Methodological proposal to remote detection and management of areas that are naturally vulnerable to floods

Abstract

Floods are the main natural disasters in Brazil, causing loss of life and socioeconomic damage. This work proposes a model for the remote detection of areas that are naturally flood-prone due to the morphometric characteristics of their relief and drainage networks in the Alto Sapucaí River in Minas Gerais, Brazil. The morphometric parameters used were the drainage density, river density, relief ratio, roughness index, maintenance coefficient, form factor and stream surface length. The risk areas had a compactness coefficient of 0.75 and a form factor of 0.56, and both were considered a high risk for floods. The obtained results allowed the identification of a significant predictive equation that suggested a cutoff value of 3.82 for the discriminant function; areas with values under this cutoff were considered naturally more vulnerable to floods occurrences. These areas were corroborated with the emergency maps of the municipalities. The map obtained by the proposed model was compared with the Civil Defense map, and its accuracy, according to the Kappa coefficient, was 0.83, indicating strong similarity between the two maps.