How parameter-set design alters parameter sensitivity of CERES-Wheat model under water stress across two growing seasons

Parameter sensitivity analysis is crucial for crop model calibration and application, yet the impact of parameter-set design (mixed- versus single-category) on parameter sensitivity under different irrigation regimes remains poorly understood. To address this, global sensitivity analyses were performed using the Extended Fourier Amplitude Sensitivity Test (EFAST) method, based on data from a two-season field experiment (2018–2019 and 2019–2020) under different irrigation regimes in the North China Plain. The analyses covered both the combination of cultivar, soil, and management parameters (mixed-category parameter set) and each of the three categories individually (single-category parameter set). Results showed that phenological sensitivity parameters remained constant and the sensitivity was dominated by direct effects, irrespective of water stress, growing season, or parameter set. While yield and evapotranspiration (ET) each exhibited consistent sensitivity to its ownspecific set of parameters, the specific parameters to which each was sensitive were entirely distinct. However, the ranking of all sensitivity parameters varied with water stress, growing season, and parameter-set design. Furthermore, under moderate water stress, yield sensitivity shifted strongly toward water-related soil parameters, whereas ET remained constantly sensitive to them. The mixed-category parameter set better captured the complex interactive effects across parameter categories that are crucial for simulating yield and ET, whereas the single-category sets more clearly isolated the primary controlling influence of core parameters within each respective category. Our findings demonstrate that parameter sensitivity is not a fixed property but is jointly determined by parameter-set design and environmental conditions.