.
Step 3;  New estimates of the missing values are calculated as forecasts .of
         the model M, = (
-------
missing values as well as in the parameters of the model.

     The above algorithm will be addressed as RAEMV-U (Recursive Algorithm
for the Estimation of Missing Values - Univariate model) and is schematically
shown in Fig. 2.  A FORTRAN program has been developed for the above algorithm
(Foufoula-Georgiou, 1982).  Input is the incomplete rainfall series and the
positions of the gaps.  Output is the final estimated complete series as well
as the final parameters of the fitted ARMA model.

Results of the Method

     Little influence of the method used to determine initial estimates of
missing values was found on the final values of parameters  and 6 and on
the final estimates of missing values computed by the recursive scheme.  All
methods that were tried yielded identical estimates of missing values and
model parameters after five iterations at 10 percent missing values ( = 0.5095,
6 = 0.4333) and eight iterations at 20 percent missing values (((> = 0.0776,
6 = -0.0293).  Moreover, by using zeroes as initial estimates the same
results were obtained, suggesting the latter as a convenient choice.

     The RAEMV-U method was assessed using the same statistical measures as
used for the four traditional techniques described previously.   Table 5 shows
the bias in the mean, standard deviation and serial correlation coefficient
for the final series  (at 10% and 20% missing values).  The bias in the mean
and correlation coefficient is not significant at the 5% significance level;
however, the bias in the standard deviation does not pass the stringent
F-test (requiring exact equality of standard deviations) and thus is
significant.

Table 5.  Bias in the Mean, Standard Deviation and Serial Correlation
          Coefficient-Univariate Model.

10%
20%
ye " ya
-0.021
-0.083
s Is
e a
0.983
0.951
T" — T"
l,e rl,
0.018
0.044
a


     The forward mean square error forecasting procedure that was used worked
satisfactorily In the sense that rapid convergence to a statistically
acceptable series occurred.  Damsleth (1980) introduced the optimal between-
forecasts as that linear combination of forecasts and backforecasts which
gives the minimum mean square error.  For the case of monthly rainfall data
the use of more sophisticated forecasts seems not to be justified.  The
parameters  and 6 of the fitted  AEMA(1,1)  model are very close to each
other and the value of § is small as compared to one, thus making the large
white noise variance the predominant term in the calculation of the mean
square forecast error (Box and Jenkins, 1976, p. 154).
                                     185

-------
So
Mo

s.
M,

S2
Mz

                                                        f MI -Mj

                                                        I §i •* Sj.
II1 l+l

  i+l
  Fig.  2.   Recursive Algorithm for the Estimation of Missing Values—
            Univariate model (RAEMV-U).  S^ denotes the series, and
            Mi  the model,  (,6)-j_,  at the ith iteration.
Fig. 3.  Recursive Algorithm for the Estimation of Missing  Values-
         Bivariate model  (RAEMV-B).  Sj_ denotes the series,  and
         KjL the model, (PjQ).^ at the ith iteration.
                               186

-------
ESTIMATION BY A MULTIVARIATE STOCHASTIC MODEL
Introduction

     When the concurrent rainfall  series of nearby stations are available,
their correlation with the series  of interest may be incorporated in the
model for an improved estimation of the missing values.  The lag-one
multivariate autoregressive model  (Matalas, 1967) is expressed as:
                          = P
                                                       (12)
where Zt and Z    are n-length vectors  of  the normalized and standardized
variables at time t and t-1, H is an n-length vector of random components
and n is the number of stations used.   The above model preserves the lag-zero
(MO) and lag-one (M^) correlation matrices when the coefficient matrices P
and Q are estimated by:
                              M, M
                                    -1
                                                       (13)
                      Q QT =
                                                       (14)
Equation (14) may be solved  for  Q  using  a  principal component analysis
(Fiering, 1964) or much  easier by  an  upper triangularization technique
(Young, 1968; Young and  Pisano,  1968).   Missing values in any of the records
may result in no solution at all or a solution that contains complex numbers
since Q Q  may not be a  positive semidefinite matrix as required for a real
solution to occur  (Valencia  and  Schaake,  1973;  Slack, 1973).

     The special case considered here is that of a bivariate AR(1) model
between the interpolation station  A and  the index station 2  (Fig. 1).  This
model is written as:
                        12
             Xt-1

             52,t-l
quO
                                                     n2,,
                                                                         (15)
Following  the Box-Jenkins  forecasting procedure,  the mean square error
forecasts  z'  (5,)  of  z1   ,Q   are:
            JL 9 U         -L j t »jlj
lll,t
                           P12Z2,t
                                                 , a = 2,
                                                                         (16)
where k  is  the number of values missing in each  gap.

Proposed Estimation Algorithm

     An  algorithm analogous to the one for the univariate  case  is  also
proposed for the bivariate case.  The procedure  is  exactly the  same,  except
that now the parameters of the model,  M = (P,  Q) , are matrices  calculated
from equations (13) and (14), and the forecasts  are calculated  from eqn .  (16)
                                      187

-------
     The algorithm will be addressed as RAEMV-B (B stands for Bivariate
model) and is shown schematically in Fig. 3.  A FORTRAN program is also
available (Foufoula-Georgiou, 1982).  Input data are: the incomplete series
of the interpolation station, the position of its gaps, and the complete
series of the index station.  Output results are: the final estimated
complete series of the interpolation station, the parameters P and Q of the
fitted bivariate model and the correlation matrices MQ and Mj_.

Results of the Method

     Again, the scheme converges rapidly and independently of the method used
to obtain initial estimates of missing values, thus suggesting their
convenient replacement by zeroes to start the algorithm.  Also, the conver-
gence of the bivariate scheme seems to be less sensitive to the percentage
of missing values as compared to the univariate one (three to four iterations
were needed in both the 10% and 20% missing values).

     Table 6 shows the bias in the mean, standard deviation and serial
correlation coefficient for the final series (at 10% and 20% missing values).
Again, the bias in the mean and correlation coefficient is not significant
at the 5% significance level, but the bias in the standard deviation is.

Table 6.  Bias in the Mean, Standard Deviation and Serial Correlation
          Coefficient—Bivariate Model.

                       y  - y             s /s             r-i    - r,
                        e'a             ea             l,e    l,a

    10%                -0.030             0.983              0.016
    20%                -0.049             0.959              0.050
CONCLUSIONS

     On the basis of the monthly rainfall data from the four south Florida
stations used in the analysis, the following conclusions can be drawn:

(1)  All the traditional estimation techniques give unbiased (overall and
     monthly) means and correlation coefficients at the 5% significance level
     even for as high as 20% missing values.
(2)  At high percentages of missing values (greater than 10%) the MV method
     gives the more biased (although not significantly so) correlation
     coefficients.
(3)  All methods give a slightly biased overall variance but unbiased monthly
     variance at the 5% significance level,  and the MV method gives the most
     biased variances for all percentages of missing values.
(4)  The NR method  gives the most and the MV the least accurate estimates,
     at almost all  levels of percent missing values.
(5)  The proposed recursive algorithm works  satisfactorily in both the
     univariate and bivariate case.  It converges rapidly and independently
     of the initial estimates and gives unbiased means and correlation
     coefficients at the 5% significance level.


                                      188

-------
(6)  The use of a bivariate model as compared to a univariate one did not
     improve the estimates except for a slight improvement at 20% missing
     values.  However, the use of a multivariate model based on three or four
     nearby stations is expected to give much better estimates.  The use of
     three adjacent stations is the main reason for the better performance of
     the NR method over the more sophisticated univariate and bivariate ARMA
     models which use only zero and one additional stations.

     If the purpose of estimation is to calculate the historical statistics
of the series (e.g., mean, standard deviation, and autocorrelations) the
selection of the method matters little, and the simplest one may be chosen.
However, if it is desired to fit an ARMA model to the incomplete series, to
be used, say, to construct forecasts, the estimation of the missing values
and the parameters of the model by the proposed recursive algorithm is
recommended.  In this case the equilibrium state (i.e., final series and
parameters of the model) achieved upon convergence is unique, depending only
on the existing information in the system (available data) and not on any
external information added to the system (by the replacement of the missing
values with some estimates derived by an arbitrary chosen method).

ACKNOWLEDGEMENTS

     I would like to acknowledge with gratitude Professor Wayne C. Huber
for the invaluable guidance and encouragement he provided along the course
of, this research.

     The study was supported in part by the South Florida Water Management
District.
                                     189

-------
REFERENCES

Box, G.E.P., and Jenkins, G.M., 1976, Time Series Analysis Forecasting and
     Control, Holden-Day, San Francisco, Revised ed.

Damsleth, E., 1980, "Interpolating missing values in a time series," Scand.
     J. Stat., 7:33-39.

Dean, J.D., and Snyder, W.M., 1977, "Temporally and areally distributed
     rainfall," J. of the Irrigation and Drainage Div..  ASCE, 103(IR2):221-229.

Delleur, J.W., and Kawas, M.L., 1978, "Stochastic models for monthly rainfall
     forecasting and synthetic generation," J.  Appl. Meteor., 17(10):1528-
     1536.

Fiering, M.B., 1964, "Multivariate technique for synthetic hydrology,"
     J. Hydraul. Div., ASCE, 90(HY5):43-60.

Fiering, M.B., 1968, "Schemes for handling inconsistent  matrices," Water
     Resour. Res., 4(2):291-297.

Foufoula-Georgiou, E., 1982, Estimation of Missing Values in Monthly Rainfall
     Series, Masters Thesis, University of Florida, Gainesville.

Kawas, M. , and Delleur,  J., 1975, "Removal of  Periodicities by differencing
     and monthly mean substraction," J. Hydrol., 26:335-353.

Kottegoda, N.T., and Elgy, J., 1977, "Infilling missing  flow data," ^Ln
     Modeling Hydrologic Processes, Ed. by Morel-Seytoux, H., Salas, J.D.,
     Sanders, T.G., and Smith, R.E., Water Resour. Res.  Publications,  Fort
     Collins, Colorado.

Matalas, N.C., 1967, "Mathematical assessment of synthetic hydrology,"
     Water Resour. Res.,  3(4):937-945.

Paulhus, J.L.H., and Kohler, M.A., 1952, "Interpolation  of missing precipita-
     tion records," Mon.  Weather Review, 80:129-133.

Roesner, L.A., and Yevjevich, V., 1966, "Mathematical models for time  series
     of monthly precipitation and monthly runoff," Hydrology paper No. 15,
     Colorado State University, Fort Collins, Colorado.

Slack, J.R., 1973, "I would if I could (self-denial by conditional models),"
     Water Resour. Res.,  9(1):247-249.

Shearman, R.J., and Salter, P.M., 1975, "An objective rainfall interpolation
     and mapping technique," Hydrological Sciences Bulletin, 20(3):353-363.

Stidd, C.K., 1970, "The nth root normal distribution of  precipitation,"
     Water Resour. Res.,  6(4):1095-1103.
                                     190

-------
Valencia, D.R., and Schaake, J.C., Jr., 1973, "Disaggregation processes in
     stochastic hydrology," Water Resour. Res., 9(3):580-585.

Wei, T.C., and McGuirtess, J.L., 1973, "Reciprocal distance squared method, a
     computer technique for estimating areal precipitation," ARS NC-8, U.S.
     Dept. of Agriculture, Washington, D.C.

Young, G.K., 1968, "Discussion of 'Mathematical assessment of synthetic
     hydrology' by N.G. Matalas," Water Resour. Res.,  4(3):681-682.

Young, G.K., and Pisano, W.C., 1968, "Operational hydrology using residuals,"
     J. Hydr. Div., ASCE, 94(HY4):909-923.
 The work described  in  this  paper  was  not  funded by the U.S. Environmental
 Protection Agency.  The  contents  do not necessarily reflect the views of the
 Agency and no official endorsement should be  inferred.

                                      191

-------
AREAL INTENSITY-DURATION-FREQUENCY  CURVES -
A POSSIBLE  SAY OF IMPROVING THE RAINFALL INPUT.


by Janusz Niemczynowicz *
 1.INTRODUCTION

Intensity-duration-frequency (i-d-f)  relationships, usually derived from
point rainfall measurements, have for a long time been used for synthesizing
so-called "design storms". Simple block rain can  easily be derived from
i-d-f curves for the desired duration and return period and then be used as
an input for simulation of runoff occurances.

Since i-d-f curves comprise the statistical properties of a long time series
of rainfall data in a comprehensive form, it is easy to believe that rain-
fall input derived from them has a good statistical justification.  As  lung
as the rational  method was used, no proof could be found that statistical
information taken from point i-d-f curves was  not always sufficient  for
design purposes. Later  on. when more  sophisticated  methods of runoff
prediction come into common use, it was realised that the single block rain
not  only gives the wrong picture of a  hyetograph, but also the wrong rain-
fall volume. (Arnell 1962)

Furthermore, it was soon realised  that the rainfall frequency given by
intensity-duration relationships not  correspond  to  the observed runoff
frequency. (Sieker 1978, Urbonas 1979. James 1981)

During recent years a number of design storms with different shapes  have
been developed. (Keifer at al 1957, Amorocho 1981, Arnell 1982) Some of them
tried to reproduce a real shape of observed hyetographs.  But these design
storms can  not reflect  the dynamics of the moving storms and  by no means
represent areal properties of the rainfall pattern.

The main reason that there is no simple linear relationship between the
frequencies of  rainfall and runoff is  probably the fact that the rainfall
frequency comes from point observations, while  runoff represents the areal
and  dynamical properties of the rainfall. In other words, a similar rain-
fall hyetograph observed by one gauge  can give a number of different runoff
occurances.  or vice versa.

Convective storms, which  are most significant for design purposes, are
rather limited in space. Design storms derived from point i-d-f curves have
no areal  dimensions, but  are nevertheless used for runoff simulations on
catchments  of different sizes. We can reduce this idea "ad  absurdum" by
trying to imagine one single design storm falling simultaneously on thou-
sands of square kilometers.
* Department of Water Resources Engineering,
 University of Lund, Sweden.
                                     192

-------
It is obvious that some kind of reduction factors, taking into consideration
the areal properties of a rainfall,  have to be used, especially while
modeling runoff from large catchments.

The coprehensivness and convinience of the statistical information in i-d-f
curves, make us to belive that design storms will  stay with us  for some
time.

One possible way of improving the rainfall input derived from i-d-f curves
is to develop AREAL INTENSITY-DURATION-FREQUENCY curves.  Another way is to
develop area-rainfall  depth  relationships  from which  factors reducing
rainfall from point to areal values can be  obtained  for different areas,
durations and return  periods. Those factors can then be used to reduce the
design storms from point to areal values.

These solutions are more practical  than complete, because neither areal
i-d-f curves nor areal reduction factors describe the dynamics of the moving
storms.

The third and most desirable solution is to  develop a  statistical model
simulating a rainfall series taking into account temporal, spatial and
dynamical variations of the rainfall pattern. (Bras et al 1976, Amorocho et
al 1977, Gupta et al 1979)

The need for reasonable rainfall input for runoff simulations in big cities
has caused a number of area-rainfall depth relationships to be developed in
different countries. (Abraham et al 1976. Bell 1976, Rodrigues-Iturbe et al
1974) One important problem  to face is the lack of  sufficiently long  time
series of observations on a dense network with good time synchronisation.

The raingauge network installed in Lund in 1978 covers approximately a 25
sqare kilometer area with 12 gauges. More than three years of registered
data  is  assumed to be sufficient to produce reasonably good statistics for
short term rainfalls. This paper describes point and areal  i-d-f curves
derived for the city of Lund. Presented areal relationships will give more
realistc design storms in comparison with design storms derived from point
relationships.

2. THE GAUGING SYSTEM AND DATA PROCESSING

Twelve automatic  tipping-bucket gauges were installed in Lund in 1978 to
cover an area of approximately 25 sq km. The depth resolution of the  gauges
is  0.035 mm per tipping, the time resolution of registration is one minute.
All gauges are connected via open telephone lines to the  receiving  station
in the laboratory of the department. Since all gauges are governed  by the
same clock,  the absolute time synchronisation is achieved. Figure 1. shows
the situation  of the raingauges in Lund. The gauging system and the data
collecting procedure have been described before. (Falk et al 1979,  Niemczy-
nowicz at al 1981). Collected data was processed according to the flow chart
shown in Figure 2.
                                      193

-------
        LUND
     AUTOMATIC RAINGAUGES
Figure 1. Situation of the raingauges in Lund.

The rainfall series from all twelve gauges were divided into rainfall events
with an arbitrarily chosen interval between the events set to 40 minutes.
Since the main goal of this study is to compare the point rainfall statis-
tics with the area! rainfall statistics for short term rainfalls, all events
with low intensity were taken out of the data base. The criterion chosen was
that all events with rainfall depth less than 0.35 mm observed during ten
succesive minutes were excluded from data base. If any of the gauges excee-
ded the criterion, data from all other gauges were accepted for  the  same

s
:
D
Dj
1IQ
*•

ENT
F1L
FWT.,,: | | „„..„ 1 |«5,
ELL
^
a
«
-*•
IL
me
CCtS
FILE
—
pU 1""
INTtMSITY
» DUfiATIO-*
TRESrtOLD
1, », 1
klk"i_l
Figure 2. Flow chart for rainfall data processing.

                                       194

-------
period of time. It was found that this procedure only eliminated rainfall
events with a very uniform  temporal and spatial distribution. All data
suspected to be wrong and all events with less than eight gauges operating
were taken away from the data base. This data base, originally consisting of
about 130 thousands lines was reduced to 20 thousand lines which made file
operations much easier. A total of 588 rainfall events were finaly included
into the data base.

Malfunction  of the gauges was observed or suspected about 15 2 of the time
on the average for the 12 gauges. Gauge No  1 was the worst with malfuncti-
ons during  38 Z of the time, gauges No 2.3 and 8 were functioning all the
time. Most of the malfunctions were caused by broken telephone lines.

3. POINT INTENSITY-DURATION-FREQUENCY RELATIONSHIPS

Point i-d-f curves were developed separately for all 12 gauges in Lund. The
one hundred maximum values of rainfall intensity were found for durations of
1,5,10,15,20.30, and 40 minutes.  Intensities for each duration were then
ranked and listed for each gauge. The return period for each intensity and
duration was calculated by dividing the total time of operation of the gauge
by the rank number. Descending maximum intensities were than plotted on a
linear scale against return periods for each duration and gauge.

Several investigators have found that the Log-Pearson typ III distribution
function fits well with the maximum rainfall intensity data. (Arnell  1981)
Since our main interest is the differences between point and area! rainfall,
no other distribution functions were  tested during  this study.  A
Log-Pearson distribution function was fitted to the observation points by
method of moments according to the computer program given by Kite. (Kite
1977)

Figure 3. shows  an example of the intensity-return period diagram with
observed and fitted distribution function for one of the gauges. Values from
the fitted  distribution  curves were then rearranged and the usual form  of
intensity-durationf requency curves were drawn as shown in Figure 4. The
described procedure was  followed for the 12 gauges and resulted in 12
complete i-d-f curves, each of them representing a point  value.  The  mean
point-value  i-d-f curve was finally calculated by averaging the values for
all durations and return periods.  Figure 5.  shows the mean point-value
intensityduration-f requency curves.
 4. SPATIAL VARIATIONS OF RAINFALL INTENSITY

 TABLE 1. gives the  range of differences in maximum intensities observed
 between 12 gauges in Lund.  The highest (Max), the lowest (Min),  the mean
 values (Mean), the standard daviations between 12 gauges (s) and the coeffi-
 cients of variation (CV = s/mean) are given for different durations and
 return periods.

 The differances between the highest and the lowest values and standard
 deviations decrease  quickly for shorter return  periods. Coefficient of

                                       195

-------
                                          £2.0
                                          i/i
                                                              3D    40
                                                            RAINFALL DURATION MIN.
FIGURE 3. Intensity-return period
diagram for gauge No 8.
    FIGURE 4 . Intensity-Duration-
Frequency curves for gauge No 8.
 variation seems to  be rather constant for all durations within the same
 return period, but decreases with shortening return period.  Obviously,
 extreme intensity values  are the most unevenly distributed in space. The
 highest values of intensity occur persistently in gauges No.  2,3,4,5 and 12
 which are situated  in the central part of the town and to the north-east
 which is the most prevaling wind direction. The lowest intensity values are
 typical for gauges  No. 9,1,11, and 6, situated outside of the town.  This
 effect can perhaps be explained by the influence of the city on precipitati-
 on.
 5. AREAL INTENSITY-DURATION-FREQUENCY RELATIONSHIPS

 In order to extrapolate rainfall data from point measurements to area!
 values, specific areas  were associated to all gauges.  The influence that
 different methods of extrapolating point values to area! means have on the
 accuracy of area! estimation was investigated  by Gottshalk and Jutman
 (1982). Results show that the magnitude of error is  to  a  small extent
 influenced  by the method of extrapolation. For simplicity of calculations,
 the method of Thiessen polygons was chosen for this study.

 Due to periodical malfunction, the gauges  represented slightly different
 periods  of observations.  In order to develop areal  relationships, mean
 rainfall values from a number of gauges had to be treated simultaneously.
 In order to avoid shortening the total length of the record by averaging the
 observation period for all gauges, a special routine for reproducing the
 missing values was developed.
                                      196

-------
            POINT VALUES
            1 GAUGE
            LUND
                         RETURN
                         PERIOD:
                                               AREAL VALUES
                                               12 GAUSES 25.2 KM2
           RETURN
           PERIOD:
                                                   RAINFALL DURATION MIN.
AREAL VALUES
5 GAUSES ICL5 KM2


LUND
                                                    RAINFALL DURATION MIN.
FIGURE 5. Point and areal itensity-duration-f requency curves
      for the city of Lund.

The average weighted values from the three nearest functioning gauges were
inserted in place of the missing values for each minute of  the data base.
After this procedure, a new point i-d-f relationships were developed for all
12 gauges. No significant changes  in results were observed after this
procedure.  All  further calculations were performed on the same data base
with reproduced missing values.

Areal i-d-f curves were developed by making calculations  (described  in
chapter 3) on mean weighted values from combinations of the gauges.  Each
combination consists of 12 groups of gauges. In each combination, the  same
gauge is represented the same number of times.
For example, a mean-value i-d-f curve representing an area of two gauges was
developed as follows:  mean value of rainfall intensity for each minute  in
the data base was calculated for the arbitrarily chusen group of two gauges.

Intensity-duration-frequency relationships were then developed for this
group. The next group of two gauges was chusen and new relationships were
developed. Calculations  proceeded until 12 groups of gauges were treated.
This resulted in  12 different, complete intensity-duration frequency relati-
onships representing  different areas of pairs  of  gauges. Finally, the
mean-value relationship was calculated representing the  mean area of 12
pairs of gauges.
                                        197

-------
TABLE 1. MAXIMUM RAINFALL INTENSITIES OBSERVED IN 12 GAUGES IN LUND DURING A
THREE YEAR PERIOD.  (The number of the gauge where the value was observed is
in parenthesis.)
MAXIMUM
INTENSITY
MM/MIN 1
D U

5
RATION

1 0
IN

1 5
M I N U

20
T E S

30


40
                         RETURN  PERIOD  3.0  YEARS
Max   3.19(7)  1.98(2)  1.36(3)  1.00(2)  0.84(3)  0.62(3)  0.55(4)
Min 1 .3
Mean 12
gauges
s
CV
1(9)

2 . 20
0.61
0. 27
1.08(11

1.61
0.51
0 . 24
) 0 . 86

1
0
0
(9)

. 20
. 43
. 26
0. 64 I

0 .
0.
0 .
!9 )

93
. 34
, 32
0.51

0
0
0
(9 )

. 76
. 26
. 37
0 . 40 (

0 .
0 .
0 .
9 )

57
1 8
4 1
0.37(9)

0 .47
0.15
0 . 37
The same procedure for calculating was carryed on for the groups consisting
of 3.4.5,6.8.10 and  12 gauges.  Intensity-duration-frequency curves for
point value, and the 10. and 20 sq.km values are shown in Figure 5.

