Modeling Nutrient Runoff Yields from Combined In-Field Crop Management Practices Using SWAT

Abstract

Cropland best management practice recommendations often combine tillage and nutrient application improvements to reduce nutrient losses with surface runoff. This study used the Soil and Water Assessment Tool (SWAT) model to evaluate nutrient runoff yields from conventional-till and no-till management practices with surface and deep-banded fertilizer application in a sorghum-soybean rotation. The model was calibrated for three field plots (0.39 to 1.46 ha) with different combinations of practices and validated for three field plots (0.40 to 0.56 ha) during 2001 to 2004. Daily performance of the calibrated SWAT model in simulating total N for all treatments was satisfactory for median-based Nash-Sutcliffe model efficiency (Ef* of 0.54 to 0.64), good to very good for percent bias (PBIAS of 31% to 7%), and satisfactory to good for median-based root mean square error to observations standard deviation ratio (RSR* of 0.72 to 0.62). Performance was slightly lower and more variable for total P calibration (Ef* of 0.42 to 0.62, PBIAS of -48% to 2%, and RSR* of 0.76 to 0.62). Monthly statistics improved for total P runoff yield compared to daily performance, but changed little for total N runoff yields, probably due to the stronger influence of outliers in the N data. Based on validation results, SWAT was more robust in simulating total N runoff yields from the treatment with less soil disturbance (NT/SB) and total P for the two treatments with more soil disturbance (NT/DB and TILL). A major concern was that SWAT predicted greater annual average total N runoff yields for no-till treatments than for tilled treatments, which was contrary to measured values at the study site. This reinforces a fundamental research issue that tillage system effects on nutrient losses are still very much uncertain and thus may not be properly modeled. The SWAT model generally underpredicted monthly total N yields for all treatments in the higher-precipitation months of May and June and overpredicted total N and total P yields from September through November. Calibration for N and P resulted in identical calibration parameters for NPERCO (1.0), RSDCO (0.05), BIOMIX (0.2), PPERCO (10), PHOSKD (175), and UBP (50) regardless of tillage practice or fertilizer application method. Together with results that calibrated parameters for runoff (CN, Ksat, AWC) and erosion (Cmin) differed among the treatments, this study found that differences in nutrient yields among tillage and fertilizer management may be adequately modeled with SWAT by calibrating runoff and sediment yields only, and that further calibration of nutrient parameters may not improve model results.