中文摘要：

精准估计农作物灾害性损失风险是农业保险开展的基础和前提。本文在比较现有估计方法的基础上，提出了非参数局部线性加权估计方法并进行了数值仿真模拟，结果表明：现有方法受模型设定或参数选择的影响较大且估计结果不同程度地偏离真实数据，改进方法则克服了模型设定和参数选择问题且估计误差低于现有方法。基于1950—2014年粮食单产数据的估计结果表明，相对于改进方法而言现有方法低估了农作物灾害性风险。该研究对丰富农作物灾害损失风险估计方法，加快农业保险供给侧改革，促进农业保险可持续发展具有重要意义。

英文摘要：

Accurate estimation of the risk of crop disaster losses is the basis and prerequisite for operating agricultural insurance. Based on the comparison of existing estimation methods, we proposed a nonparametric weighted local linear estimation method, and through the numerical simulation, we found that the estimation results of the existing methods deviated from the real data because of the parameters or the model selection, and the improved method could overcome these problems and the estimation error was smaller. Based on the estimation of grain yield data of 1950- 2014 years,the results showed that ：compared with the improved method,the existing methods underestimated the risk of crop disaster. This study is of great significance for the estimation methods of the crop disaster losses risk, and for the reform and development of agricultural insurance.