Professional Documents
Culture Documents
Abstract
Efforts to manage National Forests in the US for wood production, while protecting water quality, are
currently constrained by models that do not address temporal dynamics of variable non-point source
(NPS) areas. NPS areas are diffuse sources of contaminants contributed by runoff or percolation as a
result of different land use activities. A distributed process-based model for predicting potential NPS
areas prone to generating runoff was applied to the 72 km2 Pete King watershed located in the
Clearwater National Forest (CNF) in central Idaho, USA. This Geographic Information System gridbased modeling approach integrates the Soil Moisture Routing (SMR) model with probabilistic
analysis. The SMR model is a daily water balance model that simulates the hydrology of forested
watersheds by combining real or stochastically generated climate data, a digital elevation model, soil,
and land use data. The probabilistic analysis involves Monte Carlo simulation, which is used to
incorporate the variability of input parameters and account for uncertainties associated with the
prediction of NPS areas. A one-year simulation was performed to examine if NPS areas prone to
generating runoff would change spatially and temporally. Sample results of this integrated approach of
dynamic prediction of saturated areas prone to generating runoff suggested that due to seasonal
variability of saturated areas the potential NPS of water pollution also varies spatially and temporally.
Results from simulated model outputs should help decision-making in effective forest management
and planning by mapping or delineating NPS areas likely to transport contaminants to perennial
surface water bodies.
Key words: GIS; water quality; water balance; distributed modeling, hydrologically sensitive area; and
Monte Carlo simulation
Introduction
Unlike point source (PS) pollution discharged from defined sources such as industrial and sewage
treatment plants, in forested watersheds, NPS pollution comes from many diffuse sources initiated by
human activities such as recreation, road building, logging, and other intensive forest management
practices. As a consequence these activities have altered the natural settings of land surfaces, thus
increasing runoff and overland flow. As the runoff moves down the slope, sediments and other
dissolved harmful particles are transported to perennial surface water bodies, which impact water
quality, water quantity, and biodiversity of aquatic life (EPA, 1997).
Sound and innovative management of forest watersheds is necessary to protect water quality. In the
U.S.A., for instance, the number of impaired water bodies placed on the 303 (d) list is still growing
(EPA, 1992). The U.S. Environmental Protection Agency has identified NPS pollutants as major
contributors to the ecological impairment of these water bodies. A need for identification and
implementation of management practices capable of controlling acceptable Total Maximum Daily
Loads (TMDLs) and minimizing NPS pollution is essential to prevent environmental damage at all
stages of management implementation. Understanding, controlling and mitigating the impacts from
existing managed watersheds usually is addressed through Best Management Practices (BMP). The
initiation of BMPs by the U.S. government encompasses a variety of activities that can provide better
guidance for improving water quality and reducing runoff while meeting TMDL requirements. However,
it is not an easy task to identify dynamic and variable areas that expand and contract through time
while different possible scenarios of land use are implemented.
Walter et al. (2000) referred to variable source areas (VSA) as saturated areas that are initiated
primarily from runoff, while falling rain on these areas becomes overland flow. In each watershed,
regions exist that are more susceptible to producing runoff than others, often considered as
Information systems and decision support
BALWOIS 2004
hydrologically active areas (HAA). When a pollutant is present in these HAA, the area is considered a
hydrologically sensitive area (HSA). In this paper, the term HSA refers to regions in the watershed that
are prone to generating runoff and are hydrologically sensitive with respect to transport of sediments
and other dissolved harmful particles to perennial surface water bodies.
To quantitatively identify HSA requires a distributed process-based modeling approach that can
predict the spatio-temporal patterns of these areas. In recent years the use of Remote Sensing (RS)
and GIS for distributed process-based modeling has increased. Many different spatial approaches
have been proposed based on different assumptions (Beven & Kirkby, 1979; Grayson et al., 1992a,b;
Wigmosta et al., 1994; Boll et al., 1998; Frankenberger et al., 1999). Beven & Kirkby (1979)
developed a physically-based hydrological model (TOPMODEL) driven by topographic slope and the
transmissivity of the soil. The lateral flux of water in their model is related to the upslope contributing
area. Further development of the topographically driven model resulted in the development of steadystate and dynamic models (OLoughlin, 1986). The general assumptions for these steady-state and
dynamic models are that the water table is nearly parallel to the soil surface; the saturated hydraulic
conductivity of the soil decreases exponentially with depth; and the movement of water is spatially
uniform. An advantage of using dynamic modeling is that the input variables can be varied for different
applications and used to simulate different conditions due to management. A disadvantage is that
variables that are difficult to obtain, such as soil depth, soil porosity, and hydraulic conductivity, are
required.
The function of the SMR model developed by Boll et al. (1998) and Frankenberger et al. (1999) is
appealing because of its simplicity and the model may be parameterized using publicly available data.
