MODELING OF THE WATERSHED USING SWAT 2000

Introduction

The Genesee River is a major tributary to Lake Ontario and has a watershed area of about 2500 square miles. At its mouth the river flows thru the City of Rochester and the federal navigation channel in the city. The watershed is primarily agricultural (52%), closely followed by forest (40%). About 4.6% of the land area is classified as urban. The remaining land is split between water and wetlands (2%).

Water quality concerns of the watershed include the amount of sediment that is present in the river. Average annual volumes of dredge material from the federal channel approach 250,000 cubic yards. Disposal of this material occurs at an open lake site located approximately 1.5 miles northeast of the harbor. Two beneficial use impairments for the area of concern are the degradation of the benthos and restrictions on dredging. The degradation of benthos is primarily the result of excessive sedimentation.. The restrictions on dredging activities are, in fact, simply a mandate against overflow dredging for the purpose of reducing the impact of the release of sediment into the navigation channel.

Section 516e of the Water Resources Development Act of 1996, as amended (WRDA96) directs the United States Army Corps of Engineers to develop sediment transport models for tributaries to the Great Lakes that flow into Federal navigation channels or areas of concern. The purpose of the program is to evaluate the impact of land use and other management practices on the production and delivery of sediment on a watershed scale.

Development of the model will assist the Genesee/Finger Lakes Regional Planning Council (GFLRPC) and State and local watershed managers with a tool to aid in their evaluation, prioritization, and implementation of alternatives for soil conservation and non-point source pollution prevention in the watershed.

Methodology

The Soil and Water Assessment Tool (SWAT) is a river basin, or watershed scale model developed by Dr. Jeff Arnold for the USDA Agricultural Research Service (ARS). SWAT was developed to predict the impact of land management practices on water, sediment, and agricultural chemical yields in large complex watersheds with varying soils, landuse, and management conditions over long periods of time. SWAT has been used extensively in the United States for TDML applications. SWAT has been incorporated into the US EPA’s BASINS (Better Assessment Science Integrating Point and Nonpoint Sources) system, developed for watershed and water quality based assessment and integrated analysis of point and nonpoint sources BASINS integrates a geographic information system (GIS), national watershed and meteorlogical data, and state-of-the-art environmental assessment and modeling tools into one convenient package. The SWAT modeling work in this study was conducted within the BASINS system (version 3.0).

The Soil and Water Assessment Tool (SWAT) is a physically-based, continuous simulation erosion model designed to simulate water and sediment yield from watersheds. 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. 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 has been integrated into Better Assessment Science Integrating Point and Nonpoint Sources (BASINS) suite of models developed by the United States Environmental Protection Agency (USEPA)

Description of the Watershed

The Genesee River drains about 2500 square miles in states of New York and Pennsylvania. 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. A location map of the watershed is shown on Figure 1.

The Genesee River has a total length of about 157 miles. It rises in the Allegany Mountains in Potter County, Pennsylvania, at an elevation of about 2500 feet. It flows generally northwest to approximate river mile 106 near Houghton, New York, and then shifts to the northeast to its mouth on Lake Ontario at an elevation of about 247 feet.

The topography of the southern portion of the basin (Upper Basin), upstream of the dam, is steep and rugged, while the northern portion (Lower Basin) 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 , just upstream of the Mount Morris Dam, the river drops from an elevation of about 1080 feet to 768 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 513 to 247 feet. Between Letchworth State Park and the headwaters, the average stream slope is 8.9 feet per mile, while between Rochester and Mount Morris, the average stream slope is 0.8 feet per mile.

Figure 1 Location Map The largest tributary of the Genesee River is . It has a drainage area of 334 square miles and joins the Genesee River near Jones Bridge, just downstream of Mount Morris at approximate river mile 62. It resembles the Genesee River in that its upper reaches, above the Village of Dansville, are steep and rugged, while its lower valley is a flat alluvial plain which is frequently flooded for long durations of time. Above Dansville, the main stem of the creek has a slope of about 40 feet per mile, while from Dansville to its mouth, it has a slope of about 3 feet per mile. The Canaseraga Creek basin is roughly square in shape, about 20 miles on a side. The main stem, which rises at an approximate elevation of 1900 feet, has a length of 42 miles and joins the Genesee River at approximate elevation 548. Other tributaries of the Genesee have a wide range in size and topographic characteristics. For example, Angelica Creek, located in the upper basin, has a drainage area of 85 square miles and is topographically rugged, with a main stream slope of 38 feet per mile. Conversely, Black Creek, located in the lower basin, has a drainage area of 214 square miles. Its basin is relatively level and marshy with a main stream slope of 6.5 feet per mile.