For all intensity-duration-frequency  curves was fitted a mathematical
expression  of the form:
            a
         I =	* c
           T + b
where I - maximum area) rainfall intensity for duration T (mm/h)
    T - rainfall duration (min)
    a.b.c -  constants

The constants a.b. and c were calculated using the least squares criterion
and the  optimization routine  accordingly to Marquardts method. (Ericsson
1979).


6. AREAL VARIATIONS OF RAINFALL INTENSITY

By rearranging the values from point  and areal i-d-f relationships the
statistical areal reduction factors were developed for  different durations
and return periods. Factors for 3.0 and 1.0 years return periods are shown
in Figure 6. It is interesting to notice that  statisticaly derived areal
reduction factors not only depend on  duration, but also on the return
period, which was questioned before. (Bell 1976)
7. CONCLUSIONS

Areal intensity-duration-frequency relationships give different rainfall
values then point i-d-f relationships. Areal rainfall intensity values are
lower for all durations and return periods.

                                       198

-------
If point i-d-f curves are  used for deriving design storms, the error in
average rainfall intensity  and rainfall volume will  be introduced for
simulation of runoff from real catchments.

The magnitude of  error  depends on duration, return period and the size of
the catchment.

The most significant differences between point and areal rainfall values can
be found for short durations and long return periods.

Factors  reducing point  rainfall values to areal values are given in this
paper. The presented relationships will give a more realistic design storms
in comparison with such storms derived from point i-d-f curves.
                                20      25
                                   AREA KM2
             RAINFALL DURATION : 40 MIN.
                                         RETURN
                                         PERIOD:
                                20      25
                                    AREA  KM2
FIGURE 6. Statistical areal reduction factors.
                                        199

-------
REFERENCES
Abracham.C.,Lyons,T.,C..Schulze,K.,W.,(1976):  "Selection of a Design Storm
for Use  With Simulation  Models".  National  Symposium  on  Urban
Hydrology.Hydroulics and Sediment Control, Univ.of Kentucky, July 1976.

Amorocho,J..Wu,B..(1977): "Mathematical Models for the Simulation of Cyclo-
nic Storm Sequences and Precipitation Fields".  Journal of Hydrology, 32
1977.

Amorocho,J.,(1981): "Stocastic Modeling of Precipitation in Space and Time".
International  Symposium  on Rainfall-Runoff Modeling Mississippi  State
Univ.,May 1981.

Arnell,W.,(1982): "Rainfall Data for the Design of Sewer Pipe Systems",
Chalmers Institute of Technology, Report series A:8, Goteborg 1962.

Bell,F.,C.,(1976): "The Areal  Reduction  Factors in Rainfall Frequency
Estimation". Institute of Hydrology, Wallingford,  U.K. Report no 39 Decem-
ber 1976. Bras.R.,L.,Rodriguez-Iturbe.L,(1976): "Rainfall Network Design
for Runoff Prediction". Water Resources Research Vol.12 no 6, December 1976.

Ericsson.G.,( 1979): "Numerisk analys". Institutionen for informationsbehand-
ling. Sigma-tryck, Lund  1979.

Falk,J.,J6nsson.O.,Niemczynowicz.J.,(1979): "Measurements  of   Rainfall
Intensities in Lund". Lund  Institute of Technology, Department of Water
Resources Engineering, Report No 3023,1979.

Gottshalk,L.,Jutman,T.,(1982): "Calculation of Areal Means of Meteorologic
Variables  for  Watersheds"  Nordiske Hydrologiske Konferanse, Forde. Juni
1982.

GuptalV..K..Waymire,E.,C.,(1979): "A Stochastic Kinematic Study of Subsynop-
tic Space-Time Rainfall". Water Resources Research Vol. 15 No 3 June 1979.

James,W.,( 1981): "Kinematic Design  Storm Incorporating Spatial and Time
Aweraging". Second International Conference on Urban Storm Drainage", Univ.
of Illinois, Urbana,  June  1981.

Keifer,C.,U.,Chu,H.,H..( 1957):  "Synthetic Storm  Pattern  for  Drainage
Design". Journal of the Hydroulics Div. ASCE, Vol. 83 No Hy4, August 1957.

Kite,G.,W.,( 1977): "Frequency  and  Risk Analysis in Hydrology".  Water
Resources Publications,  Fort Collins, Colorado 1977.

Niemczynowicz,J.,Jonsson,0.,(1981):  "Extreme  Rainfall Events in Lund
1979-1980" Nordic Hydrology, 12. 1981.
                                      200

-------
Rodrigue2-Iturbe,I.,Mejia,J.,M.,(1974): "On the  Transformation of Point
Rainfall  to Areal  Rainfall". Water Resorces Research Vol.10 No 4 August
1974.

Sieker,F.,(1977): "Statistical  Simulation Model  Based  on  Analysis  of
Variance". 3-rd International Hydrological Symposium, Fort Collins, June
1977.

Sieker.F.,(1978): "Investigation of the Accuracy  of the Postulate "Total
Rainfall Frequency Equal Flood Peak Frequency"", International Conference on
Urban Storm Drainage Univ. of Southampton, April 1978.

Urbonas,8.,(1979): "Reliability of Design Storms in Modeling".  Internatio-
nal Symposium on Urban Storm Runoff. Univ. of Kentucky, July 1979.
The work  described in this  paper was not  funded  by the U.S. Environmental
Protection  Agency.  The contents do not necessarily reflect the  views of the
Agency  and  no official endorsement should  be  inferred.


                                      201

-------
HYDROLOGICAL REGIONALISATION :  A QUESTION OF PROBLEM AND SCALE

I. Simmers and E. Seyhan : Free University, Amsterdam

Extended Abstract
Quantifying catchment hydrological response and subsequent regionalisation of
resultant information have until now been largely dependent on techniques of:

(a) generalised water balance determinations
(b) regional statistical generation of descriptive variables or model
    parameters
(c) lumped catchment models
(d) complex process models which rely on detailed plot or experimental
    basin studies for input parameters.

The spectrum of available models does not reflect individual dissatisfaction
with the approach of others, but is a manifestation of two fundamental is-
sues, viz. the type of problem to be solved and the scale at which a solu-
tion is required. It is therefore to be expected that some models are ideal
for initial estimates of a water resource but are unusable in an operational
mode. Others present problems of translation to basins of varying size,
while many have computing requirements which are so great that their prac-
tical application is restricted to research areas where economic criteria
are less dominant.
The classification of an area into distinctive hydrological response units
from natural resource maps is seen as a first step to resolve the obvious
confusion surrounding data regionalisation. Unit identification may be
achieved using multivariate statistics and, dependent on the immediate
problem, can be at any scale.
Preliminary results from a continuing study in east Luxembourg are presented
to demonstrate the technique. Since the problem to be resolved involves pre-
diction of hydrological response consequent upon landuse change, the scale
is small and requires application of a physically based, dynamic contribut-
ing area approach. Data collection was initiated from a regionalised sam-
pling matrix based on topography, vegetation cover, soil and rock type.

The results from several study catchments are encouraging and adequately
define the spatial variability of hydrological response and the varying in-
fluences of physiographic parameters on mechanisms which govern streamflow
generation. Schematic models are presented for both hillslope processes and
total system operation within a larger area. Regionalisation and solution of
the present problem are thus deemed achievable by way of reconstituted re-
sponse (matrix) unit flow diagrams, requiring only verification sampling.

In areas where natural resource maps are available but quantitative data are
scarce, the described approach to regionalisation shows promise and will
certainly minimise instrumental and data collection requirements.

Authors' address
Prof.dr. I. Simmers; Dr.ir. E.  Seyhan,
Dept. Hydrogeology and Geographical Hydrology,
Institute of Earth Sciences, Free University,
P.O. Box 7161, 1007 MC Amsterdam, The Netherlands


                                     202

-------
          HYDROLOGICAL REGIONALISATION :  A QUESTION OF PROBLEM AND SCALE

              I. Simmers and E. Seyhan :  Free University,  Amsterdam
ABSTRACT

Parameter prediction for ungauged catchments and regionalisation of hydrolo-
gical data are bound by two fundamental issues - the 'problem' to be resolved
and the 'scale' at which a solution is required. The variety of current ap-
proaches to the question suggests that there is a need to develop a simple,
robust method for regionalising   hydrological information which can encom-
pass these issues. Classification of an area into distinctive hydrological
response units from natural resource maps is seen as the first step to a so-
lution; unit identification can be at any 'scale' dependent on the 'problem'.
Preliminary results from a continuing study in east Luxembourg are encour-
aging and are presented to illustrate the technique. The 'problem' required
application of a physically based, dynamic contributing area approach and
data collection was initiated from a regionalised sampling matrix technique.
The described approach to regionalisation is shown by this and other studies
to be robust and if combined with remote sensing should minimise instrumental
and data collection requirements.

INTRODUCTION

Quantifying catchment hydrological response and subsequent regionalisation of
resultant information have until now been largely dependent on techniques of:
(a) generalised water balance determinations
(b) regional statistical generation of descriptive variables or model param-
    eters
(c) lumped catchment models
(d) complex process models which rely on detailed plot or experimental basin
    studies for input parameters.
The spectrum of available response models does not reflect individual dis-
satisfaction with the approach of others, but is a manifestation of two fun-
damental issues, viz. the type of  'problem' to be solved and the  'scale1 at
which a solution is required.
For example, if a 'problem' requires generation of streamflow records from a
longer rainfall time series, a lumped model may prove adequate. However, if
it is necessary to predict the hydrological effects of localised landuse
change, the consequences of spatially variable inputs and outputs, or the
movement of pollutants and sediment through a catchment, then it is necessa-
ry to use a distributed, physically based model. Changes in catchment char-
acteristics are thus directly reflected by changes in the model parameters
(Seven and O'Connell, 1982) - such considerations cannot be satisfactorily
contemplated using spatially averaging models.
Other examples and valid applications of hydrological response models defined
by (a) -  (d) above are of course abundantly reported in the literature and
need not be elaborated on here.  However, given the heterogeneity of  'problem*
and  'scale' it is therefore to be  expected that some models are ideal for

                                     203

-------
initial estimates of a water resource but are unusable in an operational mode.
Others present problems of translation to basins of varying size, while many
have computing requirements which are so great that their practical applica-
tion is restricted to research areas where economic criteria are less domi-
nant .

Furthermore, the choice of model for even one particular 'problem' is never
simple - there is a marked 'contrast between the complexity of hydrological
reality and the important and often pressing practicality of making manage-
ment decisions based on limited knowledge of that reality'  (Beven and
O'Connell, op cit) . Model selection will be based on, inter alia, economic
constraints, hydrological considerations and data availability, with a further
essential prerequisite being to match management problem requirements with
the complexity of model used.

Parameter prediction from ungauged catchments and regionalisation of hydrol-
ogical data are also subject to the same constraints imposed by'problem1 and
'scale'. A number of methods in current use rely on statistical regression of
calibrated model parameters against catchment characteristics (eg. NERC,
1975), with obvious dangers. However, such an approach can resolve only a
limited range of 'problems' and omits the difficulties associated with data
translation at different scales due to variable storage effects.

It is evident, therefore, that there is a need to develop a simple yet robust
method for regionalising hydrological information which can acknowledge these
aspects. The classification of an area into distinctive hydrological response
units from natural resource maps is seen as a first step to resolve the ob-
vious confusion. Unit identification can be achieved by map interpretation
and multivariate statistics and, dependent on the immediate 'problem', can be
at any scale' (see, for example, Body, 1982; Krasovskaia, 1982 (a), (b) ;
Refsgaard and Hansen, 1982; Simmers, et al, 1982 •  Seyhan and Hope,1983).
Preliminary results from a continuing study in east Luxembourg are presented
to demonstrate the technique. Since the 'problem' to be resolved involves
prediction of hydrological response consequent upon landuse change, the
'scale' is small and requires the dynamic spatial variation in discharge
source areas to be taken explicitly into account (Beven, et al, 1980; Beven
and Hornberger, 1982; Freeze, 1980; Pilgrim and Bloomfield, 1980; O'Loughlin,
1981). As such, the physically based, dynamic contributing area approach
adopted by Beven and Kirkby (1979) is taken as a basis for the present study.
A full description of the research area, overall project objectives and anal-
yses of data available to date are presented elsewhere (Both and van der
Sommen, in press).

STUDY AREA
Figure 1 displays the area under investigation. Four catchments are involved
thus far; two (labelled C and D) lying within Triassic Keuper formations and
the others (A and B) in exposed or covered Lower Jurassic sandstone. Since
the Luxembourg sandstone is more important from a local water resource point
of view (v. Hoyer,1971), the research project has until now focussed prin-
cipally on the two catchments in this region - viz. Tollbach and Dosbach. The
present paper reflects this priority.
                                     204

-------
                       LJ map location

                      ^r^f drainage pattern

                       •  village

                       A  Dosbach

                       B  Tollbach

                       C  Deifebach

                       D  Briicherbach
K   -y Keuper formation
 mX  (Km) - Luxembourg
      sandstone (Li)
      boundary
  li   exposed Luxembourg
      sandstone
[:[••:•:[ Luxembourg
I : 3:  I sandstone covered
      by Arietenmarls
                                                             3     4 km
                                                              I	I
Figure 1  :  East Luxembourg research area
                                  205

-------
The Tollbach lies  entirely in exposed sandstone, has an area of 2.54  km2  and
an elevation which varies from 285 to 421 m above sea level. Conversely,
weathered and unweathered Arieten marls cover 66 % of the Dosbach,  and the
base of the underlying sandstone is exposed in the channel close  to the
lowest flow measurement point. Total catchment area is 1.98 km2 and eleva-
tion varies from 210  to 365 m.

The Arieten marl soils are poorly developed, though arable land areas have a
well ploughed upper zone of loamy clay (~ 30 cm). Beneath this are  1-3 m  of
heavy, compact, blue  to blue-grey clays,  which cap a confined aquifer in
beds of unweathered marls and karstified limestone. Weathering reduces the
Luxembourg sandstone  to a brown-yellow loamy sand, the clay content of which
reduces both down  slope and with increasing profile depth. The Luxembourg
sandstone aquifer  is  the most important in Luxembourg and is characterised by
short residence times (~ 1 year - v. Hoyer, 1971) and spring discharges which
reflect, inter alia,  the influences of tectonic control and degree  of marl
cover.
DATA NETWORK DESIGN

In order to determine the nature and dynamics of runoff production  within
each major geological unit data collection was initiated on a regionalised
sampling matrix according to topography,  vegetative cover and nature  of sur-
ficial deposits, similar to that proposed by Tricker  (1981) . The watershed
is thus regarded as  a system and is considered to consist of a series of
'homogeneous1 units,  each playing their particular role in runoff formation.
    slope position :
    vegetation :
    geology (soil type)
    slope position :
        1  - downs lope and
           riparian zone
        2 - midslope
        3 - upslope
    vegetation :
        b - beech forest
        p - pine/spruce forest
        g - grassland
        c - arable land (crop)

    geology :
        K - Keuper marls
        S - Luxembourg
           sandstone
        A - Arieten marls
Matrix unit
No. (fig. 2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Matrix
unit
(table 1)
Kb!
Sbl
Sg2
Agl
Kb 2
Sb2
Sp2
Ab2
Sg2
Sc2
Ag2
Ac 2
Ap2
Ag3
Ab3
Ap3
Sg3
Sp3
Spl
Sb3
Sc3
% Dosbach
area
0.6
2.0
0.4
1.0
4.5
5.5
2.0
1 .0
15.5
1 .6
29.7
2.5
4.4
8.7
13.7
4.5
1 .8
0.6
-
-
-
% Tollbach
area
_
1 .0
-
-
-
5.2
30.9
-
-
1.5
-
-
-
-
-
-
-
29.1
1 .0
18.5
12.8
  Table  1  : Basic  sampling matrix
Table 2  : Dosbach and Tollbach homo-
          geneous sampling  units
  Table  1 gives  the basic matrix used, while table 2 and figure 2  show the
  resultant  division of the Dosbach and Tollbach into relatively homogeneous
  sampling units.  Such a procedure obviates the necessity for excessive data
                                      206

-------
collection  and the derived results may  be objectively  incorporated into the
Beven  and Kirkby (1979)  model without  the possibility  of introduced  bias. It
is of  interest to note  that for the Dosbach eight units  allow representative
sampling  over 87 % of the total area,  and for the Tollbach this decreases to
five units  for 97 %.
     (i) SLOPE
          A:DOSBACH
                                      B: TOLLBACH
                                                        N
                                                            SLOPE

                                                            I u u u I upslope (O-3°)

                                                            I m mm I midslope (3-9°)

                                                            L. m, I spur/ridge . „..
                                                            LJ	ll positions  {'* >

                                                                 riparian zone (0-3°)
                                                            VEGETATION
                                                            I « « » I spruce/pine forest

                                                            | o o ° | beech forest

                                                            v%/%\ arable land (crop)

                                                                 grassland
                                                            GEOLOGY

                                                            I ~I I I Arieten marls

                                                            I:: :•:•:•:•:•:] Luxembourg
                                                            |x::::::::::::| sandstone

                                                                 Keuper marls
   Figure 2  : Slope (i),  vegetation (ii), geology  (iii) and sampling matrix
              units (iv)  for the  Dosbach (A) and Tollbach (B) catchments
 Data collection  was largely in accordance with the indicated division and in-
 volved measurement of normal  input-output variables,  soil moisture levels,
 surface-, spring-, well- and  soil-water chemistry, piezometric  levels and
 selected soil hydraulic properties such as infiltration rate and conductivity.
 Primary sampling sites for the Dosbach are shown  in figure 3.
                                       207

-------
    Legend:
• •••• catchment boundary
—J50-— contours (m)


——=r drainage pattern
 ffi  pond
 O*  spring
 O5  well
 ©R  pluviograph
 J  V-notch
 Wu |_  waterlevel recorder
 p  piezometer
 T  tensiometer
    Ssoil moisture
    chemistry sampler
A—A' transect
 X  throughfall troughs
                      N
Figure 3  : Dosbach field measurement  and  instrumentation network

RESULTS

Streamflow characteristics
Analyses  of  the basic discharge and precipitation records demonstrate marked
differences  in response between the Tollbach and Dosbach. Groundwater from a
number of distinct aquifers plays a dominant role in the total  runoff pro-
cess for  the Tollbach, with < 2 % of  measured precipitation  resulting in
quickflow from storms with maximum rainfall  intensities which vary from
2.8-20.3  mm/hr. For the Dosbach, with 66  % marl cover, mean  baseflow is ~ 45%
of that recorded from the Tollbach, while corresponding quickflow response
varies between 5 and 60 % of storm precipitation. The overall results, when
taken in  conjunction with hydro-chemical  investigations, confirm the appro-
priateness of present project objectives  and methods.

Modelling the runoff process
The  existence of various mechanisms  for streamflow generation  is now widely
accepted  and summaries of current knowledge  are presented by,  inter alia,
Dunne  (1978) and Freeze  (1980). All  recognise the importance of soil param-
eter heterogeneity in space, time and depth  governing these  mechanisms.
                                      208

-------
Unfortunately insufficient data have as yet been collected or processed to
allow rigorous testing of the complete Beven and Kirkby model, so emphasis is
placed here only on measurements relating to variations in soil moisture, in-
filtration rate, saturated hydraulic conductivity and water table depth for
a number of the proposed sampling units (figure 2).  Since surface processes
are relatively less important in the Tollbach principal effort to date has
been on data collection in the Dosbach, though sufficient information is now
available to propose schematic models for both hillslope processes and total
system operation in the Luxembourg sandstone region (figures 4 and 5).
Evidence to support these models is derived in part from measurement tran-
sects AA', BB1 and CC1 shown in figure 3. Piezometers have depths of 0.5 to
3.0 meters and tensiometer pairs are installed at 17 and 35 cm. Saturated
hydraulic conductivities for varying profile depths were determined using the
constant head method and infiltration parameters by sprinkling infiltrometer.
Tables 3 and 4 present a sample of results obtained.




28/9
29
30
1/10
2
3
4
5
6
7
8

D

(mm)
6
-
_
-
2
8
-
1
-
_
-
Matrix unit
A-A'


PI T2
8 s
16 s
29 s
34 s
35 s
26 s
19 s
28 s
31 s
38 s
38 s
15

P2
11
9
1?
10
15
1 1
9
10
1 1
17
20
1 1

P3 T4
1 1 s
8 s
9 s
9 s
10 s
8 s
6 s
5 s
6 s
12 s
16 s
1 1

P4
7
8
8
10
9
6
5
6
7
10
16
1 1

P5 T6
73 52.0
76 51.5
81 51.3
83 50.6
85 50.6
79
75 54.0
84 52.7
82 51.3
86 51.0
88 50.0
4
B-B'


P6
14
1 1
16
24
23
11
12
1 1
12
24
23
14

P7 T8
7 s
5 53.2
9 52.7
10 52.7
13 52.0
2 s
5 s
7 s
7 53.8
14 53.2
21 52.0
1 1

P8
6
2
9
6
6
1
4
4
5
11
14
1 1

P9 T10
44 50.0
33 49.8
48 49.4
54 49.5
60 49.4
24 49.6
36 50.0
49 49.5
43 50.0
58 49.4
66 49.4
4
C-C'


Pll
42
37
41
41
44
40
38
39
40
43
45
14

P12
12
8
13
14
17
8
7
8
7
12
16
9

P13
_
-
20
12
21
20
2
8
8
19
27
9

P14 Til
t -
39.0
1 38.5
o 38.7
o 38.5
2 38.5
A 38.7
38.7
38.6
38.2
38.7
9
 P = piezometer (water level depth in cms); T = tensiometer (values at 17 cm
 depth expressed as soil moisture content % by volume :  s (saturation) = 57%)

Table 3  : Dosbach piezometric head and soil moisture variations (28.9-8.10.81)
Matrix
unit No.
(fig. 2)
4
6
9

11

12

14
15
21
% Area
(table 1)
1.0
5.5
15.5 ]

29.7 j

25\
)
8.7
13.7
12.8
Slope
O
_
25
5
10
7
9
9
3
3
3
-
4
OFV
(mm/hr)
_
-
-~
49
-
320*
170
110
27
-
"~
SD
(mm)
_
7
33
12
14
6*
4
8
7
-
™"
i
o
(mm/hr)
_
83
92
> 105
58
51
53*
43
52
85
-
59
vertical
K
s
(m/d)
20
-
} 3

! 0.1

j
5
-
0.4
~
lateral
K
s
(m/d)
5
-
9

0.007



-
0.1
~
K
s
depth
(cm)
20
-
20

30



-
15
—
                                     209

-------
 *= top 20 cm removed; OFV = overland flow velocity parameter; ±Q = final
 constant infiltration rate; SD = maximum value of interception and de-
 pression store (see Seven and Kirkby, 1979)

 Table 4 : Selected saturated hydraulic conductivity (K ) and infiltration
                                                       s
           parameter estimates - (for table see previous page).

