地震地质 ›› 2021, Vol. 43 ›› Issue (5): 1250-1268.DOI: 10.3969/j.issn.0253-4967.2021.05.013

• 研究论文 • 上一篇    下一篇

基于GIS和logistic模型的地震滑坡致死人数快速评估方法

白仙富1,2)(), 聂高众3),*(), 叶燎原1), 戴雨芡1), 余庆坤2)   

  1. 1)云南师范大学地理学部, 昆明 650500
    2)云南省地震局, 昆明 650224
    3)中国地震局地质研究所, 北京 100029
  • 收稿日期:2020-06-02 修回日期:2021-01-07 出版日期:2021-10-20 发布日期:2021-12-06
  • 通讯作者: 聂高众
  • 作者简介:白仙富, 男, 1979年生, 现为云南师范大学地理学部山地环境与自然灾害专业公共安全与区域防灾方向在读博士研究生, 高级工程师, 主要从事重大地震灾害及其灾害链风险综合评估基础理论与关键技术、 区域地震应急救援对策与区划等研究工作, 电话: 0871-65747065, E-mail: xf_bai520@163.com
  • 基金资助:
    国家重点研发计划项目(2018YFC1504503);中国地震局地质研究所基本科研业务专项(IGCEA2106);云南省地震局科技专项(2018ZX03)

A METHOD FOR RAPID ASSESSMENT OF DEATH TOLL FROM EARTHQUAKE-INDUCED LANDSLIDE BASED ON GIS AND THE LOGISTIC MODEL

BAI Xian-fu1,2)(), NIE Gao-zhong3),*(), YE Liao-yuan1), DAI Yu-qian1), YU Qing-kun2)   

  1. 1) Faculty of Geography Science, Yunnan Normal University, Kunming 650050, China
    2) Yunnan Earthquake Agency, Kunming 650224, China
    3) Institute of Geology, China Earthquake Administration, Beijing 100029, China
  • Received:2020-06-02 Revised:2021-01-07 Online:2021-10-20 Published:2021-12-06
  • Contact: NIE Gao-zhong

摘要:

科学评估地震地质灾害可能造成的人员伤亡数量, 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。文中提出了一种基于GIS和logistic模型的地震滑坡致死人数快速评估方法, 尝试在此方面开展研究。该方法包括3个步骤: 1)在GIS中将评估区划分为1km×1km的网格单元, 把网格范围内的人口数量赋值给网格单元作为其人口属性信息; 2)基于回归的logistic模型, 根据网格单元的地震滑坡危险性属性计算各网格单元的地震滑坡致死率; 3)计算各公里网格单元的地震滑坡致死人数, 然后进行整个研究区的地震滑坡致死总人数评估。利用本方法, 以2014年“8·3”鲁甸MS6.5地震, 2012年“9·7”彝良MS5.6、 MS5.7地震和2008年“5·12”汶川MS8.0地震的3个地震灾区为研究区, 进行了地震滑坡致死人数测试。其中, 基于鲁甸震例构建地震滑坡致死人数快速评估方法并评价其有效性, 利用彝良和汶川震例测试该方法在更广范围内开展地震滑坡致死人数快速评估时的外延性和适用性。基于文中提出的模型评估得到鲁甸地震灾区因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共250人相比少17人, 死亡总数评估误差率为6.80%; 公里网格单元死亡人数评估结果的Kappa检验值为0.912, 表明用所建模型计算的死亡人数与实际死亡情况的一致性较好, 模型具有很好的统计学意义。而外延适用性分析结果显示: 汶川地震时滑坡导致的死亡和失踪合计约20 000人, 模型评估的结果为18 732人, 死亡总数评估的误差率约为6.5%; 模型评估的彝良地震灾区地震滑坡致死人数为48人, 比实际死亡人数少11人, 死亡总人数评估误差率为18.64%, 公里网格单元死亡人数评估结果的Kappa检验值为0.889。实例研究结果表明: 在允许存在一定误差的前提下, 文中所提出的地震滑坡致死人数快速评估方法可推广应用到其他区域地震滑坡灾害导致的人员死亡数量的估计中, 该方法具有一定的外延适用性。

关键词: 地震滑坡, 致死人数, 评估模型, 公里网格

Abstract:

The scientific assessment is one of the cornerstones for improving the earthquake disaster loss assessment system and the accuracy of the assessment of casualties caused by seismogeological hazards. However, this question has long been plaguing scientists engaged in the earthquake disaster risk assessment. So far there is no better solution yet. To solve this problem, the authors develop a new method for rapid assessment of the death toll caused by the earthquake-induced landslide based on GIS techniques and the logistic regression model. The method mainly consists of three procedures: 1)On GIS platform, the study area is divided into 1km×1km grid cells, and the number of the people in each cell is assigned to the cell as its population attribute; 2)based on the cell’s earthquake risk attribute, the landslide death rate in each cell is calculated by using the logistic regression model, and 3)by adding the death number in each cell, the total death toll in the study area are got.
To test our new model, we select four earthquake events for case study: the 2014 Ludian, Yunnan MS6.5 earthquake; the 2008 Wenchuan, Sichuan MS8.0 earthquake, and the 2012 Yiliang, Yunnan MS5.6 & 5.7 double earthquakes. Here, the Ludian earthquake case is for testing the effectiveness of our new model. The other two cases serve for testing the practicability of our model for the future earthquake disasters in other places in China.
Our logistic model of the landslide death-rate in the kilometer grid shows that the absolute values of the coefficients of the 2 impact factors are relatively large. These 2 factors, which indicate respectively the highest landslide-hazard grade and the lowest one, are playing critical roles in deciding the death toll from earthquake-induced landslide.
In each kilometer grid, when the impact factors of the landslide-hazard grade are 5, 4, and 3, their corresponding coefficients are respectively 0.040 77, 0.031 30, and 0.013 65. If the number of the sub-grids corresponding to these 3 impact factors rise, the death rate will accordingly rise. The more the sub-grids with high impact factors, the higher the death-rate will be. When the impact factors of the landslide-hazard grade are 2 and 1, their corresponding coefficients are respectively -0.016 66 and -0.09652. If the number of the sub-grids corresponding to these 2 impact factors rise, the death rate will accordingly fall. The more the sub-grids with low impact factors, the lower the death-rate will be.
For the Ludian earthquake, we estimate 233 deaths according to our model, with an error rate of 6.80%compared with 250 dead and missing people in reality, and Kappa coefficient is 0.912, indicating the feasibility of our model. In the 2008 Wenchuan earthquake event, a total of about 20 000 people died and went missing due to earthquake-induced landslides, while our evaluation is 18732, with an error rate about 6.5%. For the Yiliang double earthquake event, we get 48 deaths by our model, 11 less than the actual number. The error rate is 18.64%, and Kappa coefficient is 0.899. The results suggest that on the premise of the allowable error, our method is practical and can be applied to assessing the landslide death toll for other earthquake events in future.

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

中图分类号: