ASAE Conference Proceeding
This is not a peer-reviewed article.
Calibration of the GIS-SWAT Model for the Simulation of Phosphorous Export in Turfgrass Sod in the North Bosque River Watershed.
G. R. Stewart, C. Munster, D. M. Vietor, C.E. Richards, I. Choi, B. McDonald
Pp. 184-189 in Total Maximum Daily Load (TMDL) Environmental Regulations II, Conference Proceedings, 8-12 November 2003 (Albuquerque, New Mexico, USA), ed. Ali Saleh. ,8 November 2003 . ASAE Pub #701P1503
The Upper North Bosque River watershed is an impaired watershed due to high loadings of P. The watershed’s primary agricultural activity is dairy production and large quantities of manure are produced. A best management practice that has been proposed is to use composted dairy manure in commercial turfgrass operations to remove manure from the watershed. When turfgrass sod is harvested a thin layer of soil is also removed. Therefore excess P bound to the soil can be transported out of the watershed in a sustainable manner. The Soil and Water Assessment Tool (SWAT) was calibrated to simulate the effectiveness of using turfgrass sod fertilized with composted dairy manure to export P from the watershed. The model was first calibrated to predict average monthly flows and sediment loadings from 1996 to 1999. The Nash-Sutcliffe model fit efficiency was used for evaluating the model. The model fit efficiency was 0.81 for flow and 0.53 for sediment loading.KEYWORDS. TMDL, Watershed Modeling, Agricultural BMPs, SWAT Model, Turfgrass, P, Dairy Manure.
The Upper North Bosque River (UNBR) Watershed in North Central Texas is one of the most studied watersheds in the U.S. This watershed, covering most of Erath County, has one of the highest concentrations of dairies in the nation with over 38,000 cows (TWRI, 1998). The concentration of dairy manure has led to high levels of nutrients, primarily phosphorous (P), in the UNBR. An assessment of the Total Maximun Daily Loads (TMDLs) for the UNBR led to a recommendation of a 50% reduction of soluble P (TNRCC, 2001). Hydrologic simulations of the UNBR using the Soil and Water Assessment Tool (SWAT) were instrumental in this TMDL assessment. In addition, extensive research and SWAT model simulations have gone into the development of best management practices (BMPs) to reduce P loading to the UNBR.
A new BMP proposed for P reduction in the UNBR watershed is the use of turfgrass sod to export manure nutrients. The use of composted dairy manure as fertilizer for commercially grown sod would bind a high percentage of the manure’s P in the soil layer. Then the sod, soil and P would be harvested and exported out of the impaired watershed in a sustainable manner. Ongoing research at Texas A&M University is evaluating the percentage of nutrient export in turfgrass sod, at both plot and field scales. One objective of this research is to use the SWAT model to simulate water quality improvements due to the implementation of the turfgrass BMP in the UNBR watershed.
Suitable sites for turfgrass production are already determined from previous research (J. Hanzlik, 2003). Therefore, the goal of the SWAT simulations will be to assess the effectiveness of turfgrass farms to help achieve the TMDL for the UNBR.
The Upper UNBR, located almost entirely in Erath County, is part of the Environmental Protection Agencies (EPAs) hydrologic unit (HUC) # 12060204. The U.S. Geological Survey (USGS) has a gauging station (Gauge No. 08094800) in Hico, TX that serves as the watershed outlet in the model simulations. The UNBR watershed which is 93,250 ha, receives an annual rainfall of approximately 75 mm, and has a daily average temperature range from 6 to 28ºC. This watershed represents the head waters of the North Bosque River, which flows into Lake Waco and provides drinking water to over 200,000 central Texas residents.
The predominant land cover in the UNBR watershed is rangeland and the major agricultural activity is dairy production. There are approximately 127, with a recent trend towards fewer but larger dairies (TWRI, 1998). There is some row crop agriculture as well as some forested areas in the watershed, but these are a small percentage of the landuse/landcover. The main community in the watershed is Stephenville with a population of approximately 15,000(US Census, 2000). The Stephneville waste water treatment plant is the only significant point source discharge in the watershed. Stephenville and part of the town of Dublin plus other small communities make up the urban part of a 98% rural watershed.
Figure 1 . The UNBR watershed with sub-basin distribution used in the SWAT model.
The high dairy concentration produces close to 200,000 tons of manure (50% moisture content) in the watershed (Norvell, 1998). This dairy waste contributes to high nutrient concentrations in the UNBR. The UNBR was deemed an impaired water body in 1998 by section 303(d) of the Clean Water Act (USEPA, 1998). A TMDL that mandated a 50% reduction of soluble P was established by the Texas Commission on Environmental Quality (TCEQ) (TNRCC, 2001). Currently the TCEQ and the Texas State Soil and Water Conservation Board (TSSWCB) are in charge of developing and implementing a plan to achieve this TMDL.
