Sie sind auf Seite 1von 13

Copyright 2002 Water Environment Federation All Rights Reserved.

Watershed 2002
RiverSpill: A GIS-Based Real Time Transport Model for Source Water
Protection

William B. Samuels
1
, Rakesh Bahadur
1
, David E. Amstutz
1
, Jonathan Pickus
1
and
Walter Grayman
2

1
Science Applications International Corporation
1410 Spring Hill Road
McLean, VA 22102
2
W.M. Grayman Consulting Engineer
Cincinnati, OH 45229

ABSTRACT

To assist in delineating source water protection areas based on stream flow, a tool
has been developed to calculate the time of travel (based on real-time stream flow
measurements), decay, and dispersion of a pollutant introduced into surface
waters. The tool allows the user to:

! select a location on a river to introduce a chemical or biological constituent,
! track the pollutant downstream, under real-time flow conditions, to a water
supply intake,
! further track the constituent to a water filtration plant, and
! identify the overall population served by the plant.

The databases required for this tool include the Enhanced River Reach File
(ERF1), the USGS real-time stream flow measurements, and the EPA Safe
Drinking Water Information System (SDWIS). The Arcview Network Analyst
extension is used to integrate the databases and to provide the user with a tool to
quickly assess the consequences of the introduction of a chemical or biological
contaminant to the source waters (surface water) of a public water supply.

KEYWORDS

Source Water Protection, Travel Time, River Spill

INTRODUCTION

Source water is untreated water from streams, rivers, lakes, or underground
aquifers which is used to supply private wells and public drinking water. The EPA
has issued draft susceptibility guidance to States that calls for the following: The
state will segment its watershed areas based upon a four-hour travel time of
stream flow to the intake. Four hours is thought to be the emergency response
time needed to alert a water supplier of a contaminant spill in time for the supplier
to respond.

Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002
As states delineate source water protection areas for surface-water based sources
of drinking water, they may want to consider using buffer/setback zones, time-of-
travel zones and/or use modeling techniques to enhance the delineation. These
enhancements can assist states in defining "segments" for management actions.

In cooperation with the Federal Emergency Management Agency (FEMA), the
Technical Support Working Group (TSWG) has sponsored a project to develop a
software program to assess the vulnerability of public water supplies to toxic
spills based on time-of-travel calculations. The tool can be used to provide for
better protective measures; and assess and manage consequences of emergency
incidents. It can also be used to predetermine the ranges upstream from a
treatment plant that would require physical protection from an anticipated
deliberate contamination event. The prototype system, RiverSpill, has been
developed and is operational for Ohio and Utah.

METHODOLOGY

Databases

The databases required for this tool include the Enhanced River Reach File
(ERF1), the USGS real-time stream flow measurements, and the EPA Safe
Drinking Water Information System (SDWIS). The Arcview Network Analyst
extension is used to integrate the databases and to provide the user with a tool to
quickly assess the consequences of the introduction of a chemical or biological
contaminant to the source waters (surface water) of a public water supply (see
figure 1).

Figure 1 - RiverSpill components and databases

Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002
The original EPA Reach File Version 1.0 (RF1) is a vector database of
approximately 700,000 miles of streams and open waters in the conterminous
United States. The RF1 stream flow data consist of mean annual flow and 7Q10
low flow estimates made at the downstream ends of more than 60,000 transport
reaches. The digital data set ERF1 includes enhancements to the EPAs River
Reach File 1 (RF1) to ensure the hydrologic integrity of the digital reach traces
and to quantify the time of travel of river reaches and reservoirs. ERF1 was
designed to be a digital data base of river reaches capable of supporting regional
and national water-quality and river-flow modeling and transport investigations in
the water-resources community.

The stream-gauging program of the USGS is an aggregation of networks and
individual stream flow stations that originally were established for various
purposes. Approximately 5,000 of the 6,900 U.S. Geological Survey sampling
stations are equipped with telemetry to transmit data on stream flow, temperature,
and other parameters back to a database for real-time viewing via the World Wide
Web.

The locations of public water supply plants and intakes were obtained from the
EPA Safe Drinking Water Information System (SDWIS). The Safe Drinking
Water Information System (SDWIS) contains information about public water
systems including their violations of EPA's drinking water regulations. SDWIS
(Safe Drinking Water Information System) is an EPA national database storing
routine information about the nation's drinking water. The SDWIS stores the
information EPA needs to monitor approximately 175,000 public water systems.

