A GEOSPATIAL MODEL FOR DESERTIFICATION RISK MAPPING (RIDES): APPLICATION IN THE SEMIARID REGION OF CEARÁ MUNICIPALITIES, BRAZILIAN NORTHEAST

Desertification is a current environmental problem that causes the degradation of natural vegetation and soils, reducing the productive potential of agricultural lands. There are few methods in the literature for mapping desertification risk, especially for Brazilian geographical conditions. This study proposes a new geospatial model for mapping desertification risk (RIDES), tested in municipalities of the state of Ceará in the Brazilian semi-arid region. The model uses true desertification areas, mapped by 2020 Sentinel-2A/MSI orbital images visual interpretation, as a dichotomous variable, and the following environmental variables associated to desertification process: average annual precipitation (PRT) from 1990 to 2020; vegetation index (IVE) and land surface temperature (TST), calculated from 2018 to 2020; hypsometric integral of the relief (IHI); topographic position index (IPT); terrain roughness index (IRT); rural population density (DPR) in 2020; distance to river channels (DCF), distance from urban areas (DAU), and total area of agriculture and pasture (AGP). The relationship between the dichotomous and explanatory variables was analyzed using the logistic regression method, from which the probability of desertification occurrence (p) was calculated and used as a risk measure. The results showed that the variables IVE, TST, DPR, AGP, and PRT had the highest correlation with desertification and the most significant weight in estimating desertification risk. The RIDES model had an accuracy of 91.9% in mapping areas undergoing desertification processes and can be used to monitor desertification risk in the Brazilian semi-arid region.