One of the proposed BMPs to help reduce P loading is the use of commercial turfgrass farms to export excess nutrients out of the watershed. Research has determined that between 46% and 77% of the P from dairy manure is removed in the harvest and between 36% and 47% of nitrogen, depending on turf species and other environmental conditions (Vietor et al. 2002). Currently Texas A&M University is performing field and plot scale research to confirm this removal rate and to further study the effectiveness of composted dairy manure as a fertilizer.
The objective of this modeling project is to simulate the effects on water quality of large scale turfgrass production in the UNBR watershed. The SWAT 2000 model that is packaged in the EPA’s Better Assessment Science Integrating point and Non-point Sources (BASINS) software package was used for the simulation. The SWAT model is ideal for watershed scale, continuous-time simulations, and the ArcView interface allows for very detailed spatial data inputs (Neitsch et al., 2001).
In this study, the SWAT model was used to simulate monthly flows and P concentrations from 1990 to the end of 2002. Calibration of the model is done with the observed data from 1996 to 1999.
Land Use, topography and soils
A Digital Elevation Model (DEM) from the National Elevation Dataset (NED) was used to provide topographical data for the model simulations. This dataset, provided online by the USGS, has 30 m pixel resolution (USGS, 2003). The National Hydrography Dataset (NHD) stream network for the watershed was digitally burned into the DEM using BASINS. This ensures correct placement of streams for accurate water routing.
The soil types for UNBR watershed were derived from the Soil Survey Geographic (SSURGO) dataset developed by the National Resource Conservation Service (NRCS). This dataset was manipulated to link the spatial cells with the correct soil type and characteristic databases since SWAT was setup for State Soil Geographic (STATSGO) soil inputs. This ensures correct curve numbers and soil properties with much improved spatial detail.
The land use data, obtained from the USGS as part of the National Land Cover Dataset (NLCD), provided a 30 m resolution grid for land use in the area. The land use types in the dataset were not directly compatible with SWAT, so SWAT equivalent land uses were selected.
Land and Crop Management
Land management and crop cycles in the UNBR watershed were gathered from prior calibrations of the model. Philip Gassman calibrated the Agricultual Policy Environmental Extender (APEX) model with great detail in land use management, and crop rotation (Gassman, 1997). The detailed inputs were converted into SWAT crop management scenarios for each of the agricultural land use types. Gassman’s detailed land management report was complemented with evaluations by Todd Adams of the Texas Institute for Applied Environmental Research (TIAER).
For the UNBR watershed there are numerous weather stations that log daily data to the National Climatic Data Center (NCDC). In addition TIAER also has several rain gauges throughout the watershed. A total of 11 weather stations with rainfall data were used for the calibration, of which two had temperature data. Six of these stations were from the NCDC with online data (NCDC, 2003). The other five precipitation data sets were from TIAER’s monitoring of the UNBR.
Flow, Sediment and Nutrients
TIAER’s monitoring station in Hico, TX (BF040) was used for flow and water quality calibration of the SWAT model. This station provided daily flow data for the simulated watershed outlet at Hico. In addition two other monitoring stations (NF050 and BF040) upstream and downstream of Stephenville were also used for the calibration process. Along with the collection of flow data, TIAER also has a water quality program that includes storm sampling and bi-weekly grab sampling at these locations. These samples from the UNBR that were analyzed for sediment and nutrient concentrations were used to calibrate the SWAT model.
TIAER also provided both nutrient concentrations and flow rates for the effluent from the Stephenville wastewater treatment plant. The data came from the required monthly reports of the wastewater treatment plant.
Calibration of the model to match monitored data from 1996 to 1999 was very successful. The Nash Sutcliffe (Nash et al., 1970) equation was used to assess the model calibration by comparing measured and simulated values. This equation returns a model fit efficiency (E) with 1.0 representing a perfect model fit.
Initial calibration of the model for stream flow, over predicted measured flow rates. Flow rates were reduced by decreasing the curve number values (CN) by 8%, and increasing the evapotranspitration (ET) coefficients (ESCO). In addition the model over-predicted base flow. Therefore the base flow delay (Alpha_BF) and the groundwater evaporation coefficients (GWQMN) were increased (Table 1). After calibration modeled flow matched measured flow very well (E = 0.81) at the watershed outlet. A comparison between modeled and measured flow for the calibration period is shown in Figure 2.
Figure 2 . Measured and simulated average monthly flows in the UNBR
Table 1 . SWAT model variables adjusted during flow calibration.