Although statistical verification of suggested interrelationships is not yet
possible a number of points emerge. Allowing for minor surface irregular-
ities water table depths reflect the dominant influences of topographic po-
sition and lithological differences, with up- and down-slope piezometers
showing a strong, rapid head response during and after rain (figure 6). This
is readily explained by the Kg values in table 4 despite saturated soil con-
ditions in the upslope areas. Response is least in the midslope zone, with
shallow water tables and lowest Kc.
                                 o
Infiltration parameters also display considerable spatial variability, with
iQ values which exceed rainfall intensities measured during the observation
period. Marked differences between i  and Kg, particularly in unit 11, sug-
gest that despite existing vegetation cover it is unwise to assume equiva-
lence of these parameters in areas subject to ploughing (cf. Anderson and
Burt, 1978). On the Arieten marls i  appears strongly related (but not equiv-
alent) to only subsurface Kg, since similar i  values are measured for vary-
ing slope angle and landuse (units 11 and 12). However, landuse has a marked
influence on OFV. Conversely, Kg is found to be not a limiting factor for
infiltration on weathered Luxembourg sandstone, slope angle is also unimpor-
tant,  and iQ seems to be principally controlled by landuse (units 6, 9 and
21) - highest under grass cover and least with bare soil or crop as expect-
ed. The same conclusion is also reached by Tricker (1981).

DISCUSSION
The above preliminary results, supporting hydrochemical data and field ob-
servations allow a number of generalisations to be made on system operation
within the Luxembourg sandstone.

At the hillslope scale on Arieten marls (figure 4) the most important runoff
producing process is saturated overland flow, initiated in the midslope area
(cf. Krasovskaia, 1982(a),  (b)).Channel precipitation will of course con-
tribute. Since the upper weathered zone is relatively permeable infiltration
will occur until storage is filled, this occurring first in the midslope
zone where water table depths and lateral Kg values are least (Freeze, 1980;
O'Loughlin,1981). Growth of the saturated area will be predominantly in a
downslope direction as a result of sheet/rill flow and percolation, with
subsurface flow a minimal component in this rapid response system. The small
peak discharges measured in weathered sandstone areas are attributed to
saturated overland flow from a narrow zone along the channel, 'Horton1 over-
land flow on access tracks, channel precipitation and groundwater response.

Figure 5 defines the hydrological system on a regional scale. Response to
precipitation is primarily  controlled by the degree of marl cover, since
Qma + QUJJ » Qgf + Qga and Qbg > Q^sd,,) • Groundwater is the dominant runoff
component from Luxembourg sandstone, and apart from marl cover reflects the
characteristics of thickness and storage capacity of the weathered zone,
structure and local lithological differences. %ws is small, as is

                                    210

-------
                                        downslope        midslope     upslope
                                      <          X             X     >
                                           ip           .        L
  o ^o  |o ^   I   o   I   o
                                      Legend:

                                           sandstone

                                           marls with limestone

                                           weathered marls

                                      •;•;•;•.•.•'.•'.•] saturated zone
                                           unsaturated zone
 |p  precipitation

E	  flow lines
 Figure 4  : Schematic model  of runoff producing processes on hillslopes  of
            the Dosbach catchment
CONCLUSIONS

Although results reported here represent only the initial investigatory  phase
of our ongoing programme  they are a source of encouragement . The spatial
variability of hydrological response is adequately defined, as are  the vary-
ing influences of physiographic parameters on mechanisms which govern stream-
flow generation.

To solve the immediate  'problem' , testing of the Seven and Kirkby  (1979) mod-
el structure under varying input and operation conditions continues and
amendments will be made as necessary. Verification of model components and
results is being based  on further field measurements of inter- and  intra-
matrix unit input variability in space and time, hydrochemical and  isotope
studies, and geophysical  surveys.

To address the more  general problem posed by the paper title and introduc-
tion, the stratified sampling matrix adopted as a basis for the present  over-
all project is found to be an appropriate technique for regionalised data
collection. Regionalisation for purposes of assessing the hydrological ef-
fects of landuse change,  both within the same area and elsewhere, is thus
deemed achievable by way  of reconstituted response (matrix) unit flow dia-

-------
    (T) Arieten  marls

    (2) Luxembourg sandstone
               rr~?
                   	.
                 |forest  |
                                T
E
      OSQ
                          t
      rforest |
  Osf <	 I T *]
                        Omf
             E
             t
                                (Su) Surface zone
                      ma
                   Qbu
                                (Swm) Weathered marls
                 (Sm)  Marls
                 (Sws)
              Weathered
               sandstone
                   Qbm
                         \r
                         Lm
                                                  Obls(m)
                                                  Qbs(m)
                                                       i
    Notation :
    E  evapotranspiration
    L  leakage
    P  precipitation
    Q  runoff
    S  storage
    T  throughfall
  a
  b
  f
  I
 m
(m)
  s
  u
                    agricultural land
                    base
                    forest area
                    local area
                    marl formations
                    formations with marl cover
                    sandstone formations
                    surface zone
                                w  weathered material
Figure 5 : Hydrological  system of the Luxembourg sandstone region
                              212

-------
              10
            E 20
            u
            SZ 30
              40
           "§ 50
            O
60


70


80
                                             10

                                             20

                                             30
                      (C)
                  (a)-upslope ; (b)-midslope ;  (c)-downslope
                    22   24  26  28   30   2   4    6    8   10
                             September- October 1981
  Figure  6  :  Mean Dosbach piezometric head variations,  sections  AA',  BB',
             CC'  (24.9-8.10.81)

grams, a general distributed, physically-based model, and verification sam-
pling. The sampling matrix approach  to data regionalisation is also shown to
be robust by the independent successful application of the technique to a
variety of 'problems' at  'scales' which range from five to 1000 km2 (see
Krasovskaia, 1982; Refsgaard and Hansen, 1982; Body, 1982).

In areas where natural resource maps are available but quantitative data are
scarce, this approach to regionalisation shows promise and if combined with
remote sensing will certainly minimise instrumental and data collection re-
quirements .

ACKNOWLEDGEMENTS

This project would not be possible without a team effort. The guidance and
assistance given by both staff and senior students from the Department of
Hydrogeology and Geographical Hydrology, Free University, are gratefully
acknowledged.

REFERENCES

Anderson, M.G. and Burt, T.P. 1978 : The role of topography in controlling
     throughflow generation. Earth Surface Processes, 3; 331-344.
Beven, K.J. and Hornberger, G.M. 1982 : Assessing the effect of spatial pat-
     tern of precipitation in modelling streamflow hydrographs.  Water
     Resources Bulletin, 18 (5); 823-829.
Beven, K.J. and Kirkby, M.J. 1979 : A physically based, variable contributing
     area model of basin hydrology. Hydrol. Sci. Bull., 24 (1);  43-69.
                                    213

-------
Seven, K.J. and O'Connell, P.E. 1982 :  On the role of physically-based dis-
     tributed modelling in Hydrology. Institute of Hydrology,  Wallingford,
     Report No. 81; 36 pp.
Seven, K.J. et al.  1980 : SHE  : towards a methodology for physically-based
     distributed forecasting in hydrology. IAHS Publication no. 129; 133-137.
Body,D.N. 1982 :  The application of results of catchment research in water
     resources planning and management. In, Proc.  IAHS Symp. on Hydrological
     Research Basins and their use in Water Resources Planning, Berne,
     Switzerland.
Both, M. and v.d. Sommen, J. in press : Catchment  studies for the application
     and regionalisation of physically based hydrologic models :  Luxembourg
     sandstone region. Dept. Hydrogeol. and Geog.  Hydrol.,  Inst.  of Earth
     Sci.,  Free University, Amsterdam,  Working Paper no. 1.
Dunne, T. 1978 :  Field studies of hillslope processes. In,  Hillslope hydrol-
     ogy, ed. M.J.  Kirkby, Wiley Interscience. New York; 227-293.
Freeze, R.A. 1980  : A stochastic-conceptual analysis of rainfall-runoff
     processes on a hillslope. Water Resources Research, 16(2); 391-408.
von Hoyer,  M. 1971 : Hydrogeologische und hydrochemische Untersuchungen im
     Luxemburger Sandstein. Publications du Service Geologlque du Luxembourg
     vol. 21; 61 pp.
Krasovskaia, I. 1982(a) : Hypothesis of runoff formation in small watersheds
     in Sweden. SMHI FoU-Notiser 19, Norrkoping; 15 pp.
Krasovskaia, I. 1982(b) : Rainfall-runoff relationships in small  research
     basins in Sweden. In, Proc. IAHS Symp. on Hydrological Research Basins
     and their use in Water Resources Planning, Berne, Switzerland.
Natural Environment   Research Council  (NERC). 1975 : Flood Studies Report,
     London, 5 vols.
O'Loughlin, E.M.  1981 : Saturation regions in catchments and their relation
     to soil and topographic properties. J. Hydrology 53; 229-246.
Pilgrim, D.H. and Bloomfield, P.H. I'A. 1980 : Problems in determining in-
     filtration and soil parameters of runoff models. IAHS Publication no.
     129; 271-277.
Refsgaard,  J.C. and Stang, 0. 1981 : An integrated groundwater/surface water
     hydrological model. Danish Committee for Hydrology, Report SUSA H13;
     122 pp.
Refsgaard,  J.C. and Hansen, E. 1982 : An integrated surface/subsurface
     catchment model. In, Proc. IAHS Symp. on Hydrological  Research Basins
     and their use in Water Resources Planning, Berne, Switzerland.
Seyhan, E.  and Hope, A.S. 1983 : On the estimation of runoff equations and
     the classification of catchments in South Africa by multivariate statis-
     tical  techniques. Water S.A., 8(6) (in press).
Simmers, I., v.d. Sommen, J.J. and Both, M. 1982 : Regionalisation of
     hydrological data : a dynamic modelling approach. In,  Proc.  IAHS Symp.
     on Hydrological Research Basins and their use in Water Resources
     Planning, Berne, Switzerland.
Tricker, A.S. 1981 : Spatial and temporal patterns of infiltration.
     J. Hydrology 49; 261-277.
 The work described  in  this  paper  was  not  funded by the U.S. Environmental
 Protection  Agency.   The  contents  do not necessarily reflect the views of the
 Agency and  no official endorsement should-be  inferred.

                                    214

-------
               SNOWMELT INDUCED URBAN RUNOFF IN  NORTHERN  SWEDEN

                                      by

                                Lars Bengtsson
                        Computational Hydraulics Group
                    McMaster University, Hamilton,  Ontario



                                   ABSTRACT

     Snownelt and runoff data  collected  during  7  years at different sites  in
the  Lule  region,  Sweden  is  summarized  and analyzed.    The  capacity  of  a
snowpack to  hold liquid water  and the effect  of refreezing during cold nights
are included in a degree-day approach.

     The maximum observed melt flux to the base of a snowpack during one  hour
is 4.1 mm/h.  The maximum  observed  daily melt is  40 mm,  but  the daily melt  of
two year  return  period  is only 20  mm.   The runoff from  small  study plots  is
(at least in  practical  considerations)  found to be distributed over about  12
hours.    Maximum observed   runoff  from  these  study  plots of  hard  packed
gravelled or grassed  surfaces  is about  2 mm/h.   In the late phase of a  snow-
melt period  the daily runoff is shown to closely correspond  to  the daily melt.

     The very maximum runoff values are, however,  observed during  or  seme  days
after  the  snowmelt  period.   Overland  flow  or  flow just beneath  the  ground
surface is shown to  take place.

     Finally some very rough "design criteria" for  snowmelt  induced  runoff are
g iv en.


                             REVIEW OF THE PROBLEM

     Due  to  the research  during  the  last two  decades  the  frequency  of
different rain intensities over different  time  intervals is  rather  well  known
for  populated  areas  in  Europe and  North  America.   From  this  knowledge and
using  physically sound  runoff  models  the runoff due  to rain from urban  areas
can be computed  quite  accurately.   Computations may be  made for  a  single
rainfall event or as continuous simulation.

     Only few  studies of snowmelt induced runoff have been made, even if the
snowmelt rate at a popint  has  been  frequently studied  and analyzed  in detail.
The different  energy  fluxes,  which determine the  snowmelt  rate is  discussed

                                     215

-------
by, for example, Gray and Male (1981), who recently presented the state of the
art of snow hydrology and snow management.

     Before  any meltwater  can  leave  a snowpack,  the  snow must  first  be
saturated  above  its  irreducible liquid  content.   The meltwater,  which
thereafter  reaches  the  base  of  the  snowpack,  infiltrates  into the  ground
unless the  infiltration  capacity of the ground  is  reduced -iue to an  asphalted
surface  or  as a consequence  of that  the ground  is frozen  and  saturated,  in
which  case  runoff  can  take place  along  the ground  in the  snowpack  or just
beneath the ground  surface.   The theory of  percolation  of meltwater  through a
snowpack  has  been  developed to an advanced level  through the work of  Colbeck
(1978).    Once vertical  drains develop in  the  snowpack,   it   is,   however,
difficult to quantify the percolation  process.

     There  are no  systematically collected  data  of snowmelt induced  runoff or
even snownelt data  at a  point  from  urban areas  reported in the literature.   A
group  within  the Swedish Board  for  Building  Research made a  state of  the art
report of  snow management in urban areas in  Sweden,  Bengtsson  et al .  (1980).
From a continuation  of that work Bengtsson (1981) found  that the snowmelt rate
was higher  in urban  areas than in rural areas.  He also  found that in  the late
phase  of a snowmelt period runoff took place  also  from  normally  permeable
surfaces.   Also  the problem of infiltration  in  frozen  soils is  summarized  in
the snow handbook by Gray and Male.

     To be  able  to  estimate the runoff  it  is of course essential to  know  in
what way  the  snow  has been  redistributed  by man.   Snow is piled  up within the
cities, but large  snow masses are also  transported  away from  the cities.  The
degree to which  snow is  transported away depends on if  there is space  to pile
up the snow within  the city, and to a large degree  on how much money that has
been reserved for snow management.

     The above cited works and 7 years of snowmelt observations at WREL (Water
Resources Engineering,  Lulea, Sweden)  show that  at  least for periods less than
6-12 hours  rain  intensities exceed  snowmelt  intensities.   Still,  in  northern
Sweden  problems  with  damming  in  urban  areas   are much  more   severe  during
snowmelt  than  during heavy rain  storms.   After  a long  period of  snowmelt
runoff can  take  place from  normally permeable surfaces.   Rainfall during this
period or just  after the snow is completely  gone  may also cause  runoff from
these surfaces.  When retention storages, perkolation storages or infiltration
surfaces are  included  in a storm water  system,  it  is  possible  that at  least
seme parts of the system should be designed  for  snownelt conditions.

     In this paper  the theory of meltwater  flux  to  the  base of  a  snowpack  is
treated.    Account  is taken  for  refreezing  of the  liquid  water  of the  upper
part of  the snowpack during  cold  nights.   Comparison  is made  with measure-
ments.    Thereafter  snowmelt  observations  from  the  Lulea region in northern
Sweden are  presented.   The degree-day method, but  including the capacity  of
the snow  to hold liquid  water,  is  used on 25  years  of data for 13 areas  in
Sweden for estimating return periods of snowmelt intensities.   Observed  runoff
from an all  areas in  the  Lulea  region  are  also presented and  compared with the
melt rate.  Rain on  snow and runoff caused  by rainfall just after the  snowmelt
period  are discussed separately.  Finally seme rough rules of thumb are given.

                                     216

-------
     All the data presented  in  this  paper  has  been  collected  by the staff of
WREL,  when  the  author  was head  of  the  department  at Water  Resources
Engineering, Lulea,  Sweden.
             SNOWMELT AND MELTWATER FLUX TO THE  BASE OF A SNOWPACK

     The snovmelt  rate  at the snow  surface can be determined  from an energy
budget analysis.   The  energy balance for  a  snow surface is  dominated  by the
radiation balance.    A large  heat  flux  of solar radiation  and  atmospheric
radiation  is  almost  completely  compensated  for by  the  reflected  solar
radiation and by the longwave radiation  from the snow.  In Fig.  1  it is shown
how the measured  accumulated  net  radiation followed  the  accumulated snovmelt
during a snovmelt period at a site near  Lulea.   Although the radiation balance
dominates the  energy balance of  a snow cover, the convective sensible heat
flux between  the atmosphere and   the  snow cover determines  to  a large extent
how intense  the  snowmelt  will be.   In  fact the  degree-day method  is quite
accurate for determining daily snowmelt.

     When consecutive snow surveys are made or  snow plates are  used, it is not
the snowmelt  at  the snow  surface  which  is measured,  but how much water that
leaves  the  snowpack  or  the  meltflux  at  the   base  of the  snowcover.   The
snowpack can  hold liquid  water.   The  percolation  rate  in dry  snow  is much
slower  than  the  percolation  rate   in  snow  above  the  irreducible  liquid
saturation,  since rather much meltwater  is  needed  to raise the liquid content
of the dry snow to and above the  irreducible value. Since the  liquid content,
which corresponds to a  certain melt  flux,  does  not deviate very much from the
irreducible liquid content, the propagation  rate of the  wetting  front in dry
snow can as shown by Bengtsson (1982a) approximately be described by

     C = i/Q                                                              (1 )

where C  =  propagation  rate of wetting  front,  i  = snovmelt  intensity,  6^  =
irreducible liquid content  (volume/volume) .  A  typical propagation rate is of
the order cm/h or even less.

     The percolation  rate  in  that part of  a snowpack, which  initially is at
its  irreducible  liquid  content,   depends  on to  which  extent vertical  flow
channels have developed  and  on  the snowmelt intensity.   Observations at WREL
show typical  percolation  rates in  the drains to be at least 0.1 but closer to
0.5 m/h.

     Since a snow cover must be saturated above  its irreducible liquid content
before  it  can release  any water   and since  intense snovmelt  causes a faster
percolation of meltwater  than  less intense  snowmelt,  it is  often first when
the surface melt  is  as  most intense  that the snow  cover starts to  release any
water.  The melt flux to the base   is  then close to the surface melt rate.  An
example from measurements at WREL is shown in Fig.  2.

     From Fig.  2  it is  seen  that  the  peak melt  flux  to  the  base  of the
snowpack exceeds  the  peak value  of the  surface melt rate.   This might be due
to  errors  in  computing the  energy balance, but since the depth  of the snow

                                     217

-------
cover decreases less  liquid  water  can be held in the snowpack.  Liquid water
previously held in the snow is released.  At constant melt rate  the meltwater
flux to the bottom of the snow cover  is

     vb = i (1  + 6i (p/ps»                                                 (2)

where v, = meltwater  flux  to the base of the snowpack,  p = density of water,
p  = density of snow (not including liquid water).  The meltwater flux to  the
bottom of the snow cover can exceed the surface melt  rate by as much as 155&.

     During a snowmelt  period  the  air temperature falls at least during  some
nights  below  freezing.   Then  the liquid water  in  the upper  part  of a  snow
cover  refreezes.    Bengtsson  (1982b)  has   shown  that  the  depth  to  which
refreezing takes place can  be approximately  found  from the  implicit equation
               a — 0. 5 T
                        a
                                                                           (3)
where  zf  =  refreezing  depth,  t  =  time  from  when  refreezing  starts  at the
surface, k = thermal diffusivity of the snow accounting also  for  seme convec-
tive effects due to intense winds,  T  = air  temperature  above  freezing, and
                                    3

     a = e_!l e                                                          (4)
         ps ci  *

where F = latent heat of fusion,  c.  =  heat capacity of ice.

     For small  negative values of  T  the  last term of the denunerator of eq .
(3) can be neglected and the refreezing depth  is determined as

     zf = Cf N°'5                                                         (5)

where Cf = refreezing coefficient,  which is

               -5                                                         (6)
and  N  =  number  of negative  degree- seconds  (or  more  practical  negative
degree-hours if Cf is given in appropriate  units) defined as

         t
     N = / -T  dt                                                         (7)
             3
         o

     Let us now return to the degree-day method, which gives a linear relation
between melt  rate  and  air  temperature.   From  a practical point of view it is
easiest  to  work  with an  equilibrium  temperature  of  0°C.    The  degree-day
equation is then simply

     m = C Ta                                                             (8)
            d

                                    218

-------
where M  =  melt rate, C  =  degree-day coefficient,  T  = air temperature  above
freezing.

     The degree-day method  is  usually used on a daily basis  or  a 12-h  basis
but sometimes also on a monthly basis.  When an equilibrium  temperature of 0°C
is used, the value of  the  degree-day coefficient depends on the  solar radia-
tion, wind conditions,  the atmospheric emissivity and  on  the properties of the
snow.  For forested  areas  with  a  high canopy density it can be  theoretically
shown that  the degree-day  coefficient  should  be between  1.5  -2 mm/ C day,
Bengtsson (1976).  This  theoretical  value  has been  confirmed by many measure-
ments, e.g. Kuusisto (1980).

     For  open  areas  degree-day coefficients  determined  from   eq.  (8) varies
from  region  to  region,  area to  area,  and  from  year to year.   Usually the
degree-day coefficient is  observed  to  increase  in the course  of  the  snowmelt
period.   This result is  partly  due  to  that it  is not  the  surface  melt but the
amount of  water  that leaves the snow cover,  which  is directly or  indirectly
measured.  But the melt  rate does  in fact increase.   This is  primarily due  to
the decrease of the albedo of the snow, so that more of the  solar  radiation  is
absorbed in the late phase of a snowmelt  period  than  in  the early phase.  The
albedo of fresh newly fallen snow is 0.8 or more.  When the  melt  period begins
it is usually  0.6 - 0.7  and drops to 0.4  during the  snownelt.   In city  areas
the  snow albedo  drops  rather  fast  to 0.2 - 0.3  as reported  by  Bengtsson
(1981).   The absorbed  solar  energy  is  almost  twice  as large in the late  as  in
the early phase of a melt  period.

     The degree-day  coefficient as  determined  from  eq.  (8)  and using  daily
average  air temperatures can well  vary within an interval of 2-10 mm/ C day.
In Sweden  the  degree-day  coefficient  is  usually higher  in northern than  in
southern Sweden,  since the melt  period  occurs later in  spring during a higher
solar  intensity  in  northern than  in  southern  Sweden.    Since   the  rate  of
refreezing of  meltwater  in  the  snowpack  is not linearly proportional to the
air temperature below freezing, the  high  value of the coefficient  is also due
to the  frequent  large diurnal  temperature variations during  melt periods  in
northern  Sweden.   For  open  areas the  degree-day coefficient  can for most  of
Sweden be  estimated  to 3 mm/ C-day, but  in northern  Sweden values around 4-6
mm/ C-day are more reasonable.

      It  is not  obvious  how a  day  having air  temperatures  above and  below
freezing should be treated  when  using  the degree-day  method on a daily basis.
The degree-day equation  gives  a linear relation between melt and  air  temper-
ature  above  freezing,  whereas eq.  (5)  shows  that  the  refreezing  depth  is
proportional to  the  square root of the temperature below freezing.   Assuming
that  the degree-day  method is strictly correct  for determining  snowmelt, the
average  daily  air -temperature  inserted  in the  degree-day  equation will not
give  the correct melt rate, if the air  temperature  is below  freezing during
some  part of the day.

      Look  at  the following  example.   The average daily temperature is  +2  C,
but during 12  h of daytime  the average temperature is 4°C  and during  12 h  of
nighttime  it is -2°C.  The observed  snovmelt is 6 mm.   When  eq. (8)  is  applied
on a daily basis, the degree-day coefficient is calculated to  be  3 mm/  C  day.


                                     219

-------
     Wlaen  refreezing  is taken  into  account  and  choosing  Cf  r  50  mm/(°C
day) '  , the refreezing depth is computed to 50 mm.   If the irreducible liquid
content is 0.04, about 2 mm of meltwater is needed for saturating the snowpack
to  its  irreducible  value.   The actual  surface melt  is therefore the observed
snowmelt 6  mm  plus  the  above 2  mm.   The  total  of  8  mm  melts  over a  12  h
period, when the temperature  is 4°C.   The  degree-day coefficient is therefore
2 mm/ C day.