There are numerous artificial controls in the Genesee River basin. The major one is the Mount Morris Dam and Reservoir which was completed in June 1952. It is a concrete gravity dam with an uncontrolled ogee spillway 550 feet long, and a crest elevation of 760 feet, 175 feet above the streambed. The dam has an overall length of 1028 feet and a maximum height of 215 feet. Reservoir control is provided by nine 5 x 7 foot rectangular conduits, each controlled by a vertical hydraulic slide gate. The reservoir is contained within the deep, narrow valley between Mount Morris and the Lower Portage falls. At the op of the flood control pool, the reservoir has a total length of about 17 miles and a maximum width of about ½ mile. The total storage at the spillway crest (elev. 760) is 337,400 acre-feet, of which 610 acre-feet is dead storage, leaving 336,790 acre-feet or 5.86 inches of storage for flood control. The reservoir is regulated for flood control most frequently in the winter and spring months and has resulted in significant benefits in the lower basin.

Other artificial controls in the Genesee River basin include the following:

A series of run-of-river structures for hydroelectric power, developed in the falls reaches of Rochester by the Rochester Gas& Electric Company. Run-of –river structures are those that utilize the natural flow of the river and contain little or no storage capacity.

A state operated gated dam in Rochester for regulation of the elevation of the New York State Barge Canal, which crosses the Genesee River at grade just upstream of Rochester. Its elevation is maintained at approximately 513 feet during the navigation season, and it is provided with guard gates on either side of the river to prevent high flows from entering the canal.

A dam and reservoir operated by the Rochester Gas & Electric Company, on Caneadea Creek, an upper basin tributary which enters the Genesee at about river mile 108 on the main stem. Power is not produced at this dam, its purpose being to augment low flows downstream.

A dam on Hemlock Lake, in the basin, operated by the City of Rochester, to provide water supply to the city.

A dam on Conesus Lake outlet to maintain adequate lake levels for recreation on that lake.

A dam on the Genesee River, just below Mount Morris, operated by Rochester Gas and Electric Company for power. It has an uncontrolled concrete spillway 264 feet in length with its crest at 580 feet. Flashboards on the crest provide an additional three feet of surcharge. The plant is basically run-of-river, and releases from the Corps Mount Morris Dam are held at or above 300 cfs when natural flows permit, to provide flow for the RG&E Dam.

A concrete arch dam on , 3 miles upstream of the Genesee River. This dam provides storage and part of the head for a Rochester Gas & Electric power development at Wiscoy.

SWAT Input Data

GIS data for topography, soils, landuse/landcover, and hydrography were used in the SWAT model. The data used were the most current at the time of model development. Observed daily rainfall, temperature, and streamflow data was obtained and used in all modeling.

An Arcview GIS interface is available to generate the required model inputs from publicly available GIS data. These GIS datasets are processed by the interface and converted to a format usable by the model. GIS data layers of elevation, soils, and landuse are used to generate the input files. Observed temperature and precipitation are incorporated into the model. If no observed weather data are available or if there are missing segments in the observed data, weather can be generated statistically.

Elevation

Elevation data was defined by use of a Digital Elevation model (DEM). DEM’s for the entire United States are available for download from the internet. The DEM was used to calculate sub- basin parameters such as slope, slope length, and to define the stream network. The stream network was used to define the layout and number od sub-basins. Characteristics of the stream network, such as channel slope, length, and width were all calculated from the DEM.

Individual 1:24000 thirty meter DEMS were downloaded and merged together to build a DEM for the entire watershed. When the individual tiles are merged, there may be missing data at the seams of the tiles. The missing data is restored by using a 3x3 convolution filter applied to the entire DEM that results in a seamless filtered DEM. The filter replaces missing data by finding an average cell value of all the cells within the 3x3 matrix. Filtering of the DEM tends to remove both peaks and valleys in the area of the seams, therefore use of the filter should be at a minimum and selection of the size of the filtering matrix should be taken with care.

