作者: A.C. Mondini , F. Guzzetti , P. Reichenbach , M. Rossi , M. Cardinali
DOI: 10.1016/J.RSE.2011.03.006
关键词:
摘要: Abstract We present a method for the semi-automatic recognition and mapping of recent rainfall induced shallow landslides. The exploits VHR panchromatic HR multispectral satellite images, was tested in 9.4 km2 area Sicily, Italy, where on 1 October 2009 high intensity event caused landslides, soil erosion, inundation. Pre-event post-event images study taken by QuickBird satellite, information location type landslides obtained field through interpretation aerial photographs, were used to construct validate set terrain classification models. models classify each image element (pixel) based probability that pixel contains (or does not contain) new landslide. To models, procedure five steps adopted. First, pre-event pan-sharpened, ortho-rectified, co-registered, corrected atmospheric disturbance. Next, variables describing changes between attributed landslide occurrence selected. three calibrated training using different multivariate statistical techniques. then applied validation same independent variables, Lastly, combined prepared areas. performances evaluated four-fold plots receiver operating characteristic curves. proved capable detecting area. expect be analogous similar (rainfall) or (e.g. earthquake) triggers, provided slope failures leave discernable features captured are adequate quality. proposed can facilitate rapid production accurate event-inventory maps, we it will improve our ability map consistently over large Application advance evaluate hazards, foster understanding evolution landscapes shaped mass-wasting processes.