Fort Cobb Basin - Modeling and Land Cover Classification DRAFT. Submitted to. Oklahoma Department of Environmental Quality.

Transcription

1 Fort Cobb Basin - Modeling and Land Cover Classification DRAFT Submitted to Oklahoma Department of Environmental Quality Submitted by Dr. Daniel E. Storm Mr. Phillip R. Busteed Mr. Michael J. White Biosystems and Agricultural Engineering Department Division of Agricultural Sciences and Natural Resources Oklahoma State University January 19, 2006 I

2 Acknowledgment We would like to thank the Oklahoma Department of Environmental Quality, Oklahoma Conservation Commission, and the Oklahoma State County Extension Offices in Caddo, Washita, Custer County for all of their help. We would also like to thank David Nowlin, Mark S. Gregory, Dirk Webb, and Dr. Hailin Zhang for providing crop management and county soil information which was critical to developing a good model. We would also like to thank Applied Analysis Incorporated (AAI) for their hard work and dedication to this project. Their land cover image was a crucial portion of this project. II

3 Table of Contents 1 - Introduction SWAT 2000 Input Data... 2 Topography... 2 Soils... 2 Applied Analysis Incorporated Land Cover... 2 Subbasin Delineation... 3 HRU Distribution... 3 Weather Data... 3 Soil Test Phosphorus and Management Operations... 4 Cattle Stocking Rate Verification... 4 Reservoirs and Pond Input Data... 5 Concentrated Animal Feeding Operations Model Calibration and Validation Flow Calibration Flow Validation Nutrient Parameter Modifications Total Phosphorus and Sediment Loads Model Limitations and Conclusions Model Limitations Conclusions Model Update with 2005 Land Cover References Appendix A - AAI Land Cover Classification Report... A1 Appendix B - Crop Management... B1 Appendix C -Reservoir and Ponds Information...C1 Appendix D - Soil Test Data...D1 Appendix E - Stream Flow Data... E1 Appendix F - Water Quality Data... F1 Appendix G - Subbasin Properties...G1 III

4 List of Figures Figure 2.1 Ten meter USGS Digital Elevation Model with county boundaries for the Fort Cobb Basin Figure 2.2 Thirty meter Applied Analysis Incorporated Landsat derived land cover with county boundaries for the Fort Cobb Basin Figure 2.3 Center pivot irrigation systems in the Fort Cobb Basin identified from 2003 one m digital aerial photography. Figure 2.4 Land cover coverage incorporating Applied Analysis Incorporated land cover data, center pivot locations, and National Agricultural Statistics Service data Figure 3.1 U.S. Geological Survey (USGS) stream gage and Cooperative Observation Network (COOP) weather station locations in the Fort Cobb Basin Figure 3.2 General SWAT model calibration procedure Figure 3.3 Average annual flow calibration procedure used for the SWAT model Figure 3.4 Monthly flow calibration procedure used for the SWAT model Figure 3.5 Observed and calibrated SWAT simulated flow at Cobb Creek near Eakley, OK ( ) Figure 3.6 Scatter plot of daily average ( ) observed and SWAT simulated flows at Cobb Creek near Eakley Figure 3.7 Scatter plot of monthly average ( ) observed and SWAT simulated flows at Cobb Creek near Eakley Figure 3.8 Scatter plot of annual average ( ) observed and SWAT simulated flows at Cobb Creek near Eakley Figure 3.9 Validation time series for observed and SWAT predicted flow at the Cobb Creek near Eakley, OK ( ) Figure 3.10 U.S. Geological Survey, Oklahoma Conservation Commission, and U.S. Fish and Wildlife Service water quality gage locations in the Fort Cobb basin Figure 3.11 Observed total phosphorus concentrations vs SWAT model (hydrologic calibrated model with modified nutrients parameters) predictions for the Fort Cobb basin Figure 3.12 Observed total phosphorus concentrations vs SWAT model (hydrologic calibration only) predictions for the Fort Cobb basin Figure 4.1 Subbasins draining to the Fort Cobb Reservoir and to Cobb Creek below the reservoir dam only. Cobb Creek drains 853 km 2 of which 799 km 2 drains to the Fort Cobb Reservoir, and 6.3% of the Cobb Creek Basin is downstream of the Fort Cobb Reservoir dam.. 31 IV

