ASAE Conference Proceeding
This is not a peer-reviewed article.
Evaluation of SWAT on Modeling Nitrate-Nitrogen in Soil Profile and Stream discharge for Walnut Creek Watershed with Tile and Pothole
B. Du, A. Saleh, J. G. Arnold, and D. B.Jaynes
Pp. 509-517 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
Contamination of surface and subsurface waters with nitrate-nitrogen (NO3-N) has been related to agriculture across the Midwestern USA. The SWAT model was recently modified to better describe NO3-N fate and transfer within tile and pothole systems. The modified SWAT was evaluated using measured data from Walnut Creek watershed (WCW) located in central Iowa. These data include 7 years of measured NO3-N loading data in stream discharge and 4 years of soil NO3-N data at 105-cm depth. The model was calibrated to the period of 1992 to 1995 and validated to the period of 1996 to 1998. Nash-Sutcliffe E values was used to evaluate the accuracy of the model. The results show that the pattern of predicted and measured NO3-N loads in stream discharge at the center and outlet of WCW during the validation period was reasonably close (E value of 0.85 and 0.70, respectively). However, the daily prediction of NO3-N loads in stream discharge were not as good as monthly (E = 0.51 for the center and 0.37 for the outlet of WCW during the validation period). The model reasonably simulated monthly NO3-N loads in subsurface flows (E values=0.79 and 0.68, respectively for the calibration and validation periods), though improvement is needed in the simulation of daily subsurface NO3-N loads (E values = 0.47 and 0.38). A reasonably good pattern between measured and predicted soil NO3-N (for 105-cm depth) was found for all simulated soil types. The E values for predicted NO3-N in Okoboji, Clarion and Canisteo soil profiles during 1992 to 1995 were 0.61, 0.73 and 0.68, respectively.Keywords. Water quality, Nitrate nitrogen, Soil profile, Modeling, Tile drains, Pothole, Watershed, SWAT.
Higher fertilizer application rates in agricultural lands have resulted in greater concern about nitrate-nitrogen (NO 3 -N) pollution into water sources. Modeling nitrogen cycle in soils and streams is an efficient method to predict the water contamination due to nitrogen over-fertilization. Therefore, numerous models such as SWAT (Soil and Water Assessment Tool) (Neitsch et al., 2002) have been developed to simulate the nitrogen cycle at field and watershed levels. The SWAT model recently was modified to better simulate landscapes with tile and potholes drainage systems (Arnold et al., 2003). The modified SWAT was evaluated using the data obtained from Walnut Creek watershed (WCW) dominated with tile and pothole drainage systems (Du et al., 2003). The results of this evaluation showed that the overall performance of SWAT in predicting surface and subsurface flow and NO 3 -N loading in WCW was satisfactory. In SWAT the main part of NO 3 -N is: 1) consumed by plant; 2) stored in soil profiles; and 3) lost from soil profiles by means of denitrification process, leaching into underground water, and transporting into streams by means of tile drain and/or surface runoff. Therefore, during calibration and validation process one needs to consider all the above factors. However, past researchers have focused their efforts mainly on evaluation of SWAT on predicting NO 3 -N losses from agricultural lands with little consideration for NO 3 -N content within soil profiles (Arnold et al., 1999; Spruill et al., 2000; Saleh et al., 2000).
The main objective of this study is to evaluate the NO 3 -N simulation in both soil profile and stream discharge by the modified SWAT model with tile and pothole components, using measured data from WCW located in the central Iowa.
Materials and Methods
Watershed Description and Sampling
The 5130 ha WCW, located in Story county, central Iowa, is typical of the poorly drained, gently rolling landscapes of central Iowa row cropping areas. This landscape was formed on young till plane and contains numerous closed depressions or potholes as a result of a poorly-developed, geologically-young surface drainage network. These potholes often fill with water, especially during snowmelt and after heavy rainfall, which can result in a reduction in crop yields. The upland soils are underlain by a dense unoxidized till that restricts vertical drainage resulting in poorly drained soils in the lower elevation areas. A corn-soybean rotation cropping system is predominately used in this area.
The watershed has an average elevation of about 300 m above sea level. The measured average annual precipitation during the 7 years of simulation was approximately 820 mm, and the average temperature during crop growth seasons ranged from 9.0 to 23.0 o C.