     In the above example the degree-day coefficient was overestimated using  a
daily average value.   If there are  rather  large  negative temperatures during
the night, the  snowmelt can be rather much underestimated using the degree-day
method  on a daily basis  and  using  a degree-day coefficient derived in periods
of night temperatures above or only slightly below freezing.

     The re free zing-degree day method  suggested by  Bengtsson  (1982b)  was from
a  study in  Besbyn  Research Basin  found  to  be  accurate on  a  12 h  time basis,
but should be applicable also for a shorter time basis as long as the time for
the water  to  percolate through the  wetted  zone of  a snowpack is  short.   The
refreezing method can be summarized as follows:

1.   Determine  the   number  of negative degree-hours  over a short  freezing
     period, for example a 24 h period.

2.   Compute the refreezing depth using eq. (5) or eq. (3).

3.   Compute  the  amount of meltwater,  which is  refrozen, using an  a  priori
     value of the irreducible liquid content.

2-3. When eq.  (5)  is used  for  computing the refreezing depth, the  amount of
     refrozen meltwater, M», is

           Mf = 9i Cf N°'5                                                 (9)

4.   Determine  the  number  of positive degree-hours  over  a short  period  pro-
     ceeding the short period of freezing air temperatures.

5.   Compute the amount of surface melt using the  degree-day eq.  (8).

6.   Deduct the amount of refrozen meltwater, which must be remelted  and  held
     as liquid water in  the snowpack before  any water  can  leave  the  snowpack,
     from   the  computed  surface  melt  to  find  the  amount  of meltwater  that
     reaches the bottom of the snow cover.

     It was seen for Fig. 2 that the melt flux to the bottom  of  the  snowcover
was distributed over  about  12 hours.  Even if there is melt  flux  also during
nights, most of the meltwater reaches the ground during  daytime hours.  A typ-
ical runoff hydrograph observed  from a snow plate  at  WREL is shown  in  Fig.  3.
                      OBSERVATIONS OF SNOWMELT AT A POINT

     The data  presented  in  this chapter  regards  the meltwater  flux  to the

                                      220

-------
bottom of a  snow cover.   Maximun melt  intensities  for  different periods  are
given .

Snowmelt Intensities from Temperature and Precipitation Observations

     Using  25  years of temperature  and precipitation  measurements  the
frequency of different snownelt intensities has been computed for 13 areas in
Sweden.  The procedure for the computations was:

1.   Update measured snow precipitation by 30? (=  S)

2.   Determine  the  number  of  positive degree-days  during the  melt  period
     (= IT)

3.   Determine  the  degree-day coefficient for the  area  and for  the  year  as
     C r S/ET

4.   Multiply the temperature, T,  of the warmest day of the  melt  period  by  the
     degree-day coefficient.  The maximum melt rate  of that  year  is m  =  C.T.

5.   Distribute the snowmelt over 12 h so that m = cm/12 h.

6.   Update  this  value  by  10%  to  account  for the liquid  water,  which  can be
     held by the  snowpack.

     Extreme value analysis has been used in analyzing the data.   The  computed
snowmelt intensities are shown in Table 1.

     The table can be summarized:  Every second year the maximum  daily melt is
20 mm/12 h in southern Sweden and 30 in northern Sweden.

Snowmelt Intensities from Snow Surveys

     WREL is during  snowmelt  carrying  out  snow surveys  every second  day at  a
number of sites  around Lulea.   Since the energy balance  for the  snow  cover is
computed using measurements of  radiation  balance  and  computations of  sensible
and  latent  heat  fluxes  from profile measurements, eventual  observation  errors
can  be revealed  fairly well,  and  the snowmelt over  at least as  short time as
2-4  days can be determined quite accurately.

     The reduction  of the snow cover as  water equivalents  is  shown  for 1982
for  an open  field  and a  forested  area in Bensbyn  Research Basin  in Fig.  4  and
Fig. 5.  Tne  snow cover  of the forested area  is reduced  by  5 mm water  equiv-
alents per  day  during most of the melt period.   The melt rate of the snow in
the  open field is considerably faster.   All the snow disappears  in the  course
of  a week.   The melt  rate  is  somewhat more than  15  mm/day.    In  1982  the
snowmelt process  in  Lulea  was,  however, more  even than  usualy with no  single
day  of intense snowmelt.

     Observations of the snowmelt have been made  in  Bensbyn  and also at  For son
near  WREL's laboratories  since  1976.   The  observed maximum melt  rates  for
different years  are  summarized  in  Table 2.   The   shown values,  which are melt

                                     221

-------
                                   TABLE 1

      Recurrence interval  of snowmelt runoff with intensities in mm/12 h.
Region
Karesuando
Haparanda
Stensele
Harnosand
Ostersund
Falun
Stockholm
Karlstad
LinkOping
GOteborg
Visby
Vaxj'6
Lund
2 year
37
33
22
20
28
24
23
21
20
21
21
20
22
5 year
53
11
28
30
42
41
35
30
30
30
30
25
36
1 0 year
63
46
32
36
51
52
43
35
36
36
36
30
46
50 year
84
57
42
51
71
76
59
46
50
50
50
38
67
                                   TABLE 2

Maximum snowmelt runoff  (mm/day)  at  a  point having  a duration of at least 3
days as calculated  from snow surveys.
                         1976    1977    1978    1979    1980
1981     1982
Bensbyn open field
B
B
- small field with
bushes
- forested area
Porsttn open field
P
- forested area
13
17
8
25
10
24
22
9
24
10
40 15
38 13
21 8
20
21 11
15
15
10
12
9
30 15
20 13
15 11
— -
                                    222

-------
fluxes to the base of the  snow  cover,  can be assumed  to  occur during  at  least
three consecutive 12 h periods.

     As can  be  found from  Table  2 the maximum  observed  daily  melt  from  the
open  field  in  Bensbyn is 40 mm/day, but  the median value is only  15  mm/day.
For the forested area the maximum  observed  value is 21 mm/day and  the median
value 10 mm/day.

Snowmelt Intensities from Observations of Runoff from an  Asphalted Surface

     The runoff  from  a  25  m long  and  8 m wide asphalted study plot  having  a
slope of 2% has been measured with a resolution in time of 1-10  minutes during
the  snownelt periods  in Lulea  the years  1979, 1980,  1981  and 1982.    For
periods of  intense  snowmelt over  one hour  or  more  the runoff corresponds  to
the melt  flux  to  the base of the  snowpack,  and  also to  the surface melt when
the  reduction  of  the   total  liquid  holding  capacity  of  the  snowpack   is
subtracted.    In Table  3 maximum   observed  runoff over  1 h,  12 h,  24 h  and
maximum weekly  runoff is given.   The  four  highest  independent  1-h values  of
runoff for  the  different years  are given in  Table  4.  The corresponding 12 h
runoff values are given  in Table 5.  Consecutive days  have only  been included,
if the runoff ceased during the  nights in between daytime hours.

     The runoff over 12  h  is observed  to  be very close to the runoff over  24
h.  A very intense melt  flux is about 15 mm/12 h, but  in  1981  a  melt exceeding
20 mm/12 h was observed  at two occasions.  The highest observed  hourly runoff,
i.e.  approximately  the  melt flux  to  the base of the  snow  cover,   is  4.1  mm.
However, only at two occasions runoff of higher intensity than 3 mm/h has been
observed .

Observations in Downtown Lulea

     In 1980 snow surveys and runoff measurements were carried  out  and radia-
tion  measurements  were   taken in downtown  Lulea.   The degree-day coefficient
from  a small park surrounded by streets with heavy traffic was  found to be  8.5
mm/ C day and  for  a  larger  park 7  mm/  C-day.  The runoff was distributed over
the daytime  hours.  During the measurement period the maximum  12 h melt was 30
mm.   However,  for  an  early  warm period when no measurements  were made, the 12
h melt  from the two parks  was  using  the  degree-day approach estimated to  40
and 50 mm respectively.

Degree-Day Coefficient

     From the snow surveys  in Bensbyn  Research Basin  and  at  PorsOn  an average
degree-day coefficient  for each year and each site was determined.   The calcu-
lated coefficients "are  shown in Table 6.  The median value for  "open field" is
4.8 mm/°C day and for "forest"  1.8 mm/0C day.  The maximum degree-day coeffic-
ients were  found for 1979.   The "open  field"  coefficients were 8 and  12  mm/ C
day, but the "forest" coefficients did not exceed the median  value very much.

      The  refreezing-degree-day  method  was used on  Bensbyn-data for  the  snow-
melt  period of  1981 and  1982.  The degree-day coefficient was then  found  to be
2.8 mm/°C day  for  both  years for  the open field, and  1.4 and 1.5 for  the  two

                                     223

-------
                                    TABLE 3

Mafcimun observed  snowmelt  induced runoff (mm) from  a 25 m  long  asphalted  study
plot having a slope of 2%.

1 - h v al ue
12-h value
24-h value
weekly runoff
1979
3.3
16.9
17.7
84
1980
2.5
14.6
14.7
75
1981
4.1
21.3
23.2
90
1982
2.0
9.6
11.3
50
                                    TABLE 4

The  four  highest  observed  independent hourly runoff values due  to snownelt
(mm) for different years from a  25 m  long  asphalted  study plot having a slope
of 2% in Lulea and date when the runoff was observed.

1979
1980
1981
1982

3.
2.
4.
2.

3
5
1
0
1
(25/4)
(15/4)
(18/4)
(17/4)

2.
2.
3.
1.

9
1
0
8
2
(28/4)
(13/4)
(17/4)
(1/4)

2.7
1.9
2.7
1.4
3
(23/4)
(12/4)
(17/4)
(26/3)

2.
1.
2.
1.

6
9
1
4
4
(24/4)
(1 6/4 )
(19/4)
(21/4)
                                    224

-------
                                    TABLE  5

The four  highest  observed  independent 12-h runoff  values  (mm)  for different
years from a 25 m long asphalted study plot in Lulea and date when the runoff
was observed .

1979
1980
1981
1982

16.
14.
22.
9.

9
6
3
6
1
(23/4)
(15/4)
(11/4)
(1/4)

16.
12.
21.
8.

0
4
4
5
2

(28/4)
(1
(1
(1
4/4)
8/4)
7/4)

15.
10.
10.
8.

6
6
1
0
3
(25/4)
(13/4)
(14/4)
(26/3)
4
-
10.2 (7/4)
8.9 (15/4)
7.4 (16/4)
                                    TABLE  6

Degree-day coefficient (mm/°c day)  estimated  from  snow  surveys  for entire melt
periods using 0 C equilibrium temperature.
1975
Bensbyn open
field
B - an all field
with bushes
B - forested
area
Person open
field 4.0
P - forested
area 2. 1
1976
2.8
2.3
1.8
2.9
1.8
1977
5.9
3.7
2.2
6.5
2.2
1978
7
6
2.1
10
2.2
1979
8
6
2.2
12
2.5
1981
4.8
4.8
1.8
3.7
1.8
1 981 1 982
3.1 3.9
- -
1.5 1.6
-

                                      225

-------
years for the forested area.

     When  refreezing  is  accounted  for  the  degree-day coefficient  has more
physical significance than  when  it  is not accounted for.   The  coefficient  is
reduced and is more constant, which enables better  forecasts.

     The runoff data  for  1981 from the asphalted study plot at WREL  was used
by WesterstrSm (1982) for determining the degree-day coefficient using differ-
ent temperature index.  He  separated  "early"  and "late" melt.   When the daily
mean temperature was  used  as the  temperature  index  the degree-day coefficient
for early melt  was found  as  1.7 mm/  C day  and  for  late  melt (after  about  25
accumulated degree-days)  as 6.5.   When  only  the part of  the  day having  air
temperatures above freezing  was included  in the temperature index  (which then
corresponds to for example number of degree hours above freezing) ,  the degree-
day coefficients for  early and late melt were found to be  1.5  and 4.4  mm/°C-
day respectively.  The  "late"  melt  period using this temperature  index could
not be considered to  have  started  until after  about  45  accumulated  degree-days
(  C-days).

                            SNOWMELT INDUCED RUNOFF

Infiltration

     When a  snowpack  has  been saturated  above its  irredicuble  liquid  content
and meltwater reaches the base of the  snow  cover, the melt-water  can, depend-
ing on  the  conditions  of the soil,  either  infiltrate  into the  soil or  run
along  the  ground.   The soil  beneath a  melting  snowpack  is  usually  frozen.
Even if the  ground has  not been  asphalted, the  infiltration  capacity of  the
soil is reduced.   In  the early phase of  a  melt  period  all the meltwater  can
usually  infiltrate into  the  soil.    Then the  soil  moisture  is gradually
increased.  Also seme of  the meltwater refreezes, when it  penetrates  into  the
frozen soil.  The  infiltration capacity of the  soil  is reduced.  In   the late.
phase of a long melt  period  the  infiltration may cease completely.  Since  the
total melt amounts to  large quantities,  it may also be  that  the  ground water
level rises almost to the ground  surface.

     Tne theory  of imfiltration  is  not  treated in  this  paper.   It  is only
noted  that  there  is a  relation between  negative  soil  temperature  and   the
amount of  liquid  water  in  the  frozen  soil.   If the moisture  content of  the
soil is high when  the soil  freezes  in  the autumn,  and  if the ground is  cooled
down and very frozen  before it is covered by an  insulating  snow  cover, the  ice
content of the  soil  will  be large and the  infiltration capacity  reduced very
much.  Then only a small  amount of  meltwater  can be introduced  into  the soil
before runoff takes place  along the  surface.

     A measure of the infiltration capacity of the  frozen soil can  be  obtained
by measuring how much the  water equivalents of a snow cover is  reduced before
any runoff is observed  in  nearby  small perennial  streams.   Such  observations
have been made in  Bensbyn  Research  Basin.   The  observed reduction of  the snow
cover for consecutive years is shown in Table  7.  After the  runoff has started
the daily  runoff  corresponds  almost  to  the  daily  mel±  found  from  the snow
large open field,  which constitutes 30% of the  1.6  km   catchment, is meadow

                                      226

-------
                                    TABLE 7

Reduction of the water equivalents of the snow cover 4-  rain  precipitation (mm)
during  different  snowmelt  periods  up  to the  time   when  runoff  was first
ob serv ed .
                              1977     1978     1979      1980    1981     1982
Bensbyn - open field
Bensbyn - forested areas
70
70
90 90
50
100
100
10
10
grass.  There is a sandy layer of some 10-30 cm above a  silty  soil.

Runoff from Bensbyn

     The  Bensbyn  Research Basin  can  be divided  into  two parts.   The  upper
part, about 1.1 km , of the catchment is covered  with dense coniferous  forest.
The lower  part, about  0.5 km~,  is flat  and  is mainly a  large meadow.   Across
the meadow there  are two  small ditches.    Two  small  rivulets,  which are dry
most of the year,  meet at the edge of the  meadow.  The discharge  is measured  a
few hundred metres below the confluence.

     Maximum  observed  runoff  from  Bensbyn   for  different years  is  given  in
Table 8.   For daily  runoff only one of  consecutive days of high  discharge has
been included .

     From  Table 8  it  is  seen that the maximun daily runoff every year  during
the  period 1977-1982 has been at least 6 mm.  The  three  highest daily  runoff
values without contribution  from  rainfall are  18,  12 and  11  mm.   The  corres-
ponding highest 1-h  runoff values are 0.75, 0.5H  and 0.53 mm/h.  The largest
weekly runoff corresponds to 9 mm/day.

     The  very highest runoff  from  Bensbyn  has,  however, been  observed as  a
consequence of rainfall  just after  the  snowmelt  period  in  1982.    A  daily
runoff of  26 mm and an hourly peak value of 1.3 mm/h were  observed.

     The runoff given  in  Table 8  is referred  to  the area  of the  entire  catch-
ment.  However, the  snowmelt starts earlier from the open area  than from the
forest.    Many  years the snow in the open  field  has disappeared completely
before runoff commences  from  the  forested areas.   The snowmelt induced  runoff
fron  the   separate areas  is  therefore higher  than  the  average value for the
whole catchment as given in  Table 8.

     'IXiring 1978,   1980 and 1981 the maximun  discharge in  the  small stream was
not  observed  until the open   field  was  free  frcm  snow.  For these three years
it  was  possible to separate  the  runoff from  the  forest and  the  meadow.  The
maximun daily runoff for the two areas and the  three years is given in Table
9.

                                      227

-------
                                    TABLE 8

The  three  highest observed  runoff values (mm)  for different  time intervals
from Bensbyn Research Basin and date when the runoff was observed.

1-h
val ue


daily
r uno f f


weekly
runoff
1977
0.80X
(16/5)
0.75
(5/5)
0.58X
(13/4)
17.8
(5/5)
16.2X
(16/5)
11. 9X
(13/5)
8.1
(13-19/5)
1978
0.34
(16/5)
0.32
(17/5)
0.30
(18/5)
7.3
(17/5)
6.0
(19/5)
4.4
(15/5)
40
(15-21/5)
1979
-
-
-
12.2
(4/5)
9.7
(2/5)
9.7
(30/4)
63
(30/4-6/5)
1980
0.32
(23/4)
0.29
(1/5)
0.28
(4/5)
6.0
(1/5)
5.6
(4/5)
5. 1
(23/4)
36
(30/4-1/6)
1981
0.54
(13/5)
0.53
(12/5)
0.53
(14/5)
11. 1
(14/5)
10.3
(12/5)
8.0
(10/5)
68
(11-17/5)
1982
1.28Z
(29/5)
0.391
(9/5)
0.34
(25/4)
26. Oz
(29/5)
7.81
(9/5)
6.0
(25/4)
43
(4-10/5)
x)   during the period 12-14 May the rain precipitation was 12 mm.

z)   during the previous day the rain precipitation was 60 mm.

i)   during the previous day the rain precipitation was 9 mm.




                                    TABLE 9

Maximum snowmelt induced daily runoff (mm) fron the forested  area and the open
field of the Bensbyn Research Basin for the years 1978,  1980  and 1981.

Bensbyn - open field
Bensbyn - forested area
1978
10.4
10.8
1980
8.6
5.7
1981
19.9
23.0
                                      228

-------
     It  should  be   noted  that  the  design  runoff  values  for  stream  flow
generated from farming areas suggested in Swedish handbooks  is  13  mm/day.

Runoff from WREL Study Plots

     In a previous  chapter  runoff data from  an  asphalted  study plot at  WREL
was presented.   Now  this data  will be compared  with  the runoff from  two  study
plots with grassed and gravelled surfaces.  The  area of these  two  study plots
is for the  asphalted  surface 8 x 25  m  and the  slope is 2%.   The  three  plots
are next to  each other.  The grassed   surface  is  shaded by a house  towards the
evening,   so  the melt  rate  at the grassed  surface may in  late afternoon be
reduced compared to  the melt rate at  the asphalted  surface.

     The most intense observed runoff for different  years is given  in Table 10
and as runoff over 12  h  in  Table 11.   All of these runoff values  were  exclu-
sively due to snowmelt.

     The  maximum weekly runoff  varied  between  values corresponding to  7-14
mm/day for  the  asphalted  surface,  3-11 for the  gravelled surface  and 1-8 for
the grassed  surface.  In  1980 the maximum weekly runoff  corresponded  to 11
mm/day'  from the  asphalted  as  well  as  for  the gravelled   surface.   No  data
exists for  that  year  for  the grassed surface.but usually  the  runoff from the
surface  covered   with  grass  is  very  close  to  the  runoff from  the surface
covered with gravel.

     The  runoff  from  the  surface covered with gravel may be  as high as  from
the asphalted  surface  as  was  found in  1980,  when all the three daily maximum
runoff values  were equal.   All the meltwater  ran along  the ground.   Usually,
however,  the runoff  from the grassed  and  gravelled surfaces is  about half the
melt value  found as  the  runoff  from  the asphalted  surface.   When  plotting
accumulated  runoff versus accumulated degree-days Westerstrbm  (1982)  found the
("runoff"-)  degree-day coefficient for  grass and gravel to be about half of
the degree-day coefficient for snovmelt.

Runoff Observations in Downtown Lulea

     During the  snovmelt  of 1980 some  few measurements of  the runoff  from  a
6500 m   peak,  where  the  snow  had been  removed  from  the gravelled  paths,  were
made.  The maximum observed  runoff corresponded  to 2.3 mm/h, when  the area of
the gravelled  paths  was excluded.   This  peak value  is  in   agreement with the
data from the WREL study plots.

Rain During the Snowmelt Period

     Rain on  snow has  only  a minor effect on  the melting  process, but  it has
the effect  that  it  increases the degree of liquid  saturation of a  snow  cover.
If the  snow is wet  when the  rain  starts  to  fall,  all  the   rain precipitation
reaches the bottom of  the  snow cover.   In the late  phase of a  snownelt  period
all the rain  precipitation may run off as overland  flow.

     When a thick snow cover is completely gone, the infiltration  capacity of
the soil  is  still very much reduced.   Most of the rain  precipitation on  this


                                      229

-------
TABLE 10
The three yearly highest observed hourly runoff
sloping study plots, WREL. Date of observation

asphalt


gravel


grass


1979
3-3 (25/4)
2.9 (28/4)
2.7 (23/4)
2.4 (28/4)
1.6 (25/4)
0.9 (29/4)
-
-
-

1980
2.5 (15/4)
2. 1 (13/4)
1.9 (12/4)
2.2 (15/4)
2.1 (13/4)
2.0 (12/4)
-
-
-
TABLE 11
values (mm) from 25 long, 2%
is given within brackets.
1981
4.1 (18/4)
3.0 (11/4)
2.7 (17/4)
1.9 (18/4)
2.6 (20/4)
1.6 (11/4)
1.8 (18/4)
1.6 (20/4)
1.2 (11/4)

The three yearly highest observed 12-h runoff values (mm) from
sloping study plots, WREL, Lulea. Date of observation is
brackets .

asphalt


gravel


grass


1979
16.9 (23/4)
16.0 (28/4)
15.6 (25/4)
12.7 (28/4)
8.2 (25/4)
4.9 (29/4)
-
-
_
1981
14.6 (15/4)
12.4 (14/4)
10.6 (13/4)
14.2 (15/4)
12.4 (14/4)
10.6 (13/4)
-
-
_
1981
22.3 (11/4)
21.4 (18/4)
10.1 (14/4)
11.0 (18/4)
10.1 (11/4)
5.8 (19/4)
11.6 (18/4)
8.5 (11/4)
6.7 (20/4)
1982
2.0 (17/4)
1.8 (1/4)
1.4 (26/3)
1.2 (17/4)
0.8 (18/4)
0.7 (16/4)
0.2 (17/4)
0.1 (23/4)
0.1 (18/4)
25m long , 2%
given within
1982
9.6 (1/4)
8.5 (17/4)
8.0 (26/3)
4.4 (16/4)
37 (18/4)
3.4 (16/4)
1.0 (17/4)
0.7 (23/4)
0.5 (18/4)
  230

-------
wet soil, which still is frozen at sane depth below the ground,  may except for
evaporation losses run off as overland flow.

     The runoff from the Bensbyn Research Basin was observed  to  be  high during
the period 12-17 May  1977.   By  that  time the open field was snow  free.   From
energy balance computations based on measurements of radiation balance, wind-,
temperature-, and humidity profiles  and  controlled  by  snow  surveys,  the total
snowmelt  from  the  forested  areas  was for the  period  calculated  to be 47  mm
with reference to the area of the whole catchment.  The rain  precipitation was
12 mm.  The observed total runoff of 56 mm almost matched  the sun of the total
melt and the rain precipitation.