Figure 2 Genesee River DEM

Soils

Soil data is required by SWAT to define soil types and develop Hydrologic Response Units (HRU). USDA State Soils Geographic Database (STATSGO) Soil data was used in this model. STATSGO data is generalized at a 1:250000 resolution and is formatted for direct use in SWAT. Soil attributes are stored in polygon format. Each polygon includes multiple soil series with information on its areal percentage of the polygon.

An alternative to the STATSGO soils is the use of the more detailed Soil Survey Geographic Database (SSURGO). SSURGO soils are mapped out for each soil series and are available in county tiles. SSURGO soils must be processed for use in SWAT, and a methodology has been developed to accomplish that. However, the small amount of SSURGO data available within the Genesee River watershed and the difficulty in merging STATSGO and SSURGO data resulted in a decision to use only the STATSGO data.

Land Cover

Land cover can be considered the most important dataset used in the model. The land cover theme determines the amount and distribution of pasture, forest, cropland, etc. in the basin. The various land covers are very different in the way they interact in the watershed. The different land covers each contribute varying amounts of sediment to the river and both cause and prevent erosion of the landscape.

It is important that the land cover data be compatible with the other parameters so that model calibration can be performed with the most accuracy possible. While streamflow data was available in several locations within the watershed, sediment data was much more limited. Suspended sediment data was only available for the period April 1974 to September 1977. For this reason it was decided to use land cover data applicable to this time period in order to calibrate the model.

Under a separate agreement, the Genesee/Finger Lakes Regional Planning council undertook the task of updating land cover data. However, this task only covered the watershed area below Mount Morris. Since the project was to build a model for the entire watershed, the use of this data at this time would be incompatible with the other parameters used in the calibration process. If at some future date resources can be identified to complete the land cover for the rest of the watershed, it can be used to refine the model using current streamflow information.

Figure 3 Genesee River Soil Types

Figure 4 Genesee River Landuse

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. Table XX lists the gages, their locations, and type.

Data for these gages was downloaded from the internet and used in the calibration of the model.

Table XX Stream Gages in the Genesee R Basin

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

In April 1975, the USGS began to collect suspended sediment data at seven gages 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 the table below.

Table XX Sediment Gages in the Genesee R Basin

STATION WATERWAY HUC LATITUDE LONGITUDE NUMBER 4221000 Genesee R At Wellsville 4130002 42.1222 -77.9575 4223000 Genesee R At Portageville 4130002 42.5703 -78.0425 4227500 Genesee R Nr Mount Morris 4130002 42.7667 -77.8392 4228500 Genesee River At Avon 4130003 42.9178 -77.7575 4230500 Oatka Creek At Garbutt 4130003 43.0100 -77.7917 4227000 Canaseraga Creek at Shaker's Crossing 4130002 42.7369 -77.8408 4232000 Genesee R At Rochester 4130003 43.1806 -77.6278

Climate 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, and relative humidity.

There are 17 precipitation gages and 9 temperature stations located in the Genesee R basin that were used in the model. The locations of theses stations are shown in the tables below.

Table XX Precipitation Stations

ELEVATION ID NAME LATITUDELONGITUDE (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 HORNEL-ALMOND 42.35000 -77.70000 404 305597 MT MORRIS 42.73056 -77.90444 268 306745 PORTAGEVILLE 42.56667 -78.05000 356 307167 ROCHSTER 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 XX Temperature Stations

ELEVATION ID NAME LATITUDELONGITUDE (m) 300085 ALFRED 42.26083 -77.78556 540 300183 ANGLICA 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

Precipitation depths and maximum and minimum temperature data was downloaded from NOAA’s National Climatic Data Center for the period January 1, 1970 through March 31, 2005. 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.

Lakes and Reservoirs

There are 6 major lakes and/or reservoirs within the Genesee R watershed that were included in the model. Four lakes of the Finger Lakes chain are located in the basin. They are Conesus, Hemlock, Canadice, and Honeoye Lakes. In addition, two reservoirs were modeled. They are the Mt. Morris Reservoir and the Churchville Reservoir.

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. When located off-channel it is considered a pond. All of the impoundments modeled in this study are located along the stream channel and as such were modeled as dams.

