西北春麦区Apsim-Wheat模型参数全局敏感性分析

为了提高西北春麦区Apsim-Wheat模型的模拟精度及适应性，以甘肃省定西市安定区2015-2018年田间试验数据为基础，采用基于参数筛选法(Morris)和基于方差分解法(Sobol)分析了Apsim-Wheat模型中作物品种参数、土壤参数及田间管理参数对小麦产量的敏感性，并分析比较了两种方法对模型的适应性。结果表明，用Morris法得到的对小麦产量敏感的参数分别为始花期积温(TFI)、出苗到拔节期积温(TEOJ)、春化敏感指数(VS)、田间持水量(DUL)、萎焉系数(LL15)、小麦萎焉系数(WheatLL)、播种日期(sowing date)、播种密度(sowing density)、播种深度(sowing depth)；用Sobol法得到的对小麦产量敏感的土壤参数及田间管理参数与Morris法相同，而得到的敏感作物品种参数不同，分别为春化敏感系数(VS)、开花期积温(TF)、出苗到拔节期积温(TEOJ)、潜在灌浆速率(PGFR)、光周期敏感指数(PS)、灌浆期积温(TSGF)；在Morris法和Sobol法下Apsim-Wheat模型中对小麦产量最敏感的参数均是作物品种参数，且多与品种积温、春化及光周期等参数有关，分别占总敏感性的73.85%和62.9%，土壤参数及田间管理参数对小麦产量的总敏感性分别为26.15%和37.1%，占比较小。综合来看，Morris法和Sobol法均可筛选出对小麦产量敏感的参数，且在调整土壤参数和田间管理参数时，两者具有可替代性；在调整作物品种参数时，可以依据地区实际情况，结合两种方法各自优势，快速筛选出模型敏感参数。

Global Sensitivity Analysis of Apsim-Wheat Model Parametersin Northwest Spring Wheat Region

In order to improve the simulation accuracy and adaptability of Apsim-Wheat model in northwest spring wheat region in China,based on the field experimental data from 2015 to 2018 in Anding District,Dingxi City,Gansu Province,the sensitivity of crop variety parameters,soil parameters and field management parameters to wheat yield in Apsim-Wheat model was analyzed by parameter screening method(Morris) and variance decomposition method(Sobol),and the adaptability of the two methods to the model was analyzed and compared. The results showed that the parameters sensitive to wheat yield obtained by Morris method were accumulated temperature at the beginning of flowering(TFI),accumulated temperature from emergence to jointing stage(TEOJ),vernalization sensitive coefficient(VS),field capacity(DUL),wilting coefficient(LL15),wheat wilting coefficient(WheatLL),sowing date,sowing density,and sowing depth. The soil parameters and field management parameters sensitive to wheat yield obtained by Sobol method were the same as those obtained by Morris method,but the parameters of sensitive crop varieties obtained by Sobol method are different,namely: vernalization sensitivity coefficient(VS),accumulated temperature at flowering stage(TF),accumulated temperature from emergence to jointing stage(TEOJ),potential filling rate(PGFR),photoperiod sensitive coefficient(PS),and accumulated temperature at filling stage(TSGF).In the Apsim-Wheat model under Morris method and Sobol method,the most sensitive parameters to wheat yield were crop variety parameters,which were mostly related to variety accumulated temperature,vernalization and photoperiod,accounting for 73.85% and 62.9% of the total sensitivity,respectively. The total sensitivity of soil parameters and field management parameters to wheat yield was 26.15% and 37.1%,respectively. In summary,both Morris and Sobol methods could screen out the parameters sensitive to wheat yield,and they were replaceable when adjusting soil parameters and field management parameters. When adjusting the crop variety parameters,the sensitive parameters of the model were be quickly selected according to the actual situation of the region and combining the respective advantages of the two methods.