目前国内针对雪灾的研究大多是从自然致灾因子的角度出发，而忽视了灾害是地球表层孕灾环境、致灾因子、承灾体综合作用的产物，在全球变暖的背景下，为探讨灾害系统论在风险评估中的具体应用，本文以内蒙古的锡林郭勒牧区为例，按照灾害系统论观点从孕灾、致灾和承灾的角度，建立了评估雪灾风险的指标体系。利用气象、经济、畜牧业数据，以锡林浩特气象站点为例分析了该地区气候变化特征，并利用熵权法确定各气候因子对雪灾的影响权重，从而探讨了雪灾随气候的变化趋势。另外，本文引入人工神经网络模型的概念，采用Visual Basic程序语言，编程实现了雪灾风险评价模型，并应用此模型分析了锡林郭勒牧区雪灾风险空间分布差异特征。通过以上的研究，得到了如下主要结论：(1) 在全球变暖的背景下，锡林浩特市的全年平均气温以及各季节的平均气温都有所增加。其中，冬、春增温较为明显，夏、秋增温较小。锡林浩特的降水总体呈现增加趋势，并在季节分配上发生了变化，即冬季降水增加显著，春、夏增加缓慢，而秋季呈不明显的下降。另外，冬半季的平均相对湿度和日平均风速风速都有不明显的下降趋势。(2) 选定影响雪灾风险的气候因子指标分别为：冬季降水量、积雪持续时间、低温容易掉膘期的负积温、五日温差≥12℃日数、积雪期大风日数、牧草生长季降水量六因素。应用熵权法分析了各因素对雪灾风险贡献水平，得到：积雪期大风日数、积雪持续时间、冬季降水量较其它因子对雪灾贡献大，低温容易掉膘期的负积温、五日温差≥12℃日数对雪灾贡献次之，牧草生长季降水量对雪灾影响最小。(3) 随着气候的变化，研究区雪灾有减轻趋势。这说明气温的升高对雪灾带来的正效应大于由此造成的冬季降雪量增加带来的负效应。(4)分别应用基于BP人工神经网络的雪灾风险评估模型和白灾评判系数法计算锡林郭勒牧区1976～1978、1980～1996、1998～2000、2003、2004共25年的雪灾风险度，通过对比分析其结果，验证了综合考虑孕灾、致灾、承灾的BP神经网络方法相对于仅从致灾方面考虑的传统方法，在评估雪灾风险上有很大的改进。(5) 将建立的基于BP人工神经网络的雪灾风险评估模型应用于锡林郭勒五纯牧区旗县，发现东乌珠穆沁旗雪灾风险较高，其次是阿巴嘎旗。

The Research of heavy snow disaster emphasizes on natural hazard analysis at present at home; however, disaster is a system, correlative with natural, social, environmental, economic factors and so on. Taking Xilingol as an example, this study analyzed climatic change and its influence on heavy snow disaster risk, and according to district disaster system theory, an indices system containing environmental possibilities with hazard, factors of inducing disasters, bodies of bearing disasters, was proposed. Using back propagation artificial neural network, an evaluation model of heavy snow disaster risks was established, and application of this model in Xilingol was done. There are several discoveries found in the research progress as follow:(1) In the background of global warming, both annual temperature and seasonal temperature in Xilihot as a representation of Xilingol have increased, of which temperature in spring and winter increased higher than in summer and autumn. Except in autumn, seasonal and annual precipitation have increased, too.(2) It’s found that duration of wind speed ≥ 6m/s during snow cover, duration of snow cover, annual snow contribute more to heavy snow disaster risk than negative accumulated temperature when daily mean temperature below -5℃, duration of temperature drop in five days ≥12℃ and precipitation in grass growing period (from Apr. to Sept.).(3) It’s found that heavy snow have a lessening trend, showing that positive effect caused by warming is the main influencing factor counteracting negative effect caused by more snow. (4) Taking Xilingol as an example, compared with a conventional method, it’s found that evaluation model of heavy snow disaster risks based on back propagation artificial neural network can reveal heavy snow disaster risk well.(5) The study revealed that heavy snow disaster risk of Dong Ujimqin, Abag are higher than other study parts in Xilingol.