Land Management Information

Land management practices are important inputs for a SWAT model. SWAT defines management as a series of individual operations. The timing of these operations may be defined by a date, or as a fraction of the total heat units required by a crop. Each land cover is assigned a set of management operations.

Temperature is one of the most important factors governing plant growth. Each plant has its own temperature range for growth. For any plant, a minimum or base temperature must be reached before any growth will take place. Above the base temperature, the higher the temperature the more rapid the growth rate of the plant. Once the optimum temperature is exceeded the growth rate will slow until a maximum temperature is reached at which growth will stop.

Heat unit theory assumes that plants have unique heat requirements that can be quantified and linked to the time a plant take to mature. Since a plant will not grow when the mean temperature falls below its base temperature, the only portion of the mean temperature that contributes to plant growth is the amount that exceeds the base temperature.

Heat units are accumulated when the average daily temperature exceeds the base temperature of the crop. The base temperature is the minimum temperature required by the crop to grow. The amount of heat units accumulated each day is equal to the average daily temperature minus the base temperature of the plant. When no crops are growing the model uses a base temperature of 0 degrees C. and keeps a running total. The running total is used to schedule planting dates because no heat units can be accumulated until plant growth begins. This method assumes that the rate of growth is directly proportional to the increase in temperature. The heat unit method is the default in SWAT and was used in this model.

Model Development

Subbasin Delineation

The subbasin layout was defined by SWAT using the DEM, a stream theme, and a table of outlets. The stream theme is a shapefile from the National Hydrography Dataset (NHD) which is “burned” into the DEM to force SWAT to define the stream locations correctly in areas of flat topography.

Model predictions are only available at subbasin outlets, so additional outlets were added at points of interest such as gage locations and the inlet and outlet points of the water impoundments.

After the stream burn-in process, the removal of sinks is next. Sinks are anomolies in the DEM that trap water and do not allow flow to the watershed outlet. This will incorrectly define sub- watersheds. The process is to fill the sink cells until flow vectors are maintained. The next step is to compute flow direction grid. This step uses an 8-point pour point algorithm to determine flow direction on a theoretical drop of water. It compares a gridcell to the surrounding eight cells to determine which has the greatest slope. It them routes the water to that cell and continues to the watershed outlet. The next step is to calculate the flow accumulation grid. This grid sums all the cells draining to the watershed outlet in the upstream direction.

A threshold value of 4000ha (9884 acres) was used to delineate subbasins. The threshold value is the minimum contributing upland area required to define a single stream. The processing resulted in 83 subbasins. Fewer subbasins would simplify the modeling effort, but this level of detail reasonably represented the basin.

Hydrologic Response Units (HRU)

Hydrologic Response Units are areas of unique combinations of land use and soil type within a subbasin. Subdividing watersheds into HRU’s enables the model to reflect differences in hydrologic conditions for different land covers/crops and soil types. Runoff is predicted separately for each HRU and routed to obtain the total runoff for the watershed. This increases the accuracy of the predictions and provides a much better physical description of the water balance.

SWAT provides two methods to determine the HRU distribution: the first is to assign a single HRU to each subwatershed based upon a dominant landuse and soil type, and the second is to assign multiple HRU’s to the subwatershed based on a threshold value for landuse and soil type. The multiple HRU option was selected for the Genesee River watershed.

The threshold values for landuse and soil type is 20% and 10% respectively. In developing this model the value for landuse was reduced to 5% and the threshold value for soil type was left at the default. This resulted in a final number of 552 Hydrologic Response Units in the model.

Model Configuration

After the data inputs are developed a number of configuration options are available to the user. These options are dependent upon both the data available to the modeler and the characteristics of the watershed.

Potential Evapotranpiration

SWAT offers three methods to calculate potential evapotranspiration (PET). These methods are Penman-Montheith, Priestly-Taylor, or Hargreaves. The model also has the ability to read in daily PET values from an external source. The methods not only differ in the way potential evapotranspiration is calculated, but also in the input data required to make the calculations. Penman-Montheith method requires values for solar radiation, air temperature, relative humidity, and wind speed. The Priestly-Taylor method requires solar radiation, air temperature, and relative humidity. The Hargreaves method only requires air temperature.

