Modeling and spatio-temporal prediction rainfall data applied to mass movements in Belo Horizonte, Minas Gerais - Brazil

Disasters associated with natural events are the focus of many studies due to their great potential for negative impacts. Events linked to mass movements caused by rainfall have significantly increased in quantity and magnitude of impact. Rainfall is the main trigger for mass movements and, therefore, its measurement is paramount for understanding, monitoring the natural disasters caused by these events. The lack of a historical rainfall series with spatialized data on an adequate scale is a major limitation for developing risk management studies. Faced with the scarcity of rainfall data interpolated on a regular grid, geostatistics has become useful for generating climate data where observed data are insufficient or non-existent. Thus, this work aimed to apply the geostatistical modeling and spatio-temporal prediction methodology to rainfall data in Belo Horizonte, Minas Gerais - Brazil for use in municipal disaster risk management. Specifically, were evaluated: Kriging and Fixed Station Kriging (FRK). Among these two predictors, FRK obtained more robust and homogeneous results. The kriging variance was used to calculate uncertainties. The FRK method showed less variability in uncertainty estimation. The location of the maximum annual rainfall in one day and the maximum accumulated rainfall in four days were also compared with the village/slum areas and with the distribution of the main areas where mass movements happened. It was found that the mass movements recorded in Belo Horizonte are mostly in regions of villages and favelas. It was observed that even regions with lower rainfall have many occurrences of mass movements, which may indicate that these areas have some conditioning characteristics.