A Modified Soil and Water Assessment Tool (SWAT) Model for Flow and Sediment

Transport in the Genesee River Basin

Technical Report

Zachary M. Easton1, Daniel R. Fuka1, Tammo S. Steenhuis1, Byron Rupp, and Paul
Murawski2
1
Dept. Biological and Environmental Engineering
Cornell University
2
US Army Corps of Engineers
Buffalo, NY

Abstract
A modified version of the Soil and Water Assessment Tool (SWAT) model was
employed to simulate the flow and sediments in the Genesee River Basin, New York.
SWAT was modified to incorporate variable source area hydrology (SWAT-VSA) in the
prediction of flow and sediment. Water quality concerns in the Genesee Basin primarily
revolve around the amount of sediment discharged into Lake Ontario at Rochester, NY,
as well as in stream sediment levels throughout the basin. A beneficial use impairment
is in place because of degradation of the benthos, primarily the result of excessive
sedimentation. Thus, the model is developed to assist the Genesee/Finger Lakes
Regional Planning Council (GFLRPC) and state and local watershed managers in their
evaluation, prioritization, and implementation of alternatives for soil conservation and
non-point source pollution remediation in the watershed. The model was calibrated for
flow and sediment during the 1975-1977 period at 12 gauges and run from 1970-2009
for validation. Model results showed a good fit to measured data during both the
calibration and validation periods. Daily Nash-Sutcliffe Efficiencies (NS) ranged from 0.5
to 0.8. Modeled sediment predictions also showed good agreement with the short
duration of measured data. Daily model predicted flows were within 15% of measured
flows at all gauges, and considerably lower at most (<5%). While sediment export from
the various gauged subbasins was highly variable, SWAT was able to capture the
dynamics well. Daily model predicted sediment export from was generally within 25% of
the measured export at the gauges. Interestingly, the model indicated that there was a
significant amount of stream bank/channel erosion in the upland (first order subbasins)
that was subsequently deposited in lowland subbasin under flow regimes less than the
65th percentile. Under higher flows (> 65th percentile) these deposited sediments can be
mobilized. The results of the model can be used to evaluate and direct implementation
of alternatives for soil conservation and non-point source pollution remediation in the
watershed.
Introduction
The US Army Corps of Engineers (Buffalo Office) developed a SWAT model using a
previous version of the model (SWAT 2000). Presented and discussed here are
modifications and updates made to the original modeling framework. The current model
was developed using SWAT 2005 and the ARCSWAT 9.2 interface, with modifications
to incorporated variable source area hydrology (VSA) as presented in Easton at al.
(2008)

Water quality modeling is an important tool used by researchers, planners, and

government agencies to optimize management practices to improve water quality. Many
water quality models use some form of the Natural Resources Conservation Services
(formerly Soil Conservation Service) curve number (CN) equation to predict storm runoff
from watersheds. The way the CN is applied in these models implicitly assumes an
infiltration-excess (or Hortonian, i.e., Horton, 1933) response to rainfall (Walter and
Shaw, 2005). However, in humid, well-vegetated regions, especially those with
permeable soils underlain by a shallow restricting layer, storm runoff is usually
generated by saturation-excess processes on variable source areas (VSAs) (Dunne and
Black, 1970: Dunne and Leopold 1978: Beven 2001: Srinivasan et al., 2002:
Needleman et al., 2004). Thus, many of the most commonly used water quality models
do not correctly capture the spatial distribution of runoff source areas and, by
association, pollutant source areas (Qui et al., 2007). Examples of models using a CN-
type approach include the Generalized Watershed Loading Function (GWLF) model
(Haith and Shoemaker, 1987), the Soil Water Assessment Tool (SWAT) model (Arnold
et al., 1998), the Storm Water Management Model (SWMM) (Krysanova et al., 1998),
the Erosion Productivity Impact Calculator (EPIC) model (Williams et al., 1984), and the
Long-Term Hydrologic Impact Assessment (L-THIA) model (Bhaduri et al., 2000).
Although these models have routinely been calibrated to correctly predict stream
discharge and sometimes stream water quality at the watershed outlet this does not
mean that distributed hydrological processes were correctly captured (Srinivasan et al.,
2005). Indeed, lumped models, which implicitly ignore how intra-watershed processes
are distributed, can predict integrated watershed responses like stream flow as well as
models that simulate a fully distributed suite of intra-watershed processes, (Franchini
and Pacciani, 1991: Johnson et al., 2003).

Watershed managers tasked with implementing strategies for controlling non-point

source (NPS) pollution need models that can correctly identify the locations where
runoff is generated in order to effectively place management practices. Although, as
discussed earlier, most current water quality models do not have this capacity, Lyon et
al. (2004), building on Steenhuis et al. (1995), showed how CN models can be used to
predict the distribution of VSAs. Schneiderman et al. (2007) were the first to use this
body of work to modify an existing water quality model, namely GWLF (Haith and
Shoemaker, 1987), so that it explicitly simulated saturation excess runoff from VSAs.

It is challenging to modify more complex models like SWAT or EPIC because many of
their sub-models, especially those simulating various biogeochemical processes, have
been developed around the assumption that a watershed can be characterized by an
assemblage of non-interacting landscape-units delineated by land use and soil type.
While this is perhaps a reasonable assumption when considering deep soil profiles
without restricting layers or shallow water tables and, hence, areas where surface runoff
is primarily produced when the rainfall intensity exceed the infiltration capacity of the soil
(i.e., Hortonian or infiltration-excess runoff), it may not prove accurate in VSA dominated
areas.

In many areas, VSAs are formed when shallow ground water flowing within the soil from
upslope areas of the watershed accumulates and saturates lower parts of the
watershed. These VSAs become “active” when groundwater flow exceeds storage or
when precipitation falls on the saturated area, causing saturation excess runoff. This
concept necessitates recognizing interactions throughout the landscape. Thus, the
challenge of incorporating VSA hydrology into models employing CN-type functionality
is to find ways of capturing the spatial arrangement of saturation excess runoff source
areas while also retaining the soil and land use information necessary to various
nutrient and biogeochemical subroutines.

