ORIGINAL PAPER A model-based site selection approach associated with regional frequency analysis for modeling extreme rainfall depths in Minas Gerais state, Southeast Brazil L. C. Assis 1 • M. L. Calijuri 2 • D. D. Silva 2 • E. O. Rocha 3 • A. L. T. Fernandes 1 • F. F. Silva 2 Ó Springer-Verlag GmbH Germany 2017 Abstract Extreme rainfall data are usually scarce due to the low frequency of these events. However, prior knowl- edge of the precipitation depth and return period of a design event is crucial to water resource management and engineering. This study presents a model-based selection approach associated with regional frequency analysis to examine the lack of maximum daily rainfall data in Brazil. A generalized extreme values (GEV) distribution was hierarchically fitted using a Bayesian approach and data that were collected from rainfall gauge stations. The GEV model parameters were submitted to a model-based cluster analysis, resulting in regions of homogeneous rainfall regimes. Time-series data of the individual rainfall gauges belonging to each identified region were joined into a new dataset, which was divided into calibration and validation sets to estimate new GEV parameters and to evaluate model performance, respectively. The results identified two distinct rainfall regimes in the region: more and less intense rainfall extremes in the southeast and northwest regions, respectively. According to the goodness of fit measures that were used to evaluate the models, the aggregation level of the parameters in clustering influenced their performance. Keywords Regional frequency analysis  Model-based site selection  Extreme daily rainfall  Hierarchical Bayesian inference  Model-based cluster analysis  Return period 1 Introduction Heavy rainfall in Southeast Brazil is mostly associated with two types of atmospheric perturbation: Cold Front and the South Atlantic Convergence Zone (Lima et al. 2010). The processes that generate events classified as extreme are very complex and depend on scale factors linked to several features of general atmospheric circulation (Crisci et al. 2002). The estimation of extreme events is often required in hydrological practice (Viglione et al. 2007); the pre- diction of precipitation amounts and their uncertainties are important in hydrological modeling, the design of hydraulic structures, and urban and landscape planning (Kysely ´ et al. 2011). Indeed, given the scarcity of discharge data for flood events, rainfall frequency estimates are often used to describe the characteristics of such events (Nor- biato et al. 2007). However, storm data records, such as maximum daily rainfall depths, are not easily obtained, and only a few records are typically available in Southeast Brazil, where heavy rainfall events in the summer are responsible for almost all natural disasters (Lima et al. 2010). Conventional statistical methods (i.e., single-site analy- sis) are limited in providing estimates for higher return periods. For design purposes, water resource engineers often face the task of estimating the rainfall depth for an event with a return period that is longer than the length of records available (Burn 1990). Methods based on the regional analysis of hydrological variables are an alterna- tive solution. The framework of regional methods makes & L. C. Assis leonardo.assis@uniube.br 1 Universidade de Uberaba - UNIUBE, Av. Nene ˆ Sabino, 1801, Uberaba, MG 38.055-500, Brazil 2 Universidade Federal de Vic ¸osa - UFV, Av. PH Rolfs, Vic ¸osa, MG 36.570-000, Brazil 3 Fundac ¸a ˜o Estadual de Meio Ambiente de Minas Gerais - FEAM, Belo Horizonte, MG, Brazil 123 Stoch Environ Res Risk Assess DOI 10.1007/s00477-017-1481-1