中文摘要：

2010年1月12日海地Ms7．0级地震触发了数以万计的滑坡。本文应用二元统计分析模型、地理信息系统与遥感技术，开展海地地震区滑坡危险性区划研究，并对结果合理性及模型预测功能进行检验。地震前后多源高分辨率遥感数据目视解译结果表明，海地地震触发了30828处滑坡。选择高程、坡度、坡向、曲率、与水系距离、地震动峰值加速度（PGA）、与恩里基约一芭蕉园断裂距离、沿恩里基约一芭蕉园断裂距离8个地震滑坡影响因子进行海地地震滑坡危险性区划。将这30828处滑坡随机分为训练样本与预测样本两类。分别进行模型的建立与测试，训练样本包括滑坡总数的70％（21579处滑坡。覆盖面积为11．18km^2）；预测样本包括滑坡总数的30％（9249处滑坡，覆盖面积为4．56km^2）。基于滑坡训练样本、8个滑坡影响因子、权重系数模型（二元统计方法模型）和GIS技术，构建了滑坡危险性概率分布图。结果合理性检验表明了模型的正确率为84．966％。预测率为84．547％，两者都表明了实际滑坡位置与滑坡危险性结果图具有良好一致·l生。按照滑坡危险性索引值的大小，将研究区分为非常高、高、中等、低、非常低5类。本文证明了在地震滑坡危险·性区划中。权重系数模型作为一种二元统计方法具有良好的建模与预测能力，可为海地地震灾区恢复重建中滑坡的防灾减灾工作提供参考。

英文摘要：

Tens of thousands of landslides were triggered by the January 12, 2010 Port-au-Prince, Haiti earthquake （Ms 7.0）. The main purpose of this study was to apply and verify the hazard ,napping techniques of earthquake triggered landslide by Bivariate Statistics （BS） method. Geographical luformation System （GIS） and remote sensing technologies in the Haiti earthquake stuck area. A total of 30828 landslides were delineated in the study area from visual interpretation of mr,hi-source and high resolution remote sensing images preand post-earthquake. Eight faetors, ineluding elevation, slope angle, slope aspect, slope curvature, distance from drainages, Peak Ground Acceleration （PGA）. distance from the Enriquillo-Plantain Garden Fault （EPGF） and distance along the Enriquillo-Plantain Garden Fault （EPGF） were selected as impact factots; for the Haiti earthquake triggered landslide hazard mapping. The 30828 landslides were randomly partitioned into two subsets： A training dataset, which contained 70% （21579 landslides, with a total area of 11.18km^2）, was used for building the model： and a testing dataset containing 30% （9249 landslides, with a total area of 4.56km^2） was used for model testing. Landslide hazard probability index map was then generated using the training dataset, the eight landslide impact factors, the weight index modeling, a BS method, and GIS technology. The validation results showed a success rate of 84.966% between the hazard probability index ,nap and the training dataset. The predietive rate of 84.547% was obtained fi＇om comparing the testing dataset and the landslides hazard probability index map. Both the success rate and the predictive rate showed sufficient agreement between the landslide hazard map and the existiug landslides data. The resulting landslide hazard map showed five classes of landslide hazard, such as vel7 high, high, moderate, low and very low. This paper showed weight index ,nodeling, as a BS method in earthquake triggered landslide hazard map