NO 3 -N concentrations in stream discharge were intensively monitored at sites 330, 310 and 210 of the watershed from 1992 to 1998 by USDA-ARS, National Soil Tilth Lab (Jaynes et al., 1999). The monthly NO 3 -N loads were the summation of the scattered measured daily values. Soil sampling in the area covering 38 ha in the western part of the WCW for soil NO 3 -N up to a depth of 105-cm from 1992 to 1995 was conducted by Cambardella et al. (1999).
Precipitation data measured at 17 weather gages within the watershed (Figure 1) were used in SWAT. The measured daily maximum and minimum temperature at two locations within the watershed and solar radiation measured at one station were also used for SWAT simulations.
Sites 310 and 330, located at the center and outlet of WCW, respectively, were used to evaluate overall simulation of NO 3 -N in stream discharge. Subsurface NO 3 -N content
measurements at site 210 were used to evaluate SWAT’s simulation of subsurface NO 3 -N loads.
Input data and Model assumptions
The digital elevation, land use, and soils maps, and measured daily precipitation, temperature and solar radiation for the watershed were provided during the initial setup of the input data files for SWAT using the ArcView interface for SWAT2000 (AVSWAT) (DiLuzio et al., 2002). Other input data such as daily wind speed and relative humidity were generated by SWAT from long term monthly statistics. The Penman/Monteith within SWAT was selected for potential ET calculation.
Eighty-seven% [must spell out words at first of sentence] of the WCW was covered by corn and soybean crops while other land uses including croplands such as wheat and alfalfa, roads,
and forest occupied the remaining 13% of the area. Continuous corn production occurred on 15 % of the total farmland, while 85% of the area was in a corn-soybean rotation (Hatfield et al., 1999).
The seven predominated soils of Clarion, Webster, Canisteo, Lester, Harps, Okoboji and Lester were used. The very poorly drained Okoboji and Harps soils were assigned to potholes. The Clarion, Canisteo and Okoboji soils were selected to evaluate the simulation of NO 3 -N in soil profile (Table 1). Based on the Figure 2 and Hatfield et al. (1999), it was assumed that about 66% of the total watershed area was tile drained and 57% of the total surface runoff directly flowed into potholes. Total pothole area occupied 10% of the total land use. Table 1 shows the overall outline of landuse soils, tiles and potholes used by SWAT.
Most nitrogen fertilizers were applied in fall in form of anhydrous ammonia. However, in 1992 because of unsuitable weather condition, nitrogen fertilizer was applied following soybean (Cambardella et al., 1999). The fertilizer application rates from 1992 to 1998 in the entire WCW were determined from farmer surveys (Hatfield et al., 1999). The annually averaged nitrogen fertilizer rate for corn fields was 166 kg/ha. The phosphorous fertilizer rate was 46.6 kg/ha.
A standard tile drain at the depth of 1.2 m is used throughout WCW. Initially, 3 to 4 soil layers were created by AVSWAT interface for various soils covering the WCW. However, for precise description of depth of tile drains, the number of soil layers was modified to 7 for all soils. Tile drains are usually designed to reduce the soil water content to field capacity within 48 hours, so the initial value of tile drain (time to drain soil to field capacity) within SWAT was set at 48 hours.
Model evaluation and calibration methods
SWAT were calibrated for the period of 1992 to 1995 and validated for the period of 1996 to 1998 for NO 3 -N loads in stream discharge. SWAT source codes were modified in order to output NO 3 -N in a soil profile of 105-cm depth on certain days. No calibration was done especially for the simulation of soil profile NO 3 -N. In other words, the accuracy of the SWAT’s simulation of soil profiles NO 3 -N content was directly evaluated after NO 3 -N loads in stream discharge was calibrated without additional model adjustments.
The Nash-Sutcliffe model efficiency (E) (Nash et al., 1970) was used to compare the measured and predicted soil profile NO 3 -N, NO 3 -N loads in stream discharge and subsurface tile flow.
Nperco (nitrate percolation coefficient) is often used for NO 3 -N calibration. In this study, the tdrain (drain tile lag time), gdrain (tile drain lag time), and cmn (mineralization factor of active organic nutrients were also modified for the NO 3 -N calibration.
