Open AccessApplying logistic regression method to determine combinatorial optimization of landslide-related factors and construct landslide hazard map in Khanh Vinh district, Khanh Hoa Province
Author(s) -
Danh Thanh Nguyen,
Ngo Van Dau,
Dung Quoc Ta
Publication year - 2017
Publication title -
khoa học công nghệ
Language(s) - English
Resource type - Journals
ISSN - 1859-0128
DOI - 10.32508/stdj.v20ik4.1121
Subject(s) - landslide , logistic regression , goodness of fit , hazard , statistics , geographic information system , cartography , environmental science , geography , geology , mathematics , geomorphology , ecology , biology
The purpose of this study is to produce landslide hazard map in Khanh Vinh district, Khanh Hoa province using logistic regression method integrated with GIS analytical tools. The spatial relationship between landslide-related factors such as topography; lithology; vegetation; maximum precipitation in year; distance from roads; distance from drainages; distance from faults and the distribution of landslides were used in the landslide hazard analyses. Using success rate and prediction rate curve assess the fit and accuracy of logistic regression method. The results show that this method have the goodness of fit and the high accuracy (Areas Under Curves - AUC = 0.8 ~ 0.9). Bayesian Model Average (BMA) of the R statistical software was applied to identify the most influential factors and the combinatorial optimization models of landslide-related factors. There are four the most important landslide-related factors and five combinatorial optimization models of landslide-related factors. Model 3 (slope angle, slope aspect, altitude, distance from roads and maximum precipitation in year) is the best optimization.