作物模型可定量描述作物生长发育过程及其与环境因子的关系，在农业生产管理决策等方面具有重要的应用价值。模型参数调试是作物生长模拟模型进行应用前的重要步骤，且往往需要大量时间和精力进行调试，敏感性分析可以以较高的效率筛选出敏感参数，是模型本地化的重要环节，对模型的应用有重要意义。文章研究基于Morris法和EFAST法对RiceSM模型的作物参数进行了敏感性分析，筛选出输出变量中成熟期、叶面积指数、地上部生物量、产量的敏感参数，并比较分析两种方法结果的异同。结果表明，移栽至拔节阶段的基本发育系数K3、出苗到移栽阶段的基本发育系数K2、移栽到拔节阶段叶干物质的分配系数CLV1是影响RiceSM模型主要输出结果的最敏感参数，两种方法得到的敏感参数结果基本一致，但各敏感参数的重要程度略有差异。以筛选出的敏感参数为基础，基于长沙、常德两站的农气观测资料对RiceSM模型进行调参与验证。验证结果表明，早稻和晚稻叶面积指数模拟值与实测值的归一化均方根误差(NRMSE)在21.63%~47%之间，早稻和晚稻茎、叶、穗、地上部生物量和产量模拟值与实测值的NRMSE分别为4.77%~39.51%、5.46%~6.64%、3.78%~4.15%和2.78%~3.52%和9.29%~12.12%之间。调参后的模型能够较好地模拟早稻和晚稻生物量、叶面积指数的动态变化和产量形成过程。研究结果可为RiceSM模型的本地化、参数优化和推广应用提供支持。

Crop model can quantitatively describe crop growth and development processes and their relationships with environmental factors, and have important applications in agricultural production management decisions and other areas. Model parameter debugging is an important step before crop growth simulation models are applied, and often requires a lot of time and effort for debugging. Sensitivity analysis can screen out sensitive parameters with high efficiency, and is an important part of model localization, which is of great significance for model application. Sensitivity analysis was conducted on the crop parameters of the RiceSM model based on the Morris method and EFAST method to screen out the sensitive parameters of maturity, leaf area index, total biomass and yield among the output variables, and to compare and analyze the similarities and differences between the results of the two methods. The results showed that basic development factor from transplanting to jointing stage K3, basic development factor from seeding to transplanting stage K2 and dry matter distribution coefficients of leaf from transplanting to jointing stage CLV1 were the most sensitive parameters affecting the main output results of RiceSM model, and the results of the sensitive parameters obtained by the two methods were generally consistent, but the importance of each sensitive parameter differed slightly. The validation results showed that the normalized root mean square error (NRMSE) of simulated and measured values of early and late rice leaf area index ranged from 21.63% to 47%, and the NRMSEs of simulated and measured values of stem, leaf, spike, aboveground biomass and yield of early and late rice ranged from 4.77% to 39.51%, 5.46% to 6.64%, 3.78% to 4. 15% and 2.78% to 3.52% and between 9.29% and 12.12% respectively. The model was able to better simulate the dynamics of bio-mass, leaf area index and yield formation in early and late rice. The results of the study provide a reference for the localization of the model.