Spatial Modeling of Landslide Susceptibility (Case Study In West Java Province, Indonesia)

Abstract

Indonesia is frequently affected by landslides caused by both rainfall and earthquakes. During the period 1981-2007 the annual landslide frequency varies with an average of 49 events per year this was reported by (Chrisanto et al., 2008). There is also information from the global catalogue presented by (Kirschbaum et al., 2009) where in the years 2003 and from 2007 to 2009 there were reports 97 landslides in Indonesia, which produced 872 casualties. Also within the period of 2011-2012 where in 2011 there was about 29 events in West Java alone and in 2012 almost all of regions in Indonesia had a landslide event due to the high intensity of rainfall and active tectonic systems in that year. (Geological Agency, 2011 & 2012). Fatalities in Indonesia caused by landslides were also high. Study done by (Petley, 2007) showed that in 2007, total number of fatalities caused by landslides in Indonesia alone was 465 which was the second highest next to China. Due to the high occurrences of landslides and the number of fatalities, there is a need of mitigation strategies in Indonesia to prevent future landslide disasters. One of the mitigation strategy is to provide landslide susceptibility mapping. This research aims into developing a spatial modeling method in landslide susceptibility that can find landslide susceptibility area based on the data being analyzed. The result of the research can also observe the main factors that supports the events of landslides; predict the areas that are prone to landslides; and generate landslide susceptibility map of the areas that are prone to landslides. This research is done within 4 regencies in West Java Province which are Cianjur Regency, West Bandung regency, Bandung Regency, and Garut Regency. The area of this research is chosen because of the characteristic that is mountainous in the West Java Province which is the characteristic where landslides most likely occur in West Java. The materials used in this research is digital maps in vector format that is Landuse/Landcover Map, Soil Type Map, Geology Map, Climate Type Map, Annual Rainfall per Year Map, Landsystem Map, River Map, and Landslide Data. The tools used in this research are Personal Computer, GIS Software, Statistic Software (SPSS), Microsoft Office (MS Word, and MS Excel), and GPS. The factors were given rescaled scores based on the correlation between the landslide densities. The methods used in this analysis were regression analysis and correlation analysis in determining the factors to be analyzed and Principal Component Analysis (PCA) in determining the weight of each factors. From the 9 factors of landslide chosen for the analysis, based on the result of the weight determination, principal component 1 until principal component 8 which explained 96% of the total variance of the data gave the most accurate overall accuracy result for 3 susceptibility class at 68.20%. The mathematical equation obtained from this model: Landslide Susceptibility Index (LSI) = 0.231 * Landuse/Landcover + 0.188 * Slope + 0.112 * Distance to River + 0.107 * Soil Type + 0.100 * Annual Rainfall + 0.073 * Population Density + 0.068 * Geology Age + 0.066 * Climate Type + 0.054 * Geomorphology. The result of the weight determination concluded the rank of the factors in contribution to the landslide events within the research area. Based on the weight determination, Landuse contributed the greatest weight of 23.1% followed by slope 18.8%, distance to river 11.2%, soil type 10.7%, annual rainfall 10.0%, population density 7.3%, geology age 6.8%, climate type 6.61%, and geomorphology 5.4%. From the result of the LSI score, all of the factors that were within the class range of high/very high susceptibility class showed that the most dominant subfactors within the high/very high susceptibility class were dominantly areas with heavy rainfall and human cultivation activities. For the three most highest contributing factors, for the landuse factor timber plantation followed by dry land agriculture were most dominant area within the high/very high susceptibility class regions. For slope, class slope of 3 (15-25%) were most dominant followed by slope class 2 (8-15%) within the high/very high susceptibility class area. For distance to river, the distance lower than 250 m from the river were most dominant within the area of the susceptibility class. For the fourth highest contributing factors which was the annual rainfall, areas with precipitation of over 3000 mm/year were most dominant within the high/very high susceptibility class. This also showed that for this research area, human activities play a big role in the cause of the landslide mainly in land cultivation activities such as farming and plantations. This condition is very much prone to landslide disasters. Therefore, for this research area, future planning and development must be considered in areas for human activities to prevent future landslide disasters.