Binary Logistic Regression and Normalization for Landslide Hazard Analysis in Cianjur District, West Java

Abstract

Cianjur had 33 times occurrence landslides during 2002-2007. The objectives of this study were: (i) to identify the main cause of landslide hazard; and (ii) to analyze the landslide hazard areas in Cianjur. Analysis methods to identify the main cause of landslide hazards were based on binary logistic regression and normalization. Based on binary logistic regression and normalization result, rainfall is the main cause of landslide hazard in this study area. This was shown by the highest coefficient value of rainfall at the 3 equation (0.542 by using SPSS, 0.920 by using Idrisi, and 0.29 by using normalization). Positive coefficient value means that the occurrence of landslide mainly influenced by the highest class of rainfall. The three landslide hazard maps resulted from different methods showed different locations with moderate to very high-level hazards. A hazard map of binary logistic regression using SPSS showed that the moderate to very high-level hazard was found in the northwest and southeast part of Cianjur. Hazard map using binary logistic regression on Idrisi showed that the moderate to very high-level hazard was found in middle and north part of Cianjur. Hazard map using the normalization method showed the domination of moderate level hazard which was spread through all subdistricts in Cianjur. The landslide hazard map of binary logistic regression on Idrisi was better than two other maps, indicated by the highest determinant coefficient value which is 0.980.

Keywords: Binary logistic regression, hazard, landslide, normalization