Curve Number (CN)
ET Coeff. (ESCO)
Ground Water Delay (Alpha_BF)
Ground Water Evap. (GWQMN)
The second step was to calibrate the SWAT model for sediment loads. Since sediment flux is partly driven by flow, the calibration for sediment was simpler due to the availability of accurate simulated flow. Initially the model over predicted sediment load. To reduce simulated sediment loads, the variables that determine sediment transmission or settlement during routing were adjusted (Table 2). In addition, the coefficients of the Universal Soil Loss Equation (USLE) were also decreased. Routing of sediment was decreased by lowering the sediment transmission coefficients of the main channels along with increasing the channel cover and decreasing channel erodability factors. To further decrease sediment, the average slope and slope length of the hydrologic response units were decreased and both the USLE crop and practice coefficients were decreased. In the end a good match (E = 0.53) was achieved for sediment over the calibration time period. The results of the calibration can be compared to measured values in figure 3.
Table 2 . SWAT model variables adjusted during the sediment calibration
Slope Length (SLSUBBSN)
Channel Cover (CH_Cover)
Channel Erodability (CH_EROD)
Management Practices Coeff.(USLE_P)
Crop Coefficients (USLE_C)
Linear Channel Transmission (SPCON)
Exp. Channel Transmission (SPEXP)
Figure 3 . Measured and simulated average monthly sediment concentrations in the UNBR.
The model calibration will serve as the basis for the simulations to test turfgrass sod BMPs in the UNBR watershed. The extensive model calibration ensures should lead to accurate simulations of the UNBR to assess water quality improvements due to the implementation of turfgrass BMPs.
The model fit efficiencies (E) for flow rate and sediment were all greater than 0.53. This high model fit efficiency ensures accurate simulation of the runoff and erosion processes within the UNBR watershed. Additional calibration will ensure accurate modeling of P loadings from the watershed. P loading will be calibrated for both mineral-P (PO 4 ) and organic-P. Latter validation with crop yield data will refine this calibration. Make the SWAT model a solid base for the simulation of water quality impact due to the establishment of turfgrass operations in the UNBR watershed. In addition, to improve model algorithms for slow release organic fertilizers should also increase model accuracy.
Special thanks to Dr. Anne MacFarland, Dr. Ali Saleh and Todd Adams at TIAER. Plus Jason Afinowicz and Chad Richards for help with the SWAT Model. In addition Dr. Clyde Munster and Dr. Don Vietor at TAMU for guidance.
Census. 2000. Erath County QuickFacts from the US Census Bureau. Available at: http://quickfacts.census.gov/qfd/states/48/48143.html. Accessed: July 2003.
Gassman, Philip W. 1997. NPP integrated modeling System: Environmental Baseline Assumptions and Results for the APEX Model. Staff Report 97-SR 85. Ames, Iowa: Center for Agricultural and Rural Development, Iowa State University.
Hanzlik, J.E. 2003. Location of Turfgrass Production Sites for Phosphorous Removal from an Impaired Watershed. Unpublished. M.S. Texas A&M Department of Biological and Agricultural Engineering.
NCDC. 2003. National Climatic Data Center. Available at: http://www.ncdc.noaa.gov/oa/ncdc.html. Accessed: Jan 2003.
Nash, J.E., and J.E. Sutcliffe. 1970. River Flow Forecasting through Conceptual Models. Part 1—A discussion of Principles. discussion of Principles. J. Hydrol. (Amsterdam) 10:282-290
Neitsch, S.L., Arnold, J.G.. Kiniry, J.R. Williams, J.R. 2001. SWAT: Soil and Water Assessment Tool (User Manual). Temple, Texas: Grassland, Soil and Water Research Laboratory, USDA Agricultural Research Service.
Norvell, S. 1998. An Overview of the Dairy Industry: Erath County, Texas. TIAER Information Bulletin 98-02. Texas Institute for Applied Environmental Research. Stephenville Texas.
TNRCC. 2001. Two Total Maximum Daily Loads for Phosphorous in the North Bosque River for Segments 1226 and 1255. Austin: State of Texas.
TWRI. 1998. Texas Water Resource Institute, Upper North Bosque River Project. Available at: http://twri.tamu.edu/twripubs/WtrResrc/v21n2/text-1.html. Accessed: July, 2003.
U.S. EPA, 1998. Section 303(d) List Fact Sheet for Watershed NORTH BOSQUE. Available at: http://oaspub.epa.gov/pls/tmdl/huc_rept.control?p_huc=12060204&p_huc_desc=NORTH%. Accessed: July, 2003.
Vietor, D.M., Griffith, E.N., White, R. H., Provin, T.L., Muir, J. P., Read, J.C. 2002. Export of Manure P and N through Turfgrass Sod. Journal of Environmental Quality . 31:1731-1738. College Station, TX: Texas A&M University.