Implementation of SMR to assess water quality risk in the context of HSA has shown satisfactory
results in studies reported by Boll et al. (1998), Frankenberger et al. (1999), Walter et al. (2000), and
Brooks (2003). The SMR model is a dynamic process-based hydrologic model that implements a
water balance at each grid point through space and time. Soil moisture content is predicted for inputs
that are either individual rainfall events or long-term sequences of storms. The inputs can be actual
precipitation records or predicted records. Moisture above saturation results in surface runoff.
However, modeling spatially distributed processes requires knowledge of the regional heterogeneity of
the physiographical catchment characteristics such as climate, soil, topography and land use.
Because detailed mapping of regional heterogeneity in forested and mountainous areas is rare, an
alternative representation of the uncertainty and variability inherent in these parameters is needed.
In this study, we expand on the previous work that modeled HSA to assessing water quality risk by
using a more probabilistic framework for input parameters that are highly variable in space. Therefore,
this study implements distributed process-based modeling for predicting potential NPS areas prone to
generating runoff by coupling the SMR model with probabilistic analysis. Required input variables that
exhibit various levels of uncertainty due to measurement logistics and scale dependencies are
estimated by Monte Carlo simulation (Burrough & McDonnell, 1998; Malczewski, 1999). This
incorporates uncertainty exhibited by distributed input variables and allows estimation of probabilities
of occurrence. A brief overview of the components used in this integrated distributed process-based
model with probabilistic analysis is presented.
d i
= P (t ) i ET (t ) i + Qin.i Qout .i Li Ri
dt
(1)
BALWOIS 2004
BALWOIS 2004
1.5
6 Kilometers
Figure 1. Study area and vegetation classes in the Pete King watershed
SMR Model Input
The SMR model was calibrated for a period of 14 months (July 1995 to September 1996) and then
used to simulate the HSA for the same year. Input for the SMR model consisted of the 7.5-minute U.S.
Geological Survey DEM, land cover derived from a satellite image classification, observed daily
weather data, and soils data from the Soil Survey Geographic Database (SSURGO) (USDA-NRCS,
2001).
The soil type for the entire study area was loam, and the soil profile was divided into three horizons (A,
B, and E). Soil moisture data were taken from Flanagan & Livingston (1995) and Saxton et al. (1986)
(see Table 1). Spatially distributed soil depth values did not exist for the Pete King watershed.
Therefore, to more realistically represent soil depth across the watershed the compound topographic
index (CTI) was used to generate spatially distributed soil values (Gessler et al., 1995; 2000). The
CTI, often referred to as the steady-state wetness index (Beven & Kirkby, 1979; Moore et al., 1991), is
defined as:
CTI = ln (As/tan )
(2)
2
Soil Layer
A
B
160
50
E
30
Reference
Calibrated
0.04
0.04
0.04
Calibrated
0.58
0.37
0.11
0.52
0.28
0.09
0.42
0.27
0.08
BALWOIS 2004
4
3
2
0
Rainfall (cm/day)
Daily weather data from Fenn Ranger Station (4608'39" N 11535'41" W; elevation 487 m), Idaho
were used for the calibration and the simulation period (Figure 2). The temperature was adjusted for
elevation by a lapse rate of 0.64 C per 100 meters. A total of 20 missing values for the precipitation
data and 13 missing values for the temperature data for 1995-96 were interpolated for distance and
elevation differences from nearby weather stations (Dworshak Hatchery - Idaho, Kamiah - Idaho,
Kooskia - Idaho). Daily weather inputs (NOAA, 2001) consisted of minimum and maximum
temperature, and total precipitation.
09/01/1995
11/13/1995
01/25/1996
04/07/1996
06/19/1996
08/31/1996
Time in days
40
a)
20
10
0
-20
Temperature (C)
30
Maximum
Minimum
09/01/1995
11/13/1995
01/25/1996
04/07/1996
06/19/1996
08/31/1996
b)
0.6
0.4
0.0
0.2
PET (cm/day)
0.8
Time in days
09/01/1995
11/13/1995
01/25/1996
04/07/1996
Time in days
06/19/1996
08/31/1996
c)
Figure 2. Climatic data used in the SMR model: a) Hyetograph for the Pete King watershed from the
storm events 1995-96 b) observed maximum and minimum temperature, and c) potential
evapotranspiration (estimated by the method of Hargreaves (1995))
Potential evapotranspiration was estimated using the Hargreaves method (Hargreaves & Samani,
1985). A cloud-free Landsat Thematic Mapper image from 1995 was used to extract simple vegetation
classes for the study area to assign evapotranspiration coefficients. The land cover was classified
using an unsupervised clustering algorithm, which incorporated an iterative process of randomly
selecting cluster centers. A simple three-vegetation cover type classification was developed: clear-cut
areas, regenerated-young forest areas, and mature forest areas (Figure 1). Based on the classification
in 1995, the area consisted of 44% clear-cut, 28% regenerating young forest, and 28% mature forest.