This study specifically focused on SWAT, one of the most commonly used and well
supported water quality modeling systems available. The strengths of SWAT are its use
of readily available input data and its process based nutrient biogeochemistry sub-
models (Santhi et al., 2001: van Griensven and Bauwens, 2003: Borah and Bera, 2004:
Ramanarayanan et al., 2005). Storm runoff is predicted with the CN equation, which we
have already noted can be used in a VSA context. SWAT divides sub-basins into
hydrological response units (HRUs) but currently does not allow water flow among
HRUs and, therefore, cannot simulate the formation of VSAs. In this investigation we
propose using a topographic wetness index to redefine HRUs so that spatial runoff
patterns would follow those observed in VSA dominated landscapes. Indeed, Lyon et al.
(2004) found that topographic indices described the evolution of a shallow water table in
the Catskill Mountain of New York and that this shallow water table was the primary
control on VSA formation. One of the earliest uses of topographic wetness indices to
simulate VSA hydrology was TOPMODEL (Beven and Kirkby 1979) and the topographic
index concept has since been applied outside of TOPMODEL and found to effectively
predict VSAs for many watersheds dominated by saturation-excess runoff (Western et
al., 1999, Lyon et al., 2004; Agnew et al., 2006; Schneiderman et al., 2007). This re-
conceptualization of SWAT allows users to model and predict saturation excess runoff
from VSAs, and thus provides a simple means of capturing spatially variant saturation
excess runoff processes from the landscape. The re-conceptualized version of SWAT,
referred to as SWAT-VSA (Easton et al., 2008).

Summarized SWAT Description

The SWAT model is a river basin or watershed scale model created to run with readily
available input data so that general initialization of the modeling system does not
require overly complex data gathering or calibration. SWAT was originally intended to
model long-term runoff and nutrient losses from rural watersheds, particularly those
dominated by agriculture (Arnold et al., 1998). It was developed by the USDA-ARS to
provide a tool for predicting the impact of land management practices on water,
sediment, and agricultural chemical yields in large complex watersheds with varying
soils, land use and management conditions over long periods of time. SWAT is
considered to be an acceptable a tool for development of total maximum daily loads
(TMDLs) by the US Environmental Protection Agency (USEPA). The model contains
components of both the Universal Soil Loss Equation (USLE) and the Modified
Universal Soil Loss Equation (MUSLE). The model can be applied to large watersheds
and complex landscapes. It uses a grid-cell characterization of the landscape to
represent the spatial variability across watersheds or regions. Input information is
grouped into categories consisting of weather or climate, land cover, soil, and land
management. It has the capability of analyzing the above categories for sub-
watersheds, ponds/reservoirs, groundwater, channels, or reaches. The model can be
extended to include nutrients and pesticide loadings.

SWAT requires soils data, land use/management information, and elevation data todrive
flows and direct sub-basin routing. While these data may be spatially explicit, SWAT
lumps the parameters into hydrologic response units (HRU), effectively ignoring the
underlying spatial distribution. Traditionally, HRUs are defined by the coincidence of soil
type (Hydrologic Soil Group, USDA 1972) and land use. Simulations require
meteorological input data including precipitation, temperature, and solar radiation. In the
present study, model input data and parameters were initially parsed using the
ARCSWAT interface. The interface assimilated the soil input map, digital elevation
model, and land use coverage.

CN Equation Applied to VSA Theory

SWAT-VSA capitalizes on the re-conceptualization of the CN equation to capture VSA
storm hydrology (Steenhuis et al., 1995, Lyon et al., 2004, Schneiderman et al., 2007),
which we will briefly summarize the background of here. The CN equation was originally
developed by Mockus (Rallison, 1980) and estimates total watershed runoff depth Q
(mm) (both overland flow and rapid subsurface flow) for a storm (USDA-SCS 1972):

2
Pe
Q
Pe Se (1)

where Pe (mm) is the depth of effective rainfall after runoff begins, i.e. rainfall minus
initial abstraction, Ia, (rainfall retained in the watershed when runoff begins) and Se (mm)
is the depth of effective available storage in the watershed, or the available volume of
retention in the watershed when runoff begins. Ia is estimated as an empirically-derived
fraction of available watershed storage (USDA-SCS 1972).

Steenhuis et al. (1995) showed that Eq. 1 could be interpreted in terms of a saturation-
excess process. Assuming that all rain falling on unsaturated soil infiltrates and that all
rain falling on areas that are saturated becomes runoff, then the rate of runoff
generation will be proportional to the fraction of the watershed that is effectively
saturated, Af, which can then be written as:

Q
Af (2)
Pe

where ΔQ is incremental saturation-excess runoff or, more precisely, the equivalent

depth of excess rainfall generated during a time period over the whole watershed area,
and ΔPe is the incremental depth of precipitation during the same period. Thus, by
differentiating Eq. 1 with respect to Pe, the fractional contributing area for a storm can
be written as (Steenhuis et al., 1995):

2
Se
Af = 1 - 2 (3)
Pe + S e

According to Eq. 3, runoff only occurs from areas that have a local effective available
storage, e, (mm), less than Pe. Therefore, by substituting e for Pe in Eq. 3 we have a
relationship for the fraction of the watershed area, As, that has a local effective soil
water storage less than or equal to e for a given overall watershed storage of Se
(Schneiderman et al., 2007):

2
Se
As = 1 - 2 (4)
e + Se
Note that that both Se and Ia are watershed properties while e is defined at the local
level. Solving for e gives the maximum effective local soil moisture storage within any
particular fraction, As, of the watershed area (Schneiderman et al., 2007):

1
e Se 1 (5)
(1 As )

For a given storm event with precipitation P, the fraction of the watershed that saturates
first (As = 0) has local storage e = 0, and runoff from this fraction will be P – Ia.
Successively drier fractions retain more precipitation and produce less runoff according
to the moisture – area relationship of Eq. 5. The driest fraction of the watershed that
saturates during a storm defines the total contributing area (Af). As average effective
soil moisture storage (Se) changes through the year, the moisture-area relationship will
shift accordingly (Eq. 5); Se is constant once runoff begins.

According to Schneiderman et al. (2007), runoff for an area, qi (mm), can be expressed
as:

qi = Pe – e for Pe > e, (6)

For Pe≤ e, the unsaturated portion of the watershed, qi =0. We propose approximating
Eq. 6 with the CN equation (Fig. 1):

2
Pe
qi (7)
Pe e

This approximation gives the same result when P→∞ satisfying the boundary condition.
However, runoff starts earlier as shown in (Fig. 1). Thus, unlike Eq. 6, Eq. 7 predicts
some flow before the local soil deficit is filled, which may be interpreted as early, rapid
subsurface flow or as an uneven distribution of storages ( e) within the runoff producing
area.

Implementation for Variable Source Areas (VSAs)

The most significant part of our re-conceptualization of SWAT lies in re-defining HRUs.
Following Schneiderman et al. (2007), we use a topographic wetness index in
combination with land use to define the HRUs. Specifically, we use a soil topographic
index (STI) (e.g., Beven and Kirkby, 1979):

a
STI ln (8)
T tan

where a is the upslope contributing area for the cell per unit of contour line (m), tan is
the topographic slope of the cell, and T is the transmissivity (soil depth x saturated soil
hydraulic conductivity) of the uppermost layer of soil (m2 d-1) (Lyon et al., 2004). The
local storage deficit of each wetness class ( e,i) is determined by integrating Eq. 5 over
the fraction of the watershed represented by that wetness index class (Schneiderman et
al., 2007):

2Se ( 1 As,i 1 As,i 1 )

e,i Se
( As,i As,i )
1 (9)

where the fractional area represented by each index class is bounded on one side by
the fraction of the watershed that is wetter, As,i, i.e., the part of the watershed that has
lower local moisture storage, and on the other side by the fraction of the watershed that
is dryer, As,i+1, i.e., has greater local moisture storage. Index classes are numbered from
1, the driest and least prone to saturate, to n, the most frequently saturated.

