Forward logistic regression has allowed us to derive anearth-flow susceptibility model for the Tumarrano river basin,which was defined by modeling the statistical relationships betweenan archive of 760 events and a set of 20 predictors. For eachlandslide in the inventory, a landslide identification point (LIP)was automatically produced as corresponding to the highest pointalong the boundary of the landslide polygons, and unstable conditionswere assigned to cells at a distance up to 8m. An equalnumber of stable cells (out of landslides) was then randomlyextracted and appended to the LIPs to prepare the dataset forlogistic regression. A model building strategy was applied to enlargethe area included in training the model and to verify thesensitivity of the regressed models with respect to the locations ofthe selected stable cells. A suite of 16 models was prepared byrandomly extracting different unoverlapping stable cell subsetsthat have been appended to the unstable ones. Models were finallysubmitted to forward logistic regression and validated. The resultsshowed satisfying and stable error rates (0.236 on average, with astandard deviation of 0.007) and areas under the receiver operatingcharacteristic (ROC) curve (AUCs) (0.839 for training and0.817 for test datasets) as well as factor selections (ranks andcoefficients). As regards the predictors, steepness and large-profileand local-plan topographic curvatures were systematically selected.Clayey outcropping lithology, midslope drainage, local andmidslope ridges, and canyon landforms were also very frequently(from eight to 15 times) included in the models by the forwardselection procedures. The model-building strategy allowed us toproduce a performing earth-flow susceptibility model, whose modelfitting, prediction skill, and robustness were estimated on the basis ofvalidation procedures, demonstrating the independence of theregressed model on the specific selection of the stable cells.
|Numero di pagine||15|
|Stato di pubblicazione||Published - 2014|
All Science Journal Classification (ASJC) codes