5 List of Tables Table 2.1 Applied Analysis Incorporated land cover by percentage within the Fort Cobb Basin based on 2001 TM Imagery Table 2.2 Crop and management breakdown based on the 2005 Fort Cobb Basin Agricultural Field Management Survey for the period Table 2.3 Minimum C factor by crop and tillage for agricultural HRUs used in the SWAT Model Table 2.4 Comparison of the number of cattle simulated in the SWAT model and estimates derived from the Oklahoma Agricultural Statistics Service Table 2.5 Concentrated Animal Feeding Operations in the Cobb Creek Basin. Derived from the Oklahoma Department of Agriculture, Food and Forestry database Table 3.1 Drainage area and flow per unit area from to for available USGS stations in the Fort Cobb Basin Table 3.2 Parameter values use to calibrate flow and modify total phosphorus load for the Fort Cobb SWAT model Table 3.3 Parameter values use to calibrate the Fort Cobb SWAT model for flow only Table 3.4 Summary statistics for SWAT model hydrologic calibration for flow at Cobb Creek near Eakley gage for the period 1/ / Table 3.5 Summary statistics for SWAT model hydrologic validation for flow at Cobb Creek near Eakley gage for the period 1/ / Table 3.6 Observed and predicted total phosphorus concentrations using three different averages for both the hydrologic calibration SWAT model and the hydrologic calibration with modified nutrient parameters SWAT model shown Table 3.7 All water quality sites in the Fort Cobb Basin and the number of samples collected during baseflow, high flow, and recession flow Table 4.1 SWAT simulated loads by land cover for the Fort Cobb Basin for the period 1/ / Table 4.2 SWAT simulated loads by year for the Fort Cobb Basin for the periods of 1/1995 to 12/ Table 6.1 SWAT simulated winter wheat sediment loading based on tillage implementation and acreage in the Fort Cobb basin Table 6.2 SWAT simulated winter wheat total phosphorus loading based on tillage implementation and acreage in the Fort Cobb basin V

6 Table 6.3 SWAT simulated peanut sediment loading based on tillage implementation and acreage in the Fort Cobb basin Table 6.4 SWAT simulated peanut total phosphorus loading based on tillage implementation and acreage in the Fort Cobb basin Table 6.5 SWAT simulated loads per unit area for all peanut and winter wheat scenarios Table 6.6 SWAT simulated loadings for all land covers in the updated model runs (except peanut and winter wheat) in the Fort Cobb basin for Table 6.7 SWAT simulated loads per unit area for all land covers for the previous (hydrological calibration only) SWAT model for Table 6.8 Comparison of simulated sediment and total phosphorus loads for all SWAT model runs from VI

7 1 - Introduction The Oklahoma Department of Environmental Quality (ODEQ) is developing a Total Maximum Daily Load (TMDL) for the Fort Cobb Reservoir/Cobb Creek Basin. The Fort Cobb Basin is located in Southwestern Oklahoma in Caddo, Washita, and Custer Counties. The basin area is 314 square miles and the surface area of the Fort Cobb Reservoir is 4,100 acres. The Fort Cobb Reservoir and six stream segments in its basin are listed on the Oklahoma 303(d) list as being impaired by nutrients, pesticides, siltation, suspended solids, and unknown toxicity. The purpose of the project is to estimate total phosphorus loads to the reservoir using the Soil Water Assessment Tool (SWAT) 2000 model (Arnold et al., 1998; Arnold et al., USDA, Agricultural Research Service. Grassland, Soil, and Water Research Laboratory, 2002) for the time period 1996 to