     The  peak  runoff  from  Bensbyn  during  the  snowmelt  period of 1982  was
observed on  9 May.   At  that  time the open field was free  from  snow.   The  melt
rate in the forest was during 8-10 May about 7-8  mm/day, which  with  the total
area as reference corresponds to 5 mm/day.  The rain precipitation  was  6 mm on
8 May  and 3 mm  the day  after.   The mean  runoff  of  9-10 May was almost  7
mm/day, which should be compared  to  the  mean melt rate and  precipitation, 9.5
mm/day.  Most of the rain ran off rather fast.

     The highest discharge ever in the small Bensbyn stream  was  recorded on 29
May 1982.  Tne open field had been snow  free for  one month, and the  last  snow
in the  forest  had  melted  two weeks  earlier.   During the night  between 28 and
29 May  there fell 61  mm of rain.  At 9  o'clock in the evening of 28  May the
low discharge of  21  I/sec in the  stream began to increase.   A peak value of
570 I/sec  was observed  at  4  o'clock  the  following afternoon.    By  noon 31 May
the discharge had decreased  to  80 I/sec.  The total runoff during  the two days
29-30 May only corresponded  to 38 mm.

     From  the above rather brief analysis of the combined  snowmelt-rain runoff
it can  be concluded that  rain  on snow  or  rain directly  on frozen  soils can
give rise  to fast runoff from normally permeable surfaces.
                          SUMMARY OF THE MEASUREMENTS

     The melt rate of  2 year  return  period  was from  precipitation and  temper-
ature  observations estimated  to 20 mm/12 h  for  open  areas  in  southern Sweden
and  to 30 in northern  Sweden.  The values for 10 year recurrence interval were
30 mm  for southern and 40 for northern Sweden.

     From snow  surveys over 7  years around  Lulea the median value  of  maximum
daily  melt  was  found  to be 15 mm  for open  fields and 10 mm  for dense  forest.
The  corresponding maximum observed values were 40 and 21 mm/day, respectively.

     The degree-day" coefficient  for  forested  areas is about  4  mm/°C-day.   The
degree-day  coefficient for  open  areas  may  vary between 2-10  mm/ C-day.   By
accounting  for  refreezing of  the liquid  water  held by the  snowpack and apply-
ing  a  refreezing-degree-day approach,  a  rather constant degree-day coeffic-
ient,  3-4 mm/ C-day, can be used, which enables more  accurate forecasts.

     The hourly maximun runoff observed from a small  impermeable study  plot at


                                      231

-------
WREL is  4.1  mm/h.   However, only twice a melt  flux  exceeding  3  mtn/h has been
observed.  Most of the melt flux reaches the bottom  of a  snow  cover  over a 12
h  period.    The  12 h  runoff  is very  close  to the  total daily runoff.   The
maximun observed runoff over 12 h is 22 mm.

     In the  late phase of a melt  period  the  infiltration capacity of normally
permeable  soils  may be  almost completely reduced.   Overland  flow can  take
place  also  from  surfaces  covered   with  grass  or  gravel.    In  the  Bensbyn
Research Basin runoff  from  an  open  field is  usually  not  observed  until  on the
average 70 mm of meltwater  has infiltrated  into the  soil, and  runoff from the
forest not until about 40 mm  has infiltrated.   Thereafter  the  runoff  corres-
ponds almost to the melt  rate.   The values given above vary from year  to year
and depend on the soil conditions by the time of the  first snow cover in early
winter .

     Tne maximun  observed  daily runoff from  the  open   field  in the  Bensbyn
Research Basin corresponds  to  23 mm and  that from  the  forested area  to 16 mm.
From an  analysis of  the  diurnal discharge fluctuations in  the  Bensbyn  stream
Bengtsson  (1982  c)  suggests that half of  the daily runoff is  contributed  to
groundwater .

     The measurements  of runoff from  the  WREL study plots show that  in  the
late  phase of  a  melt period  the  runoff  from packed grassed  and  gravelled
surfaces may be very close  to   the runoff from asphalted  surfaces.

     The maximun observed  runoff  from  the  study plots covered  with  grass and
gravel is  2.4 mm/h, which is  in agreement  with measurements from a  city park
in downtown Lulea.   The maximun observed daily runoff from the  study  plot with
gravel is  14 mm, but   almost all the meltwater  ran off  during  a  12  h  period.
The total weekly runoff for that period was 80  mm.

     When rain falls on snow or just after  a snownelt period almost all  of the
rain may run off  as  overland  flow also  on surfaces covered  with  grass  or
gravel .
                       SOME RULES OF THUMB FOR  ENGINEERS

     There are no design criteria for snownelt.   From  the experience gained  at
WREL  the  following  very  rough  rules  of  thumb are   suggested  for   Swedish
conditions.

     A design melt rate of 1 year return period  is for open areas in Sweden  20
mm/12  h  and  for a  shorter interval 3  mm/h.   For urban  areas  a value of  at
least  30 mm/12 h is  suggested.   For  forested  areas  the  value  should be  10
mm/12 h.   Rain during the snownelt should be added to  the melt.  The  melt  is
linearly distributed over 12 h.

     If  there  is much rain in the autumn  and  if the  soil temperature is low
before a  snow cover is  formed,  overland  flow can be  expected  to  take place
from all  surfaces  in  the late snowmelt period.   If the soil is  packed hard,
the runoff may correspond to the melt rate.

                                     232

-------
        2.00 r mm
        100 ,.
         SO -
                           10     15     ZO
Figure 1   Observed accumulated  snownelt  1976  (solid  line)  from  the  open field
          in  Bensbyn  Research  Basin and  the measured  radiation balance  as
          corresponding to melt (dashed  line).
                                      233

-------
                                        15
20     hours
Figure 2  Snownelt intensity (solid line) from measured at WREL,  24 April 1979
          and  observed  runoff from a snow plate (dashed line).
        7   13  19    1    7    13  19   I    7   13   19   1 hour*
Figure 3  Runoff 16-18 April 1981 frcm the short asphalted  surface  study plot
          at WREL Water Resources Engineering, Lulea.
                                     234

-------
            mm
            140
              20   "     30
                march
april
Figure  5  Water  equivalents of  the  snow cover in  the open  field  in  Bensbyn
         Research Basin 1982.
             mm
            14Q.-
             80
             20
                20   30   5
                mars       april
           30  5      15
               may
Figure 6  Water  equivalents of  the  snow  cover  in  the  forest  in  Bensbyn
          Research Basin 1982.
                                   235

-------
                                  REFERENCES

Bengtsson,  L.  (1976),  "Snownelt estimated  for energy budget  studies",  Nordic
     Hydrology 7,  PP. 3-18.

Bengtsson,  L. (1981), "Snownelt generated  runoff in  urban  areas",  Proc.  Second
     Int.  Conf. Urban Storm  Drainage,  pp.  444-451,  Urbana,  Illinois.

Bengtsson,  L.  (1982  a), "Percolation of meltwater  through a  snowpack" ,  Cold
     Regions  Science and Technology,  6,  pp.  73-81.

Bengtsson,  L. (1982 b) , "The importance of refreezing on  the  diurnal  snowmelt
     cycle",  Nordic Hydrology,  13, pp.  1-12.

Bengtsson,  L. (1982 c) , "Ground and meltwater  in the  snovmelt induced  runoff" ,
     Hydrological  Sciences Journal 27,  pp.  147-158.

Bengtsson,  L. ,  A.  Johnsson, P. A.  Malmquist,  E. S3rner,  J.   HSllgren  (1980),
     "Snow management  in  urban areas",  Swedish  Council  of Building  Research,
     Report R27,  1980.

Colbeck, S. C. (1978), "The physical aspects of  water  flow  through  snow",  Adv.
     Hydrosci. 11, pp.  165-206.

Gray, D. M.  and D. H. Male (1981), Handbook  of Snow,  Pergammon  Press, Toronto.

Kuusisto,   E.   (1980), "On  the   values  and  variability  of degree-day melting
     factor in Finland", Nordic Hydrology  11,  pp. 235-242.

Westerstrom,  G.  (1982),  "Estimating  snowcover  runoff  by the  degree-day
     approach", Nordic Hydrologic Conf., Forde,  Norway,  WREL,  Ser.  A,  No.  107,
     TULEA,  1982:  29.
The work described  in this paper was? not funded by  the U.S.  Environmental
Protection Agency.  The contents do not necessarily reflect  the  views of  the
Agency and no official endorsement should be inferred.

                                    236

-------
                 AN ADVANCEMENT IN HYDRAULIC MODELING
                       OF POROUS PAVEMENT FACILITIES
                                 Gary F. Goforth
                         Espey, Huston & Associates, Inc.
                                   Austin, Texas
ACKNOWLEDGEMENTS

            The  study  addressed in  this paper was part of a comprehensive investi-
gation of  operational  characteristics of  porous and  conventional paving  systems
conducted  by Espey, Huston  & Associates, Inc. and the City of Austin, Texas.  The
project was  executed in fulfillment  of  Grant No. R806338-01-Z with  the  City of
Austin, under the sponsorship of the  U.S. Environmental Protection Agency.  A  report
documenting the  entire project will be completed during the summer of 1983.

INTRODUCTION

            The  liabilities  of urban development to indigenous water resources are
generally accepted to be manifested  in stormwater runoff peaks and event volumes of
greater magnitude than in the predeveloped state, often occurring in association with
a degradation of receiving  water quality.  The increase in impervious areas such as
roofs,  streets, and  parking lots in urban  areas  reduces the  infiltration capacity of
urban watersheds and produces a corresponding increase in runoff rates and volumes.
Stormwater runoff from developed areas has been recognized as a source of contami-
nant loading to surface and  ground  water bodies.  Impervious areas generally have
limited assimilative properties and in some cases tend to yield contaminants that are
not amenable to  control and removal using standard treatment procedures(l). Storm-
water  flows transport contaminants which have accumulated on  the watershed  during
dry weather, however,  the  total mass transported is a function  of the  contaminant
accumulation rate, the number of antecedent dry days, the intensity of the rainfall,
the  velocity  and  volume  of surface flow  and other physical properties of  the
catchment.   Heavy metals,  exhaust products,  oils  and other hydrocarbons  from
automobiles  and  machinery, suspended solids from dust and  dirt  accumulation and
airborne  pollutants  washed out during precipitation events are typical contaminants
present in urban stormwater runoff(4).

            Stormwater management generally consists of collecting and transporting
overland runoff in a  conveyance system of storm sewers and possibly channels  which
are tributary to  a nearby stream or lake. Although local flooding problems may be
solved by this system, the shorter time of concentration and higher  peak flows  which
are  generated may  tend to  create  more severe flood  problems downstream.   The
increase  in flow  velocities in the improved channels creates a high  erosion and scour
potential, thus exacerbating the problem of pollutant transport to receiving bodies of
water.

                                       237

-------
POROUS PAVEMENTS

            A porous pavement  facility  is an innovative solution to the problem  of
stormwater drainage from parking and other low  density traffic areas in the urban
landscape.  This type of pavement uses the natural infiltration capacity of the soil to
absorb rainfall and local runoff after accumulation  in a porous base, consisting of sand
or large diameter open graded gravel. If  infiltration into the soil is undesirable or not
practical, lateral drainage to  a sump or channel may be provided.  Porous pavement
systems can  be designed  to  minimize  changes in  the runoff quantity  and quality
characteristics  of  a  watershed during and after development.  A cross-section of a
typical porous pavement facility is presented in Fig. 1.

            In regular applications  for highway and airport  runway  construction, a
commonly used porous pavement surface  has been referred  to as plant mix seal coat,
open graded mix,  gap  graded mix,  popcorn mix, or porous  friction  course(2).  This
material  consists  of an open  graded  asphalt  and  concrete mixture with a high
percentage by weight of aggregate larger  than a number four seive, laid to a thickness
of /i to 1 inch.  The resultant  paving has a coarse surface texture and a high void ratio
resulting in temporary storage of surface water  while maintaining the coefficient  of
friction between a vehicle tire and pavement at values comparable to the coefficient
under dry conditions.  The  open graded  asphalt mixture  underlain by a gravel base
course  with appreciable  storage  capacity is the most frequently used  type of porous
pavement.   The  whole  system  may be  isolated from  the natural  ground  by  an
impermeable  membrane  such as a  polyethylene liner, in which  case some type  of
artificial drainage would be needed; or, the porous pavement  system may be allowed  to
drain into the natural ground at all points of contact.  The  latter arrangement does not
preclude the use of artificial drains, as in  the case of highly impervious natural ground.
Where storage is  provided,  flow  control  devices can  be incorporated to regulate the
release rate,  e.g.,  to prevent  discharge during a predetermined period after a storm
event.

            By design, several inches of rainfall  and runoff can be stored  within a
porous pavement system  prior to discharge.  The pavement  can  be designed to retain
all of the runoff with no drainage from the site, to retain a sufficient volume of runoff
to reduce the after-development hydrologic conditions to  predevelopment conditions,
or to delay runoff from  the site, thus attenuating peak discharges  and reducing the
impact  of associated pollutant transport.  Any combination of these properties can be
incorporated into an overall project drainage design to satisfy municipal or watershed
drainage management criteria.

            Pollutant removal mechanisms in a porous pavement system have not been
fully documented.  The  relatively slow  hydrodynamics may  allow some  settling of
suspended  matter.   Adsorption  to  and  absorption in  the  base media may also be
realized.  Although transport  of  soluble constituents into the ground via  infiltration
removes them from the porous pavement  system, subsequent groundwater transport to
receiving waters may result.
SIMULATION OF STORMWATER HYDRAULICS

            Prediction of hydraulic characteristics is a valuable tool for assessing the
performance  of  stormwater  runoff  control  strategies.    Stormwater  hydraulic

                                        238

-------
                                                                         Collection Drain
L   length of pavement
W   width of pavement
D   depth of base  layer
Sb  slope of base  layer
ir   rainfall intensity
is   infiltration rate
dD  depth of water in base layer
Qs  surface discharge
Qb  collection drain discharge
   Fig.  1..   CROSS SECTION OF TYPICAL POROUS PAVEMENT  FACILITY
                                      239

-------
characteristics of porous and nonporous pavement study sites were evaluated using a
revised version of the computer model PORPAV, initially developed for incorporation
into the EPA  Storm  Water  Management  Model  (SWMM)(3).   PORPAV is  a  two-
dimensional dynamic water budget analysis of a pavement facility.  The utilization of
PORPAV allows a comprehensive analysis of flow and storage in porous and nonporous
pavement facilities, facilitating comparisons of the hydraulic response of alternative
pavement designs.  The computational scheme of PORPAV is described below.

            The rate of inflow to the pavement facility from rainfall and, if present,
any contributing area is compared to the permeability of the porous pavement for each
time interval.  For  porous systems  in general, the permeability is much  greater than
the  inflow rate and all of the water moves into  the pavement control volume.  For
nonporous pavements the  permeability  is  generally less  than the inflow rate  and
limited portion of  the inflow moves into the pavement. The excess is stored on the
surface of  the  pavement  for subsequent computation of  surface  runoff from  the
facility.

            The  inflow into the pavement  control volume is  added  to  the existing
storage and then compared to the permeability of the base.  If the base  permeability is
greater than the stored  volume in the pavement, all of  the flow is transferred into the
base  control  volume.   This is true  for most  porous  pavement  systems operating
according to design.  In those instances where  the base permeability is less than the
inflow volume, the inflow into the base is computed as the  vertical seepage into the
base, at a rate limited by the smaller  of  the pavement or  base permeabilities. The
lateral  outflow from the pavement top layer volume is assumed to be negligible as
compared to the vertical flux.  The difference between the flow into the pavement and
the transport to the base layer is stored in the pavement control volume.

            The  inflow to  the base  control volume is added to  the existing base
volume. The revised PORPAV includes a routing procedure to account for  the vertical
transport of water  within the layers, simulating the vertical movement of the wetting
front as it passes through the pavement system.  In essence,  volumes of water, defined
by the  permeability of  the layer and the length of the computational time step, are
routed through the depth of the layer. PORPAV allows the option of utilizing single or
multiple collection  drains for discharge from  the base layer.   An  expression  for
estimating the horizontal discharge  from the porous pavement base was developed to
reflect  the nonsteady flow regime in the porous media.  Darcy's Law was employed as
the governing flow equation(S),

                                 Qb  =  KbA dh/dx

            where      Q, is the average horizontal discharge;
                       K^ is the permeability of the base media;
                       A is the cross-sectional area of flow; and,
                       dh/dx is the  energy gradient.

The energy gradient was approximated by

                                   dh/dx = H/L

            where      H is the total elevation potential, equal to d^ + LS^j
                       djj is the depth of water in the base layer;

                                       240

-------
                       L is the normal length of the base layer; and,
                       S,  is the slope of the base layer.
The cross-sectional area of flow was approximated as

                                     A = wd/2

            where     w is the width of the layer.

This yields

                         Qb = &W V2 + {Kbw)
or on a unit area basis,

                                 % = Cldb + C2db2
where      c,    =     K,S, /2L; and,
            When  there  is  no  impermeable seal  present  to  restrict  flow,  some
horizontal discharge  will occur to the adjacent soil.  However, this horizontal flux is
generally negligible when compared to the vertical component leaving the layer via
infiltration because  of  the much smaller  cross-sectional area  of  flow.  Also,  the
moisture  content of the surrounding soil  increases  during the storm event, thereby
reducing the hydraulic energy gradient between  the  porous media and the soil.   In  a
narrow, high-wall trench without an impermeable seal or a drain pipe, the horizontal
flux to the soil during the initial period of the  storm, before  the  hydraulic  gradient
between the base and the soil diminishes,  may be of the  same order of magnitude as
the infiltration flux.  By neglecting the horizontal flux when there is no impervious
seal present  this  assumption  represents  a  conservative case  with regard to  the
effective storage of the porous pavement system.

            PORPAV incorporates Horton's equation to describe the variable rate of
infiltration during and subsequent to a precipitation event. This was expressed as

                                       ,.     . \  -kt
                               ls = lt + (lo -  af} e

       where     i  is the infiltration rate  at  time t;
                  i  is the infiltration capacity (minimum rate) of the soil;
                  i  is the initial  infiltration rate;
                  k is the  first-order decay coefficient; and,
                  t is the elapsed time.

            The initial infiltration rate is  dependent on the initial  moisture condition
of the soil.  The difference  between  the  initial and minimum infiltration rates can
result in significantly greater vertical  transport calculated during the storm event.  If
a constant infiltration  rate  is preferred,  the initial rate may  be replaced by  the
infiltration capacity (minimum  rate)  of  the  soil.   If the bottom is sealed with an
impermeable membrane, no flow  is discharged to the natural ground.  The difference
                                        241

-------
between the inflow to the base layer and the outflows (vertical and lateral) from the
base is stored in the base control volume.

            All stored  volumes within each  layer are  compared to maximum void
volumes.  If the storage volume in the base is exceeded, the excess is stored in the
pavement; if the storage volume in the pavement is exceeded, the excess  is addded to
the surface storage on the pavement, if any exists. Surface runoff is then computed as
broad channel flow from the pavement using Manning's equation.

            Provisions were added to PORPAV to calculate the theoretical detention
time provided by a pavement facility.  This  duration was  calculated as the elapsed
time between the center of mass of  the inflow hydrograph and the center of mass of
the discharge  hydrograph.   Additionally, average and cumulative inflow, peak and
cumulative discharge and other discharge hydrograph characteristics are compiled for
each simulation.
DESCRIPTION OF THE MONITORING SURVEYS

            An extensive monitoring program was initiated to document hydraulic
characteristics of several pavement systems.  A monitoring network of three parking
lots was selected,  representing a variety of pavement surfaces.  The following text
discusses the physical characteristics and sampling procedures for each of the study
lots.

Porous Asphalt Lot

            A plan drawing of the porous asphalt  study lot is presented in Fig. 2. The
porous asphalt lot  consisted of three layers  of stone  and asphalt  constructed on  an
impervious limestone  bedrock base.  The lowest  layer was made up of  a stone  base
course with rocks ranging from 1.5 to 2.5 in (3.0-6.5 cm) in diameter. This base ranged
in depth from approximately 4 in (10 cm) on  the upslope end to 42 in (107 cm) on the
low end and provided a void space of approximately 40 percent of its volume for water
retention.  The second layer averaged 2 in (5  cm)  in depth and consisted of a stone top
course (filter course) with material ranging from 0.4 to 0.6 in (1.0-1.5 cm) in diameter.
This intermediate layer  was selected to provide a uniform surface for the application
of the porous asphalt.  The final layer consisted  of 2.5-in (6.4 cm) of porous asphalt
mix with 5.5 to 6.0  percent asphalt content.

            The original design called for runoff  to be monitored within a collection
basin located at the downslope corner of the lot.  A 6-mil polyethylene  impermeable
liner was installed along the above-grade downslope width and side to prevent seepage.
However, this  seal  leaked and  the  base discharge did not flow into  the  monitoring
barrel.  Small trenches and berms were constructed along the periphery of the lot to
ensure  all runoff  was  captured.   However, these  trenches  were  not  lined  and,
consequently, some percolation and erosion inevitably occurred.  A 90-degree V-notch
weir was installed below the lot in order to measure the discharge rates. Visual stage
readings at this weir by the field crew were used as a basis for the runoff calculations.
                                       242

-------
Fig.  2.
                POROUS ASPHALT
JJUT
>-.
•OUECTICH .>
BIT 	 	 ^
4 DEEP 	 --""
ENTRANCE
RftMP ,
'SENCH^
I c.^.rn;'
.MUM
FLO* /
/ J!0'
', \ ' \ i \ "Ji 1
' : '
5 . n,.1 , ' ± \
* !^~~; 7 ; I
30TEI ' ; \ • 1
/It • GRAVEL TRENCH U0r h [
'. S 51 i ,
-< \ " '. \ \
\ N ,' '
^ \ 1 f*l ItNTW. »« • 1
! '""*" ^ i j
"LOW i i i " - ^ t
;;';/ '; '1
2' DEEP ^_^_ 	 	 — ' [
-j-, . Q AUSTIN MIMCIPftl. AIRFO"
"•*-&• J • 3HAVEL TRENCH LOT
H
1
s
cmi.p»8
'o»e»
NiaawTE
                  243

-------
Gravel Trench Lot

            A plan drawing of the gravel trench lot is presented in Fig. 3.  The study
area consisted of a conventional asphalt lot with a 4-ft (1.2-m) wide and 3-ft (0.9-m)
deep  drainage trench  at the  downslope  end.   The  trench  was lined with a  6-mil
polyethylene  impermeable  membrane  and  filled with  1.5  to 2.5 in  (3.8-6.4 cm)
diameter crushed stone, cleaned and washed. This base was topped with approximately
1 ft (30 cm)  of  smaller than 1-inch (2.5-cm)  diameter  gravel.   The  trench was
subsequently  flushed with several volumes of water  to rinse out construction  fines.
Stormwater  flows within the  drainage  trench were  monitored within a  55-gallon
(208-1) barrel. A 4-in (10-cm) diameter pipe was used as the discharge control.  All
discharges were  calculated based on readings of the water level within the barrel by
the field crew.