In the standard version of SWAT land use and soil type define the area of each HRU. In
SWAT-VSA the area of each HRU is defined by the coincidence of land use and
wetness index class determined from the STI (Eq. 8); Figure 2 shows how HRUs are
delineated in the SWAT-VSA framework. The additional information necessary for
SWAT‟s nutrient and biogeochemical subroutines, specifically, topology (e.g. slope
position and length, etc) and various soil physical and chemical properties, are
averaged within each wetness index class. We do not anticipate serious problems due
to taking average soil properties within an index class because there is evidence that
soil variability roughly correlates with topographic features, probably because soil
genesis is to a great extent driven by hydrology and topology (Page et al., 2005:
Sharam et al., 2006: Thompson et al., 2006).

In SWAT, runoff is calculated for each soil/land-use defined HRU using Eq. 1. In SWAT-
VSA, the runoff for each wetness-index-class/land-use defined HRU is calculated with
Eq. 7, which is approximately equivalent to Eq. 1 for large events. The difference is that
in SWAT-VSA, the effective storage is associated with each HRU‟s wetness index (Eq.
9) and in SWAT the effective storage is based on land use and soil infiltration capacity.
Note that in SWAT-VSA, runoff depth within a wetness index class will be the same
irrespective of land use while nutrient dynamics will vary with land uses and, thus, can
differ within index classes. Nutrient loads from each wetness-index-class/land-use HRU
are tracked separately in SWAT-VSA, but are otherwise processed similarly to SWAT.
For the entire watershed, runoff depth is the aerially-weighted sum of local runoff
depths, qi, for all HRUs.

Although topographic wetness indices capture the spatial patterns of VSAs, the SWAT
code does not allow water flow among HRUs. Thus, to compensate for this we propose
adjusting the available water content (AWC) of the soil profile so that higher wetness
index classes retain water longer, and the lower classes dry quicker. In VSA theory
runoff production is related to saturation dynamics, accordingly there needs to be higher
water contents in high runoff producing areas, thus we relate local soil water storage,
e,i, to AWC with the following relationship:

b 0.4 clay b 254

AWC 1 0.05 0.01 (10)
2.65 100 e,i 24.5

where ρb is the soil bulk density (g cm-3) and clay is the soil clay content (cm3 cm-3).
Equation 10 makes conceptual sense in the SWAT-VSA framework because e,i is
directly related to the saturated area by Eq. 4. Notice that the form of last term of Eq. 10
is the same as the relationship between Se and the CN. Using this relationship the soil
moisture in runoff producing areas is forced to a higher level. This calculation is derived
and adapted directly from SWAT‟s own soil water calculations, its soil physics routine
(AWC = Field capacity – Wilting point).

HRUs dominated by impervious surfaces are simulated identically and with the same
infiltration-excess approach as used in SWAT.

In the standard version of SWAT land use and soil type define the area of each HRU. In
SWAT-VSA the area of each HRU is defined by the coincidence of land use and
wetness index class determined from the STI (Eq. 8); The additional information
necessary for SWAT‟s nutrient and biogeochemical subroutines, specifically, topology
(e.g. slope position and length, etc) and various soil physical and chemical properties,
are averaged within each wetness index class. We do not anticipate serious problems
due to taking average soil properties within an index class because there is evidence
that soil variability roughly correlates with topographic features, probably because soil
genesis is to a great extent driven by hydrology and topology (Page et al., 2005:
Sharam et al., 2006: Thompson et al., 2006).

In SWAT, runoff is calculated for each soil/land-use defined HRU using Eq. 1. In SWAT-
VSA, the runoff for each wetness-index-class/land-use defined HRU is calculated with
Eq. 7, which is approximately equivalent to Eq. 1 for large events. The difference is that
in SWAT-VSA, the effective storage is associated with each HRU‟s wetness index (Eq.
9) and in SWAT the effective storage is based on land use and soil infiltration capacity.
Note that in SWAT-VSA, runoff depth within a wetness index class will be the same
irrespective of land use while nutrient dynamics will vary with land uses and, thus, can
differ within index classes. Nutrient loads from each wetness-index-class/land-use HRU
are tracked separately in SWAT-VSA, but are otherwise processed similarly to SWAT.
For the entire watershed, runoff depth is the aerially-weighted sum of local runoff
depths, qi, for all HRUs.

Genesee River Watershed Description

The Genesee River a major tributary to Lake Ontario and drains an area 6420 km 2 the
majority of which lies in Western New York State (6100 km 2) Fig 1. The watershed is
primarily agricultural (52%), closely followed by forest (40%), 4.6% of the land area is
classified as urban and the remaining land is split between water and wetlands (2%)
(Fig. 2). Its drainage area encompasses parts of nine counties in New York and one in
Pennsylvania. The basin is roughly elliptical in shape, with a major north-south axis of
about 100 miles, and a maximum width of about 40 miles. The basin lies generally
between 41o 45‟ and 43o15‟ North Latitude and between 77o25‟ and 78o25‟ West
Longitude. The basin is split into two hydrologic units at Mount Morris Dam, built and
operated by the Corps of Engineers. The drainage area above the dam is about 1080
square miles.
Figure 1. Genesee River Watershed location in New York State. The Genesee River
flows into Lake Ontario in Rochester NY.
Figure 2. Land use in the Genesee River Watershed.
Figure 3. Digital Elevation Model (DEM) of the Genesee River Watershed. Elevations
range from 66 to 797 m above sea level. DEM has grid resolution of 25 m.

The Genesee River has a total length of about 253 km. Its source is in the Allegany
Mountains in Potter County, Pennsylvania, at an elevation of 797 m. It flows generally
northwest to approximate river mile 171 near Houghton, New York, and then shifts
northeast to its mouth on Lake Ontario at an elevation of about 66 m (Fig. 3).
The topography of the southern portion of the basin (Upper Genesee), upstream of the
dam, is steep and rugged, while the northern portion (Lower Genesee) is gently rolling.
Geologically, the upper basin is in a stage of young maturity, while the lower basin has
reached a geologically old stage with much meandering, a wide flood plain, and
numerous oxbows. In Letchworth State Park, just upstream of the Mount Morris Dam,
the river drops from an elevation of about 329 feet to 234 feet, over three successive
falls, flowing through a deep gorge cut in rock. It then flows through narrow valleys and
gorges to enter the broad lower Genesee Valley in the village of Mount Morris. From
this point to the City of Rochester, the river valley is a flat alluvial plain up to three miles
wide and was subject to frequent flooding before the construction of the dam in 1952. At
Rochester, the river drops over three falls from elevation 156 to 75 m. Between
Letchworth State Park and the headwaters, the average stream slope is 1.6 m per km,
while between Rochester and Mount Morris, the average stream slope is 0.15 m per
km. Slopes range from 0 to 56% with a median slope of 7.9% (Fig. 4).