8 2 - SWAT 2000 Input Data The SWAT 2000 model was used to estimate erosion and total phosphorus loads from the upland areas of the basin. SWAT is a distributed parameter basin scale model developed by the USDA Agricultural Research Service at the Grassland, Soil and Water Research Laboratory in Temple, Texas (Neitsch et al., 2001). SWAT is included in the U.S. Environmental Protection Agency s (EPA) release of Better Assessment Science Integrating Point and Nonpoint Sources (BASINS). Because SWAT is a distributed model, data requirements are vast and data manipulation is extensive. These requirements are met using an ArcView GIS interface, which generate model inputs using commonly available GIS data. These GIS data are summarized by the interface and converted to a form usable by the SWAT model. Topography Topography was defined by a digital elevation grid (Figure 2.1). Seamless elevation grids for the United States are available for downloading via the USGS Seamless Data Distribution System ( A 10 meter digital elevation model (DEM) was used in this study, which was a higher resolution than the common 30 meter DEM. The high resolution 10 m DEM was used to calculate subbasin parameters, such as slope, slope length, and to define the stream network. The resulting stream network was used to define the layout of the subbasins. Characteristics of the stream network, such as channel slope, length and width, were derived from the DEM. Soils Soil characteristics were defined by SWAT using soil GIS data. SWAT uses either STATSGO (State Soil Geographic Database), NRCS (Natural Resources Conservation Service) MIADS, or SSURGO (Soil Survey Geographic Database) data to define soil attributes for each soil. SSURGO data were not available for Caddo County. Therefore, NRCS MIADS (200 m) data were selected over the STATSGO data because of the higher level of detail. Applied Analysis Incorporated (AAI) Land Cover A 30-meter land cover data layer was supplied by Applied Analysis Inc. (AAI) (Appendix A). The data layer was created from Landsat TM imagery collected on June 10, Seven land covers were defined with the imagery. The land cover classes and percentages are presented below and given in Appendix A: # Barren (Bare Soil) - 0.2% # Forest - 6.7% # Pasture % # Planted/Cultivated 1-46% # Planted/Cultivated 2-5.0% # Urban - 0.5% # Water - 1.9% Planted/Cultivated categories 1 and 2 differ by the amount of vegetation present. Planted/Cultivated 1 was sparely vegetated, and Planted/Cultivated 2 was bare. The Fort Cobb Basin was primarily crop land with several different types of crops grown (i.e. corn, wheat, rye, peanuts, and others). To adequately model the basin, more detail was needed about agricultural land types. A detailed 2

9 survey was given in 2005 to Oklahoma State University (OSU) Cooperative Extension Service Agents and Specialists to gain an understanding of agricultural practices and land covers that occurred from 1996 to This survey went into great detail about the different types of crops in the basin along with different tillage practices, common double crops, fertilization rates, cattle stocking rates, and harvest dates. A copy of the survey is given in Appendix B. Results from the survey indicated that over thirty different agricultural land covers/practices occurred in the basin. With so many different agricultural land covers, SWAT would have a difficult time modeling the basin at this level of detail due to the complexity of the input files. As a result, land cover classifications were consolidated on the basis of similar agricultural practices, such as tillage type, harvest and plant dates, irrigation practices, and fertilizer rates. The final product produced twelve separate agricultural land covers, which are presented in Tables 2.1 and 2.2 along with their area percentages within the basin. The two cultivated land covers provided by AAI needed to be modified to represent all twelve crops grown in the basin. The first step was dividing crop land into irrigated and non-irrigated land, and locating center pivot irrigation locations. Center pivot irrigation locations were tagged from 2003 aerial photography to aid in accurately identifying irrigated fields (Figure 2.3) (ftp://okmaps.onenet.net, Digital Orthographic Photography, dates vary). Since the exact location of these different crops were unknown, the twelve cultivated land covers were randomly assigned throughout the basin based on a uniform distribution (Figure 2.4). It should be noted that there were other types of irrigation that could not reliably be identified from aerial photography. However, the land area from these other irrigation types was relatively minor compared to the center pivot systems. Subbasin Delineation The subbasin layout was defined by SWAT using the DEM, a stream burn-in theme, and a table of additional outlets. The stream burn-in theme consisted of digitized streams; its purpose was to help SWAT define stream locations correctly in flat topography. A modified REACH3 file from the US Environmental Protections Agency's BASINS model was used. A stream threshold value of 750 ha was used to delineate subbasins. Threshold area was the minimum contributing upland area required to define a single stream. The result was 90 subbasins. Fewer subbasins would simplify the modeling process, but this level of detail was needed to adequately represent the basin. HRU Distribution Each of the 90 subbasins were split into HRUs (Hydraulic Response Units) by SWAT. The land use [%] over subbasin area threshold was changed from the default 20% to 0%. This threshold determined the minimum percentage of any land cover in a subbasin that will become an HRU. The soil class [%] over subbasin area was also reduced from its default value of 20% to 0%. By reducing these thresholds to 0%, all land cover and soil combinations were represented. The total number of HRUs was 2,599. Weather Data Observed daily precipitation, and minimum and maximum temperatures were used in the SWAT model. Tabular weather data from the NOAA Cooperative Observation Network or COOP data (Surface Data, Daily, NOAA National Climatic Data Center, 2003) were used in all modeling. A total of six weather stations were used in the model. COOP data were seldom continuous for long periods of time. Missing days and even months were common. The period of record at the stations were typically inconsistent, so the number of active 3

