资源环境科技发展态势分析平台

This study assessed the sensitivity and uncertainty interval of the Noah land surface model with multiparameterization (Noah-MP) using observed meteorological data from eight sites with different snow climates. A total number of 20,736 Noah-MP physics-ensemble simulations was conducted for each site by combining different parameterization schemes of physical processes. A natural selection approach and Tukey's test were used to analyze the sensitivity of simulated snow depth to parameterization schemes. The sensitivity of parameterizations at each site was discussed, and the uncertainty interval of sensitive parameterizations was further explored. The results of sensitivity analyses showed the following. The parameterizations for the first-layer snow or soil temperature time scheme, the lower boundary condition of soil temperature, and the partitioning of precipitation into rainfall and snowfall showed sensitivity at all sites. Snow surface albedo parameterizations showed sensitivity at all sites except for two sites in China. Parameterizations for vegetation canopy and canopy stomatal resistance showed sensitivity at some sites. Further comparative results indicated that uncertainties in multiparameterization ensemble simulations were mainly due to sensitive parameterizations under the condition of disregarding the uncertainties from forcing and parameters. After removing the sensitive parameterization schemes that were notably poor-performing, the uncertainty interval in the ensemble simulations decreased significantly. Finally, we concluded that an optimal combination group of well-performed sensitive parameterization schemes can be configured based on the results of sensitivity analysis.