基于概率统计方法的承德市农业旱灾风险评估

根据旱灾风险的统计特征，提出2种基于概率统计方法的旱灾风险评价方法：1）旱灾损失的概率分布曲线，2）旱灾损失与干旱概率的关系曲线。以滦河上游承德市辖区的农业旱灾风险为例进行应用研究。根据1990－2007年历年农业旱灾损失数据计算粮食因旱减产率，拟合其概率分布。利用分布式水文模型模拟得到该区域1956－2005年0～50 cm的月平均土壤含水率，识别农业干旱事件，计算农业干旱烈度概率，选用对数函数拟合因旱粮食减产率和农业干旱烈度概率关系曲线。采用因旱粮食减产率期望值表征区域旱灾风险，2种方法结果一致性较好，研究区域农业旱灾风险上游大于下游，各个县的多年平均因旱粮食减产率为7%～15%。

Assessing agricultural drought disaster risk in Chengde city using stochastic method

Abstract: Drought is one of the widespread natural disasters affecting agricultural production, and causes the uncertainty and vulnerability to food security in the world. While rising temperature and less precipitation have contributed to increasing drought, increases in the intensity, duration and area of each drought event have been observed. Therefore, it is of guiding significance to evaluate agricultural drought risk. There are two common methods used for drought disaster risk assessment. One is the fuzzy evaluation method, which is commonly used even though it is weak of objectivity and facticity. Another is the statistical analysis method, which is seldom used in drought disaster risk evaluation due to the data limit. In order to give a quantitative assessment of agriculture drought, this study proposed two statistical analysis methods for drought disaster risk assessment. The first one is the probability distribution curve of drought loss. This method is based on the assumption that drought loss is a random variable and has the same probability distribution as the drought event. The second one is the regression curve between the drought loss and the drought probability, which implies that a regression relationship exists between the scale of the drought event and drought loss. The two methods were applied to the agricultural drought disaster risk assessment in the upper Luan River basin within the administrative boundary of the Chengde city, Hebei Province. Firstly, the annual crop yield loss rate caused by drought was estimated from the historical drought disaster data from 1990 to 2007. The probability distribution curve of the crop yield loss rate was gained by the frequency analysis. Secondly, the agricultural droughts were detected from the long-term soil moisture data, represented by the soil moisture of the top 50 cm soil layers over the agricultural land, which was simulated by the GBHM (Geomorphology-Based Hydrological Model). The probability of each agricultural drought event was calculated using the time series of monthly soil moisture storage anomaly. The logarithmic function was used to fit the regression curve of crop yield loss rate and the exceedance probability of drought. Finally, the regional agricultural drought disaster risk map was represented by the expected yield loss rate calculated by the two methods. In the risk map, the agricultural drought disaster risk was classified into 5 levels. It was found that the results of the two methods showed good consistency. The agricultural drought disaster risk increasing from the downstream to the upstream, and the drought caused crop yield loss rate ranged from 7% to 15% in the study region. The rationality and the reliability of the two methods were also discussed in this paper. From this study it can be seen that drought disaster risk analysis based on the historical data is practically useful. It is necessary to take human's resilience to drought disaster into account for the drought disaster risk assessment.