基于GPD模型的粮食作物巨灾的定量界定——以我国稻谷巨灾界定为例

Research on Statistical Definition of Food Crops Catastrophe Based on GPD Model:Take the Case of Rice Catastrophe Definition in China for Example

粮食作物巨灾的定量界定,是确定粮食作物巨灾保险的"触发条件"、产品研发以及巨灾管理等方面的现实需要。粮食作物的巨灾发生是由极端气候事件导致的,所以,应当基于极值理论对巨灾进行界定。首先分析灾损数据厚尾性的分布特征,根据极值理论PBDH定理,应当以GPD作为灾损数据的尾部极端值的分布;再以样本经验平均超出函数法、正态近似法、峰度法初步估计分布的门限值;然后对初估的门限值及附近可能的门限值,逐一以超额灾损数据建立GPD模型并进行检验,检验通过的GPD模型所对应的门限值即为最终确定的门限值,即粮食作物单产视角的巨灾界定值。以我国稻谷为例,基于1979—2015年各省(市、区)各年的单产数据,对我国稻谷巨灾的定量界定值进行了研究。检验表明,该稻谷巨灾界定值不仅在理论上得到了验证,而且实践上也是正确合理的。

The statistical definition of food crops catastrophe is the trigger condition to decide catastrophe insurance,and the realistic need to research and develop products and manage catastrophe. The catastrophes of food crops are caused by extreme climate events,so we should define the catastrophes statistically based on the extreme value theory. Initially,the fat-tailed distribution characteristic of the disaster data is analyzed. According to the PBDH theory of the extreme value theory,GPD should be used as the distribution of the tail extreme values of the disaster data. Furthermore,the threshold value of the distribution is preliminarily estimated by means of Empirical Mean Excess Function,Approximate Normal Distribution Method and the Kurtosis Method. Then,we establish and test the GPD models based on the excess disaster data for the first estimated threshold value and its nearby possible threshold values one by one. The threshold value corresponding to the adopted model is the final threshold value,that is,the statistical definition value of food crops catastrophe in the view of per unit yield. The statistical definition research of rice catastrophe was done based on the yield data of each province in China from 1979 to 2015. And the checkout shows that defining the value of the rice catastrophe is not only verified theoretically,but is also correct and reasonable practically.