Figure 4. Slope classes used in the SWAT Model. 7.9% is the median slope value.

Input Data
Spatial Data: Required landscape data includes tabular and spatial soil data, tabular
and spatial land use information, and elevation data. All soils data were taken from the
SSURGO soil database (USDA-NRCS, 2000) (Fig. 5); arithmetic means were used for
all soils properties for which SSURGO contained a range of values. For SWAT-VSA, we
substituted the STI for the soils map to create the HRUs. However, since SWAT
requires many soil properties for hydrologic, sediment, and biogeochemical sub routines
we areally weighted the soils map with the STI using GRASS (U.S. Army CERL, 1997)
and extracted the associated soils properties from the SSURGO database. These
values were then integrated into look up tables and linked to the map in the ARCSWAT
9.2 interface. For SWAT-VSA, we lumped the watershed‟s STI into 10 equal area
intervals ranging from 1 to 10, with index class 1 covering the 10% of the watershed
area with the lowest STI (i.e. lowest propensity to saturate) and index class 10
containing the 10% of the watershed with the highest STI (i.e. highest propensity to
saturate) (Fig. 6). These wetness index classes were intersected with the land use (Fig.
2) to create 962 HRUs in 26 sub basins . Sub basins were defined in the ARCWAT 9.2
interface by setting a minimum area threshold of 10 km 2 and creating a dense stream
network, which was subsequently thinned manually to force subbasin outlets (Fig. 7) to
correspond with locations of USGS gauges in the watershed (Fig. 8). A 25 x 25 m and
0.1 m vertical resolution digital elevation model (DEM) of the basin was obtained. The
DEM was used to calculate sub-basin parameters such as slope, slope length, and to
define the stream network. Characteristics of the stream network, such as channel
slope, length, and width were all calculated from the DEM. Land cover was chosen to
correspond to land cover existing in the 1970s to allow proper calibration of sediment
parameters.

Figure 5. SSURGO and STATSGO soil coverage‟s for the Genesee River Watershed.
Figure 6. Wetness Index Classes used in the SWAT-VSA model. The wetness index
classes are derived from an areally weighted Soil Topographic Index. SSURGO soil
coverage was nested

Required Meteorological Data

Climate data in the form of precipitation, temperature, solar radiation, and humidity are
used in the model to predict crop growth, evapotranspiration, and snowmelt. Daily
climate data inputs required in the model were precipitation depths, maximum and
minimum temperature, solar radiation, wind speed and relative humidity. Meteorological
data can be downloaded at: http://www.ncdc.noaa.gov/oa/climate/stationlocator.html.
There are 17 precipitation gages and 9 temperature stations located in the Genesee
River Basin that were used in the model. The locations of these stations are shown in
the Tables 1 and 2 and Fig. 8.
Table 1. Precipitation Stations

COOP-ID NAME LAT LONG ELEV(m)

300085 ALFRED 42.26083 77.78556 540
300183 ANGELICA 42.30139 77.98917 441
300343 AVON 42.92083 77.75556 166
300443 BATAVIA 43.03028 78.16917 278
300766 BOLIVAR 42.08056 78.17556 482
301974 DANSVILLE 42.56639 77.71833 201
303065 FRIENDSHIP 42.13333 78.23333 500
303722 HASKINVILLE 42.42056 77.56750 503
303773 HEMLOCK 42.78333 77.61667 275
303983 HORNELL 42.35000 77.70000 404
305597 MT MORRIS 42.73056 77.90444 268
306745 PORTAGEVILLE 42.56667 78.05000 356
307167 ROCHESTER 43.11667 77.67667 183
307329 RUSHFORD 42.40000 78.26667 469
308962 WARSAW 42.68333 78.21667 555
309072 WELLSVILLE 42.12194 77.95639 460
309425 WHITESVILLE 42.03833 77.76194 522

Table 2. Temperature Stations

COOP-ID NAME LAT LONG ELEV (m)

300085 ALFRED 42.26083 77.78556 540
300183 ANGELICA 42.30139 77.98917 441
300443 BATAVIA 43.03028 78.16917 278
300766 BOLIVAR 42.08056 78.17556 482
301974 DANSVILLE 42.56639 77.71833 201
303773 HEMLOCK 42.78333 77.61667 275
305597 MT MORRIS 42.73056 77.90444 268
307167 ROCHESTER 43.11667 77.67667 183
308962 WARSAW 42.68333 78.21667 555
Figure 7. Subbasin delineation for the watershed

Precipitation depths and maximum and minimum temperature data was downloaded
from NOAA‟s National Climatic Data Center for the period January 1, 1970 through July
31, 2009. Daily datasets were obtained, assembled, and processed for each station to
form the SWAT weather input files. Data processing included unit transformation and
database file development.
SWAT has the ability to generate values for missing data from a supplied database of
weather statistics from stations across the United States. The values for missing data in
the downloaded datasets were derived from the statistical information. Daily potential
evapotranspiration rates were calculated in the SWAT model using the Penmen-
Monteith method. Solar radiation and wind speed were calculated in the model using
the weather generator. Meteorological stations were geo-referenced (latitude, longitude,
and elevation) and the variables adjusted in SWAT using lapse rates in the watershed.
Streamflow Data
There are 14 stream gages operated by the USGS located within the Genesee River
watershed; 12 of these gages monitor flow and 2-monitor stage. Data is available at :
http://waterdata.usgs.gov/ny/nwis/ . Figure 8 shows their location in the watershed.

