干旱区生态环境知识资源中心

Reliable estimation of ET is especially important in semi-arid conditions where ET prediction is instrumental in cost-effective management of scarce water resources and crop production. In addition to the role of climate variables and soil characteristics, the actual ET is influenced by dynamic crop characteristics during the growing season. Therefore, accounting for the interaction between crop growth and actual ET can significantly improve the performance of models. In this study, an efficient approach is presented for simultaneous calibration of ET and crop growth parameters for the Agricultural Policy/Environmental eXtender (APEX) model for daily, weekly, and monthly ET. The proposed approach involves the development of an objective function based on a compounded time series comprising of scaled annual ET and crop yield data. An efficient search algorithm based on the DiffeRential Evolution Adaptive Metropolis (DREAM) algorithm was implemented in R language to find the parameter values that minimizes the defined objective function. The simultaneous calibration approach, which utilized annual data, improved ET prediction for annual and finer time scales (25% in RMSE reduction for 3-month, 18% for monthly, 17% for 2-week, 19% for weekly, 17% for 3-day, and 13% for daily time scales). The average of absolute relative error for crop yield predictions was also reduced from 43% to 16% for the calibrated model. The simulated leaf area index (LAI) for the calibrated model (i.e. calibrated using annual ET and crop yield data) was also consistent with the measured LAI values, confirming the validity of the calibrated parameter values.