The Penman-Montheith method was not used as the complexity of the interaction of the variables can cause significant errors when using mean daily data. Initial model runs were made using both the Priestly-Taylor and Hargreaves methods, and even though the Priestly-Taylor requires additional data, it showed better agreement with the observed data than Hargreaves. Therefore, Priestly-Taylor was used in the development of this model.

Rainfall Distribution

SWAT offers two options in the way the model distributes rainfall. They are skewed normal and mixed exponential. The skewed distribution calculates the amount of rainfall by comparing the rainfall on a given day to the mean daily rainfall for the month. It computes a standard normal deviate and skew coefficient for the day and then distributes the rainfall. This is the default method. The exponential method also compares the daily rainfall to the mean daily rainfall for the month, but uses an exponential distribution to determine rainfall. It is commonly used where precipitation data is limited. The skew normal method was used in this study.

Surface Runoff

Surface runoff occurs when the rate of water application to the ground surface exceeds the rate of infiltration. Initially, when water is applied to a dry soil, the rate of application and infiltration will be nearly equal. As the soil becomes wetter, the rate of infiltration will decrease until the point is reached where the application rate is greater than the infiltration rate. When this occurs, surface depressions will fill, and once all surface depressions are filled, surface runoff begins.

SWAT provided two methods to estimate surface runoff: the SCS curve number and the Green& Ampt method. The SCS method is an empirical model that uses data from soils, landuse, slope, and soil moisture to compute a curve number (CN) for the day. The Green & Ampt method predicts infiltration assuming excess water at all times. It assumes the soil profile is homogeneous and moisture is distributed uniformly in the soil profile. As water infiltrates into the soil, the model assumes the soil above the wetting front is completely saturated and there is a sharp break in the moisture content at the wetting front.

Channel Routing

SWAT simulates the physical process that affects the flow of water and the transport of sediment in the channel network of the watershed. SWAT assumes the main channels, or reaches, have a trapezoidal shape with 2:1 (run to rise) side slopes. SWAT offers two options of routing flow and sediment in the model. They are the Muskingum and variable storage methods. Both are variations of the kinematic wave model.

The Muskingum method models the storage volume in a channel segment as a combination of wedge and prism storage. As a flood wave enters a channel segment, inflow exceeds outflow and a wedge of storage is produced. As the flood wave leaves the channel segment, a negative wedge is produced. In addition to wedge storage, a prism of storage is formed by a volume of constant cross-section along the channel reach.

The variable storage method is based on the continuity equation, where, for a given time step, the outflow volume is subtracted from the inflow volume to give the volume stored in a specified reach segment. Travel time is computed by dividing the volume of water in the channel by the flow rate. Variable storage is the default option and is the method used in this model.

Data inputs for channel routing include the channel slope, channel length, Manning’s “n” value, channel width at top of bank, and channel depth at top of bank. Initial estimates for these variables were calculated from GIS datasets used to develop the model.

In July 2003, Dr. Richard Young of the State University of New York (SUNY) College at Geneseo, studied the recent and long-term sedimentation and erosion of the lower Genesee Valley (below Mt. Morris). During this study, laser measurements were made of the bank widths at selected point along the river. These widths were incorporated into the model when defining the channel.

Model Calibration

After initially configuring SWAT, model calibration is performed. Calibration of the model refers to the adjustment or fine-tuning of the model parameters to reproduce observed values. Calibration is performed at several locations throughout the watershed, at the gage locations. The model was calibrated for flow and sediment.

Calibration and validation were completed by comparing time-series model results to observed monitoring data. Output from the model in the form of monthly average flow and monthly average sediment were compared to the observed data. Graphical comparisons of the time series plots between observed and simulated values, and residual scatter plots (observed vs. simulated) were made to visually illustrate model agreement.

Modeling covered the time period January 1970 to December 1980. This time period was selected because it was the only time interval where all the required inputs were present. A modeling time of 10 years is sufficient to determine trends within the watershed. It included the period between April 1975 and September 1977 where the USGS monitored sediment at several gage locations.

Calibration and verification of the model was performed during the period of sediment monitoring. The time period was split into two periods for performing the calibration and verification of the model. Calibration was performed between April 1, 1975 and July 31, 1976. The model calculations were began at January 1, 1970 to give the model time to “ramp up” prior to the time period of interest. This is due to the fact that the model starts with virtually all parameters at essentially zero; soil is dry, rainfall is null, etc.