Figure 8. Location of precipitation, temperature, flow, and sediment gauges in the

watershed.
Table 3. Stream Gages in the Genesee River Basin

GAUGE WATERWAY HUC LAT LONG YEAR EST. TYPE

4221000 Genesee River Wellsville 4130002 42.1222 77.9575 1956 F
4223000 Genesee River Portageville 4130002 42.5703 78.0425 1902 F
4224775 Canaseraga Creek Above Dansville 4130002 42.5356 77.7044 1972 F
4227500 Genesee River Mount Morris 4130002 42.7667 77.8392 1890 F
4228500 Genesee River Avon 4130003 42.9178 77.7575 1956 F
4229500 Honeoye Creek Honeoye Falls 4130003 42.9572 77.5892 1945 F
4230380 Oatka Creek Warsaw 4130003 42.7442 78.1378 1963 F
4230500 Oatka Creek Garbutt 4130003 43.0100 77.7917 1945 F
4230650 Genesee River Ballantyne Brdg 4130003 43.0922 77.6806 1973 S
4231000 Black Creek Churchville 4130003 43.1006 77.8825 1945 F
4227000 Canaseraga Creek Shaker's Crossing 4130002 42.7369 77.8408 1916 F
4232000 Genesee River Rochester 4130003 43.1806 77.6278 1904 F
4227995 Conesus Creek Lakeville 4130003 42.8556 77.7167 1996 F
4224000 Mt Morris Lake Mt Morris 4130002 42.7333 77.9111 1952 S

Sediment Data
In April 1975, the USGS began to collect suspended sediment data at seven gauges in
the basin. This program continued until September 1977. The data collected was
downloaded and used to calibrate the model for sediment. The locations of the
sediment monitoring gages are shown in Table 4 and Fig. 8

Table 4. Sediment Gages in the Genesee River Basin

GAUGE WATERWAY HUC LAT LONG
4221000 Genesee River Wellsville 4130002 42.1222 77.9575
4223000 Genesee River Portageville 4130002 42.5703 78.0425
4227500 Genesee River Mt Morris 4130002 42.7667 77.8392
4228500 Genesee River Avon 4130003 42.9178 77.7575
4230500 Oatka Creek Garbutt 4130003 43.0100 77.7917
4227000 Canaseraga Creek Shaker's Crossing 4130002 42.7369 77.8408
4232000 Genesee River Rochester 4130003 43.1806 77.6278

Lakes and Reservoirs

There are six major lakes and/or reservoirs within the Genesee River. Four lakes of the
Finger Lakes chain are located in the basin. They are Conesus in Subbasin 17, and
Hemlock, Canadice, and Honeoye Lakes in Subbasin 16. In addition, two reservoirs
were modeled. They are the Mt. Morris Reservoir in Subbasin 22 and the Churchville
Reservoir in Subbasin 5.

SWAT models lakes and reservoirs based on their proximity to the stream channel.
When an impoundment is located along the main stream channel, it is considered as a
dam, and when located off-channel it is considered a pond. All of the impoundments
modeled in this study are considered as dams. Locations can be seen in Figure 8.
.
Calibration
A two-step procedure was used to calibrate the CNII values for SWAT-VSA‟s wetness
index classes based on 1975-1977 data (when corresponding sediment data was
available). A basin wide storage (Se) of 16 cm calculated from baseflow separated
runoff plotted against effective precipitation (Pe) used to distribute the e,i values in the
basin according to the wetness index (Eq. 9), which results in a basin-wide average
CNII = 73 per the method outlined in the model development section (i.e., VSA CNII
method). This first step established the relative distribution of e,i. Because the CN
equation is non-linear, the flow predicted using a lumped Se would not necessarily be
the same as that predicted as the sum of multiple contributing areas, even if the
average storage in both cases is the same. Therefore, we re-ran the calibration
uniformly adjusting all the CNII (by CN-units of 0.25) to minimize the RMSE between
predicted and measured runoff. The optimized average CNII, over all wetness index
classes, was 70, which was similar to that found in an earlier study using SWAT by the
US Army Corps of Engineers for the same calibration period, CNII = 70.1. The
differential between the CNII value of 73, determined using the VSA methodology
described above, and the CNII value of 70, determined using the minimization method,
is due to the different runoff calculations used (i.e., Eq. 6 vs. Eq. 7). To insure good
calibration, we also made sure that our calibrated result maximized the coefficient of
determination (r2) and the Nash-Suttcliffe efficiency (E) (Nash and Suttcliffe, 1970).
While the AWC was not „calibrated‟, it was adjusted via Eq. 10 to account for the fact
that we recognize that water accumulates in low-lying areas or areas that drain large
upslope expanses, and since SWAT does not route water laterally from HRU to HRU,
this adjustment was needed to compensate. First order subbasins were calibrated first
(i.e.(subbasins #26,25,24,16,5, and 23) followed in increasing order by higher order
subbasins (i.e., 20 followed by 18 followed by 10, followed by 13, followed by 2).
Parameter sets were not altered after the original calibration.

Sediment parameters were calibrated on 1975-1977 data as well. Channel sediment

erosion parameters were set to values that allowed the stream channel bed and bank to
erode. Upland erosion is modeled using MUSLE. In the MUSLE, the rainfall energy
factor is replaced with a runoff factor. This serves to improve the sediment yield
prediction, eliminates the need for delivery ratios, and allows the equation to be applied
to individual storm events. Sediment yield is improved because runoff is a function of
antecedent moisture conditions as well as rainfall energy. Since MUSLE is at best an
empirical relationship between physical parameters and erosion, MUSLE sediment
parameters were allowed to vary around the tabulated value by 40% during
optimization.

The calibration objective for flow and sediment was to maximize the NS coefficient while
simultaneously trying to minimize the sums of squares error (SSE). The model was
calibrated first for flow, then summer sediment load, and then winter sediment load. The
baseflow separation technique of Hewlitt and Hibbert (1967) was used to estimate the
baseflow component of measured and calibrated model flows at each gauge site.
Subbasins containing no measured flow data were calibrated simultaneously with
subbasins containing measured data. For instance subbasins 22, 21, and 19 have no
gauge at the outlet, but ultimately drain out of subbasin 18 (Mt Morris), thus they were
included in the calibration for the Mt Morris gauge. For sediment calibration, we utilized
a parameter transfer methodology from calibrated subbasins. Sediment parameter sets
from calibrated subbasins (subbasins 26, 25, 20, 19,18,13, and 2) were transferred to
adjacent or otherwise similar (in terms of soil, land use topography) subbasins. For
instance sediment parameters from subbasin 25 (Portageville) were transferred to
subbasin 23 (Oatka Creek at Warsaw) because they were both adjacent and have
similar land use topography and soil properties.

Validation
We validated SWAT-VSA during the 1977-2009 periods by comparing daily, event and
monthly runoff and sediment export predictions to direct measurements. While the
model was run on a daily time step, integrated validation was performed on an event
basis. For each constituent, model performance was evaluated with four methods:
qualitatively using time series plots and quantitatively using the coefficient of
determination (r2), the root mean square error (RMSE), and the Nash-Suttcliffe
coefficient (NS).

