Identification of the Most Sensitive Parameters of Winter Wheat on a Global Scale for Use in the EPIC Model

Abstract

Core Ideas Using computer technology, crop models are widely used in agricultural research and applications. Sensitivity analysis can screen sensitive parameters to calibrate the EPIC model for different regions. WA, HI, and TBS had a greater influence on simulated wheat yield than other parameters. Elevation had a higher impact on the sensitivity of crop parameters, especially on TBS (0.53) and HI (–0.47). These results can increase the efficiency of regional calibration of the EPIC model on a large scale. The purpose of this study was to identify the most sensitive parameters (MSPs) of the Environmental Policy Integrated Climate (EPIC) model for winter wheat ( Triticum aestivum L.) simulations across the globe. First, 48 crop parameters related to the growth process of winter wheat were selected from the EPIC model. The range and distribution were assigned to each crop parameter. Second, a number of samples of parameter combinations (SPCs) were generated, containing 48 random values of model parameters for each grid. Winter wheat yields were simulated to correspond with the SPCs. Then the SPCs and 10‐yr average yields were used in a sensitivity analysis that utilized the Extended Fourier Amplitude Sensitivity Test approach. The results showed that harvest index (HI), biomass/energy ratio (WA), and base temperature for plant growth (TBS) were the common MSPs worldwide, and they affected 73, 71, and 70% of the grids, respectively. Sensitive parameters vary among different regions and may be affected by environmental factors. Elevation influences the winter wheat sensitive parameters greatly. The results can serve as a reference for calibrating the EPIC model in different regions to increase the calibration’s efficiency on a large scale. The sensitivity analysis of crop parameters in different regions at a global scale is especially conducive to model calibration in regions lacking experimental data or a priori knowledge and can improve the accuracy of the model when adjusting different parameters in different regions.