Results and Discussions
The distinct patterns of NO 3 -N in a soil profile of 105 cm depth related to the crop, time of season, and soil type produced by the model (Figure 3) were relatively close to those reported by Cambardella (1999). The model simulation showed that soil profile NO 3 -N was usually less than 50 kg/ha during the soybean planting years (1993 and 1995) and much higher with strong seasonal changes during the corn planting years (1992 and 1994) (Figure 3). The predicted amounts of NO 3 -N in the soil profiles on the day 139, 150 and 163 of 1992 and the day 138 and 152 of 1994 were much higher than those predicted on other days during the early and late corn-growing season. During the fall of 1992 (day 311), the predicted NO3-N (20 kg/ha averaged for the three soils) by SWAT was significantly less than the observed value (averaged 68 kg/ha) (Figure 3). . These unexpected high values of NO 3 -N in the soil profile, as compared to that of similar days in 1994, could be due to favorable soil temperature and moisture conditions for increasing N mineralization process (Cambardella et al., 1999). E values for the predicted NO 3 -N for Okoboji, Clarion and Canisteo soils from 1992 to 1995 were 0.61, 0.73 and 0.68, respectively, which showed that the model predicted the NO 3 -N contents within profiles of three soils reasonably well.
The model under-predicted some peaks of NO 3 -N in Clarion and Canisteo soil profiles during the corn cropping years. In 1994 during the soybean-cropping year, the predicted NO 3 -N in Okoboji soil profile was lower than the measured values and higher in Canisteo soil profiles (Figure 3). The model prediction of soil NO 3 -N was much lower than the measured for all three soils during day 91 of 1992. This under prediction could be the result of inaccurate initial values of soil chemicals being set in SWAT. The properties of Okoboji soil were quite different from that of other two soils (Table 1). Okoboji contained less sand and more clay, which resulted in faster cumulating of ponded water on Okoboji surface than other two soils during significant precipitation events. Consequently, it produced a different pattern of NO 3 -N in Okoboji soil than in Clarion and Canisteo. This was verified by the fact that the measured NO 3 -N contents in Okoboji soil during corn growth season (days 138 and 152) of 1994 were much lower than that in other two soils (Figure 3). The model predicted this trend with NO 3 -N mean of 102.4 kg/ha in Okoboji soil profile, which is lower than 136.2 and 130.9 kg/ha for Clarion and Canisteo soils, respectively.
Although anhydrous ammonia was applied in November of 1993, both field measurements and SWAT simulations show that the significant nitrification of applied fertilizer occurred by the middle of May (day 138) of 1994 (Figure 3). This is probably as expected mainly due to increasing soil temperature during the spring of 1994. Figure 3 also indicates the precision of SWAT model in predicting NO 3 -N within soil profiles during the nitrification period. (Day 306 of 1993 though day 138 of 1994).
Stream NO 3 -N loads at sites 310 and 330 of WCW
Figures 4a-b showed the monthly simulated vs measured NO 3 -N loads from the center (site 310) of WCW during 1992 to 1998. The patterns of predicted and measured soil NO 3 -N during both calibration and validation periods look very good. This was verified by the high E values up to 0.82 and 0.85 for monthly NO 3 -N for the calibration and validation periods, respectively. At the outlet (site 330) of WCW, although the model underestimated the monthly NO 3 -N loads during the calibration period (Figure 4c), its E value still reached 0.60. Both the figure (Figure 4d) and monthly E values (0.70) for the validation period for the outlet of WCW indicated that the monthly loadings of NO 3 -N were fairly satisfactorily predicted. The model’s predictions of the daily NO 3 -N loads was less accurate than those predicted at the monthly level (the E values of daily NO 3 -N loads at the center and outlet of WCW were only 0.51 and 0.37 during the validation period,
NO 3 -N loads in subsurface flow of site 210
Figures 5a-b shows that the simulated subsurface NO 3 -N loads were higher than the measured values during some months of calibration and validation periods and lower during the other months. However, the model seems to predict the monthly subsurface NO 3 -N loads from site 210 reasonably well (E values for the calibration and validation periods were 0.79 and 0.68, respectively). Lower daily E value of 0.42 averaged for the calibration and validation periods indicated the further improvement of SWAT model is needed for the simulation of daily NO 3 -N loads in subsurface flow. The daily simulated and measured NO 3 -N loads in subsurface flows at site 210 are shown Figures 5c-d.
