Detection of landslides from aerial and satellite images with a semi-automatic method. Application to the Barcelonnette basin (Alpes-de-Hautes-Provence, France)

Until now, visual photo-interpretation techniques combined to ground survey remains the most used method to locate and characterize landslides. New perspectives in using remote sensing for landslides location are now offered by the availability of new very high spatial resolution images and by the development of object-oriented image analysis. In this context, the aim of this paper is to propose a semi-automatic method to locate landslides based on very high spatial resolution (aerial and satellite) images and by using expert knowledge on landslides. This approach is based on (1) a calibration step which consists to translate qualitative indicators derived from expert knowledge (by a photo-interpretation technique) in quantitative indicators and (2) a validation step which allows testing the relevance of these indicators to detect landslides by using an object-oriented method and by only using (aerial or satellite) images with different spatial, spectral and temporal resolutions. Results are (1) a formal and generic grid characterizing landslides and (2) the identification of relevant criteria to extract landslides