10 stations may change with time. When SWAT detects missing data at a station, SWAT generates simulated weather. Therefore, gaps in a station s record are filled using interpolated data from surrounding stations. Shepherd s weighted interpolation was used, because it is computationally efficient. Shepherd s method uses weighting factors derived from the distance to nearby stations within a fixed radius: where Z o was the precipitation at the station of interest in mm, Z i was the precipitation at station i in mm, and W i was the weighting factor at station i. Weighting factors were calculated using the distance between stations: for And for where R was the radius of influence in meters, and d i was the distance from station of interest to station i in meters. Soil Test Phosphorus and Management Operations Land cover specific data, such as soil test phosphorus (STP) and fertilization practices, from were not widely available. STP for common agricultural land covers were derived from OSU Soil, Water and Forage Analytical Laboratory county level averages for the period STP data are given in Appendix D. Fertilization and management practices from were based on OSU recommendations and knowledge from local OSU Cooperative Extension Service and Conservation District personnel (Appendix B). Adjustments were made to several parameters to better represent the basin in SWAT. Minimum C Factors, part of the MUSLE equation for soil erosion, were defined based on the crop and type of tillage (Table 2.3). Also, the type of tillage implemented for each crop can have a direct effect on surface runoff and erosion. The NRCS Curve Number for each crop and tillage method was adjusted accordingly based on Rawls and Richardson (1983) and Soil Conservation Service (1972). Rawls and Richardson (1983) created a new set of Curve Numbers for conservation tillage based on residue cover and soil disturbance. For conventional tillage operations, the SCS (1972) Curve Numbers were used. Cattle Stocking Rate Verification To determine the average stocking rate used for pastures in the SWAT model, we estimated the total number of cattle in the basin and divided it by the area being grazed. County level National Agricultural Statistics Service (NASS) cattle estimates for the period were combined with land cover data to estimate the number of cattle within the basin. We assumed that cattle were evenly distributed across all agricultural land in each county. From these data we estimated the total number of cattle and calves in the basin to be 38,700 head. The SWAT model does not simulate individual cattle. Instead SWAT uses a daily biomass removal and manure deposition to represent the presence of grazing cattle. The amount of forage cattle will consume depends on the type and growth stage of the animal in question. Because there are many different types of cattle in the basin, we used the animal unit concept to define stocking rates. One animal unit can be expressed as a cow and calf pair or two-400 lb stockers; both would 4

11 consume a similar amount of forage. Because of the short duration grazing typical for small grains, we adjusted the estimate by including the duration of grazing. An animal unit*year was defined as one animal unit grazing for 365 days. The total number of animal units*years simulated in the model was 21,248. Since the NASS derived estimate is the number of cattle and calves, these estimates were not directly comparable without assuming a specific type of animal and how long the cattle were kept in the basin (Table 2.4). It is important to note that SWAT prevents grazing when the available biomass is less than the parameter BIOMIN. This parameter was set such that overgrazing was not allowed. Therefore, the actual amount of grazing in the model was less than 21,248 AU*yr. In the context of the SWAT model, it was better to overestimate stocking rates and control grazing using BIOMIN to ensure full forage utilization. This was the approach used in this study. Reservoir and Pond Input Data The size and locations of large reservoirs (> 50 acres) were taken from the U.S. Army Corps of Engineers National Inventory of Dams (NID). The total surface area and volume of these water bodies are listed by subbasin in Appendix C, and were included as reservoirs in the SWAT model. Small reservoirs and ponds (< 50 acres) were taken from USGS 7.5 minute quad maps. All known ponds and reservoirs were located and digitized into a GIS layer. The total surface area and approximate volume are listed by subbasin in Appendix C. These smaller water bodies (< 50 acres) were included as ponds in the SWAT model. Concentrated Animal Feeding Operations Only three Concentrated Animal Feeding Operations (CAFOs) were located in the Cobb Creek Basin. Approximate CAFO locations and animal numbers were taken from an Oklahoma Department of Agriculture, Food and Forestry coverage available at the Oklahoma Department of Environmental Quality website. These metadata were listed at the following web address: Details of each CAFO are given in Table 2.5. Due to the limited number and size of the CAFOs in the basin, we did not account for these in our analysis. 5

