Reducing parameter uncertainty in the Noah-MP-Crop model through global sensitivity analysis and targeted optimization

Abstract. Accurately representing crop growth in land surface models (LSMs) is crucial for capturing cropland–atmosphere interactions, but in practice this remains challenging because many crop-related parameters are poorly constrained. In this study, we focus on reducing parameter uncertainty in the Noah-MP-Crop model by combining global sensitivity analysis with a targeted parameter optimization approach. We first assess the sensitivity of simulated crop yield and leaf area index (LAI) using the Morris screening method and Sobol' variance decomposition at nine agricultural experimental sites across China. Across sites and crops, the analysis consistently points to parameters associated with CO₂ assimilation and carbon use efficiency as the primary controls on crop growth simulations. In contrast, parameters describing crop structure and phenology tend to matter most at specific growth stages rather than throughout the entire season. Guided by these sensitivity results, we then apply a targeted optimization strategy, in which the ranges of key parameters are progressively refined using an interval-based search framework. Comparison with field observations shows clear improvements after optimization: yield RMSE is reduced by 25–86 %, LAI-related objective function values decrease by 49–90 %, and mean absolute errors drop by 40–80 % across major crop types. Overall, our results suggest that sensitivity-guided parameter optimization provides a practical and efficient way to reduce parameter uncertainty in Noah-MP-Crop. The optimized model better reproduces observed LAI and yield dynamics across different climatic regions, offering useful insights for improving crop representation in land surface models.