Global Sensitivity Analysis and Parameter Calibration for a Mechanism-Based Forecasting Model of Rice Sheath Blight Disease

Variations in environmental conditions and disease epidemic characteristics across regions pose challenges for effectively applying disease forecasting models. To address this issue, this paper proposes a novel approach combining global sensitivity analysis and uncertainty quantification for model calibration: (1) The extended fourier amplitude sensitivity test (EFAST) is employed to identify key parameters of the crop disease forecasting model; (2) These critical parameters are calibrated using the generalized likelihood uncertainty estimation (GLUE) method to adapt the model for regional-scale forecasting. The results showed that the sensitivity of the output variables did not change with the changes of crop parameters, and the basic infection rate (b) and average latency time (w) showed sensitivity to the plants' infection rate relative to other parameters when the parameter fluctuations ranged from 5% to 25%. Secondly, the actual and forecasted plants' infection rate simulated by the model after calibration was in good agreement, with the RMSE decreasing to 8.71 and R2 increasing to 0.7 compared to the model calibrated by the genetic algorithm.