12 Figure 2.1 Ten-meter USGS Digital Elevation Model with county boundaries for the Fort Cobb Basin. 6

13 Figure 2.2 Thirty meter Applied Analysis Incorporated Landsat derived land cover with county boundaries for the Fort Cobb Basin. (Planted/Cultivated categories 1 and 2 differ by the amount of vegetation present. Planted/Cultivated 1 was sparely vegetated, Planted/Cultivated 2 was bare). 7

14 Figure 2.3 Center pivot irrigation systems in the Fort Cobb Basin identified from 2003 one m digital aerial photography. 8

15 Figure 2.4 Land cover coverage incorporating Applied Analysis Inc. land cover data, center pivot locations, and National Agricultural Statistics Service data. 9

16 Table 2.1 Applied Analysis Inc. land cover by percentage within the Fort Cobb Basin based on 2001 TM Imagery. Fraction of Basin Land Cover Type (%) Alfalfa 1.1 Bare Soil 0.2 Forest 6.2 Pasture 41.0 Peanut with Double Crop Winter Wheat (Conventional Tillage) 2.9 Peanut with Double Crop Winter Wheat (Conservation Tillage) 1.0 Peanut Winter Fallow 3.8 Rye (Conventional Tillage) 7.6 Rye (Conservation Tillage) 3.2 Grain Sorghum w/ Double Crop Winter Wheat 4.8 Grain Sorghum Winter Fallow 5.2 Urban 0.1 Water 2.4 Corn With Double Crop Winter Wheat 1.2 Winter Wheat for Grain (Conservation Tillage) 4.4 Winter Wheat for Grain (Conventional Tillage) 9.2 Winter Wheat for Pasture 5.6 Table 2.2 Crop and management breakdown based on the 2005 Fort Cobb Basin Agricultural Field Management Survey for the period Cropland Type Crop Type Farming Method % of Cropland Irrigated Cropland Peanuts Winter Fallow with a Conventional Tillage Practice 38% Peanuts Double Crop with Winter Wheat and Conventional Tilllage 29% Peanuts Double Crop with Winter Wheat and Conservation Tilllage 10% Corn Double Crop with Winter Wheat and Conventional Tilllage 12% Alfalfa Standard Farming 11% Non-Irrigated Cropland Grain Sorghum Winter Fallow with a Conventional Tillage Practice 12% Grain Sorghum Double Crop with Winter Wheat and Conventional Tilllage 13% Wheat for Grain Summer Fallow with a Conventional Tillage 23% Wheat for Grain Summer Fallow with a Conservation Tillage 11% Wheat for Pasture Summer Fallow with Conventional Tillage (Graze Out) 14% Rye Summer Fallow with a Conventional Tillage 19% Rye Summer Fallow with a Conservation Tillage 8% 10

17 Table 2.3 Minimum C factor by crop and tillage for agricultural HRUs used in the SWAT Model. Crop Tillage Min C Factor Source Wheat Conservation 0.01 NRCS, USDA and Wischmeier and Smith 1978 Wheat Conventional 0.03 NRCS, USDA and Wischmeier and Smith 1978 Peanuts Conservation 0.12 NRCS, USDA and Wischmeier and Smith 1978 Peanuts Conventional 0.19 Mutchler et al Sorghum Conventional 0.16 NRCS, USDA and Wischmeier and Smith 1978 Corn Conventional 0.20 NRCS, USDA and Wischmeier and Smith 1978 Alfalfa Conventional 0.00 NRCS, USDA and Wischmeier and Smith 1978 Rye Conservation 0.01 NRCS, USDA and Wischmeier and Smith 1978 Rye Conventional 0.02 NRCS, USDA and Wischmeier and Smith 1978 Table 2.4 Comparison of the number of cattle simulated in the SWAT model and estimates derived from the Oklahoma Agricultural Statistics Service. Used in SWAT (Animal Unit Years) Type of animal Animal Units Per Animal Duration of Grazing in the basin (Days) Equivalent Animals in SWAT NASS Estimate (Animals) Difference 21,248 Adult Cow ,500 38,700-37% 21, lb stocker ,000 38,700-10% 21,248 Cow calf pair ,500 38,700-37% 21, lb stocker ,250 38,700 58% 21,248 Adult Cow ,681 38,700 28% 21, lb stocker ,972 38,700 83% 21,248 Cow calf pair ,681 38,700 28% 21, lb stocker ,201 38, % Table 2.5 Concentrated Animal Feeding Operations in the Cobb Creek Basin. Derived from the Oklahoma Department of Agriculture, Food and Forestry database. Company Address Type Animal Units FARMERS F & F FARMS INC RT 2 BOX 37 Cattle 750 HARVEY FARMS RT 2 BOX 140 Cattle 2700 LIERLE, TERRY RT 2, BOX 143A Swine

