关键词: 干旱, 综合指标体系, 风险评估, 稳健性检验, 聚类分析

Abstract: Droughts are increasing in frequency and intensity as the global climate changes. Risk assessment for agricultural drought disasters is an important approach for disaster prevention and mitigation. Based on the formation mechanism of natural disaster risk, the index system model of agricultural drought risk assessment in North China was constructed, and the robustness of the established risk assessment model was tested by using the internal test method. Then based on the characteristics of agricultural drought risk formation in North China, the paper proposes corresponding risk prevention recommendations. The conclusions were as follows: when adjusting the indices and methods in the risk assessment model, the range of changes in the agricultural drought risk ranking for each city was mostly 1−2.5 places, and the model was relatively robust. With the exception of Beijing and Tianjin, which had the highest disaster prevention and mitigation capacity, central and northern China had the highest agricultural drought risk, which was previously linked to the high risk and vulnerability of these regions. The agricultural drought risk in northern Henan and southern Shandong is low, mainly due to the low risk. The 34 cities in North China can be divided into 6 agricultural drought risk clusters, which are not dispersed in space and cross-city distribution, and have good aggregation. Cluster 4, located in the northernmost mountainous areas, has the highest risk of drought, which is associated with high environmental vulnerability and low capacity for disaster prevention and mitigation, so agricultural areas should be reduced and agricultural input should be increased. Clusters 2, 5 and 6, located in the central and southern regions, have higher grain acreage and yields due to drought, so agricultural drought risk can be further reduced by increasing agricultural input and improving irrigation guarantee rate.

Key words: Drought, Comprehensive index system, Risk assessment, Robustness test, Cluster analysis