This paper describes the application of airborne remote sensing to the study of landslides on the clay-dominated slopes of the Cotswolds Hills between the towns of Broadway, Worcestershire, and Snowshill, Gloucestershire, in the UK. The project involved an initial desk study, airphoto interpretation and field survey in order to provide detailed information about the nature and extent of the landsliding in the area. High-resolution Airborne Thematic Mapper (ATM) imagery was acquired by the NERC Airborne Remote Sensing Facility, which was subsequently processed in order to develop a remote sensing method for landslide identification using airborne multispectral data. A range of image processing methods are described including colour composite enhancement and thermal imaging, while the focus of the paper will be on the development of a semi-automated method of landslide identification using image classification and texture analysis.
Results from the first stage of the study have shown that the use of image processing techniques such as colour composites and thermal imaging can provide information on the ground surface not visible in conventional aerial photography. In this case study, this has included more detailed geomorphological information on landsides in the area and the nature of the cambering and gulls at the top of the escarpment. The second part of the study has investigated the use of image texture enhancement as a method of landslide identification, applied in isolation and incorporated into an image classification scheme as a semi-automated method of landslide identification. Results from this investigation indicate that landslides can be identified automatically with a high degree of accuracy (83%) and that by using image texture, the image classification technique is able to successfully differentiate between areas of landslide activity and stable slopes in the airborne imagery.
Its is clear from the results of this study that in order to identify landslides in these types of clay-dominated terrains, image texture must be used. Inland landslides, like those on the Cotswolds escarpment, do not have a spectral signature but they do exhibit a distinct spatial signature that allows them to be identified in airborne imagery using textural analysis. The semi-automated method of landslide identification described in this paper represents a rapid method of terrain evaluation and landslide hazard assessment, which can be undertaken prior to more detailed field mapping.