RiverSpill Model

The integrated system calculates, locates, and maps the population at risk from the
introduction of contaminants to the public water supply (see figure 2). The Real
Time River Spill Model calculates the time of travel based on real time stream
flow measurements, decay, and dispersion of a contaminant introduced into
surface water. Prototype systems have been set up and tested for Ohio and Utah.
The databases and software are directly extensible to the rest of the US. The GIS-
based tool developed for this project consists of three primary components:

Time of Travel calculation using real-time stream flow data

The following relationship seems to be a generally accepted form for relating
velocity to flow:

V= aQ
b


This relationship is used in the QUAL2E model (EPA, 1996) and has been
adopted by several others (Leopold and Maddock, 1953). The technical issue
Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002
becomes one of determining parameters ("a" and "b") for the relationship. Several
sources were examined that provide information from which parameters could be
derived (Leopold & Maddock paper, 1953; W.E.Gates and Associates, 1982 and
Maidment, 1993). After examining the literature and fitting relationships to the
flow-velocity data in the original study by Gates(1982), it appears that a
reasonable range of values for parameter b runs from 0.15 to 0.3. This is actually
a fairly narrow range so that arbitrarily selecting a mid-value of b=0.225 would
not be unreasonable. With this value of "b" and the mean Q and V data in ERF1,
values of "a" were computed for each reach. Given the Enhanced Reach File (see
figure 3), for each reach we know: Qmean (mean flow), Vmean (mean velocity),
Tmean (mean travel time), and Length.


Figure 2. RiverSpill system architecture






Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002
Figure 3. Schematic of the Enhanced Reach File and RiverSpill Network



The steps to calculate the time of travel, based on real time flow data are as
follows:

1. Find nearest gage - retrieve real-time flow (Qrt)

2. Calculate real-time velocity (Vrt): V
rt
= aQ
rt

b
; coefficient "a" calculated
based on an analysis of USGS and EPA flow data associated with ERF1

3. Calculate Velocity Factor (VF): VF = V
rt
/V
mean


4. Scale T
mean
for each reach by VF to yield T
rt
: T
rt
= T
mean
/VF

5. Calculate time-of-travel (TOT) to intake - sum T
rt
for all reaches between
input location and intake

Time decay of the constituent based on time-of-travel to the intake

The initial mass (M
o
) of substance will be reduced to a secondary mass (M) using
the temporal decay parameters assigned to that substance.

M = (M
o
)exp
-kt


where k = decay coefficient ( = 0.693/t
1/2
), and t
1/2
= half-life of the pollutant
Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002

Dispersion of the constituent based on convective-diffusion equations for
turbulent flow

The distribution of a substance introduced to a river is governed by the
convection-diffusion equations for turbulent flow (Taylor, 1954). The equations
are simplified by assuming the molecular diffusion to be negligible, the density
gradients are small compared with the concentration gradients, and the mass flux
is proportional to the mean concentration and in the direction of decreasing
concentration. The remaining terms represent the change in concentration of the
substance with time, the convective mass transfer associated with the fluid
velocity, and the turbulent mass transfer.

Further assumptions can be made by considering the concentration of substance to
be a function only of time and along the stream dimension. The along stream
velocity is considered constant (the average flow over the uniform cross section).
These simplifying assumptions are wholly consistent with making use of the
available data from the stream gauging stations. In summary, the assumptions are
as follows:

! The distribution of a substance introduced to a river is governed by the
convection-diffusion equations for turbulent flow.

! Consider the concentration of substance to be a function only of time and
along the stream dimension.

! An instantaneous release is modeled by introducing a finite amount of
substance at a selected location.

! If the initial mass (M) of the substance is released over the cross section
(A) and allowed to spread diffusely, the central concentration (C) of
substance will be given by:

C = 0.003* M/(A(UR
5/6
t)
1/2
)

where U = velocity (V
rt
), R = hydraulic
radius, t = travel time

! the hydraulic radius (R) is equal to river stage

RESULTS

Modeling of river flow, and more especially the dispersion of substances within
river flow, is a challenging task. The task is challenging because there are to few
observations to support the equations representing the involved physics.
Observations are lacking largely because of the complex stream geometries and
variations in streambed characteristics. The difficulties in modeling river flow are
summarized in several references (Novak, 1985).
Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002

Model results were compared to field observations to test the validity of the flow-
velocity relationship and the predicted travel times from contaminant source to
water supply intake.

Velocity-Flow Relationship

The skill of the flow-velocity relationship was tested by accessing real-time flow
date for a USGS gage on the Colorado River and calculating the resulting velocity
using the power function described previously. The coefficients (a and b)
used in the equation were specific for the ERF1 reach associated with the gage
location. The velocity factor was calculated as described previously and this
factor was used to scale the mean velocity associated with a downstream reach
where another USGS real-time gage was located. This results in an estimate of
the real-time velocity at that gage. This value was input to the power function to
calculate the real-time flow. A comparison was then made between this predicted
flow and the measured flow at the gage. The difference was less than 8 percent
(see figure 4)


Figure 4. Comparison of predicted and observed flow for the Colorado
River.




Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002
Time-of-Travel Validation

The analytical model we use characterizes one-dimensional turbulent diffusion in
constant density flow. The concentration is considered to be a function only of
time and the distance along the longitudinal axis. Average velocities, calculated
from real-time values reported from river gauging stations, are applied over the
uniform cross sections. The initial condition specifies that a finite mass of
substance is released instantaneously and uniformly over the cross section.
Mathematically the initial release is from a Dirac delta function. The boundary
condition specifies that the concentration remains zero for all time after release at
infinite distances along the longitudinal axis (Ippen, 1966).

We can compare the model with the findings of a few carefully conducted dye
studies and with documented pollution events. Results from the present model
could also be compared with those from numerical models, where numerical
models have been applied. Numerical models are finely tuned to the specific
stretch of river to which they are applied. Therefore, one should not expect our
generic model, which is being applied to all rivers, to yield results which agree in
detail with the numerical models. If the models did agree then we would conclude
that the numerical model was very likely not needed in the first place.

Selection of Rivers
Because the proposed model is to be applied throughout the United States, its skill
must be assessed in the broadest sense. To accomplish this we selected rivers with
a wide geographic distribution and with a variety of flow conditions. It is
recognized of course that no single model will thoroughly address all of these
circumstances. Nevertheless, our analysis was aimed at determining just how well
or how poorly the model performed.

The US Geological Survey has assembled a large set of river observations
(Jobson, 1996) which reflect the broad geographic distribution and widely varying
flow conditions desired for model skill assessment. The data however, do not
represent the most western and southwestern states. River observations reported
by Jobson (1996) are tabulated in Appendices to the report and include original
references. The observations represent 102 cases where dye was injected in 38
rivers. Multiple dye injections were made at the same, and in some cases,
differing river locations. Dye injections were also made under varying flow
conditions. From the variety of rivers investigated, 11 were selected for the model
skill assessment study. The 11 were selected to meet the requirements for wide
geographic representation (Hydrologic Regions), and varying flow conditions.
Where possible, rivers with replicate dye injections were selected preferentially.
The rivers used for skill assessment, are shown in Figure 5. The number of
separate dye injections for each are: Amite (1), Antietam (6), Bighorn (2),
Chattahoochee (2), Clinch (4), Embarras (3), Hoosic (1), Kaskaskia (3), N. Platte
(1), Sabine (5), and Souris (6). Thus, 34 separate cases, representing 11 rivers are
used in the river model skill assessment. The number of rivers and separate cases
Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002
from each, used in the skill assessment, are considered appropriate to the purposes
of the prototype model.

The observations provided by Jobson (1996) for each dye injection site include
the distance(s) traveled, river discharge, time when dye began to arrive, time of
arrival of the peak concentration, and time of arrival of the trailing edge of dye.
For purposes of the skill assessment, the measured river discharge was compared
with the historic discharge data in the River Reach Files to determine the
coefficient "a" to be used to calculate river velocity.

Time of travel for the peak or central concentration was calculated for each
reported stretch of river. Comparisons were made for observed and modeled
travel times for the peak concentrations, versus distance traveled. The results are
remarkably very good. For example, the modeled and observed travel times for
the Antietam River are shown in figure 5. The results for the Amite River shows
the convergence of the model and observed travel times with a subsequent
divergence after 100 km distance.

Figure 5. Model skill assessment - comparison of observed and modeled
results

The results from the comparison of observations and model output are
summarized in the Regression Diagram of the observed and modeled peak
concentrations . The regression diagram illustrates excellent agreement ( r = 0.87)
between the observed and modeled data sets. The 34 separate cases, from the 11
rivers represent more the one hundred separate distances traveled (see figure 5).
Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002

DISCUSSION

The identification of potential contaminant sources and determination of the
significance of those sources to a public water supply, in terms of spatial location
and transport of the contaminant to the intake, is an important component of the
State Source Water Assessment and Protection Programs Guidance Document
promulgated by the EPA. This program is necessary to meet requirements of the
Safe Drinking Water Act Amendments of 1996 (P.L.104-182).

According to Appendix C (of EPAs Guidance document, EPA, 1997), Enhancing
Topographic Delineations for Source Water Protection Areas, the streamflow
time-of-travel (TOT) approach facilitates heightened management of those stream
reaches most critical to protecting drinking water intakes from upstream
significant potential sources of contamination. This method also enhances
delineations of source water protection areas by facilitating spill- and other
emergency-response activities. This method does not delineate protection zones;
rather, it calculates the TOT of flow in a stream between a drinking water intake
and a point(s) upstream. It is the streamflow TOT between the intake and the
upstream point of interest that provides the opportunity for managers to enhance
protection management of long-term potential contaminant sources and to
respond to a contamination event. Use of this method would be of greatest
importance for drinking water utilities tapping rivers or reservoirs designated for
commercial transport and municipal and industrial wastewater discharges. Water
quality flow models provide a means through which specific hydrologic,
geographic, and water quality parameters can be used to estimate the travel time
for a contaminant introduced into a river to reach a drinking-water intake and to
estimate the level of contamination at that intake.

