Modelling of future landslide exposure in Sukabumi, Indonesia in two scenarios of land cover changes

Abstract

Modelling future landslide exposure is essential for better planning for disaster minimisation. Despite the importance, little attention has been paid to study how land cover change in the future can potentially affect the exposure of an evolving population to landslide risk. This research aims to assess landslide exposure in 2030 in the Sukabumi region, Indonesia based on two land cover change (LCC) scenarios: Scenario A (baseline), and Scenario B (growth governed by land use zoning). Also assessed in this thesis is the importance of ten geo-environmental variables representing topography (slope, aspect, curvature), hydrology (slope loading, stream power, distance to rivers), anthropogenic (land cover and distance to roads), soil, and lithology and their association with past landslide events using the Weights of Evidence (WOE) method. The assessment is implemented in four steps: First, the land cover in 2030 was modelled based on the 1993-2013 land cover changes. Second, the studied predictors were assessed to evaluate their relationships with landslide events. Third, all the important predictors of landslides were used to model landslide susceptibility levels by weighting them using the importance factor based on WOE. Finally, the spatial distribution of population was estimated via dasymetric cartography modelling and overlaid with the modelled landslide susceptibility in both scenarios. It is found that lithology, curvature, soil types, and slope gradient play a critical role in assessing landslide susceptibility in the study area. WOE analysis also found that the Badui formation, Citorek tuff, Cikotok formation, Batuasih formation, young terrace deposits, and older volcanic deposits are favourable for landslide occurrence. As for curvature, more concave or convex slopes are more closely associated with the landslide occurrence. For soil, a good predictor of landslide presence are sulfic endoaquepts and typic hapludands. The relation between landslide and gradient becomes positive at 18.7° and closer as the slope becomes steeper. LCC exerts impact on landslide susceptibility zoning, hence the landslide exposure. In 2017, the two largest areas of susceptibility levels were high (120,418 ha) and medium (105,714 ha). The other classes are much less common, in the descending order of very high (85,551 ha), low (78,263 ha), and very low (30,228 ha). In 2030, in Scenario A, the very high susceptibility class is predicted to increase by 5,231 ha in Scenario A and 3,572 ha in Scenario B. In 2017, the number of people exposed to different landslide susceptibility follows the descending order of high (1,433,045 people or 51.63% of the total population), medium (20.13%), very high (18.60%), low (7.91%) and very low susceptibility (1.73%), which is the same in both scenarios. However, both the absolute population and population growth (%) vary between the two scenarios. For example, 19.66% more people will be living in high susceptibility areas in Scenario A, but only 17.19% in Scenario B, a difference of approximately 3,841 people. These findings demonstrate the critical role of land use planning in reducing landslide exposure.