Conventional Asphalt Lot

            A plan drawing of the conventional asphalt  lot  monitored in the Austin
study is presented in Fig. 4.  Runoff discharge estimates were estimated  from water
levels  ina3x3xl.5ft (0.9 x  0.9 x 0.5 m) 90-degree V-notch box weir.
SPRINKLER-GENERATED RUNOFF EVENTS

            Sprinkler-induced "storms" provided the ability to control the intensity,
duration and timing of the inflows at the study sites. Impact-type sprinklers, supplied
by the City of Austin Parks and Recreation Department, were used during the tests
with the City's fire hydrants used as the source of water.  Field observations indicated
a spectrum of spray droplet sizes, ranging from fine mist at the periphery of the spray
stream to  large droplets  in the center. A similar range of drop sizes  were observed
during natural storm  events.   The number of sprinkler heads were varied for each
induced  storm and  care  was  taken in placement  of  the heads to  provide uniform
coverage of the  lot.  A schematic of  the sprinkler application is presented in Fig. 5.
Equivalent rainfall estimates were obtained by placing eight wedge type rain gages on
wooden stands around the test  lot. During the tests, readings of the rain gages were
made at  regular intervals (every 15 to  30 minutes) and at the conclusion of the event.
Once the individual rainfall totals were compiled, the values were averaged to provide
an approximation of the total event volume.

            The gravel trench lot was  too large for sprinkler coverage, so 2000-gallon
(7600-1) capacity rear-end-dispensing water trucks provided by the City of Austin were
used.  The trucks drove slowly  across the upper end  of  the  lot releasing water at
approximately  300 gallons  per  minute (19 1/s).   Different  event intensities  were
obtained by varying the number  of trucks used, trips made, and number  of trucks
releasing at one time.
RESULTS OF THE MONITORING SURVEYS

            Hydraulic performance results  obtained in the stormwater surveys are
discussed below.  A summary of hydraulic  characteristics of the pavements during
each runoff event is presented in Table 1.
                                       244

-------
                           FIREHOSE-
                                      FIRE
                                      HYDRANT
Fig.5.   Schematic  of Sprinkler  Application
                 245

-------



•












«
w
S
oi
W
^
g:
S
CM
O
- H
W OT
rJ [I.
w c
*-c >*
P:
-. K —
0



c ^- 	
> *J s

k-

c
o _
S g
O~


— S
•2 .2 "c
r-< C ^^
^

t
m
"o —

Z 'E
0.




*•
> a



o
&

c
c
e
m
0,
rr C
O 0
o c



ro O
O O
o o

r- ro
o o

CO O

o c

oo ro
in in


o r-
ro ro
O O





O --




O \f-
-o m




o in
c ~-





CO 00






ro ro
00 00
ro t-4
ro o
0 0


13
.0
o*
w

w
I
O
0,
ro ro ^™
C O rr
Tf 0 ~



0 O — i
,)< O C

r- ^o r-
o o o

CN -« O"
T!" rf •<)<
o o o

o ^> in


c o r-
•^ CO s£)

o o —





ij* in -^
000




•*• o o




^0 ^5 xO
O O O





ro ^C TC






ro ro rO
OO OO CO
O <-i O
o o o


o
c

H
"3
A
u
0 r-
ro c-
r- »—



CO CN
T* O-
ro O
CO -•
**. *^
— 0

O in
* "^
o o

in


TJ* ro
cc ro
O O






T}* ro
O t—




^ ^-i




ro ro
0 0





* CO






—i ro
CC CO
O -H
xo in
0 0




.„
"3
"1


















tn
re
•e
QC
C
•p
OJ
C'
Cl,

c
C'
£
£
^
c
TS
It
C
1 I
S i
JS •-
H -2
»- &
^^ ^
1 S
« 0,
s "•
v-t ^
° •-
® 'H

E <
C ^
f £

** «
LJ *o
1
"6 o
^ k 6 b
•5. 'B' *" S »
o « u 1 * §
^ ro Tf rt M O
^ oo "3 fe £ E
ro ro rsl , c rt
M 'S g
1! II 11 *•* ^ .2
** o *J

tfi Jj^ 'p.
"£ 111
.S t3 .£ rt J "

































o
J2
O
q
i
i
00
c

"3
CP
_c
e
_o
«
_b
*-«

£
^
^3
•s
o
c
*c
"""

in
ui
i'
b

c.

"^
^v
•0
^c
s
1
u.
o

0>
c
-2
[5,
£
p^
_
b
3


-------
Porous Asphalt Lot

            A  maximum intensity of  1.67 in/hr (3.5 cm/hr)  was achieved with no
resulting surface runoff at the porous asphalt lot.  As presented in Table 1, the total
discharge volume, the time to peak flow and the  peak discharge rates were similar for
each event although the inflow varied from 0.94 in (2.4 cm) to 1.53 in (3.8 cm) and the
nominal  intensity  varied  from  0.94 in/hr  (Z.4 cm/hr)  to  1.67 in/hr  (4.2 cm/hr).
Observed base runoff  ranged from  37 to  73 percent  of recorded  sprinkler inflow.
Runoff ratios less than unity were attributed to wetting of the base  media, storage
within the base layer and percolation along the trenches.

Gravel Trench Lot

            Table 1 includes a  summary of  the sprinkler events monitored at  the
gravel trench  lot.    The  application  rates  were  not  varied enough  to  produce
significantly different  discharge characteristics.   Observed runoff ranged from  64 to
77 percent of recorded inflow, with an average  of  72  percent for the three events.
Observations made  during storm events indicated the  small diameter  surface gravel
was impeding the vertical flux of water, i.e., runoff  was flowing across the top of  the
trench.

Conventional Asphalt Lot

            A sprinkler-generated runoff event and a natural precipitation event were
monitored  at the conventional asphalt study  lot, and the results are  summarized in
Table 1.  Estimated runoff volumes ranged from 71 to 118 percent of recorded rainfall.
The runoff ratio greater than unity was attributed to rainfall measurement error.
STORMWATER SIMULATION RESULTS

            Stormwater  hydraulics  for  each  pavement type  were  simulated  with
PORPAV.  PORPAV was calibrated for each lot using the initial set of observed runoff
data.  The remaining events were  subsequently simulated as verification.  A list of
PORPAV  input data  is  presented  in Table 2.   Pavement characteristics  such  as
pavement length, width and depth were obtained from onsite or construction measure-
ments.  Other parameters such as the Manning's roughness coefficient, volume of dead
storage on the pavement  and pavement porosity were estimated. Records of observed
inflow were input to PORPAV.  Calibration of the model was  initialized by  varying
values  of  the  estimated  parameters to  reproduce the  observed  runoff  volume.
Generally  this was accomplished by adjusting  the  base  void volume  (the product of
depth and  porosity) for  the pervious lots and  the volume of surface storage for the
impervious lots. The second objective was to reproduce the observed peak runoff rate.
Variations  in runoff rates were effected by varying the  estimates  of  average surface
slope and the  roughness coefficient.  For the porous asphalt and gravel trench lot the
coefficient of permeability  for the base layer was varied to reproduce  the observed
peak  base  discharge  rate.   Results  of  the  simulations are  discussed  below.   A
comparison of simulated and  observed  hydraulic characteristics is summarized in
Table 3.
                                       247

-------
                 TABLE Z

INPUT DATA FOR PORPAV SIMULATIONS
Element
Rainfall rate
Pavement surface






Pavement base



Natural soil
Parameter
Magnitude
Length
Width
Slope
Area
Permeability
Depth
Porosity
Initial dead storage
Manning's n coefficient
of roughness
Permeability
Depth
Porosity
Initial and dead storage
Collection drain capacity
Initial and final infiltration
rate
Unit
in/hr
ft
ft
ft/ft
ft2
in/hr
in
ft3/ft3
in
f«1/6/s
in/hr
in
«3/f,3
in
in/hr
in/hr
. -1
          Horton's infiltration decay
            coefficient
hr
                   248

-------
                                            TABLE 3
                                      SIMULATION RESULTS
                                      POROUS ASPHALT LOT

POROUS ASPHALT LOT -
Observed
Simulated
Deviation
POROUS ASPHALT LOT -
Observed
Simulated
Deviation
GRAVEL TRENCH LOT -
Observed
Simulated
Deviation
GRAVEL TRENCH LOT -
Observed
Simulated
Deviation
GRAVEL TRENCH LOT -
Observed
Simulated
Deviation
Peak Flow
(cfs)
- STORM 1
O.Z7
0.273
+0.003
- STORM 2
O.Z4
0.514
+0.274
STORM 1
0.44
0.497'
+0.057
STORM 2
0.58
0.487
-0.093
STORM 3
1.67
0.472
-1.198
Time to Peak
(rain)

58
50
-8

53
55
+2

60
90
+30

66
70
+4

55
60
-^5
Runoff Volume
(cubic feet)

745
745
0

721
1,409
+688

1,960
2,107
-147

1,693
1,650
-43

1861
1,448
-413
Detention Time
(min)

42
28
-14

42
25
-17

29
24
-5

24
24
0

19
23
+4
CONVENTIONAL ASPHALT LOT - STORM 1
Observed
Simulated
Deviation
0.34
0.297
-0.543
53
50
-3
368
269
-99
1
5
+4
CONVENTIONAL ASPHALT LOT - STORM I
Observed
Simulated
Deviation
0.22
0.257
+0.037
7
10
+3
138
140
+2
5
8
+3
1 cfs = 28.32 Ips
1 ft3 = 28.32 1
                                         249

-------
Porous Asphalt Lot

                  The discharge hydrograph for the  calibrated data set, presented in
Fig. 6a, accurately resembles the observed one, but is advanced about fifteen minutes,
as is reflected  in the difference in detention times.  The calibrated coefficients  of
dead storage  and base permeability were held constant during the simulation of the
remaining event.   Observed and simulated  discharge hydrographs for the final event
are displayed in Fig. 6b.  The significant overestimation  of peak discharge rate and
volume probably  resulted  from an incorrect sprinkler inflow  measurement (used  as
input to  PORPAV) or  an  inconsistent  hydraulic  response,  possibly  increased  base
storage or percolation.

Gravel Trench Lot

            Three artificial rainfall events were simulated for the gravel trench lot.
For  each  case  the runoff  from the  conventional  asphalt lot  was  simulated  with
PORPAV and subsequently used  as  input  to the  gravel trench simulation,  hence,
characteristics  for both facilities  had to be  determined.   A  summary  of  simulation
results is presented in Table 3.   Coefficients of base  permeability and  dead storage on
the asphalt lot and in  the gravel trench were determined by calibrating PORPAV with
the  initial  runoff  event data.   The  calibration hydrograph  is presented in Fig. 7a.
Runoff characteristics were reproduced quite well for the  second  event, as displayed
in Fig. 7b.  Incorporating the same physical characteristics of the asphalt and gravel
trench for the final event simulation yielded a less satisfactory comparison, presented
in Fig. 7c.  The major discrepencies in the two hydrographs of the third event occur
during the  periods of water release  from  the water trucks,  possibly an artifact of
utilizing a constant inflow rate during this period in the simulation.

Asphalt Lot

            Simulation  results for the  conventional asphalt  lot are  compared  with
recorded values in Table 3.  The second event was used to  calibrate PORPAV and was
accurately  simulated, as shown in Fig. 8a.  The simulated hydrograph for the  second
event is compared to observed results in Fig. 8b. The simulation did not reproduce the
peak discharge  rate,  possibly a result of  a  short,  intense burst of  rain which was
undetected in the rainfall  data.  Both simulations depicted the  very  rapid detention
times associated with the asphalt lot events.
CONCLUSIONS

            The revised  PORPAV satisfactorily simulated sprinkler-generated storm-
water  hydraulics of  both  porous and nonporous pavement  facilities.    Sprinkler
application rates ranged from 0.4 to 1.7 inches per hour.  The favorable simulation
results obtained hi this study suggest  that PORPAV can be used to assess the relative
hydraulic performance of pavement facilities available for urban runoff control.  The
option of using  a drain pipe for the base layer discharge has not been evaluated.  A
future study should be conducted at an existing pavement facility utilizing a collection
drain system to  assess the ability of PORPAV to simulate such a control strategy.
                                        250

-------
                                g
                                f/l
                                cfi
251

-------
          2*   '-if)   99   30   tac  12$  me  ie<»  i%»  ze«
               ELPPSED  TCME  (minufss)
FT5. / cLsiMULPTED  VS.  OBSERVED  HYDftOGROPHS FOR 3/5/S2.
                                                                              9    29   U0   59   89   199. 129  149  169  189  299
                                                                                        ELPPSED TIME 
-------
                                 OO
   i«0i 39SHHOSIO
253

-------
REFERENCES

1.    Biggers, D. J., J. P. Hartigan, Jr. and H. A. Bonuccelli, Urban Best Management
      Practices (BMP's):  Transition from Single-Purpose to Multipurpose Stormwater
      Management.   Conference  Paper, International Symposium on  Urban  Storm
      Runoff, Univ. of Kentucky, Lexington, Kentucky - July 28-31, 1980.

2.    Diniz, E. V.   Porous Pavement  Phase I - Design  and Operational Criteria,
      EPA-600/2-80-135, May 1980.

3.    	.  Storm Water Management Model  Supplement to the User's Manual, EPA
      Project No. CR-805664, March 1981.

4.    Heaney, J. P., W. C. Huber, M. A. Medina, Jr., M. P. Murphy, S. J. Nix and S. M.
      Hasan.   Nationwide Evaluation  of Combined Sewer Overflows  and  Urban
      Stormwater Discharges,  Volume II:  Cost Assessment and Impacts, EPA-600/
      2-77-064, March 1977.

5.    Israelson, O. W. and  V.  E.  Hansen, Irrigation Principles and Practices, 3rd
      Edition, John Wiley and Sons, Inc., New York, N.Y., 1962.
The work  described in this paper was not funded by  the  U.S.  Environmental
Protection  Agency.  The contents do not necessarily reflect  the views of  the
Agency and  no official endorsement should be  inferred.

                                     254

-------
            PLANNING AND IMPLEMENTATION OF REGIONAL STORMWATER
           MANAGEMENT FACILITIES IN MONTGOMERY COUNTY, MARYLAND

                               JANUARY 1983

    John M. Crouse,  P.E.,1  Vincent  H.  Berg, P.E.,2 and Linda J.D. Mitchell3
                              I.   INTRODUCTION

     Montgomery County, Maryland,  is  located to the  north  and the  west  of
Washington,  D.C.   The County  is geologically  located  just  above the  fall
line  between the Piedmont  Plateau  and  the  Coastal  Plain.   The County  is
characterized  by  gently rolling  topography  with  steep  slopes adjacent  to
the  major stream  valleys.    Most  of  the major  streams originate  in  the
County and drain into the Potomac River.   The watersheds generally  range  in
size  from 20 square miles  to approximately 130  square miles.  The  County
contains  approximately 500 square miles.

     Being adjacent  to  the Nation's  Capital, the  County has  experienced  a
major  growth in population  and  development over  the  last  30 years.   The
population in  1950  was  approximately 160,000 and  today  it  is  approximately
600,000.   The  majority of  the population is concentrated  along the  major
transportation corridors extending out  like  spokes of a wheel  from  Washing-
ton,  D.C.  These  corridors do not  follow watershed  boundaries, but  extend
across several watersheds as they radiate out from the District of  Columbia.

     Each year, approximately  2,000 to  2,500 acres are converted from rural
to urban  land  use  to house the  population increase and  satisfy demographic
changes.  As  a result of this urban  growth,  it is estimated  that  approxi-
mately  200  miles  of the  County's  1,015 miles  of  streams  have  suffered
deterioration  in  the form of  increased flooding,  accelerated  channel  ero-
sion,  and reduced  water quality.  An additional  eight  miles  are  estimated
to be affected each year.
    Project  Manager,   Water  Resources   Section,   Greenhorne   &   O'Mara,
    Incorporated, Riverdale, Maryland.
    Senior Environmental  Engineer,  Water  Resources  Section,  Department  of
    Environmental Protection, Montgomery County,  Maryland.
    Former Project Engineer,  Water  Resources  Section,  Greenhorne &  O'Mara,
    Incorporated, Riverdale, Maryland.


                                    255

-------
            II.   STORMWATER  MANAGEMENT  (SWM) PROGRAM DEVELOPMENT

     Stormwater  management  was first required  as  an  outgrowth  of  sediment
control requirements.   In July 1961, the  State  Attorney General  interpreted
the powers  and  duties  of the State Water  Pollution  Control  Commission  to
include regulation of  sedimentation  resulting from  land development activi-
ties as an industrial pollutant.

     Throughout  the  late 1960's  and early 1970's  local jurisdictions  and
the State  General  Assembly  adopted  legislation requiring  sediment control
during land development  operations.  In 1971, the Maryland  Attorney General
made  another  landmark  interpretation  which  said that  the  state and  local
jurisdictions, under  sediment control  powers,  could require the  control  of
runoff after development in  a manner to prevent  off-site erosion.

     In July  1971, the  Montgomery  Soil Conservation District  (MSCD) adopted
its first stormwater management policy.   This policy  required  the detention
of  stormwater runoff  from   the  two-year   storm event  on  each  development
site.   The release rate could  be no  greater than the  runoff  that would have
left the site prior  to development.   To  achieve these  criteria,  the major-
ity  of  the   developers  used  on-site  stormwater   detention  ponds  as  the
simplest and  least expensive  means of  compliance.   However,  where the site,
due to size, topography, or  the value of the land, did  not  lend  itself to a
surface impoundment, other  means have  been used, such as rooftop detention,
underground vaults and  pipes,  infiltration pits, surface storage on parking
lots and pervious parking lots.

     The action  of the  MSCD  in requiring  on-site  detention  facilities  in
conjunction with new  development  was a significant step forward  in achiev-
ing  better control  of  storm  runoff.   Requiring  developers  to  plan  for
stormwater detention  recognizes  the public responsibility  to regulate  the
allowable peak runoff  associated  with  land development.  However,  reliance
on  only  this  approach  to  achieve the stormwater  management  objectives  is
viewed as only a partial solution for the  following  reasons:

     1.  In many instances,   individual  development proposals have site con-
         ditions which preclude the  installation of truly effective on-site
         controls.   However, until recently, no mechanism existed to permit
         a developer  to contribute,  as an  alternative  solution, to  a more
         strategically  located facility   that could  provide  a much  higher
         level of stormwater control  and which  is more cost  effective.

     2.  In some  instances,  greater  stormwater  control  is  desirable  at
         particular  locations then  can   be  required  under  existing  MSCD
         criteria.

     3.  While certain types  of control structures are  effective  in meeting
         present MSCD requirements for on-site controls, they may be highly
         undesirable  from  other  standpoints   such  as  public   safety  or
         aesthetics.
                                   256

-------
     4.  By themselves,  present  on-site control requirements  are  incapable
         of handling  runoff  volume  and  velocities  accompanying  storms  of
         greater intensity than the two-year frequency storm.

     5.  No reliable  mechanism for  insuring  perpetual  maintenance  of  pri-
         vate, on-site  facilities  exists.  Therefore, many  facilities  fail
         to function as designed due to a lack of maintenance.

     In addition  to the above considerations,  a review of  stormwater  man-
agement plans for  on-site  facilities,  reviewed  and approved  by the  MSCD
during  the 12-month  period  from  October  1,   1978,  through September  30,
1979, indicated that  78 developments on 1,584  acres would expend  over  $3.5
million to provide on-site stormwater management.

                      III.  CURRENT PROGRAM DIRECTION

     In recognition  of  the growing  problems  created by uncontrolled  urban
storm runoff  and  the  limitations of  the ongoing program of on-site deten-
tion of the two-year  frequency storm,  the  County embarked upon  an  aggres-
sive program  to  provide stormwater  management protection  on  a  watershed
basis.    In May  1973,  the  stormwater  management  program  was  established
within the Montgomery County  Department  of Environmental  Protection  (DEP)
to develop and  implement a County-wide stormwater management program.   The
program  consists  of   two  major  thrusts,  first,  the  implementation  of
remedial  projects,  and,  second,  the  development  of preventive  stormwater
management projects on  a watershed basis.

A.   Remedial Projects

     The  remedial  focus  of  the  County  SWM  program  includes  projects  to
correct  existing problems  caused  by uncontrolled  runoff  from  previously
developed  areas.  These projects are designed  to  help  stabilize  conditions
in critical deteriorating stream reaches.

B.   Preventive Programs - On a Watershed Basis

     While  the remedial  efforts  serve  to  control  or  eliminate  existing
problems,  they  do little to fulfill  the  need  for  comprehensive,  long-range
stormwater management.   Experience  has  taught us  that  controls  for  storm
runoff never  realize  the optimum potential until  we consider  the watershed
as  a whole and  evaluate the  effects that proposed  land  use  changes  will
have on the quantity  of storm flows  throughout  the watershed.   In an effort
to fulfill this need, Montgomery County has embarked upon  an ambitious  pro-
gram of planning stormwater mangement on a watershed basis.

     In 1973, the County Council  authorized funds  for consultant  studies of
three  major  watersheds  within  the County.   Several   common  conclusions
emerged from  the three  studies:

     1.  Control of runoff  should  be accomplished  near  the  source either on
         each development  site,  or  near  the  site where several  individual
         developments could share one facility.

                                    257

-------
     2.  The  level  of control should  be expanded to  include not  only the
         2-year storm,  but  also  the 10-year storm,  and,  where appropriate,
         up to the 100-year storm.

     3.  Needs for  non-point  source pollution  control  should be  evaluated
         and appropriate controls implemented.

     These studies also  indicated that  stormwater  management  controls (pri-
marily detention  impoundments) were most effective when  located  in or near
the headwaters.   Conversely,  detention  structures  placed on  tributary sub-
watersheds near  the mouth  of the  watershed  had  little  benefit and  could
actually reinforce peak flows due to timing of  the release  of the detention
structure to coincide with the peak flow of the main stream.   For this rea-
son it has become evident that the  watershed  must be considered  as a whole
when planning an effective program.

     As  a  result,  the on-site stormwater  management program  has  undergone
several  changes.  The major  change  is  to  allow a developer  the option  of
contributing towards  the  costs  of  a central  (public)  facility when  an on-
site facility is deemed  less appropriate  than a central  facility.   This
approach allows  the developer more flexibility in  designing his  proposed
development  and   enhances  the   County's  ability  to  pool   resources  from
several  developments  for a more cost-effective solution,  while  providing
the County with  the  flexibility in selecting  the  best  sites for central
stormwater management facilities.

                    IV.   STORMWATER  MANAGEMENT ORDINANCE

     The culmination  of these long years  spent in  developing the  County's
SWM program came on March 3,  1980,  with the  signing  of the new  County Code
amendment, which  became  effective on June  2,  1980.  The  Ordinance  assigned
to Department of Environmental Protection  (DEP)  the  responsibility  of coor-
dinating SWM in the County.

A.   Legislative Intent

     1.  All developers are intended to  be held responsible for  the impacts
         of  storm runoff  created  by  their  development.   This  has  been
         accomplished by requiring  the  installation  of on-site detention  on
         all  new  development  or,  in  the event  the  on-site  requirement  is
         waived,  through the  payment of  a  "contribution"  or fee  not greater
         than the cost of on-site SWM.

     2.  The  contribution  structure is  such  that  lighter-density  develop-
         ment  is  rewarded.    The  Ordinance  provides  for  the  collection  of
         SWM fees based  on  typical  impervious  areas  and  storage  quantities
         required for excess  runoff rates,  which  are proportionately  lower
         for  less dense  development.    In  fact,  construction  of  single-
         family  residences  on  lots of  two  acres or more  are  completely
         exempt from the provisions of the  Ordinance.
                                    258

-------
     3.   The planning and building of small, on-site structures  is  discour-
         aged.   Many of  these on-site detention  structures  have proven  to
         be relatively  expensive to  build,  and  most  take  up  usable  land
         area.   Most  of these  structures have  been transferred  to  Home-
         owners  Associations basically uninterested and incapable of  proper
         maintenance, and consequently they  have  not received the  required
         maintenance.