Different evapotranspiration coefficients were used in the simulation for each landcover type.
SMR Model: Calibration
The SMR program first required a 60 to 90-day period to assure the SMR model output was
insensitive to the initial soil moisture content (Boll et al., 1998). An arbitrary 10% soil moisture content
for the watershed was assigned to initialize the moisture content (starting in July 1995). Initially, using
SSURGO soil depth, peak discharges were greatly over predicted for the calibration period. Since
SSURGO provides soil depth up to 150 cm, soil depths were increased (by 10cm increments) to
Information systems and decision support
BALWOIS 2004
10 12 14
4
Predicted
Observed
Discharge (cms)
improve the fit of the peaks between the observed and actual measurements. This process was
guided by using the CTI values described above. The final soil depth values used in the SMR model
ranged from 14 to 288 cm with a mean of 145 cm and standard deviation of 69 cm. Figure 3 shows
stream flow for the final calibrated model that was used for the simulation. Almost all stream discharge
peaks were still slightly over-predicted by the SMR model, but the general fit of the model is in close
agreement with the measured data.
09/01/1995
11/13/1995
01/25/1996
04/07/1996
06/19/1996
08/31/1996
Time in days
Figure 3. Predicted versus observed stream flow for the Pete King watershed
Mar 1996
10
20
Apr 1996
30
May 1996
10
20
Jun 1996
30
Jul 1996
10
20
Aug 1996
30
40
30
20
10
Sep 1995
Oct 1995
Nov 1995
Dec 1995
Jan 1996
Feb 1996
40
30
20
10
0
0
10
20
30
10
20
30
10
20
30
Time in days
BALWOIS 2004
SEP
OCT
NOV
DEC
JAN
FEB
MAR
APR
JUN
MAY
AUG
JUL
0.0
0.5
1.0
4 Kilometers
Probability
BALWOIS 2004
60.00%
% of HSA's
50.00%
Percent (%)
40.00%
30.00%
20.00%
10.00%
0.00%
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Month
Figure 6. Monthly percent of HSAs and runoff originating from these areas
The spatial changes of HSA by different vegetation classes (i.e., clear-cut, regeneration, forest) are
shown in Figure 7. The vertical bars in the figure represent the percentage of HSA generated from
total area of each vegetation class, while the lines represent the percentage of HSA generated from
the total watershed area. The y-axis scale on the left hand side is associated with the percentage of
HSA generated from total area of each vegetation class. The y-axis scale on the right hand side is
associated with the percentage of HSA generated from the total watershed area. The percentage of
HSA from the clear-cut is the highest in most of the months except April and May when the
percentage of HSA associated with the forest class is the highest. This is due to higher temperatures
that influence snowmelt in forested areas in late spring.
60.00%
30.00%
Clear-cut
Regeneration
Forest
% of Clear-cut area
% of Regeneration area
% of Forest area
Percent (%)
50.00%
25.00%
40.00%
20.00%
30.00%
15.00%
20.00%
10.00%
10.00%
5.00%
0.00%
0.00%
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Month
Figure 7. Monthly percent of HSAs organized by vegetation classes
Conclusion
A distributed process-based model for predicting potential NPS areas was described and applied to an
example watershed. The modeling approach was based on integration of the SMR model and Monte
Carlo simulation for generation of model inputs such as soil depth. Application of the current modeling
Information systems and decision support
BALWOIS 2004
approach in the Pete King watershed in north-central Idaho provided reasonable results. The results
confirmed findings reported by Frankenberger et al. (1998) where similar methodology was used,
however in this study we used Monte Carlo simulation and showed a new methodology of an
alternative representation of distributed parameters that contain various levels of uncertainty due to
measurement logistics or scale dependences to be used in areas where detailed mapping or regional
heterogeneity is not existent.
Results from the modeling approach showed that due to seasonal variability of saturated areas the
HSA also vary spatially and temporally. Mapping the extent of these specific locations that are
potential NPS areas of water pollution at one time of the year but which are insensitive at another may
help decision-makers to select the best combination of BMPs through optimization techniques (i.e.,
linear, goal, or integer programming) while meeting TMDL requirements.
The example application demonstrated the potential of this approach in yielding valuable site-specific
information especially in areas where distributed parameters are infrequent. We believe that this tool
will assist decision-makers to identify critical source areas that are potential NPSs of water pollution,
however, further testing of the modeling approach and field verification is necessary.
Reference
Beven, K. & Kirkby M.J. (1979) A physically based, variable contributing area model of basin
hydrology. Hydrological Sciences Bulletin, 24:43-69.
