GA SAKe is a linear, steady-state hydrological model. It employs Genetic Algorithms to predict the timing of triggering of landslides, based on rainfall and previous activations. Calibration provides discrete kernels, that allow to analyse even complex geohydrological conditions, with a high level of detail. Once validated, the mobility function of the model allows to predict landslide activations. The release 2.0, here presented, introduces a powerful computational engine, and a couple of relevant modules: a validation tool, and a unique non-linear regres-sion tool, allowing to analyse even complex cases in near-real time. In particular, validation allows to quantify model's ability to predict further activations, by employing distinct fitness approaches at once. Regression provides a continuous representation of discrete kernels (usually, difficult to be interpreted), using analytical expressions, providing insights on groundwater dynamics responsible for slope instabilities. An application to the Uncino landslide in Southern Italy, characterized by 6 reactivations, is described. Calibration confirmed the same optimal fitness results already obtained with the previous version. Through an extended sensitivity analysis, the potentialities of the new release could be highlighted, providing numerous improved solutions thanks to additional fitness criteria (base time, safety margin, first-order momentum), as further confirmed through validation. A set of regression experiments was performed, by employing 6 gamma distributions, obtaining satisfactorily results.
GA SAKe 2.0-An Advanced Hydrological Model for Predicting the Timing of Landslide Activations
De Rango A.;D'ambrosio D.;Lupiano V.
;Mendicino G.;Iovine G.
2025-01-01
Abstract
GA SAKe is a linear, steady-state hydrological model. It employs Genetic Algorithms to predict the timing of triggering of landslides, based on rainfall and previous activations. Calibration provides discrete kernels, that allow to analyse even complex geohydrological conditions, with a high level of detail. Once validated, the mobility function of the model allows to predict landslide activations. The release 2.0, here presented, introduces a powerful computational engine, and a couple of relevant modules: a validation tool, and a unique non-linear regres-sion tool, allowing to analyse even complex cases in near-real time. In particular, validation allows to quantify model's ability to predict further activations, by employing distinct fitness approaches at once. Regression provides a continuous representation of discrete kernels (usually, difficult to be interpreted), using analytical expressions, providing insights on groundwater dynamics responsible for slope instabilities. An application to the Uncino landslide in Southern Italy, characterized by 6 reactivations, is described. Calibration confirmed the same optimal fitness results already obtained with the previous version. Through an extended sensitivity analysis, the potentialities of the new release could be highlighted, providing numerous improved solutions thanks to additional fitness criteria (base time, safety margin, first-order momentum), as further confirmed through validation. A set of regression experiments was performed, by employing 6 gamma distributions, obtaining satisfactorily results.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
