The northeast of Brazil is characterized by a semiarid climate, in which there is a considerable temporal and spatial variability in the distribution of rainfall. In this region, the occurrence of intense rainfall events (IREs) causes severe damage to the population, given that the rampant urbanization and land use processes make cities more susceptible to harmful consequences. Thus, the objective of this study is to estimate the period and the level of return of IREs in the capital cities of the northeast Brazil. For that end, observed data from weather stations of the National Institute of Meteorology for the period from 1988 to 2017 were used. We defined thresholds based on the excesses above the mean and inserted the data in the generalized Pareto distribution in order to obtain diagnostics of future estimates. Results showed that return periods were surpassed in seven of the nine capital cities in 2018. Furthermore, the estimates for IREs with a return period of 5, 10, and 100 years were higher if compared to the return period of 1 year. The results found in this study are extremely relevant for the understanding of the occurrence of these events and also serve as a tool for decision-making and the elaboration of policies aimed at minimizing the impacts of such events. K E Y W O R D S extreme value theory, generalized Pareto distribution, natural disasters

Extreme rainfall and streamflow events are becoming increasingly more relevant to discussions concerning potential climate changes on planet earth. In most cases, impacts related to the occurrence of such phenomena can be amplified due to the lack of planning and structure. In this context, the main objective of the study is to estimate the period and level of return of intense rainfall events in homogeneous subregions of the São Francisco hydrographic region, as well as to define which of these subregions are most vulnerable to the recurrence of these episodes. We used rainfall data from 105 rain gauges and streamflow data from 103 stream gauges comprising the years from 1988 to 2017. Through cluster analysis, four homogeneous rainfall subregions were defined and validated by the silhouette index, capable of identifying misallocated stations in the clustering process. A generalized Pareto distribution from the extreme value theory was used to calculate the period and level of return of intense rainfall events in each subregion. We then related the retrieved estimates with information about the drainage network, streamflow, land use and occupation, and socioeconomic and demographic factors in the watershed region. Results indicated the existence of subregions vulnerable to hydrological and geological processes intensified due to human activity and the recurrence of hydrometeorological phenomena. The scenarios found for the recurrence of intense rainfall events indicated the severity to which some subregions may be subjected, moreover, they reinforce the need for the development of public policies for the prevention and minimization of impacts.