Integration of radar and optical remote sensing for landslide investigation - case study of Koslanda area in Sri Lanka

Abstract

Koslanda, in Sri Lanka is an area that remains in the memories of people due to frequently occurring landslides as the area is made vulnerable by both climatic and geomorphological settings. Additionally, the aftermath of the landslide, i.e. the debris flow, causes more damages when compared to the landslide itself. As such, this study focuses on the integration of radar and optical remote sensing for landslide investigation with inclusion of debris flow. The significance of the data types derived from radar and optical images are examined in terms of sensor characteristics and spectral information. Radar and optical images before and after the event, geometrically registered and radiometrically normalized, are used to delineate the landslide area by different change detection techniques. Detected landslide areas are compared with the area determined by GPS field surveying. At the comparison stage, landslide detection capacity of the optical images was 76% while it was 86% with the radar images. This is mainly due to inherent nature of radar being able to collect data under any climatic condition. The Information Value method uses bivariate analysis without radar induced factors (BiNR), and bivariate analysis with radar induced factors (BiWR), while the Multi Criteria Decision Analysis based on AHP uses multivariate analysis without radar induced factors (MNR), and multivariate analysis with radar induced factors (MWR). When utilizing the multivariate method, an increase in the area showing high and moderate susceptibility to landslides was observed as 5% and 3% from the total area, respectively. With the inclusion of radar induced factors (surface roughness, near surface soil moisture from delta index, and forest biomass), high and very low susceptible regions to landslide increased by 7% and 4% when using the bivariate method, while it was 3% for both cases when using the multivariate method. Landslide prediction analysis is enhanced by incorporating debris flow analysis with DEM derived factors, as appropriate for a country like Sri Lanka, where data scarcity of acceptable accuracy is high for smaller scale studies.