18 3- Model Calibration and Validation Calibration is the process by which model parameters are adjusted to make its predictions agree with observed data. SWAT was designed for use on large ungaged basins and can be used without calibration. However, calibration generally improves the reliability and reduces the uncertainty of the model predictions. Validation is similar to calibration except the model is not modified. Validation tests the model with observed data that are not used in the calibration process. Flow Calibration Few stream gage data were available to calibrate the Fort Cobb Basin SWAT model for the period January 1995 to December The only suitable gage was Cobb Creek near Eakley (USGS , Appendix E). The hydrologic calibration was performed almost entirely with data from this gage. Another gage down stream of the Fort Cobb Reservoir was also utilized to corroborate the calibration (Figure 3.1). Calibration parameters from the Cobb Creek watershed were applied to all ungaged areas since older USGS stream gage data ( ) indicated that runoff volume per unit area was similar in other parts of the basin (Table 3.1). Note that Cobb Creek Near Fort Cobb is downstream of the reservoir and is subject to additional water losses (evaporation, seepage, etc.) that occur in reservoir, and therefore it is expected to have a much lower flow per unit area. Calibration was an iterative process that progresses from one model parameter to another and from course to fine parameter modifications until the difference between the model predictions and observed data met a pre-determined goodness of fit criterion. Block diagrams of these procedures are shown in Figures 3.2 to 3.4. These block diagrams are general at best, since it is not possible to represent all the specific details or decisions made by the modeler. If we could, models would be objectively calibrated via software. Many of the blocks actually represent many highly specific decisions based on many aspects of the model predications and observed data comparisons. The SWAT model was calibrated on an annual and monthly basis for flow for the period of January 1995 to December The results of the flow calibration are shown in Table 3.4 and Figures 3.5 to 3.8. As expected, R 2 decreased with shorter comparison intervals and all were considered an excellent fit to the observed data. R 2 for daily comparisons fell to 0.52, which is acceptable considering the model was not calibrated to daily flow. The Nash-Sutcliffe efficiency was used as an indicator of goodness of fit (Nash and Sutcliffe, 1970). For the calibration period, the Nash- Sutcliffe efficiencies were between 0.38 to 0.48, which was acceptable. Relative errors for daily, monthly and annual total flow were between 0 and -0.9 percent, which was also acceptable. Relative error for daily flow at the Cobb Creek near Fort Cobb gage, which was downstream of the reservoir, was 5.9%. Reservoir permeability was adjusted slightly to improve the calibration at this gage. Based on relative errors, coefficients of determination, Nash-Sutcliffe efficiencies, and the graphs the flow calibration was considered acceptable. Flow Validation The SWAT model was validated for daily, monthly and annual flow at the Cobb Creek near Eakley gage for the period January 1980 to December The hydrologic parameters used in the calibration period were used in the validation period. Flow validation indicated whether the model functions properly and predicted reasonable results under conditions outside the calibration period. The summary statistics are given in Table 3.5 and the time-series for the validation period is shown 12