Summarized Model Results

Streamflow
Figures 9-19 show the modeled and SWAT modeled daily streamflow for both the
validation period (1975-1977) and validation period. SWAT-VSA simulated streamflow
did not differ significantly (p-value<0.05 paired t-test) from measured streamflow at 11 of
the 12 gauges. Only modeled streamflow at the gauge on Oatka Creek in Garbutt
differed from measured flow (Fig. 14). The difference is mainly a result of over
prediction of baseflow. Despite this, the overall streamflow dynamics are well captured,
that is predicted runoff peaks generally coincide with measured peaks. No model
predictions varied significantly from measured data at the event or monthly time steps.
No systematic bias was evident in the results, with predicted vs. observed data evenly
scattered around the 1:1 line The largest deviations were generally found during
snowmelt or rain on snow events, which are not well captured by SWATs simplistic
temperature index snowmelt method, and should warrant father study.
Table 5.
Genesee River Basin Flow Calibration Period (1975-1977)
Daily Event Monthly
2 2 2
Site Name Stream Subbasin NS R NS R NS R

Wellsville Genesee R. 26 0.67 0.69 0.75 0.71 0.85 0.75

Portageville Genesee R. 25 0.71 0.73 0.79 0.88 0.85 0.87
Mt. Morris Genesee R. 18 0.70 0.73 0.83 0.87 0.91 0.89
Avon Genesee R. 13 0.73 0.74 0.72 0.86 0.82 0.91
Rochester Genesee R. 2 0.74 0.75 0.86 0.82 0.89 0.84
above Dansville Canaseraga Ck. 24 0.64 0.80 0.67 0.80 0.73 0.88
Shaker's Crossing Canaseraga Ck. 20 0.69 0.81 0.67 0.85 0.76 0.89
Warsaw Oatka Ck. 23 0.68 0.72 0.73 0.70 0.86 0.85
Garbutt Oatka Ck. 10 0.52 0.61 0.59 0.68 0.75 0.81
Honeoye Falls Honeoye Ck. 16 0.69 0.70 0.72 0.69 0.79 0.80
Churchville Black Ck. 5 0.71 0.70 0.76 0.74 0.77 0.79

Table 6.
Genesee River Basin Flow Validation Period (1977-2009)
Daily Event Monthly
2 2 2
Site Name Stream Subbasin NS R NS R NS R

Wellsville Genesee R. 26 0.65 0.67 0.75 0.74 0.85 0.74

Portageville Genesee R. 25 0.70 0.70 0.77 0.82 0.85 0.85
Mt. Morris Genesee R. 18 0.71 0.71 0.80 0.79 0.87 0.84
Avon Genesee R. 13 0.68 0.71 0.72 0.74 0.83 0.87
Rochester Genesee R. 2 0.66 0.70 0.85 0.71 0.85 0.86
above Dansville Canaseraga Ck. 24 0.65 0.73 0.69 0.72 0.76 0.80
Shaker's Crossing Canaseraga Ck. 20 0.67 0.69 0.65 0.80 0.77 0.85
Warsaw Oatka Ck. 23 0.67 0.70 0.70 0.66 0.83 0.83
Garbutt Oatka Ck. 10 0.58 0.67 0.67 0.73 0.72 0.82
Honeoye Falls Honeoye Ck. 16 0.71 0.68 0.75 0.69 0.76 0.88
Churchville Black Ck. 5 0.66 0.66 0.70 0.70 0.75 0.77

Sediment
Figures 9-19 show the modeled and SWAT modeled daily streamflow for both the
validation period (1975-1977) and validation period. The model generally predicted daily
sediment export form the gauged basins with good levels of accuracy (NS 0.5 to 0.7),
somewhat surprising given the limited time scale of the data. Some interesting trends
can be observed in the data and model results. For instance, if we are to trust the model
predictions to look at long-term sedimentation patterns, the two subbasins on Oatka
Creek have rather different results. The first order subbasin (at Warsaw, subbasin 23)
has relatively high sediment export, with an average of 166 tons d -1 (Fig. 15). While the
subbasin it drains into (Subbasin 19 at Garbutt) has a comparatively much smaller
sediment export (47 tons d-1). This is particularly intriguing considering the difference in
the size of the subbasin areas, 104 km2 for Warsaw vs. 423 km2 for Garbutt. This
indicates that there is substantial deposition of upstream erosion occurring before
Garbutt. This might be the result of the reduced stream power unable to keep sediment
entrained as the terrain flattens approaching the Garbutt gauge. Indeed, this is the case,
as sediment yield in the Warsaw subbasin was an order of magnitude greater than in
the Garbutt subbasin. These types of results point to differing sediment transport
mechanisms in different areas of the basin. For instance, there was very little landscape
based erosion in the subbasins containing the greater Rochester area. Model results
indicated sediment yields from the subbasin of approximately <0.5 ton ha-1, while there
was considerable channel or bank erosion (average of approximately 2000 tons d -1).
This might be a result of increased runoff from the expanses of impervious surfaces
resulting in an increased stream power scouring the more easily erodible loess and
fluvial/alluvial deposits in the near lake region.

Model results also point to Lower Genesee subbasins having reduced sediment export
since the 1970s. The major gauges in the Lower Genesee at Rochester, Avon, Mt
Morris, Canasegera Creek at Shakers Crossing, Oatka Creek at Garbutt, and Black
Creek at Churchville were all predicted to have decreasing sediment export since the
late 1970s. It is not clear what the cause of this reduction is, as there has not been a
clear decline in yearly rainfall over that period, but it is a relatively strong trend. This is in
contrast to the stable (Canasegera Creek at Dansville) or increasing sediment export
(Portageville, Wellsville, Honeoye Falls, Oakta Creek at Warsaw) in Upper Genesee
subbasins. Perhaps rainfall patterns within a year are causing the pattern, (e.g., more
intense thunderstorms, or reduced snowfall/snow pack). Another potential cause might
have been due to several large events in the 1970s, (Hurricane Agnes June 1972
among others), which might have scoured deposited sediment from the Lower
Genesee, essentially resetting the system. This would result in increased sedimentation
following the event and reduced sediment export. It appears that the discharge at
several of the gauges in the Lower Genesee (Rochester, Avon, Mt Morris, and Black
Creek at Churchville) has been decreasing as well (Figs 9, 10, 11, and 12), although,
again it is unclear why, perhaps increased withdrawals. Intuitively this could be
responsible for the reduced sediment export. Conversely, the Upper Genesee gauges
do not appear to have experienced reduced discharge, and thus would not be expected
to have reduced sediment export.
Table 7.
Genesee River Basin Sediment Calibration Period (1975-1977)
Daily Event Monthly
2 2 2
Site Name Stream Subbasin NS R NS R NS R

Wellsville Genesee R. 26 0.63 0.72 0.68 0.74 0.73 0.75

Portageville Genesee R. 25 0.68 0.61 0.69 0.67 0.74 0.73
Mt. Morris Genesee R. 18 0.73 0.81 0.78 0.83 0.82 0.87
Avon Genesee R. 13 0.72 0.77 0.74 0.79 0.76 0.81
Rochester Genesee R. 2 0.68 0.65 0.72 0.68 0.77 0.69
Shaker's Crossing Canaseraga Ck. 20 0.51 0.58 0.62 0.67 0.60 0.65
Garbutt Oatka Ck. 10 0.54 0.69 0.58 0.65 0.65 0.69

Conclusions
We modified and developed a SWAT model for the Genesee River Basin in New York
State. The model incoporates VSA hydrology in an attempt to better predict runoff and
sediment source areas. Observed streamflow at 11 gauges and sediment export at 7
gauges was well captured by the model. Analysis of model data provides insight into
mechanisims of erosion and sediment export in the basin. For instance, channel and
bank erosion appear to be significant sources of sediment in the Lower Genesee, while
both in stream erosion and landscape erosion are important sources in the Upper
Genesee subbasins.

