基于公里网格单元的地震滑坡人员死亡率评估模型——以2014年鲁甸MS6.5地震为例*

(1.云南师范大学地理学部,云南 昆明650500; 2.云南省地震局,云南 昆明 650224; 3.中国地震局地质研究所,北京 100029)

地震滑坡; 人员死亡率; logistic模型; 公里网格; 鲁甸地震

Modeling and Testing Earthquake-induced Landslide Casualty Rate Based on a Grid in a Kilometer Scale:Taking the 2014 Yunnan Ludian MS6.5 Earthquake as a Case
BAI Xianfu1,2,NIE Gaozhong3,DAI Yuqian1,YU Qingkun2,LUO Weidong2,YE Liaoyuan1

(1.Faculty of Geography Science,Yunnan Normal University,Kunming 650500,Yunnan,China)(2.Yunnan Earthquake Agency,Kunming 650224,Yunnan,China)(3.Institute of Geology,China Earthquake Administration,Beijing 100029,China)

earthquake-induced landslides; casualty rates; logistic model; grid in a kilometer scale; Ludian MS6.5 earthquake

备注

为提高地震人员伤亡预评估的准确性,完善地震灾害损失评估模型,科学评估地震地质灾害可能造成的人员伤亡数量,以2014年鲁甸MS6.5地震滑坡人员死亡数据为样本,建立了一种基于公里网格单元的地震滑坡人员死亡率logistic回归模型。采用F检验法对所建模型的合理性进行检验,计算得到的F值无限接近于1,表明模型无限接近于完全模型,具有极好的数学统计意义。根据模型评估的死亡率反演得到鲁甸地震灾区滑坡致死人数为233人,比实际少17人,总精确度为93.20%,实际死亡人数与模型识别人数在空间上也有很好的一致性,说明计算得到的地震滑坡人员死亡率是实际死亡人数的良好指标。

To improve the accuracy of earthquake casualty assessment,we should improve the earthquake damage assessment system and scientifically assess the possible casualties caused by seismogeological disasters.For this purpose,we developed a logical regression model for earthquake landslide mortality based on 1 km×1 km grid cells using the death data in the landslides of the Yunnan Ludian MS6.5 earthquake in 2014.F-value test was used to examine the model's statistical portability.In the Ludian study area,the computed F-value was infinitely close to 1.This result proved that the model we developed is close to the real model in the Ludian area.According to the field investigation,the earthquake-induced landslides in the Ludian disaster area claimed 250 deaths,while we get 233 deaths by our new model,leaving the accuracy as 93.20%.Furthermore,the death places triggered by landslides are nearly consistent with the ones indicated by our model.This suggests that the computed earthquake-induced landslides mortality values are good indicators of real casualties.