Global sensitivity analysis of crop and soil parameters of CERES-Maize model under different irrigation and fertilization regimes based on the EFAST method

Sensitivity analysis is a key approach to identify the influential parameters in crop models and improve calibration efficiency. This study employed the extended Fourier amplitude sensitivity test (EFAST) to evaluate the sensitivity of 10 crop parameters (cultivar and ecotype) and 22 soil parameters in the DSSAT-CERES-Maize model under different irrigation and nitrogen fertilizer regimes. The outputs included the state variables: anthesis date (ADAP), maturity date (MDAP), yield, grain protein content (GPC), maximum leaf area index (LAIX), and maximum aboveground biomass (CWADX), and the time-series variables: LAI, above ground biomass (CWAD), and nitrogen accumulation (CNAD). Results showed that the cultivar parameters, P1, P2, PHINT and P5, consistently dominated the simulation of anthesis and maturity stages, and their sensitivities remained stable across different water and nitrogen treatments. For LAIX and CWADX, the P1, PHINT and SRGF are the most influential parameters under different irrigation regimes. Under different nitrogen fertilizer regimes, the RUE is the most sensitive ecotype parameter, and the SDUL and SLLL show heigh sensitivity. Time-series variables display distinct stage-specific sensitivity patterns. Under different irrigation regimes, the PHINT is the most sensitive parameter for LAI, CWAD, and CNAD before jointing. The influence of P1 on CNAD approachs to peak after jointing, and the SRGF is highly sensitive for CNAD before tasseling. In comparison with water treatments, the sensitive period of PHINT to time-series variables is significantly shorter under nitrogen treatments, whereas RUE remains sensitive to LAI and CWAD throughout the growing season. For CNAD, the TSI of cultivar, ecotype and soil parameters are the highest at the seedling stage under both water and nitrogen stresses. The findings provide a theoretical basis for optimizing model parameterization and simulating crop responses, especially for precision-agriculture applications under different water and nitrogen regimes.