Acta Geographica Sinica ›› 2017, Vol. 72 ›› Issue (5): 906-917.doi: 10.11821/dlxb201705011

• Land Use and Environmental Change • Previous Articles     Next Articles

Spatial pattern and influencing factors of casualty events caused by landslides

Ying WANG1,2(), Qigen LIN1,2, Peijun SHI1,2   

  1. 1. Key Laboratory of Environmental Change and Natural Disaster of Ministry of Education, Beijing Normal University, Beijing 100875, China
    2. Academy of Disaster Reduction and Emergency Management, Beijing Normal University, Beijing 100875, China
  • Received:2016-08-01 Revised:2017-01-15 Online:2017-05-20 Published:2017-05-20
  • Supported by:
    National Natural Science Foundation of China, No.41271544, National Key Research and Development Program Project, No.2016YFA0602403, National Key Technology R & D Program of the Twelfth Five-Year Plan of China, No.2012BAK10B03

Abstract:

The economy of China has maintained rapid growth with an average annual GDP growth rate of 10.14% (in comparable price) from 2000 to 2012. During this period, China witnessed frequent landslide disasters, including 338,964 identifiable individual landslide disasters that resulted in 45,381 casualties, including 9,928 deaths. Analysis of the casualty events caused by landslides from 2000 to 2012 revealed that the spatial pattern of the casualty events was affected by terrain and other factors of the natural environment, which resulted in the distribution of casualty events being higher in the south region than in the north region. Hotspots of casualty events caused by landslides were in the western Sichuan mountain area and the Yunnan-Guizhou Plateau region, the southeast hilly area, the northern part of the loess hills, and the Qilian and Tianshan Mountains, among some others. However, their local distribution pattern indicated that they were also influenced by economic activity factors. To quantitatively analyze the influence of natural environment factors and human-economic activity factors, the binary logistic regression model was applied. The binary logistic regression model is a type of probabilistic nonlinear regression model describing the relationship between a binary dependent variable and a set of independent variables (explanatory factors). The explanatory factors used in this study included relative relief, mean annual precipitation, vegetation coverage, fault zones, lithology, soil type, GDP growth rate, industry type, and population density. The dependent variable used in this study was the presence (1) or absence (0) of casualty events caused by landslides in the county. For the logistic regression analysis, the continuous variables of relative relief, mean annual precipitation, vegetation coverage, GDP growth rate, and population density were substituted into the model. The categorical variables of fault zones, lithology, soil type, and industry type were transformed into binary dummy variables and then substituted into the model. The Probability Model of Casualty Events Caused by Landslide in China (CELC) was built based on the logistic regression analysis, and the confusion matrix and the receiver operating characteristic (ROC) curve were applied to assess the model performance. The results showed that all explanatory variables in the model were selected based on a significance level of 0.05. The coefficients of the explanatory variables showed that relative relief, GDP growth rate, mean annual precipitation, fault zones, and population density have a positive effect on casualty events caused by landslides. In contrast, vegetation coverage has a negative influence on casualty events caused by landslides. More specifically, the results showed that in terms of the influence degree of casualty events caused by landslides, the GDP growth rate ranks only second to relative relief. The probability of occurrence of casualty events caused by landslides will be 2.706 times that of the previous probability with an increase of GDP growth rate of 2.72%. In the evaluation of the model performance, the correct percentage in the confusion matrix is 75 % and the area under the ROC curve (AUC) is 0.826, revealing that the CELC model has good predictive ability. The CELC model was then applied to calculate the occurrence probability of casualty events caused by landslides for each county in China. The results showed that there are 27 counties with high occurrence probability but zero casualty events caused by landslides. The 27 counties can be divided into three categories: poverty-stricken counties, mineral-rich counties, and realty-overexploited counties, which are the key areas where great emphasis should be placed on landslides risk reduction.

Key words: landslide, casualty event, spatial pattern, influencing factors, counties, China