TY - JOUR T1 - Use of soft data for multicriteria calibration and validation of Agricultural Policy Environmental eXtender: Impact on model simulations JF - Journal of Soil and Water Conservation SP - 623 LP - 636 DO - 10.2489/jswc.73.6.623 VL - 73 IS - 6 AU - A.M. Nelson AU - D.N. Moriasi AU - M. Talebizadeh AU - J.L. Steiner AU - P.H. Gowda AU - P.J. Starks AU - H.K. Tadesse Y1 - 2018/11/01 UR - http://www.jswconline.org/content/73/6/623.abstract N2 - Use of soft data and multiple model performance criteria in model calibration and validation are critical to ensure that environmental models capture major hydrologic and water quality processes. The Agricultural Policy/Environmental eXtender (APEX) is a hydrologic and water quality model widely used for evaluating the effect of agricultural production management practices on the environment. However, there are few studies on the impact of soft data on APEX model outputs. This study sought to determine the impact of soft data (i.e., crop yields, evapotranspiration [ET], and drainage) used to constrain calibration on APEX model to simulate outputs (streamflow, total nitrogen [TN], and total phosphorus [TP] losses), as well as the impact of tile drainage on the model outputs to help identify issues with APEX. Long-term measured water quality and quantity data from Rock Creek watershed, northern Ohio, were used for this purpose. Four calibration scenarios were used: (A) streamflow statistics only, (B) streamflow statistics plus soft data, (C) streamflow plus water quality, and (D) streamflow plus water quality plus yields plus soft criteria. The number of parameter sets that meet the criteria decreased from 3,325 to 2, as the number of calibration criteria increased. In general, when soft data were used to constrain the model, the Nash-Sutcliffe efficiency (NSE) values decreased for all components. For example, NSE for streamflow decreased from 0.79 for scenario A to 0.56 for scenario B when the soft data were used, and NSE for TP decreased from 0.34 for scenario C to 0.30 for scenario D. However, the scenarios that used soft data with lower NSE values simulated the biophysical processes more accurately. For all calibration/validation scenarios, streamflow, TN, and TP values were drastically reduced when the model was run without tile drainage. However, there was hardly any tile drainage impact on ET, surface runoff, or crop yields for all scenarios, indicating that improvements in tile drainage, water table, and crop growth routines may be needed. Overall, the results show the importance of constraining the model parameters to produce reasonable values of crop yield and water balance closure to obtain realistic simulations of various management practices. ER -