Thes results and subsequent scenario analysis with the model can help planners and
regulatory personell identify important processes in theGenesee Basin and provide
support for management decisions.

Issues to Consider
Due to the size of the basin and the number of HRUs the distributed output of the model
at the HRU level were not tested here. The HRU output file is over 9 gigabytes for the
entire 40 yr model run, so to effectively utilize these data one would need to run the
mode for a shorter time span, or on a longer time step. One of the supposed strengths
of the SWAT-VSA methodology is that it better captures the spatial extent of runoff and
sediment source areas. This needs to be further tested and validated. The integrated
subbasin results appear accurate, and the underlying assumptions of VSA hydrology
clearly apply to the majority of the Genesee River Basin, thus we expect the distributed
results from SWAT-VSA to be more accurate and underpin better management
decisions than results from the standard SWAT model. That being said there is a need
for distributed data to test these results, and further refine model methodology.

We also feel that it would be worth testing alternative snowmelt/snow accumulation

models in the basin, as temperature index method performed poorly.
 1200
Calibration 1975-1977
Observed Flow at Rochester
USGS Gauge #4232000
1000 SWAT

800
Discharge (m 3 s-1)

600

400

200

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

350000
Calibration 1975-1977
SWAT
300000
Observed Sediment at Rochester
Sediment Export (ton d-1)

250000 USGS Gauge #4232000

200000

150000

100000

50000

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 9

.
 500
Calibration 1975-1977 Observed Flow at Avon USGS
Gauge #4228500
400 SWAT
Discharge (m 3 s-1)

300

200

100

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

100000
Calibration 1975-1977 SWAT
90000
Observed Sediment at Avon
80000 USGS Gauge #4228500
Sediment Export (ton d-1)

70000
60000
50000
40000
30000
20000
10000
0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 10
.
 500
Calibration 1975-1977 Observed Flow at Mt Morris
USGS Gauge #4227500
400 SWAT
Discharge (m3 s-1)

300

200

100

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

160000
Calibration 1975-1977 Observed Sediment at Mt Morris
USGS Gauge #4227500
140000
SWAT
120000
Sediment Export (ton d-1)

100000

80000

60000

40000

20000

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 11.
 160
Calibration 1975-1977
Observed Flow at Black Creek
140 Churchville USGS Gauge #4231000

SWAT
120
Discharge (m 3 s-1)

100

80

60

40

20

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

200
SWAT Sediment at Black Creek
180
Churchville USGS Gauge
160 #4231000
Sediment Export (ton d-1)

140

120

100

80

60

40

20

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 12.
 160
Calibration 1975-1977 Observed Flow at Honeoyee Falls
USGS Gauge #4229500
140
SWAT
120
Discharge (m3 s-1)

100

80

60

40

20

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

9000
SWAT Sediment at Honeoyee
Falls USGS Gauge #4229500
8000

7000
Sediment Export (ton d-1)

6000

5000

4000

3000

2000

1000

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008
Figure 13.
 250
Calibration 1975-1977 Observed Flow at Oatka Creek
Garbutt USGS Gauge #4232500
200 SWAT
Discharge (m3 s-1)

150

100

50

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

5000
Calibration 1975-1977
4500 SWAT

4000 Observed Sediment at Oatka Creek

Sediment Export (ton d-1)

3500 Garbutt USGS Gauge #4232500

3000
2500
2000
1500
1000
500
0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 14.
 70
Calibration 1975-1977 Observed Flow at Oatka Creek
Warsaw USGS Gauge #4230380
60
SWAT
50
Discharge (m 3 s-1)

40

30

20

10

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

20000
SWAT Sediment at Oatka Creek
18000 Warsaw USGS Gauge #4230380

16000
Sediment Export (ton d-1)

14000

12000

10000

8000

6000

4000

2000

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008
Figure 15.
 400
Calibration 1975-1977
Observed Flow at Canasegera
350 Creek Shakers Crossing USGS
Gauge #4227000
300
Discharge (m3 s-1)

250

200

150

100

50

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

40000
Calibration 1975-1977 SWAT
35000
Observed Sediment at Canasegera Creek
Shakers Crossing USGS Gauge #4227000
30000
Sediment Export (ton d-1)

25000

20000

15000

10000

5000

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 16.
 140
Calibration 1975-1977 Observed Flow at Canasegera Creek
Dansville USGS Gauge #4224775
120
SWAT
100
Discharge (m 3 s-1)

80

60

40

20

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

20000
SWAT Sediment at Canasegera Creek
18000 Dansville USGS Gauge #42247750

16000
Sediment Export (ton d-1)

14000

12000

10000

8000

6000

4000

2000

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 17.
 2500
Calibration 1975-1977
Observed Flow at Portageville USGS
Gauge #4223000
2000
SWAT
Discharge (m3 s-1)

1500

1000

500

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

700000
Calibration 1975-1977
SWAT
600000
Observed Sediment at Portageville USGS
Gauge #4223000
Sediment Export (ton d-1)

500000

400000

300000

200000

100000

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 18.
 450
Calibration 1975-1977 Observed Flow at Wellsville
400 USGS Gauge #4221000
SWAT
350

300
Discharge (m3 s-1)

250

200

150

100

50

0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

100000
Calibration 1975-1977
90000 SWAT
80000
Sediment Export (ton d-1)

70000
60000
50000
40000
30000
20000
10000
0
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