Burrough, P.A. & McDonnell R.A. (1998) Principles of Geographic Information Systems. Oxford
University Press. Oxford.
Bresler, E., Russo D., & Miller R.D. (1978) Rapid estimate of unsaturated hydraulic conductivity
function. Soil Sci. Soc. Am. Proc. 42:170-172.
Boll, J., Brooks E.S., Campbell C.R., Stockle C.O., Young S.K., Hammel J.E., & McDaniel P.A.
(1998) Progress toward development of a GIS based water quality management tool for small rural
watersheds: modification and application of a distributed model. Presented at the 1998 ASAE Annual
International Meeting in Orlando, Florida, July 12-16, Paper 982230, ASAE, 2950 Niles Road, St.
Joseph, MI 49085-9659, USA.
Brooks, E. S. (2003) Distributed hydrologic modeling of the eastern Palouse, Ph.D. dissertation:
University of Idaho, Moscow, USA
Environmental Protection Agency, (1992) National Water Quality Inventory: 1990 Report to
Congress. Office of Water. EPA 503/99-92-006.
Environmental Protection Agency, (1997) Non-point pointer No 7: Fact Sheet. EPA841-F-S-00-001.
Flanagan, D. C., & Livingston S. J. (eds.). (1995) WEPP user summary. NSERL Report No. 11, W.
Lafayette, IN: National Soil Erosion Research Laboratory.
Frankenberger, J.R., Brooks E.S., Walter M.T., Walter M.F., & Steenhuis T.S. (1999) A GIS-based
variable source area hydrology model. Hydrol. Processes, 13:804-822.
Gessler, P.E., Moore I.D., McKenzie N.J., & Ryan P.J. (1995) Soil-landscape modelling and spatial
prediction of soil attributes, INT. J. Geographic Information Systems, 9:421-432.
Gessler, P.E., Chadwick O.A., Chamran F., Althouse L., & Holmes K. (2000) Modeling soillandscape and ecosystem properties using terrain attributes. Soil Sci. Soc. Am. J. 64:2046-2056.
Gorsevski, P.V. (2002) Landslide hazard modeling using GIS. Ph.D. Dissertation: University of Idaho,
Moscow, Idaho, USA
Grayson, R.B., Moore I.D., & McMahon T.A. (1992a) Physically based hydrologic modeling 1. A
Terrain-Based Model for Investigative Purposes. Water Resources Research, 28:2639-2658
Grayson, R.B., Moore I.D., & McMahon T.A. (1992b) Physically based hydrologic modeling 2. Is the
Concept Realistic? Water Resources Research, 28:2659-2666
Hargreaves, G.H., & Samani Z.A. (1985) Reference crop evapotranspiration from temperature. Appl.
Eng. In Agric., 1:96-99.
BALWOIS 2004
Hammond, C., Hall D., Miller S., & Swetik P. (1992) Level I Stability Analyses (LISA) documentation
for version 2.0. Gen. Tech. Rep. INT-285, Ogden, UT: U.S. Department of Agriculture, Forest Service,
Intermountain Research Station, 190.
Malczewski, J. (1999) GIS and Multicriteria Decision Analysis. John Wiley & Sons, Inc. New York, NY
Moore, I.D., Grayson R.B., & Ladson A.R. (1991) Digital terrain modelling-a review of hydrological,
geomorphological and biological applications, Hydrol.Processes, 5:3-30.
National Oceanic and Atmospheric Administration, (2001) Climatological Data, Idaho, Fenn
Ranger Station. Accessed 20 November 2001, http://www.noaa.gov/
OLoughlin, E.M. (1986) Prediction of surface saturation zones in natural catchments by topographic
analysis. Water Resources Research, 22:794-804.
Quinn, P., Beven K.J., Chevallier P., & Planchon O. (1991) The prediction of hillslope flow paths for
distributing hydrological modelling using digital terrain models. Hydrol. Processes, 5:59-79.
Saxton, K.E., Rawls W.J., Romberger J.S., & Papendick R.I. (1986) Estimating generalized soilwater characteristics from texture. Soil Sci. Soc. Am. J., 50:1031-1036.
Thornthwaite, C.W. & Mather J.R. (1955) The water balance. Laboratory of Climatology, Publ. No. 8,
Centerton NJ.
Walter, M.T., Walter M.F., Brooks E.S., Steenhuis T.S., Boll J., & Weiler K. (2000) Hydrologically
sensitive areas: variable source area hydrology implications for water quality risk assessment. Journal
of Soil and Water Conservation. 3:277-284.
Wigmosta, M.S., Vail L.W., & Lettenmaier D.P. (1994) A distributed hydrology-vegetation model for
complex terrain. Water Resources Research, 3:1665-1642.
10