After successful calibration and verification runs, the entire model period was run to provide long term insight to watershed trends.

A number of statistical tests are available for watershed model evaluation and optimization. Two of the more common tests are the Nash-Sutcliffe efficiency (NS) and the Coefficient of Correlation (R2). The criteria for goodness of fit were used for calibration: graphical comparison of the data and the relative error method. Graphical comparisons are useful to judge the results of the model runs; time-variable plots of observed versus modeled flow give insight to the model’s representation of the flow hydrographs. The model’s accuracy was primarily assessed through interpretation of the time-series plots. The relative error method was used to support the goodness of fit evaluation through a quantitative comparison. A small error indicates a better fit for calibration.

Goodness of fit measures were evaluated to test model accuracy. These measures include percentage differences in average and standard deviations over the simulation period, coefficient of correlations (R2), and the Nash-Sutcliffe measure (NS).

The Nash-Sutcliffe values can range from negative infinity to 1, with 1 being a perfect model to data agreement and zero indicating that the fit is as good as using the average value of all the measured data. All of these goodness-of-fit measures were calculated for the stream gages in the watershed. In addition, model-to-data plots were analyzed for possible trends. The comparison of both average and standard deviations investigates whether the model stream flow frequency distribution is similar to the measured one. Although R2 values have been used in the past to compare model results to data, the recommendations of ASCE (1993) indicate that the Nash- Sutcliffe measure is a better representation for model goodness-of-fit. As a result, these results will focus on NS, while only providing R2 to compare this effort to previous applications of the SWAT model. In general, a NS value greater than or equal to 0.5 and an R2 value greater than or equal to 0.6 is considered acceptable for calibration. Figure X shows the calibration process is SWAT.

The large amount of data provided confidence in the model calibration exercise because it supplied enough data on varying spatial and temporal scales. As a result, an analysis of the impact of parameter adjustments at a number of calibration stations was possible.

Figure X. SWAT Calibration Process

Hydrology

Hydrology was the first model component to be calibrated, and it involved a comparison of observed data from USGS stream gages to modeled in-stream flow and an adjustment to key hydrologic parameters.

Key considerations in the hydrology calibration were the overall water balance, peak flow, and low flows. The stream gages are located in different part of the watershed encompassing different spatial and temporal scales. The gage locations are spread out within the watershed and drain various amounts of watershed area. This provides a basis to ensure that the model accurately simulates the watershed on different scales. Because the ultimate use of this model is to investigate the long-term trends on the watershed, it is felt that the model needs to represent long-term average flows and sediment loads to understand the impacts of watershed management practices. In addition, because the model uses a daily time step, it is difficult to accurately capture daily results at this watershed scale due to possible uncertainties in time series data. Consequently, model results were captured and compared on a monthly basis to determine if there were trends on model output or error.

Figures X to Y show the monthly calibration results for the calibration period at the USGS stream gage locations. Plots of simulated monthly flow versus observed monthly flow are also provided at these locations.

A summary of statistical results for the hydrology calibration at the gage locations is shown in Table X. The monthly NS values range from 0.26 to 0.99.

The availability of climate data plays a part in the accuracy of the model simulation. The USDA has cited spatial variability with precipitation data as one of the “major limitations to large area hydrologic modeling.” For the Genesee River watershed, precipitation from 17 stations is distributed to the 2500 square mile watershed. As a result, it is possible that the rainfall over the entire watershed is being misrepresented for the model input.