Figure 19.
References
Agnew, L.J., S. Lyon, P. Gérard-Marchant, V.B. Collins, A.J. Lembo, T.S. Steenhuis,
M.T. Walter. 2006. Identifying hydrologically sensitive areas: Bridging science and
application. J. Envir. Mgt. 78: 64-76.
Arnold, J.G., Srinivasan, R., Muttiah, R.R. and Williams, J.R., 1998. Large area
hydrologic modeling and assessment part I: model development. J. Am. Water Resour.
Assoc., 34(1): 73-89.
Arnold, J.G., and P.M. Allen. 1999. Automated methods for estimating baseflow and
ground water recharge from streamflow records. J. Am. Water Resour. Assoc., 35: 411-
424.
Beven, K. 2001. Rainfall-Runoff Modeling: The Primer. John Wiley & Sons, LTD.
Chichester, England. 360pp.
Beven, K.J. and M.J. Kirkby. 1979. A physically-based, variable contributing area model
of basin hydrology. Hydrol. Sci. Bull. 24: 43-69.
Bhaduri, B., J. Harbor, B. Engel, and M. Grove. 2000. Assessing watershed-scale, long-
term hydrologic impacts of land-use change using a GIS-NPS model. Environ. Manag.
26:643-658.
Borah, D. K., and M. Bera. 2004. Watershed-scale hydrologic and nonpoint-source
pollution models: Review of applications. Trans. ASAE 47:789-803.
Dunne, T. and R.D. Black. 1970. Partial area contributions to storm runoff in a small
New England watershed. Water Resour. Res. 6: 1296-1311.
Dunne, T. and L. Leopold. 1978. Water in Environmental Planning. W.H. Freeman &
Co., New York. 818pp.
Easton, Z.M., D.R. Fuka, M.T. Walter, D.M. Cowan, E.M. Schneiderman, and T.S.
Steenhuis. 2008. Re-conceptualizing the soil and water assessment tool (SWAT) model
to predict runoff from variable source areas. J. Hydrol. 348:279– 291.
Franchini, M., and C. Pacciani. 1991. Comparative analysis of several rainfall runoff
models. J. Hydrol., 122:161-219.
Haith, D.A. and L.L. Shoemaker. 1987. Generalized Watershed Loading Functions for
stream flow nutrients. Water Resour. Bull. 23(3):471-478.
Hewlett, J.D. and A.R. Hibbert. 1967. Factors affecting the response of small
watersheds to precipitation in humid regions. In: Forest Hydrology (W.E. Sopper and
H.W. Lull, eds.). Pergamon Press, Oxford. pp. 275-290.
Horton, R. E. 1933. The role of infiltration in the hydrological cycle. Trans AGU. 14:446-
460.
Johnson, M.S., W.F. Coon, V.K. Mehta, T.S. Steenhuis, E.S. Brooks, and J. Boll. 2003.
Application of two hydrologic models with different runoff mechanisms to a hillslope
dominated watershed in the northeastern US: a comparison of HSPF and SMR. J.
Hydrol. 284:57-76.
Krysanova, V., D. I. Muller-Wohlfeil, and A. Becker. 1998. Development and test of a
spatially distributed hydrological water quality model for mesoscale watersheds. Ecol.
Model. 106:261-289.
Lyon, S. W., M. T. Walter, P. Gerard-Marchant, and T. S. Steenhuis. 2004. Using a
topographic index to distribute variable source area runoff predicted with the SCS
curve-number equation. Hydrol. Proc. 18:2757-2771
Nash, J. E., and J. V. Sutcliffe. 1970. River flow forecasting through conceptual models.
Part I a discussion of principles. J. Hydrol. 10:282-290.
Page, T., P. M. Haygarth, K. J. Beven, A. Joynes, T. Butler, C. Keeler, J. Freer, P. N.
Owens, and G. A. Wood. 2005. Spatial variability of soil phosphorus in relation to the
topographic index and critical source areas: Sampling for assessing risk to water
quality. J. Environ. Qual. 34:2263-2277.
Qui, Z, M.T. Walter, and C. Hall 2007. Managing variable source pollution in agricultural
watersheds. J. Soil Water Conserv., 63(3): 115-122.
Rallison, R.E. 1980. Origin and evolution of the SCS runoff equation. Symposium on
Watershed Management. ASCE. New York, NY. pp. 912-924.
Ramanarayanan, T., B. Narasimhan, and R. Srinivasan. 2005. Characterization of fate
and transport of isoxaflutole, a soil-applied corn herbicide, in surface water using a
watershed model. J. Agric. Food Chem. 53:8848-8858.
Santhi, C., J. G. Arnold, J. R. Williams, L. M. Hauck, and W. A. Dugas. 2001.
Application of a watershed model to evaluate management effects on point and
nonpoint source pollution. Trans. ASAE 44:1559-1570.
Schneiderman, E.M., T.S. Steenhuis, D.J. Thongs, Z.M. Easton, M.S. Zion, G.F.
Mendoza, M.T. Walter, and A.L. Neal. 2007. Incorporating variable source area
hydrology into the curve number based Generalized Watershed Loading Function
model. Hydrological Processes 21:3420-3430.
Srinivasan, M.S., W.J. Gburek, and J.M. Hamlett. 2002. Dynamics of stormflow
generation - A field study in east-central Pennsylvania, USA. Hydrol. Proc., 16(3):649-
665.
Srinivasan, M. S., P. Gerard-Marchant, T. L. Veith, W. J. Gburek, and T. S. Steenhuis.
2005. Watershed scale modeling of critical source areas of runoff generation and
phosphorus transport. J. Am. Water Resour. Assoc. 41:361-375.
Steenhuis, T. S., M. Winchell, J. Rossing, J. A. Zollweg, and M. F. Walter. 1995. SCS
runoff equation revisited for variable-source runoff areas. J. Irrig Drain. Eng. 121:234-
238.
Thompson, J. A., E. M. Pena-Yewtukhiw, and J. H. Grove. 2006. Soil-landscape
modeling across a physiographic region: Topographic patterns and model
transportability. Geoderma 133:57-70.
U.S. Army CERL (1997), GRASS 5.3 users manual, Construction and Engineering
Laboratory, Champaign IL.
USDA-NRCS. 2000. Soil Survey Geographic (SSURGO) database for Delaware
County, New York, http://www.nrcs.usda.gov/products/datasets/sssurgo/.
USDA-SCS (Soil Conservation Service). 1972. National Engineering Handbook, Part
630 Hydrology, Section 4, Chapter 10.
van Griensven, A., and W. Bauwens. 2003. Concepts for river water quality processes
for an integrated river basin modeling. Water Sci. Tech. 48:1-8.
Walter, M.T. and S.B. Shaw. 2005. Discussion: “Curve number hydrology in water
quality modeling: Uses, abuses, and future directions” by Garen and Moore. J. Am.
Water Resour. Assoc. 41(6): 1491-1492.
Western, A.W., R.B. Grayson, G. Bloschl, G.R. Willgoose, T.A. McMahon. 1999.
Observed spatial organization of soil moisture and its relation to terrain indices. Water
Resour. Res. 35(3): 797-810.
Williams, J. R., C. A. Jones, and P. T. Dyke. 1984. The EPIC model and its applications.
Proc. ICRISAT-IBSNAT-SYSS Symp. on Minimum Data Sets for Agrotechnology
Transfer.