Abstract: Scientific assessment of casualties caused by seismic geological hazards is one of the key factors to improve the accuracy of earthquake casualties assessment and to improve the earthquake disaster loss assessment system. However, this problem has long plagued scientists of the earthquake disaster risk assessment and post-earthquake accurate and rapid assessment, so far there is no perfect solution. In this paper, the authors try to develop a new evaluation method based on GIS techniques and the logistic regression model for assessing the earthquake-induced landslide casualties, including three procedures: (1) study area being divided into1㎞×1km 1㎞grid cells, and the population information being assigned to the grid cell as its population attribute; (2) death rate being calculated in each grid cell using the logistic model, and (3) then the death number being added in each grid cell of the study area and the total deaths being evaluated. After that, three earthquake events are selected as case study: the 2014 Ludian, Yunnan Ms 6.5 earthquake area, the 2012 Yiliang, Yunnan Ms 5.6&5.7 double-earthquake area, and the 2008 Wenchuan, Sichuan Ms 8.0 earthquake area. Among them, the Ludian earthquake data is selected to establish and test the evaluation method for landslide deaths. The other two is used to test the practicability of new method. For the Ludian earthquake, 233 deaths are calculated based on the new model, with an error rate of 6.80 percent compared with 250 actual deaths and disappearances, and the Kappa coefficient is 0.912, which indicates that the establishment of the model is relatively reasonable. For the Yiliang disaster earthquake event, landslides claimed caused 48 deaths based on the new model, which is 11 less than the actual number, with the error rate of 18.64%, and the Kappa coefficient of 0.899. As for the 2008 Wenchuan earthquake, a total of about 20000 people died and disappeared due to earthquake-induced landslides, the result of model evaluation is 18732, and the error rate is about 6.5%. The results presented in this paper suggests that on the premise of the allowable error, the new evaluation method are practical and can be applied to calculate the landslide deaths in other earthquake events in future.

Key words: earthquake-induced landslides, deaths, evaluation model, one-square-kilometer grid

摘要： 科学评估地震地质灾害可能造成的人员伤亡，是提高地震人员伤亡评估准确性和完善地震灾害损失评估体系的重要因素之一，然而，这一问题却至今没有完善的解决方案。本文提出一种基于GIS和logistic模型的地震滑坡人员死亡数量评估方法，尝试开展这方面的研究。该方法包括三个步骤：（1）在GIS中将研究区划分为1㎞×1㎞的网格单元，把网格内的人口数量信息赋值给网格作为它的人口属性；（2）基于回归的logistic模型计算各网格单元的地震滑坡死亡率；（3）公里网格单元的地震滑坡人员死亡数量和整个研究区地震滑坡死亡总人数评估。基于本文提出的方法，对鲁甸、彝良和汶川三个地震灾区进行了实例研究。其中，鲁甸震例用于地震滑坡人员死亡数量评估方法的建立和方法有效性的评价，彝良和汶川震例用于所建方法在更广范围内开展地震滑坡人员死亡数量评估外延的适用性评价。本文模型评估鲁甸震区由地震滑坡导致的死亡人数为233人，与实际因地震滑坡导致的死亡和失踪250人相比，少17人，误差率为6.80%， kappa检验值为0.912，表明建立的模型较为合理。而外延适用性分析结果显示：模型评估彝良地震灾区的地震滑坡死亡人数48人，比实际少11人，误差率为18.64%，kappa检验值为0.889；汶川地震时滑坡导致的死亡和失踪合计约20000人，模型评估的结果为18732人，误差率约为6.5%。实例研究结果表明：本文所建立的地震滑坡人员死亡数量评估方法在允许一定误差的前提下，可以推广应用到其他区域的地震滑坡灾害导致人员死亡人数的计算中，有一定的外延可移植性。

关键词: 地震滑坡, 死亡人数, 评估模型, 公里网格