Summary and Conclusions
The SWAT model with modified tile drain and pothole components was evaluated at a watershed scale using the 7 years of measured data of NO 3 -N in stream discharge and 4 year NO 3 -N data in a soil profile of 105-cm depth in WCW. The centrally located site (310) and outlet site (330) of the watershed were selected to investigate overall performance of the SWAT model, while site 210 was used to evaluate the model’s capability of simulating subsurface NO 3 -N loads. The high E values for monthly NO 3 -N loads in stream discharges at the center and the outlet of WCW (E = 0.85 and 0.70, respectively) indicates that SWAT’s predictions were close to the measured values. Nevertheless, the predicted daily NO 3 -N loads were less accurate than the monthly predictions as compared to measured values (E values of daily NO 3 -N loads at the center and outlet of WCW were only 0.51 and 0.37 during the validation period, respectively). The results from site 210 indicated that the model predictions of the monthly NO 3 -N loads in subsurface flow (E values=0.79 and 0.68, respectively for the calibration and validation periods) were better than of the daily predictions (E values = 0.47 and 0.38, respectively for the calibration and validation periods).
The simulated NO 3 -N at 105-cm depth of soil profile was similar to those reported by Cambardella (1999). The E values for the predicted NO 3 -N contents in Okoboji, Clarion and Canisteo soil profiles from 1992 to 1995 were 0.61, 0.73 and 0.68, respectively, indicating that the modified SWAT was able to predicted the daily NO 3 -N at the three soils’ profiles reasonably well
Arnold, J.G., R. Srinivasan, R.S. Muttiah, and J.R. Williams. 1998. Large area hydrologic modeling and assessment Part I: Model development. J. Of the American Water Resource Assoc. 34(1): 73-89.
Arnold, J.G., P.W. Gassman, K.W. King, J.R. Williams, A. Saleh, and U.S. Tim. 1999. Validation of the subsurface tile flow component in the SWAT model. ASAE/CSAE-SCGR Annual International Meeting.
Arnold, J.G., B. Du, A. Saleh, D.B. Jaynes. 2003. Application of Soil and Water Assessment Tool (SWAT) to landscapes with tiles and potholes. Part I. Development of new procedures. (In processing).
Cambardella, C.A., T.B. Moorman, D.B. Jaynes, J.L. Hatfield, T.B. Parkin, W.W. Simpkins, and D.L. Karlen. 1999. Water Quality in Walnut Creek watershed: Nitrate-nitrogen in soils, subsurface drainage water and shallow groundwater. J. of Environ. Qual. 28(1):25-34
DiLuzio, M., R. Srinivasan, J. Arnold. 2002. ArcView interface for SWAT2000: user’s guide. TX, Temple: Blackland Research Center.
Du, B., A. Saleh, J.G. Arnold and D.B. Janyes. 2003. Application of Soil and Water Assessment Tool (SWAT) to Landscapes with Tiles and Potholes. Part II. Validation of New Procedures for Case of Walnut Creek Watershed (Iowa). ASAE annual international meeting.
Hatfield, J.L., Jaynes, D.B., Burkart, M.R., Cambardella, C.A., Moorman, T.B., Prueger, J.H. and Smith, M.A. 1999. Water Quality in Walnut Creek watershed: Setting and farming practices. J. Environ. Qual. 28(1): 11-24.
Jaynes, D.B., J.L. Hatfield, and D.W. Meek. 1999. Water Quality in Walnut Creek watershed: Herbicides and nitrate in surface waters. J. Environ. Qual. 28(1): 45-59.
Jaynes, D.B. and John G. Miller. 1999. Evaluation of the Root Zone Water Quality Model Using Data from the Iowa MSEA. Agronomy J. 91:192-200.
Neitsch, S.L., J.G. ARNOLD, J.R. KINIRY, R. SRINIVASAN, J.R. WILLIAMS. 2002. Soil and Water Assessment Tool User’s Manual. TX, Temple: Blackland Research Center.
Nash, J.E., and J.E. Sutcliffe. 1970. River flow forecasting through conceptual models. Part1 – A discussion of principles. J. Hydrol. (Amsterdam) 10:282-290.
Saleh, A., J.G. Arnold, P.W. Gassman, L.M. Hauck, W.D. Rosenthal, J.R. Williams, A.M.S. McFarland. 2000. Application of SWAT for the upper north Bosque river watershed. Trans.ASAE.43 (5): 1077-1087.
Spruill, C.A., S.R. Workman, J.L. Taraba. 2000. Simulation of daily and monthly stream discharge from small watersheds using the SWAT model. Trans.ASAE. 43(6): 1431-1439.