     4.   The Ordinance  provides  for  DEP  to  develop operational  plans  for
         centralized SWM  structures  to  be placed  in  strategic  locations.
         It also provides for County  participation to  assist developers  of
         strategic   sites  to  design  and  build  regional   structures  with
         greater than required storm runoff controls.

     5.   To assure fulfillment of SWM controls  within  a watershed with  the
         onset  of development in a given  watershed, a program for construc-
         tion  of public  SWM  facilities  has  also  been  implemented  through
         the Capital Improvement  Program  (CIP).

     6.   The Ordinance  provides  for  the  coordination  of  proposals  for
         potential   and  needed central SWM sites  to be  incorporated  into
         updated versions  of master  planning  documents.


B.   Regional  Stormwater Management

     The new legislation, while  encouraging SWM on  all  developments,  relies
heavily on  the  use of centralized  SWM facilities, each  of which serves  a
number  of  developments and  possibly  provides  multi-use  opportunities  for
the community.   The SWM legislation has had  a marked effect  on  reducing the
number  of   small  on-site SWM structures.   Under  the  County's  stormwater
management  ordinance small   on-site  structures  can  be waived,  and  larger
centralized SWM  structures  can  be  built, either  through  the County's  CIP
Program or  through  developer-constructed  SWM projects  with  County  partici-
pation.  This combination system of regional and on-site SWM structures is
most effective  when SWM facilities are  strategically  placed in  watersheds
which will  have the  greatest benefit for  the  environment  as  well   as  the
community.

     When  development  pressure  is  increasing within  a major  subwatershed
the Department  of  Environmental  Protection can  conduct a detailed  study of
a subwatershed utilizing CIP  funds  under  the  Preliminary  Stormwater  Manage-
ment  Investigation  Project.   The  major  purpose  of  these  studies   is  to
determine  the  stategic  location of  regional  stormwater management  facili-
ties and to develop a detailed operational plan for the  subwatershed.  The
subwatershed studies will  generally  be  for  areas  of  two  to  four  square
miles.

     To date, five  of  these studies have been completed  and two additional
studies  are now being  conducted.  The  Cabin  Branch  study  is  typical  of
these preliminary stormwater management studies.
                                    259

-------
                       V.  CASE STUDY - CABIN BRANCH
A.   Introduction
     Cabin Branch is a tributary of  Great  Seneca  Creek  in the Potomac River
basin.   Cabin  Branch  is located  on the  north   side  of  Gaithersburg,  the
second largest city in Montgomery County.   The  location of  the  study area
is shown  on  Figure  1.   The drainage area of  the studied portion  of Cabin
Branch is approximately 4-1/4 square miles.

     The Gaithersburg  area  has  been and is  expected  to  be an  area  of high
growth in the County.  The  effects of such  growth have  been  partially con-
trolled  by  sediment control  and  required  on-site  stormwater  management
facilities.

     The  Montgomery  County  Department  of  Environmental  Protection  con-
tracted with  the consulting firm  of Greenhorne  & O'Mara,  Incorporated,  to
perform  a  preliminary  investigation of stormwater  management,  flood con-
trol,  and erosion   control  requirements  for  the  existing  and  projected
conditions of the  Upper  Cabin Branch watershed.   The purpose  of  the study
was to determine the larger-scale  effects  of on-site SWM  facilities  on the
receiving  stream,  to evaluate  and  locate regional  SWM  facilities,  and  to
evaluate the impact of multiplefrequency control  on stream discharges.
                                         MINCI«IOR*riCO.
                          Figure 1.  Vicinity Map

                                    260

-------
     Excerpts of  the report  on  the  investigation  presented in  this  paper
illustrate the process of  analyzing  watersheds  and  locating  regional  storm-
water management facilities.

B.   Data Acquisition

     Extensive data from such County agencies as  the  Office  of  Planning and
Capital Programming and Department of Transportation  along with information
from the Montgomery Soil Conservation District, the Maryland-National  Capi-
tal Park and  Planning  Commission,  the Washington Suburban Sanitary Commis-
sion,  and  the  Federal  Emergency  Management  Agency  was  assembled.   This
information  included proposed  capital   improvements,  bridge  and  culvert
crossings,  hydrologic  soil groups,  existing  stormwater management  facili-
ties, master  plans,  flood  plain  delineation,  and utility  locations.   Other
data such  as  aerial  photography taken  at various periods,  tax  maps,  topo-
graphic maps,  and historical  sites  atlases proved  valuable in  developing
predeveloped  land use (used  as  baseline data),  existing land use,  and pro-
jected land use mapping.   The information obtained from these  data sources
was checked  with  the field conditions,  and modifications  or  additions  to
the data were made where necessary.


C.   Methodology

     1.  Hydrology model

     The hydrologic analysis  of the  watershed was accomplished  by utilizing
the Soil  Conservation  Service Technical  Release  Number 20  Computer  Model.
The hydrologic  condition  of  the  soils was taken from  the  SCS  Soil  Survey
for Montgomery  County.   Cover conditions were  determined for  predeveloped
(circa 1942),.existing,  and projected land use conditions through  the data
acquisition process previously described.

     The Cabin Branch study area  was approximately  one  percent  developed in
the early  1940's, consisting  mostly  of farmhouses and farmsteads.  Approxi-
mately seventy percent of  the area was in cultivation,  approximately twenty
percent was  forested,  and  the rest  of  the watershed  was  pasture or open
space.

     Currently, the  watershed is undergoing rapid  residential  development,
generally  of  high density.   Approximately  fifty  percent  of the  study area
is now developed.  About twenty-five percent is still in  cultivation.   Much
of the stream valley of Cabin Branch contains open space.

     Approximately  eighty   percent  of  the  study  area  is  projected   to  be
ultimately developed.  The  remainder is  zoned to  be park  land,  golf course,
and other  open  space.   The area in  the  vicinity  of the airpark  is planned
to be  developed  as commercial/light industrial  sites.   The extreme  north-
east  corner  of   the   study  area   will   be   developed   as  single-family
residential.
                                    261

-------
     2.  Subarea delineation criteria

     In  order to  provide  stream flow  estimates  at  several  points  along
Cabin  Branch,  the watershed was  divided into subareas  according  to runoff
flow paths and velocities and  points of discharge into  the  stream.   Drain-
age  area  divides  were  first   drawn between  direct  tributaries  to  Cabin
Branch which  were  shown on the  USGS 7  1/2-minute quadrangles- as  perennial
streams.  If forks or  tributaries of the direct  tributaries  were also indi-
cated  as perennial streams, the  subdivides  between them were then  delineat-
ed.  If  very  dissimilar land uses were  encompassed  in a subarea, then  the
area was further subdivided.   Discharge points used  in other  studies  and
such landmarks as major  road crossings  were also used to separate the dif-
ferent  subareas,  as  well  as  major  existing  SWM facilities  or  potential
sites for facilities.  No  subarea was to be  greater  than fifteen percent of
the study area.

     The final analysis  for Cabin Branch was performed using eighteen basic
drainage subareas as shown on  Figure 2.  These  subareas  were further subdi-
vided  as necessary to  account  for additional road  crossings  and significant
SWM facilities in existence.   A schematic  diagram of  the study  area hydro-
logy is  shown in Figure  3.

     3.  Key point selection

     Seven sites of  interest called  key points were located along the main
stem of  the study stream to  serve as convenient points  at which to  compare
the predeveloped,  existing,  and  projected  discharges.  Easily  recoverable
locations such as road crossings  and confluences of  the  main stem  and major
tributaries were designated as  key points  as well as  other  strategic loca-
tions  such as  areas  where land use  has  undergone  or  will undergo signifi-
cant change.

     4.  Existing SWM facility screening criteria

     Data on  the  existing SWM  facilities were compiled  from the previously
mentioned  data files  and field   reconnaissance.  An  extensive  amount  of
information was available because the watershed has undergone  urbanization
since  the  inception  of  the  sediment control  and  SWM regulations  and  the
requirement for  submission of  plans  and computations.  Screening criteria
were developed to remove from  the investigation  certain  SWM  facilities that
were believed to have little impact for the  scale of  the  investigation.

     Dry and  wet ponds  were considered  in  the  hydrologic analysis.  Other
SWM measures  (such as  rooftop  detention, gravel  trenches, and  seepage pits)
found  in the  watershed were not  of  sufficient  size  to  achieve  significant
stormwater management.

     It  is interesting  to note  that  the  two  SWM  ponds  in subarea  10 are in
a  subdivision  currently  under  construction  and the ponds  are  being  built
under  the County's participation  program.  Wet and dry ponds were  consider-
ed  not significant  and  were excluded  from  the  hydrologic  analysis   of  the

                                    262

-------
                                                   3)  Sutwren Number
                                                       Kcu Point
Are* = 2,7/3 Ac.

              L
            Limif of
                   Figure 2. Drainage Areas and Subareas
                                  263

-------
             CABIN BRANCH
                  PACUJTV oocc «ND DRAINAGE
               *m*CAcnn) OK SUBAREA
               ftoum coot
                    OK CULVBUT CROSSING
               /WJALV^eO POK HYDHOLO9IC
                  Figure 3.   Schematic Diagram of Hydrology
watershed  if  the pond drainage area was  less  than five percent  of  the sub-
area in which  the facility was located and less  than  five-tenths percent of
the  entire study area.   If  the  drainage  areas  for the  ponds which were
screened out by  these criteria summed to more than five percent  of  the sub-
area  in  which the  ponds were  located  or to  more than one percent  of  the
total  study  area,  the  largest  excluded  structure  was  reincluded.   An
excluded facility may be assumed to have  little  impact on  the  watershed as
a whole  but does  provide some  protection  against erosion  and  flooding  in
the  immediate downstream vicinity.   Eighteen  ponds  in  the  Cabin  Branch
study  area were determined  to  be  significant;  eight smaller  ponds  were
dropped from consideration.

     5.  Peak discharge  determination

     Estimates of the peak  discharges in the watershed for  the  2-,  10-,  and
100-year storms  for  predeveloped,  existing,   and  projected conditions were
produced using the SCS TR-20 computer model  with the input data previously
described  and  rainfall from the  U.S.  Weather Bureau  Technical  Paper  Number
40.   The peak discharges  are  summarized  in  Table 1 for key  points  along  the
stream.

                                    264

-------
                  Table 1.  Peak Discharges at Key Points

Key      	2-Year	   	10-Year          	100-Year	
Point*   Predev.  Ex.   Proj.   Predev.  Ex.   Proj.    Predev.   Ex.   Proj.
A        154      219   326      495      609   763     955     1110  1297
B        210      304   364      654      811   905    1246     1455  1586
C        331      626   837     1030     1542  1906    1976     2683  3205
D        420      720   923     1252     1666  2184    2479     3221  3910
E        440      718   866     1209     1786  2246    2619     3616  4320
F        482      771   920     1281     1916  2373    2774     3819  4506
G        500      793   941     1313     1918  2343    2837     3661  4230
         *Key Point locations are as follows:
     A - downstream of subareas, 1,2, and 3, in headwaters
     B - downstream  of  subareas,  1,2,3, and  4,  north of  Montgomery County
         Airpark, near Green Farm
     C - downstream of subareas, 1-9, at Snouffers School Road
     D - downstream of subareas, 1-12, at Goshen Road
     E - downstream of subareas, 1-16
     F - downstream limit of subarea 17, at Montgomery Village  Avenue
     G - downstream limit of study area


      6.   Runoff increase  criteria

      To  determine sites where  SWM  basins  may be  most beneficial, the  per-
 cent  change in discharge from existing to projected conditions was  examined
 for each key  point.   Subareas with relatively  high increases in  discharge
 were  considered for SWM facility  sites.

      7.   Site  considerations

      The percent increase analysis  indicated that key points A and  C  should
 be explored for  SWM  sites.   A significant  increase  in  peak discharge  from
 existing to projected conditions  was also  shown at key point D, but no  fea-
 sible basin sites  were apparent in  that  area.   Five  sites were  analyzed:
 site  1  was  located at key point A;  sites 2, 3, and  5 were situated  upstream
 of key  point C on tributaries  to  Cabin Branch; and site 4 was at  a  location
 for which a preliminary design had  been completed  for  another  study.

      Potential sites were  evaluated in terms of  the downstream  land uses,
 the timing  of the peak discharge relative to the  times  of downstream  peaks,
 and the natural  storage  available  at  the  site without unduly infringing on
 the surrounding  land  uses.   Soils   in  the immediate  vicinity  of  the sites
 were  examined  for  suitability as  embankment material  and  for   reservoir
 lining.

      Factors  other than  land  use,  soil  type,  hydrologic  response,   and
 hydrologic   performance  serve   an  important role  in  SWM  facility  location
 study.   Acquisition  of  property,  construction cost,  and utility  conflicts
 are three such factors with potentially high  impacts.   The  impacts  of these
 factors for the  five proposed sites  were evalutated and  are indicated in
 Table 2.

                                    265

-------
                      Table 2.  SWM Site Considerations
Factor
Area Property Owners


Construction Cost
Project Area

Access Right-of-Way

Utilities
Specifics
 Comments
             Site 1
MNCP&PC Hadley Farms Dairy, Inc.   Favorable
and First Citizens Development
Corp. et al
$111,300
11.18 acres (MNCP&PC); 2.68 acres
(Hadley); 1.47 acres (First Citiz
0.35 acres (MNCP&PC);
1.29 acres (First Citiz.)
8-inch, 10-inch and  12-inch
sanitary sewers (1450 L.F.)
             Site 2
Area Property Owners  H.J. Bobys, et al
Construction Cost
Project Area
Access Right-of-Way
Utilities

Area Property Owners

Construction Cost
Project Area

Access Right-of-Way
Utilities
$92,000
11.03 acres
0.35 acre
Apparently none
             Site 3
Montgomery County Airpark, Inc.
and A.W. & E. Stang
$83,000
11.49 acres (MCA); 0.80 acres
Stang
0.41 acres (Stang)
Apparently none
             Site 4
Area Property Owners  C.J. and I.B. Savage

Construction Cost     unknown
Project Area
Access Right-of-Way
Utilities
Area Property Owners

Construction Cost
Project Area

Access Right-of-Way
Utilities
 Moderate

.)
 Requires  relocation
 or  protective  measures
unknown
unknown
10-inch sanitary sewer

             Site 5
MNCP&PC, J.E. Richardson Trus.,
and Montgomery Co.  Revenue Auth.
$93,800
8.46 acres (MNCP&PC);  3.22 acres
(Richardson); 1.30  acres (MCRA)
0.06 acre (J.E.  Richardson Trus.)
8-inch sanitary sewer  (1920 L.F.)
 Favorable  if  developed
 by  owner
 Moderate
 Very  favorable

 Favorable  if  developed
 by  owner
 Moderate
 Very  favorable

 Favorable  if  developed
 by owner
 Pond   privately  pro-
 posed
Requires     relocation
or protective measures

Favorable  if developed
by owner
Moderate
Requires    relocation
or protective measures
                                      266

-------
     8.  Proposed pond development

     Preliminary designs for  sites  1 and  2  were based on  providing  reduc-
tion of the 2-,  10-,  and  100-year storms from the projected  peak  levels to
approximately the  lesser  of predeveloped or existing  levels.   The  basin at
site 3  was  sized  based  on the  intents  of  the  potential  developer  of  the
area as discussed  with the  engineer retained by the developer.  Site 4  was
based on file  data as obtained  from the Montgomery Soil  Conservation Dis-
trict.   Site 5 design was based  on  providing  2-  and  100-year  control  to  the
predeveloped conditions and control  to near-predeveloped conditions  for  the
10-year storm.

     Storage requirements at  a  site were estimated  by  the  methodology pres-
ented  in the  SCS  Technical  Release number  55,  "Urban Hydrology for Small
Watersheds."   Initial  peak discharge reduction  was typically  accomplished
with a  low  flow orifice/riser/barrel  arrangement,  with computations based
on  the  Maryland SCS  publication,  "Standards and  Specifications  for  Soil
Erosion and Sediment  Control in  Developing  Areas."   The  size  and  location
of the outlet works were  adjusted as necessary to attain  the desired goals
based on the initial design as suggested by the above references.

     9.  Scenario  development and testing

     Upon obtaining satisfactory peak discharge  reductions for  the  individ-
ual sites by utilizing the  TR 20 program,  the watershed hydrology  model  for
the  projected  land use  conditions  was  modified  to reflect  the individual
and combinations of individual  proposed  sites.   Ten  scenarios were  develop-
ed to determine  the effects of  the five proposed SWM  sites.   The  scenarios
and the sites comprising the  scenarios were:
Scenario
Scenario
Scenario
Scenario
Scenario
Scenario
Scenario
Scenario
Scenario
1
2
3
4
5
6
7
8
9
- SWM
- SWM
- SWM
- SWM
- SWM
- SWM
- SWM
- SWM
- SWM
Site
Site
Site
Site
Site
Sites
Sites
Sites
Sites
1
2
3
4
5









1
1
2
1
Only
Only
Only
Only
Only
and
and
and
, 2,





2
3
3
and
                    Scenario 10 - SWM Sites 1, 3, and 5

     The discharge values at the  key  points for  the  projected land uses are
shown in Table 3 for scenarios  1  to 5.   Table 4  presents  the percent change
in  estimated peak  discharges  from  predeveloped, existing,  and  projected
conditions to the proposed  conditions  at key  points  for selected scenarios.
The more desirable scenarios  produce  a smaller percent change in discharges
for a greater number of key points.
                                   267

-------
            Table 3.  Key Point Discharges for Scenarios 1 to 5
Key Point   Existing Structures Only
            Predev.
Two-year storm
A
B
C
D
E
F
6
 154
 210
 331
 420
 440
 482
 500
           Ex.
219
304
626
720
718
771
793
Ten-year storm
A
B
C
D
E
F
G
 495
 654
1030
1252
1209
1281
1313
One hundred-year storm
A
B
C
D
E
F
G
326
364
837
923
866
920
941
609
811
1542
1666
1786
1916
1918
763
905
1906
2184
2246
2373
2343
 955       1110   1297
1246       1455   1586
1976       2683   3205
2479       3221   3910
2619       3616   4320
2774       3819   4506
2837       3661   4230
                               Proposed Land Use Conditions,
                                 with SWM Scenario Number:
                                         1
               Basin
               1  only
                  2
                Basin
                2 only
155**
185
701
786
763
821
844
321
357
778
860
818
870
890
                507**   755
                599      889
               1501     1696
               1682     1908
               1822     1989
               1945     2093
               1954     2078
                963**   1287
               1180     1560
               2496     2907
               3213     3544
               3582     3916
               3747     4004
               3626     3800
  3
Basin
3 only
 321
 357
 716
 807
 782
 832
 853
                         755
                         889
                        1666
                        1907
                        1991
                        2086
                        2074
                        1287
                        1560
                        2935
                        3573
                        3953
                        4042
                        3840
  4
Basin
4 only
 321
 357
 823
 905
 807
 857
 878
         755
         889
        1875
        2130
        2060
        2167
        2148
        1287
        1560
        3151
        3814
        4174
        4301
        4046
  5
Basin
5 only
 321
 357
 596
 704
 714
 766
 789
         755
         889
        1451
        1670
        1790
        1897
        1900
        1287
        1650
        2593
        3245
        3565
        3703
        3557
*At  key  points above  the proposed  SWM  sites, the  listed peak  discharges
with and without  the proposed SWM may differ  slightly  due to the use  of  a
larger time increment in modeling the proposed SWM Scenarios.

**0utflow from proposed SWM Basin 1.
                                     268

-------









CO
J_
ro
c:
O)
o
oo
T3
O)
O
Ol

0)
oo

.
£r
M-
1/1
QJ
CD
fO
(j

.C
^
C
vU
O
U
0)
o_


.
"^
OJ

-Q
i_
r~~
















































^ — .
>*j
,—
C
O

- '
£
•r—
ro
u_

oo

-a
O)

o
a.
o
S-
Q-
«^»x

, —

O
H

C
0)
o
oo




£
"^
LU
cr
LU
£
Q;
^c
LU
i~
o
o
"~








>-
^
LU
cy
LU
fV*
^
(•y
^t
UJ
^
O










>-

LU

__j
o-
LU
o;
u_

ry"
5c
LU
>-
1








>-
LU
\S

O-
O
S_
D_
O
S-
Q-

0-
0
CJ-
X
LU

a.
o
s-
Q_
^
D_


Q-
O
s_
CL,
o
Q_

Q.
0
s_
0-
X
LU

Q-
0
S-
Q_

01
S-
Q_


a.
0
a.
•"-3
o
Q-


Q.
O
$-
o.
1
x
LU


a.
o
^-
D-
1
Ol
^_
0-

1—

) — 1
o
Q-




<& VO CM CO I-~ ^ ^J-
CM CM CM i — i — i — i —
1 1 1 1 1 1 1



CO CTl r- O i — CM i —
1 — 1 — 1 III
1 1



i — LO IO O 1 — LO CO
1 CM CO CO OO CM






^j- *d~ i — co  oo r-~
CO CO CM CM r— ,— i—
1 1 1 1 1 1 1




1^. <£> CO i — CM CM CM
i — CM 1
1 1


^-^
CO

-a
CMCOlO^ti — CMCTi C
i "st co LO LO •=}- ro

1 —

to
Ol
£
CMCTllDLOCMi — O -i—
i i i i i i i ro
u_

s:
oo

0*1 o^ c\j ^^ vo ^o kD ~a
CM CO i — O)
1 1 CO
0
Q.
o
s-
Q.
• — ^

i — CM CM r . ro O cri r —
i — i — oo r~~ r^- vo
1 i— O
'£
ro
c
<;CQOQLULl_CJ3 O)
o
oo




<-D
CM
1



CO
r—
1



,_
1






^J.
CO




(^
^~
1







00





C\J
LO
1




CTl
CM
1






, —






CO
1






CM
^~
1



CO






^
CM
1



CO
r—
1



00







^.
CO




CTi
I—
1






^~
CM





CO






r^
1






LO
f^




c_^






CO
CM
1




r^«
1



CM







o
CO
1





co
1






CM
CM





^o
C\J
1





10
1






CM
vo




Q






*d~ co
CM CM
1 1




CTi CTi
1 1



LO LO
CM CM







VD LO
CM CM





r^ oo
1 1






1^ 00
CO CO
1




O CTi
CM r—
t 1





^J- CO
1 1






r>- LO
LO LO




LU 1_I_






0
CM
1




1^-.
1



o
CM







^.
CM





f^^
1






^o
CO





00
r7





CO
1






LO
LO




CD


































f — V
LO
•a
c~
ro

CO

r>
r—

t/1
Q^
• ^

•r"
O
U-

««r-
-^-
3
OO
T3
O)
CO
O
a.
0
O-
•v — '

o


0
S-
ro
^
O)
o
oo
                                 <-o ID LO CM *a-  CM CO CM CM O~i
                                 i— i— CM i— CM CM i—
                                  I    I   I    I   I   I   I
                                 I— LO VO
                                               CO CO LO
                                 *± ^1- CO CTi 00 OO VO
                                 CO CO «3- CO CO CO CO
                                  I    I   I   I   I   I   I
                                 I--. t£l CT) O CM CO CM
                                 i— CM CM CVJ CM CM C\J
                                  I   I   I   I   I   I   I
                                 CM CO <£>
                                               LO LO
                                 C\J CTl l~~ O~l  C^ LO •*
                                 LO ^J- LO «d-  CO CO CO
                                  I    I   I   I   III
                                 CT> OI CO -vT  ^± CM i—
                                 CM CO •* CO  CM CM C\J
                                  I    I   I   I   III
                                    CM CO CO «3- <=T LO
                                    i—    r— CM CM CM
                                    CO  C_3 Q
-a
 c:
 ro
                                                              O)
                                                              o
                                                              O)
                                                              o
                                                              s_
                                                              Q.
•1-3.—^
 O Ol
 S_ (J
D_ ro

    a.
 • rv
 t/1 C
 C~ *r~-
 o
••- o
-t-> •!-
•i- S-
T3 ro
 c: c
 o oi
 o o
    co
 c  QJ
•r-  (/I
-M  O
 10  a.
•r-  O
 x  s_
 O)  Q-
 I +->
   •r—
    S

LU  O)
    CO
    3
 • ri
 CO T3

 O    o
 Ol  S-
T3 0.
 O)

 Q. CO
 O) ~O
 s-  c
Q-  O
269

-------
D.   Conclusions and Recommendation

     1.  Conclusions

         a.  From  predeveloped  to  existing  conditions,  discharges at  key
             points in the Cabin  Branch watershed  have  increased  as much as
             89  percent  for the  2-year  frequency, 50  percent for  the  10-
             year frequency, and 38 percent for the 100-year frequency.

         b.  From  existing  to  projected  conditions,  discharges  have  been
             estimated to  increase as much  as  34  percent, 31  percent,  and
             21 percent for the same frequencies.

         c.  The  largest  percent  increases  generally  occur  in  the  upper
             portion of the watershed.

         d.  The  Snouffers  School  Road  and Goshen  Road  crossings of  the
             main  stem  have  a  significant  impact  on the  peak of  the more
             frequent discharges.

         e.  Proposed stormwater management  facilities  located  in  the  upper
             portion of the  watershed generally reduce the  peak  discharges
             throughout the main stem to the downstream limit of the study.

         f.  Scenario  7 provides  control  of  the  2-,   10-,  and  100-year
             frequency discharges  to  existing  conditions  from  the  proposed
             facilities  to  the watershed  outfall  and  hydraulically  out-
             performs the other scenarios.

         g.  Scenario 5 provides more control for  the  2-year  frequency than
             Scenario  1.   The  control  is approximately  equal  for  the  10-
             and  100-year  frequencies.   However,   Scenario  1  protects  an
             additional reach  of  about  one-half  mile  of  Cabin  Branch  as
             compared to Scenario 5.

         h.  Scenario 10 provides  control to a  point  about half-way between
             existing  conditions  and predeveloped conditions  for   2-,  10-
             and  100-year  frequencies  for  key points  downstream  of  key
             point A.

         i.  Discharge  control   to at  least  the  existing  conditions  is
             desirable for  Cabin  Branch  to prevent  higher  velocities  and
             erosion potential.

