Statistical Modeling and Trend Investigation in Precipitation Data of UGRHI-17, Médio Paranapanema - SP

Abstract

This study synthesizes the analysis of precipitation dynamics in the UGRHI-17 (Middle Paranapanema) from 1945 to 2021, integrating three methodological approaches: hierarchical clustering, simple linear regression, and the Pettitt test for change-point detection. Initially, each station was grouped into three clusters based on its monthly climatology. The first cluster, comprising low-relief sites (Cabrália Paulista, Rancharia, and Ribeirão do Sul), showed a non-significant increasing trend of 1.6 mm yr⁻¹. The second cluster (Botucatu and Gália) exhibited large fluctuations between 1 000 mm and 1 900 mm yr⁻¹, with a slight positive bias (\~2 mm yr⁻¹) that was also not significant. In the third cluster, higher-relief stations (Óleo, Ourinhos, Cândido Mota, Avaré, and Ipaussu, there was a statistically significant increase of approximately 3.2 mm yr⁻¹ (p ≈ 0.005) and extreme variability, with annual totals ranging from under 800 mm to over 2 000 mm. The Pettitt test identified a common regime shift across all clusters between 1969 and 1971, marking a pluviometric jump that slightly raised annual means without attenuating interannual amplitude. These findings corroborate previous multifractal and extreme-event analyses in neighboring basins, demonstrating that although natural variability driven by frontal systems and convective currents remains dominant, there is a clear bias toward gradual intensification of precipitation, especially in higher-relief areas.