19 in Figure 3.9. Even though the weather was much drier during the validation period, the model provided reasonable predictions for total, surface and base flows. Relative errors for daily, monthly and annual flow were all below 4.1 percent for total flow. Based on relative errors, coefficients of determination, Nash-Sutcliffe efficiencies, and the graphs the flow validation was considered acceptable. Nutrient Parameter Modifications Two separate models were developed, one was adjusted to more closely match observed nutrient concentration data and the second was not. Insufficient water quality and flow data were available to perform a traditional nutrient calibration. If sufficient observed water quality data were available, nutrient load estimation software, such as Loadest2 or Estimator, would have been implemented. These programs use both observed nutrient concentration and flow data to estimate nutrient load (kg/day or lbs/day). SWAT would then be calibrated by comparing these load estimates to SWAT s predicted loads. SWAT nutrient parameters were adjusted to better match total phosphorus concentration data collected throughout the basin. These nutrient parameters were modified by comparing individual instantaneous water quality observations to daily model predictions at the same location. There are several concerns when calibrating the SWAT model using daily concentration comparisons as opposed to more traditional load comparisons. These observed water quality data were grab samples, not flow weighted daily mean concentrations. SWAT predicts average daily concentrations. This difference may have a significant impact when calibrating nutrients during storm events or high flow periods immediately following a runoff event. In summary, observed instream total phosphorus concentration may vary dramatically in a single day especially when there is significant surface runoff, and SWAT predicts only the average daily concentration. There is uncertainty in calibrating a model on a daily basis for both stream flow and nutrients. The variation in observed vs predicted flow is evident in Figures 3.6 to 3.8, which increases from monthly to daily predictions. Accurate daily predictions of stream flow time requires detailed observed data at multiple locations throughout the basin, which was not available. In other words, prediction of annual or monthly stream flow will be more accurate than daily stream flow predictions. The same is true for nutrient concentrations, since calibrating the model on a daily basis for nutrient concentration is heavily dependent on stream flow. The locations of the water quality sampling sites is given in Figure 3.10 and Table 3.7. The vast majority of these observed samples were taken under base flow conditions. Base flow samples are typically not indicative of in-stream concentrations during high flow events and thus their utility are limited. A total of 231 total phosphorus samples (Table3.7) were used in this phase of the project over the period of However, only 39 samples from four days were collected during high flow events. Of these 39 samples, only nine samples obtained on 6/17/2000 may have been collected during the rising limb or near the peak of the hydrograph. The other 30 samples were collected one to four days after the storm event. These water quality data, obtained from various State and Federal agencies, are listed in Appendix F. In-stream nutrient processes may be very important during base flow conditions. SWAT does not have a fully tested in-stream model, and thus we disabled it for our study (Houser and Hauck 2002). Without in-stream processes, SWAT produces wide variations in daily nutrient concentrations governed by surface runoff events. Without the in-stream model, the streams are treated basically as pipes and all nutrients are conservative. While in-stream processes are not as important during runoff events, this results in base flow comparisons having limited utility. 13

20 It is important to note that stream flow data were not collected with most of these water samples. For example, we can only assume that the sample collected on June 17 th 2000 was during high flow and not before the storm event. This is crucial because samples may have been collected during base flow conditions or sometime before or after peak flow. The total phosphorus concentrations of these nine samples, discussed earlier, were very similar to concentrations collected during base flow conditions during the same time of the year. Also, adjusting a model with water quality data collected one to four days after a storm event can present another issue. Past studies have shown that during a runoff event total phosphorus concentrations peak before stream flows peak. After the initial flush of sediment and nutrients, the concentration of both sediment and total phosphorus quickly fall back to levels similar to base flow conditions (Baker et al., 2003; Correll et al., 1999; Lerch et al., 2001). Correll et al. (1999) and Baker et al. (2003) found that concentrations of total phosphorus fell sharply within a few hours after peak discharge. As a result of these findings, we elected to generate two models, one with a hydrologic calibration with modified nutrient parameters, and a second with a hydrologic calibration only. The nutrient parameters were modified within the SWAT manual recommended parameter ranges. Caution should be exercised when utilizing nutrient loads from either of these models. The model with the hydrologic calibration and modified nutrient parameters was not truly calibrated or validated for nutrients due to limited available data. Parameters for the hydrologic calibration with modified nutrient parameters are given in Table 3.2. Parameters for the hydrologic calibration model only are given in Table 3.3. Observed vs predicted daily total phosphorus concentrations are given in Tables 3.11 and 3.12 for the hydrologic calibrated SWAT model and the hydrologic calibrated SWAT model with modified nutrient parameters, respectively. 14