     2.  Recommendation

         It  is  recommended  that three SWM  facilities  be  designed  and con-
structed in the Cabin Branch watershed.   The facilities would be  located at
sites 1, 3,  and  5.   Such a  combination  of facilities would control  the  2-,
10-, and 100-year discharges to a  point about half-way  between  existing  and
predeveloped conditions for all  key points downstream of key point  A.


                                    270

-------
     If  for  any reason  only  one SWM facility  can  be constructed  in Cabin
Branch,  the facility should be  located at site  1.

E.   Epilog

         The  development process  is dynamic  and the  implementation  of a
comprehensive  stormwater management system  must be  flexible.   The Cabin
Branch  study  recommended construction of three  regional  facilities  and  one
regional facility  was being  designed  while the  study was  being conducted.
To  date two  regional  SWM facilities have  been  constructed  (10A  &  Site  3),
Site 5  is  about to be designed as a County CIP  project and  Site  1  is being
planned  as on-site controls.
The work described in this paper was not funded  by  the  U.S.  Environmental
Protection Agency.  The contents do not necessarily reflect  the  views  of the
Agency and no official  endorsement should be inferred.

                                    271

-------
            STORM   SEWER OPTIMUM   DESIGN


                        DONG  HOANG M-Sc.P-E.
 INTRODUCTION.
 One of the main challenge the engineer is facing  is  to  insure
 that the design is technically sound and competitive  in  terms
 of costs. In other words, the optimization of a design  is  one  of
 his major concerns.The  optimization is becoming urgent and cri-
 tical  in the today world where the resources are  scarce.

 This paper presents the highlights of a computer  model called
 the STORM SEWER OPTIMUM DESIGN whose objectives are  to provide
 the least expensive design.

 To evaluate the water  runoff  for  storm  sewer  design the rational
 formula < Q=CiA ) was  the most often used.

 In order to use the  formula a set  of rainfall  intensity—duration
 -frequency  < IDF  ) curves relating  the rainfall  intensity to the
 time and return period  should be developed.

 The IDF curves are constructed based on the extreme values of
 the partial duration series and the rainfall record  length*con-
 sequently they yield conservative values/in term  of  peak dis-
 charge rate/ when compared to the actual storms.

 The IDF curves shape corresponds to the immediate peak  situation
 which  unlikely happens  in reality. The shape of the  IDF curves
 has also the effect of reducing  the flow. As matter  of fact/more
 water  is absorbed  by  the infiltration when the overland  water  is
 flowing  at  peak over  a soil  whose infiltration potential  is  re-
 latively  high  at the  beginning of the rain.

 Besides, the  runoff coefficient*C» is too simplistic to reflect  a
 very complicated situation.

 In  general/the  method  of evaluating the  overland   flow with the
rational formula is too simple to  reasonably  represent the real
s i tuat ion.


                               272

-------
 Finally, the  design process, including  the  cost computation, is  la-
 borious  to  the extent that  it  discourages the effort  to  explore
 all the  situations in order  to  arrive  to  the optimum  design.

 Considering  all  the above facts,the  ideal  design method  should:

 1. be capable  of  making use of the real rainfall  instead  of  eva-
   luating it  from  the IDF curves
 2. combine the  engineering and economic aspects  simultaneously
 3. use  method  of  computation which results  in  realistic estima-
   tion of the  overland flow and the pipe  flow
 4. be capable  of  exploring all the situations  quickly to arrive
   to the optimum design.

 The  STORM SEWER DESIGN OPTIMIZATION is developed  according  to
 the  four points exposed  above.

 The model is  composed of 2 parts. l.PART ONE  : HYDROGRAPH COMPU-
 TATION 2. PART TWO  :  DESIGN.
               PART  ONE :  HYDROGRAPH COMPUTATION
 Once the sewer  system  is  laid out, it is necessary  to  compute
 the runoff from each subcatchment tributary to the correspon-
 ding sewer line.The runoff is computed for each rainfall in-
 crement in the  form of  hydrograph based on the infiltration,
 the imperviousness condition,the ground slope,the  Manning's
 coefficient of  the subcatchment and  the? shape, slope, length,
 Manning's coefficient  of  the  gutter  conveying the  runoff to
 the manhole.

 In  the  case of designing  the  extension of an existing  sewer sys-
 tem the model computes  the combined  hydrographs which  are gene-
 rated by the combination  of the hydrographs of the  upstream
 basins  and  the hydrographs of the area whose seu»er  system is to
 be  designed.

 Another task  performed by  the model  is to create the node and
 line numbers  system.
                    PART TWO  : DESIGN
Using the hydrographs just created,  the  second  part of the  model
will compute  the design-flows.

                               273

-------
  Given the ground profile,t he minimum  cover-, the  maximum invert
  elevations, the underground constraints, the  minimum pipe size S<
  slope.- the sewer construction elements  prices, the  design objec-
  tives are to find out the opt; muni  s 3 ope, the  pipe  size, the- pipe
  class &( bedding types corresponding to  the  least  sewer cost to
  carry the design flow. The model makes  sure  that  the design
  flow at the  current time step is the greatest design flow.

  The  design process of a line  consists of varying  the slope by a
  constant  amount corresponding to the fixed drop increment a num-
  ber  of  times  consistent with  the drop availab 1e. The drop  avail-
  able  is the  difference  between the  downstream invert elevation
  of the  line  (if a  minimum slope  is  passed through  its upstream
  invert) and  its maximum allowable downstream invert elevation.

  For  each  slope  a  diameter is  calculated to carry  the design
  flow  .Corresponding  to  the  computed diameter/two  immediate
  surrounding  commmercial  pipe  diameters are selected.With  each
  commercial pipe selected recompute  the slope to carry  the  design
  flow. This slope should  create the downstream invert elevation
  falling in between  two  values: one is the upper limit  (if  speci-
  fied), the other is  the  maximum allowable downstream invert ele-
 vation.  If this  condition is not  met/ using the slope which  makes
 the downstream  invert elevation  match  with the maximum  allowable
 one. Under either cireumstances, onee the commercial  pipe diameter
 and the corresponding slope to carry  the design  flow  are  found*
 without conflicting  the  underground constraints»the  construction
 cost of the line is  computed.  By  comparison the cost  correspon-
 ding to two commercial pipe diameters>the selection  of  the  com-
 mercial pipe diameter corresponding to  the least cost can  be
 mad e.

 The same process is  repeated  with a different slope which  is
 equal to the  previous slope plus  the slope increment corres-
 ponding  to the  drop  increment. The least  cost  between the pre-
 vious and  the actual and the  associated, elements are saved.
 After varying the slopes a number of times/the optimum slope
 and diameter  can be  found to  carry  the  design flow without
 conflicting  the underground constraints.

 The model  use the Manning formula for design. For a time step/ the
 model  computes the flow out and the flow  stored  in the line by
 solving  for the  depth such that the change in the  storage  is
 equal  to the  difference  in the flow in and flow  out. The Newton-
 Raphson  technique  is again used to find the water  depth and
 flow.

The model computes  the transition width/the earth load/ the  traf-
 fic load/ if any/ to  figure outvthe type  of pipe and the  type  of
bedding. Comp1ying with the design procedure in force/and the
unit prices list/the  model computes  the  sewer construction  cost.
                               274

-------
Once  the  design is done, it is advisable to check the  behaviour
of  the  system  using the SWMM model, or any similar model with
the same  storm used for design.Flows adjustment factors are
provided  to  adjust the design flows  and consequent1y,the  slope
and pipe  diameter, in any  desired proportion to the maximum
flows generated  by the checking  model (SWMM or any similar
mode Is).
   MODEL APPLICATION
 The basin served by the sewer  system  to  be  designed  covers an
 area of 87.72 Ac with the imperviousness  factor  equal  to 397.,
 and the average ground slope of O. O15. The Manning's  factor*
 surface storage and Morton' coefficients  are as  follows:
 Manning factor   Surface storage
 Imper-  Per-
 vious   vious
 area     area
                                Horton  coefficients
Imper-
vious
area
IN
Per-
vious
area
IN
Max imum
rate
IN/HR
                                   Infi1trati on
                                   Minimum Decay
                                   rate     rate

                                    IN/HR   I/SEC
                                        Max imum
                                        allowed

                                          IN
  O13
. 025
. O62
184
3. OO
. 52  . 00115   12. 00
The  total  length of the system is 67OO Ft.Other informations
concerning  the sewer system can be found in the output  listings*
The  rainfall  used for design is from the 1O—year recurrency  IDF
curves  of  the City of Portland.Rainfal1 intensities are  recorded
for  every  5-minutes-Only the first 1O rainfall steps  are used
for  design. The rainfall is labelled:  1OIDF6-OO-5. It stands  for
1O year recurrency storm*6 hours duration*immediate peak &  5
minutes interval.

For  comparision purpose (with the rational  method) synthetic
rainfall is used instead of real rainfall. The output  listings
are  self explanatory.
  THE OPTIMUM DESIGN  METHOD VERSUS THE RATIONAL METHOD
To avoid spending  time  to  find the slopes to be used in the
storm sewer design with  the  Rational  Method*the optimum slopes
system created by  the Optimum Design  Method  is used.
*See Figure 1, Tables 1-3.
                               275

-------
By  doing this/a  significant  amount of
design  time with  the Rational  Method.
time is  saved from  the
In reality.it  is  impossible  that the optimum slopes  system can
be found  in the practice of  the Rational  Method because the
design  process/inc1uding the  cost computation/is  laborious to
the extent that it  discourages the effort of exhausting all
the situation  to  reach the optimum design.
           LEGEND

     JK32O - MANMCX.C NUMKft • J1OZO
     Km - scvuoi Lint Nuuae* • 2020
     K>,t - BHANO4 OAOCft • IO. M1CAAACHY
                   Figure 1.  Storm Sewer System

                                  276

-------
          Table 2.  Elemental Quantity of the Sewer System

            DIAMETER                 LENGTH
                  CLASS I CLASS 3 CLASS II CLASS III CLASS IV CLASS V
                 (FT)    (FT)    (FT)     (FT)    (FT)
              8
              IO
              12
              15
              18
              24
              27
              30
              36
 200.
 750.
  O.
  O
  O
  O.
  O
  O.
  O
0.
0
O
O
0
O.
O
0
0
0.
O.
4OO
&5O
O
O
0
O
0
O.
0.
750.
75O.
aoo.
O.
O.
0.
0
0
0
0
0
250
150
200.
1 5OO
3OO
         0
         0
         O
         0
         O
         O
         O
         0.
         O
        PAVEMENT PAVEMENT EXCAVATION DACKFILL
               BASE          CBANOLAH
         VD2    Y03     VD3     VD3
          3043
                      2? 4 52
                 BEDD1NC
               D-c    A
               V03   V03
 FOUNDATION
STADIL1ZATION
   VD3
 MANHOLE
3FT-SECTN
   EA
The C—factor of the Rational  Formula associated  with the combi-
ned area  is  a weighted  factor between 2 constituant areas (right
& left) with 2 associated  imperviousness  factors.Examp1e:the
area  served  by manhole  3036  is equal to 1. 52 Ac  =  83 •«- . 69
The C-factor of area 1. 52  Ac  is:
. 83X. 35
                               69X. 4O
                                     -=. 37
                           1. 5;
For comparisi on purpose,the  sewer system characteristics desig-
ned by the  Rational Method and  the Optimum Design  Method are
checked by  the SWMM model with  the same rainfall  1OIDF6-OO-5
used for  design.
 CONCLUSION
 1. The Optimum Design Method, based  on  the principle of  the
   least sewer construction cost, produces the optimum sewer
   system characteristics which  is  never realized in the  prac-
   tice of storm  sewer design with  the Rational Method

 2. The Optimum Design Method can  make  the sewer system  fully
   used. Meanwh i le» even using the  optimum slopes system/  the  Ra-
   tional Method  yields a sewer  system which is over designed
   about 67X  (average percentage  of  flow occupancy is 33%)
                                278

-------
Table 1.  Sewer System Characteristics  and  Sewer  Cost
LINE
NUMBER
DO36
2O36
1036

2034
1034

2032
1032

5O3O
403O
3O3O
SO3O
103O

4O24
3O24
2O24
1O24

3028
2028
1O28

3038
2O38
1O38

2O26
1O26

2O22
1O22

302O
2020
1O2O

PIPE PIPE
DIAMETER SLOPE
(INI (FT/FT)
IO O OO264
12 0 OO933
15 O. OO391

12. O. 002OO
15 O. OO4O9

18 O. O0817
IB. O. 01271

12 O. OO536
15 O. OO572
15 O. Ol 1O1
15 O. O1927
18. O. OO93O

24. O. OO872
27 O. OO678
30 O. OO339
3O O. OO633

10 0. OI480
12 O. O1972
13. O. O1332

8. 0. O1482
IO O O1424
12. O. Oil 36

18. O. OI488
18. O. 02786

3O. O. O1339
36. O. OO776

3O. O. O1373
3O. O. O1372
3O. 0. O17OO

INVERT PIPE TRENCH DESIGN VELOCITY
ELEVATION LENCTH WIDTH FLOW FLOW-FULL
(FT) (FTI (FT) (CFS) (FPS)
14821 3OO 00 2.5 1 13 206
146.34 2OO OO 25 3 44 4.38
145. 16 2OO OO 2. 5 4 97 4 05
BRANCH NUMBER 36 COST - 2O963
144. 5O 25O. OO 2. 5 I. 59 2. O3
143.07 35000 2.5 4.13 3.37
BRANCH NUMBER 34 COST - 221OO
141.43 2OO.OO 2.9 9. 5O 5.37
138.26 250. OO 2.9 11.84 6. 7O
BRANCH NUMBER 32 COST - 21O94.
149.93 20O OO 2.5 2 61 3.32
148. 78 2OO OO 2. 5 4 89 3. 98
146 O3 25O. OO 2. 3 6. 7B 5 52
142.18 2OO OO 2.5 8.97 7.31
140.28 200.00 2.9 IO. 24 5.79
BRANCH NUMBER 3O COST - 38O23.
136.93 ISO. OO 3.3 21.13 6.72
135. 59 2OO OO 3. 8 25. 3O 6. 41
133. 97 3OO. OO 4. 1 3O. 12 6. 13
132.71 2OO.OO 4.1 32.63 6 64
BRANCH NUMBER 24 COST - 66270
151.04 2OO.OO 23 2 67 4.89
146. 11 23O OO 2 3 5. OO 6. 37
143.41 20O.OO 25 7 31 6 12
BRANCH NUMBER 28 COST - 2O796
149.04 2OO.OO 23 1.47 4.21
143.48 25O. OO 25 2.61 4.79
142.64 23O OO 25 3. 8O 4.83
BRANCH NUMBER 38 COST - 21O14
139.66 2OO.OO 29 12.81 7.23
134. Of 2OO.OO 2.9 17.33 9.92
BRANCH NUMBER 26 COST - 17233.
128. 69 3OO. OO 4. 1 47. 46 9. 67
126. 36 3OO. OO 4. 7 38. 73 8. 31
BRANCH NUMBER 22 COST - 33491.
122. 24 3OO. OO 4.1 48. O9 9. 79
119. O9 2OO.OO 41 31.43 IO. 48
113.69 2OO.OO 41 33.48 IO. 89
BRANCH NUMBER 2O COST - 36O81
PIPE TYPE OF SEUER
CLASS BEDDING COST
2 C 7656
II 0 6424
II n 6883

III B 9113
III B 12987

III B 9233.
IV B 11839

I I B 63O5
II B 6856
II B 8371
III B 7181.
Ill B 9110

IV B IOO43.
IV B 15O66
IV B 24623
IV B 16539

2 C 5368
III B 8137
III B 7271

2 C 3305.
2 C 73O2
III B S4O7

III B 8318.
Ill B 8913.

IV B 24313.
IV B 28976.

IV B 23487.
IV B 16O78.
IV B 16316.

           SEWER SYSTEM TOTAL COST -
                      277

-------
     FROM
NODE  NUMBER   LINE
     TO
NODE  NUMBER   NUMBER
                         Table 3.  Comparison

                      PIPE DIAMETER  
-------
  3. The Optimum  Design Method can use the real  rainfall  for
    design

  4. The Optimum  Design Method which calculates  the  costs of the
    sewer lines*sewer  branches and the sewer system as a whole
    makes the Cost-Benefit Analysis of alternative  routes easy
    and effective

  5. The Optimum  Design Method Which computes and  lists the ele-
    mental quantities  regarding  the sewer construction makes the
    preparation  work for  bidding accurate and fast

  6. The Optimum  Design Method is a very  effective tool  for plan-
    ning as well

  7. The Optimum Design Method with 2 incorporated features :
    1. the automatic way to  number manholes and  sewer lines 2. the
    capability to design  the  extension of an existing  sewer sys-
    tem/makes the preparation work less  extensive and  the  storm
    sewer design more  flexible

  B. The Optimum Design Method  is  very  fast and  effective. It can
   be  handled by people whose engineering knowledge is not nec-
   cessarily as high  as the  knowledge of the people who use the
   Rational  Method for storm  sewer  design.
The work described  in this paper was not funded by the U.S. Environmental
Protection Agency.  The contents do not necessarily reflect the views of the
Agency and no official endorsement should be  inferred.
                                280

-------
                                ATTENDEES
                           USERS GROUP MEETING
                           January 27-28, 1983
Thomas Barnwell, Jr.
     EPA
     Athens, GA
Kenneth G. Eggert
     Simons,  Li & Associates
     Fort Collins, CO
Lars Bengtsson
     McM as t er Uni ver s i ty
     Hamilton, Ontario

Vincent Berg
     Montgomery County
     Rockville, MD

Patrick Beron
     Ecole Polytechnique
     Montreal, Quebec

John Capece
     University of Florida
     Gainesvi lie
Raymond A. Ferrara
     Princeton University
     Princeton, NJ

David Firmage
     Colby College
     Waterville, ME

Efi Foufoula
     University of Florida
     Gainesville

Yvonne Froscher
     Miller & Miller
     Orlando, FL
Ivan Chou
     ESE, Inc.
     Gainesville, FL

John M. Grouse
     Greenhorne & O'Mara
     Riverdale, MD
Robert Frost
     University of Florida
     Gainesvi lie

Gary Goforth
     Espey, Huston and Associates
     Austin, TX
Robert Dickinson
     University of Florida
     Gainesvi lie

Hoang Dong
     City of Portland
     Portland, OR

Debbie Dunnam
     University of Florida
     Gainesville
Philip Gronstal
     Dallas Water Utilities
     Dallas, TX

Brendon M. Harley
     Camp, Dresser & McKee
     Waltham, MA

Bruce W. Harrington
     Maryland Water Resources
     Administration, Annapolis, MD
David Dwornik
     University of Florida
     Gainesville
James P. Heaney
     University of Florida
     Gainesville
                                      281

-------
Michael Helfrick
     Montgomery County Environmental
     Protection, Rockville, MD

Wayne C. Huber
     University of Florida
     Gainesville

Maine Hutchison
     Oklahoma Water
     Resources Board
Cheri Porter
     Miller & Miller
     Orlando, FL

Mark Robinson
     McMaster University
     Hamilton, Ontario

Ian Simmers
     Free University
     Amsterdam
William James
     McMaster University
     Hamilton, Ontario

Peter Jaffe
     Princeton University
     Princeton, NJ

Terrie Lee
     University of Florida
     Gainesville

Khlifa Maalel
     University of Florida
     Gainesville

Stanley I. Mast
     Howard, Needles, Tammen and
     Bergendoff, New York, NY

Ed McBean
     University of Waterloo
     Waterloo, Ontario

Janusz Niemczynowicz
     Lund Institute of Technology
     Lund, Sweden

Alexander Padva
     Environmental Quality Lab
     Port Charlotte, FL
Charles Simon
     Nat'l Council for Air & Stream
     Improvement, Gainesville, FL

Gary Trott
     Wiedeman & Singleton Engineers
     Atlanta, GA

William F. Walker
     Environmental Engineer
     Concord, MA

Flora Wang
     Louisiana State University
     Baton Rouge, LA

Roger K. Wells
     HMM Associates
     Raleigh, NC

Stanley Wong
     Maryland Water Resources
     Administration, Annapolis, MD

Grace Wood
     Dames & Moore
     Washington, DC

Jy S. Wu
     University of North Carolina
     Charlotte, NC
                                       282
 *U.S. GOVERNMENT PRINTING OFFICE 1963 - 659-095/0745

-------

-------
   3  ±-
   Q)  o
                     > m c
               i.3
               3
               3
               3
               CD
               3
   <
   01
   c
   u>
   o>
  00
  o
  O
               T)
               O
               S
               o

               o
                    3' 3 »
                    3 Q)

                    S.6-S
                    O 3 m
                    I    3
                    01    o
                    NJ    3

                    £    3
                    03    (t
                          3

                          S
   T3
   (D
   O
   O
CD C
O  3
O  3"
*"  '
   n
Ticc  ;:
> (B  2
 '  3  (D
CO o  n
oo <  ~
ui    o
      3
D  
-------