MODELLING LANDSLIDE SUSCEPTIBILITY USING THE WEIGHT OF EVIDENCE METHOD IN A TROPICAL MOUNTAIN REGION
Modelling landslide susceptibility using the weight of evidence method in a tropical mountains region
DOI:
https://doi.org/10.5016/geociencias.v40i1.14992Abstract
Landslides are natural phenomena that take place in mountainous regions and can raise large socio-economic disasters. The recognition of landslide susceptible areas is a fundamental step in risk management to carry out projects that aim to mitigate, or even avoid damages caused by these events. This work evaluates landslide susceptibility by Weight of Evidence method, as well as the individual control of eight landslide predisposing factors in two watersheds in the Tijuca’s massif, in Rio de Janeiro city (Brazil). Five susceptibility models were computed by the combination of the predisposing factors. The models had their predictive capacity measured by the area under the curve (AAC) test. Most of the recognized landslides took place at headwaters, on steep slopes between 34° and 51 °, mainly supported by granitic intrusions and covered by preserved forest. The final map shows five landslide susceptibility classes and proved to be effective in predicting shallow landslides. The method is robust and avoids subjectivities while very versatile. Although little used in Brazil, the method is suitable to the context of complex geodiversity and lack of geotechnical data at the regional and local scales.