The RiverSpill software tool has been developed to allow analysts to rapidly
assess the time-of-travel and concentration of a toxic substance to a public water
supply intake. An example of the output of this software tool is described in the
following paragraph.

Figure 6 below shows the RiverSpill output for a hypothetical spill on the
Maumee River in Western Ohio. Using the Network Analyst, "Find Closest
Facility" function, the nearest gage to the spill site is determined. The time-of-
travel for each reach, based on the real-time velocity is calculated. The "Find
Closest Facility" is invoked again, using the public water supply intakes as the
facilities and the time-of-travel attribute as the cost field. Public water supplies
along the flow path are displayed on the map and in an associated table. Personnel
responsible for operation of the water supply facilities, and their telephone
numbers, are also then the intake and its associated table record are colored red to
indicate a warning. Figure 7 shows a similar RiverSpill output for a hypothetical
spill on the Provo River, Utah.

Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002
Figure 6. RiverSpill output for the Maumee River, Ohio


Figure 7. RiverSpill output for the Provo River, Utah




Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002
CONCLUSION

Under the SDWA, states are required to develop comprehensive Source Water
Assessment Programs (SWAP) that will:

! identify the areas that supply public tap water;
! inventory contaminants and assess water system susceptibility to
contamination;
! inform the public of the results.

The work described here provides a tool (RiverSpill) to analyze the susceptibility
of surface water intakes to contamination from toxic spills. RiverSpill is a
Geographic Information System (GIS)-based software tool with integrated data
base capability that is used to track and model the flow and concentration of
contaminants in surface waters. A prototype system has been developed Ohio and
Utah and is fully operational for public water supplies in these two States. The
system can be extended for the rest of the US.

A companion model, PipelineNet, has been developed to trace the path and
concentration of toxic substances released from any point within a water pipeline
network. PipelineNet has been applied successfully to Salt Lake City.

ACKNOWLEDGEMENTS

This project was funded by the Technical Support Working Group (TSWG),
Infrastructure Protection (IP) sub-group. Martha Snyderwine is the TSWG
program manager for the IP sub-group. Paul Bryant, Federal Emergency
Management Agency was the government technical manager for this project. Dr.
Rolf Deininger, University of Michigan School of Public Health provided
technical guidance on potential sources of public water supply contamination.

REFERENCES

EPA, 1997., State Source Water Assessment and Protection Programs Guidance,
United States Environmental Protection Agency, Office of Water, EPA 816-R-
97-009, August 1997

Ippen, Arthur T. Estuary and Coastline Hydrodynamics, Engineering Societies
Monographs, McGraw-Hill Book Company, New York, 1966. 744pp.

Jobson, Harvey E. Prediction of Traveltime and Longitudinal Dispersion in Rivers
and Streams, USGS Water-Resources Investigations Report 96-4013, 1996, 60pp.

Leopold, L.L. and T. Maddock, Jr., The Hydrologic Geometry of Stream
Channels and Some Physiographic Implications, USGS Professional Paper No.
252, 1953.
Copyright 2002 Water Environment Federation All Rights Reserved.
Watershed 2002

Maidment, David, ed., Handbook of Hydrology, McGraw-Hill Book Company,
New York, 1993.

Novak, P. (ed.) Developments in Hydraulic Engineering-3, Elsevier Applied
Science Publishers, LTD. Essex, England, 1985, 316pp.

Taylor, G. I., Dispersion of Matter in Turbulent Flow through a Pipe, Proc. Roy.
Soc. London (A), Vol. 223, 1954.

Wahl, K.L. Wilbert O. Thomas, Jr., and Robert M. Hirsch, Stream-Gaging
Program of the U.S. Geological Survey, U.S. Geological Survey Circular 1123,
Reston, Virginia, 1995

W.E. Gates and Associates, Inc., Estimation of Streamflows and the Reach File,
Prepared for the U.S. Environmental Protection Agency, Monitoring Branch,
Washington, DC

http://www.epa.gov/QUAL2E_WINDOWS Enhanced Stream Water Quality
Model, Windows (QUAL2E)

http://baltimore.umbc.edu/mdnsdi/metadata/rf1.html U.S. EPA Reach File 1
(RF1) for the Conterminous United States in BASINS, Metadata Record

http://www.epa.gov/owow/wtr1/NPS/gis/reach.html Office of Water River Reach
Files

http://www.epa.gov/OGWDW/datab/sfed.html SDWIS home page

Das könnte Ihnen auch gefallen