Figure X Sub-basin #61 USGS Gage at Wellsville

Figure X Sub-basin #62 USGS Gage at Portageville

Figure X Sub-basin #69 USGS Gage at Jones Bridge

Figure X Sub-basin #71 USGS Gage at Avon

Figure X Sub-basin #74 USGS Gage at Rochester

Figure X Sub-basin #83 USGS Gage at Churchville

Figure X Sub-basin #59 USGS Gage at Garbutt

Figure X Sub-basin #60 USGS Gage at Warsaw

Figure X Sub-basin #67 USGS Gage at Dansville

Figure X Sub-basin #68 USGS Gage at Shaker’s Crossing

Figure X Sub-basin #72 USGS Gage at Honeoye

Table X Flow Calibration Summary

Genesee River Basin Flow Calibration Results

Site Name Stream Sub-basin NS COD

Wellsville Genesee R. 61 0.52 0.61 Portageville Genesee R. 62 0.80 0.81 Mt. Morris Genesee R. 69 0.95 0.95 Avon Genesee R. 71 0.93 0.94 Rochester Genesee R. 74 0.94 0.95 above Dansville Canaseraga Ck. 67 0.68 0.86 Shaker's Crossing Canaseraga Ck. 68 0.66 0.82 Warsaw Oatka Ck. 60 0.50 0.62 Garbutt Oatka Ck. 59 0.65 0.72 Honeoye Falls Honeoye Ck. 70 0.68 0.76 Churchville Black Ck. 83 0.66 0.67

Sediment Calibration

Similar to hydrology, the sediment results were desired to understand the long-term average trend of the system to management issues. As a result, the monthly sediment loads were a focus for calibration, as opposed to daily results. In addition, considerable uncertainty exists in the estimated measured sediment loads, therefore, it did not seem practical to compare dailty model results to daily measured data. Finally, sediment erosion and transport modeling is highly uncertain and accurate simulation of sediment processes on the land surface is difficult to capture due to the nature of the watershed and the relatively unrefined methods used to explain these processes. Therefore, it is typical; that a model may perform well for hydrology, it may have limitations in capturing sediment loads. Consequently, this model will focus on the ability of the model to capture long-term trends.

SWAT used the Modified Universal Soil Loss Equation (MUSLE) to predict soil erosion. In the Universal Soil Loss Equation, average gross erosion is predicted as a function of rainfall energy. 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.

Figures X to Y show the monthly calibration results for the calibration period at the USGS stream gage locations. Plots of simulated monthly flow versus observed monthly flow are also provided at these locations.

A summary of statistical results for the hydrology calibration at the gage locations is shown in Table X. The monthly NS values range from 0.59 to 0.78.

Figure X Sub-basin #61 USGS Gage at Wellsville

Figure X Sub-basin #62 USGS Gage at Portageville

Figure X Sub-basin #59 USGS Gage at Garbutt

Figure X Sub-basin #68 USGS Gage at Shaker’s Crossing

Figure X Sub-basin #71 USGS Gage at Avon

Figure X Sub-basin #69 USGS Gage at Jones Bridge

Figure X Sub-basin #74 USGS Gage at Rochester

Table X. Sediment Calibration Summary

Genesee River Basin Sediment Calibration Results

Site Name Stream Sub-basin NS COD

Wellsville Genesee R. 61 0.73 0.83 Portageville Genesee R. 62 0.78 0.85 Mt. Morris Genesee R. 69 0.68 0.73 Avon Genesee R. 71 0.67 0.73 Rochester Genesee R. 74 0.68 0.73 Shaker's Crossing Canaseraga Ck. 68 0.68 0.82 Garbutt Oatka Ck. 59 0.59 0.66

Model Verification

Verification of the model after adjustment of the parameters was accomplished to determine if the adjustments were valid. Verification took place for the time period August 1, 1976 to September 30, 1977. This is the remaining time period where sediment monitoring took place.

The same locations were compared.

Figures X to Y show the monthly verification results for the verification period at the USGS stream gage locations. Plots of simulated monthly flow versus observed monthly flow are also provided at these locations.

A summary of statistical results for the hydrology calibration at the gage locations is shown in Table X. The monthly NS values range from 0.26 to 0.99.

