Modelando eventos extremos de precipitação no semiárido nordestino utilizando a Distribuição ZIGEV

Abstract

The semiárido region of Northeast Brazil (NEB) is characterized by irregular rainfall patterns and high evapotranspiration rates, leading to water scarcity. Additionally, the region is prone to extreme precipitation events, which can trigger socio-environmental disasters with severe impacts. Despite the development of various models for modeling extreme events, their application in the semi-arid region of the NEB is limited by the lack of detailed data and methodologies tailored to its local characteristics. This study aims to model extreme precipitation events in this region using the Zero-Inflated Generalized Extreme Value (ZIGEV) distribution. The methodological approach is based on Bayesian inference, which incorporates uncertainties and prior information into the parameter estimation process. The analysis focuses on a sub-region of the NEB semi-arid zone, marked by extremely low annual precipitation and high susceptibility to extreme precipitation events. Daily precipitation data from 16 (sixteen) localities within the study area were used to fit the model and estimate the return levels of extreme events. The results demonstrated the effectiveness of the ZIGEV model in predicting extreme precipitation events, providing reliable parameter estimates and outperforming the standard GEV model, which faced convergence issues in many datasets. The ZIGEV model proved suitable for data with a high proportion of zeros and also converged to the GEV model when the proportion of zeros was lower. Additionally, the estimated return levels for each locality provided valuable insights into the occurrence and intensity patterns of these events, contributing to water resource management and the development of disaster mitigation strategies for the NEB semi-arid region.