GIS based rockfall hazard map for Hong Kong

This paper presents a framework of generating rockfall hazard map based on rockfall inventory and GIS. A total of 374 rockfall data 1984 to 1998 are used for the analysis. The slope angle, slope aspect and elevation at each of these rockfall locations are generated using GIS-based digital elevation model (DEM). In order to conduct a logistic regression analysis for the rockfall hazard, another 374 locations without prior rockfall occurrence were selected randomly over the territory of Hong Kong. Correlations of rockfall occurrence with various factors were conducted using logistic regression through the use of SPSS. To check the validity of the final logistic regression model of rockfall hazard, the likelihood ratio test, Cox and Snell R2and Nagelkerke R2, and "Percent Correct Prediction Test" (PCPT) are used. In terms of the PCPT, the prediction successful rate for all 748 points is of 74.3%, which is considered acceptable. In addition, four additional logistic regression analyses, for which each of independent variables of geology, slope aspect, slope angle, and elevation is ignored separately, are conducted.