Figure X Sub-basin #61 USGS Gage at Wellsville

Figure X Sub-basin #62 USGS Gage at Portageville

Figure X Sub-basin #69 USGS Gage at Jones Bridge

Figure X Sub-basin #71 USGS Gage at Avon

Figure X Sub-basin #74 USGS Gage at Rochester

Figure X Sub-basin #83 USGS Gage at Churchville

Figure X Sub-basin #59 USGS Gage at Garbutt

Figure X Sub-basin #60 USGS Gage at Warsaw

Figure X Sub-basin #67 USGS Gage at Wellsville

Figure X Sub-basin #68 USGS Gage at Wellsville

Figure X Sub-basin #72 USGS Gage at Wellsville

Sediment Verification

Figure X Sub-basin #61 USGS Gage at Wellsville

Figure X Sub-basin #62 USGS Gage at Portageville

Figure X Sub-basin #69 USGS Gage at Wellsville

Figure X Sub-basin #59 USGS Gage at Garbutt

Figure X Sub-basin #68 USGS Gage at Wellsville

Figure X Sub-basin #71 USGS Gage at Avon

Figure X Sub-basin #74 USGS Gage at Rochester

Model Validation

Figure X Sub-basin #61 USGS Gage at Wellsville

Figure X Sub-basin #62 USGS Gage at Portageville

Figure X Sub-basin #69 USGS Gage at Jones Bridge

Figure X Sub-basin #71 USGS Gage at Avon

Figure X Sub-basin #74 USGS Gage at Rochester

Figure X Sub-basin #83 USGS Gage at Churchville

Figure X Sub-basin #59 USGS Gage at Garbutt

Figure X Sub-basin #60 USGS Gage at Warsaw

Figure X Sub-basin #67 USGS Gage at Dansville

Figure X Sub-basin #68 USGS Gage at Shaker’s Crossing

Figure X Sub-basin #72 USGS Gage at Honeoye

Sediment Validation

Figure X Sub-basin #61 USGS Gage at Wellsville

Figure X Sub-basin #62 USGS Gage at Portageville

Figure X Sub-basin #59 USGS Gage at Garbutt

Figure X Sub-basin #68 USGS Gage at Shaker’s Crossing

Figure X Sub-basin #71 USGS Gage at Avon

Figure X Sub-basin #69 USGS Gage at Jones Bridge

Figure X Sub-basin #74 USGS Gage at Rochester

Management Practices

Sediment Yield Sub-basin 25

350.000

Sed w/o Filter Strip 300.000 Sed w/ Filter Strip

250.000

200.000

150.000 Yield Sediment

100.000

50.000

0.000 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 Time in Years

Sediment Yield - Sub-basin 59

0.030

0.025 Sed w/o Filter Strips Sed w/ filter Strips 0.020

0.015

Sediment Yield Sediment

0.010

0.005

0.000 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 Time in Years

Conclusions

This section describes some of the results of the model and some key limitations of this study. Due to its agriculture-dominant nature, the Genesee river watershed is well suited to be modeled by SWAT. The results and experience gained from this study will become a valuable source of information for future management decisions and modeling work.

This study opted to build the watershed model with 83 subwatersheds and 583 HRU’s. Other work (Jha et. Al. 2004) has suggested that optimal subwatershed size range between 2 and 5 percent of the total basin area. While this study meets that criteria, the size of the watershed does make it difficult to evaluate management decisions on a watershed scale. Increasing the number of HRU’s by using a more refined combination of soil type and landuse could possibly provide a way to evaluate management practices at the field level. However, considering the amount of time and effort required to develop the model with the increased number of HRU’s for their particular management types and to change those types during the model simulations; would consume intense CPU resources and the time needed to process and analyze the model outputs would not be practical. A better course of action would be to create smaller models of the subwatersheds themselves where the number of HRU’s would not be so overwhelming. Management decisions can be implemented to these “mini- models” and then inserted into the overall watershed model to determine it effect on larger landscape.

The scope of this model was to focus on sediment, therefore a rigorous examination of urban processes was not undertaken. SWAT has the ability to simulate pollutant loadings damaging to the streams resulting from urban use. If these loading are to be incorporated at some future date, further examination of the urban areas is necessary.

There are some limitations to the model itself. Since SWAT was developed mainly in southern locales, its snowmelt routines have some limitations in northern climates. The model is continually being refined and updated, and future releases will contain more efficient algorithms to model snow processes.

Conclusions

This study has developed a working SWAT model for the Genesee River watershed, given the limited scope and data available. The calibrated model was able to simulate sediment loading in the watershed at several locations widely distributed throughout the basin.

The model was able to identify a mechanism to reduce sediment loading to the streams which in turn reduces the amount of sediment finding its way into the federal channel in Rochester.

In summary, given the limited scope of the study parameters, and the limited amount of sediment data available for model calibration, this modeling study has provided a valuable tool for local interests to make watershed management decisions and to evaluate the effectiveness of land management practices to reduce pollutant loadings and improve water quality in the Genesee river basin.