Intensity-Duration-Frequency Equations for Brazil by the Maximum Daily Precipitation Transformation Method: Does Hydrological Regionalization Improve the Estimates?

Intensity-duration-frequency (IDF) equations are essential for hydraulic design but depend on pluviographic records that are scarce in Brazil. The Transformation of Maximum Daily Precipitation (TMDP) method addresses this scarcity by converting a rain gauge station’s maximum daily rainfall into subdaily intensities, using a transformation coefficient from the nearest pluviographic station. This study tested whether restricting this transfer to hydrologically homogeneous regions would improve accuracy. Homogeneous regions were delineated by principal component analysis and hierarchical clustering under three variable sets: reference intensities derived from IDF equations, atmospheric reanalysis variables, and their combination. Each configuration was compared with conventional TMDP and disaggregation by national coefficients, through leave-one-out cross-validation and a paired Wilcoxon test. At the national scale, regionalization did not improve the estimates; instead, all three configurations produced a small but statistically detectable reduction in performance, with median KGE decreases ranging from 0.035 to 0.055. The nearest-neighbor criterion already incorporates the hydrological similarity required to convert daily depths into subdaily intensities, so rigid clustering boundaries added cost without benefit, making rigid regional boundaries redundant at the national scale. The Northeast was a relevant exception: OBS significantly outperformed TMDP, while ENV showed the highest frequency of first-place results among the regionalized configurations, although its difference from TMDP was not statistically significant. Relative to national-coefficient disaggregation, TMDP eliminates a systematic 26.5% underestimation that compromises design safety. Applying the best configuration, IDF parameters were made available for 5594 rain gauge stations, one equation per 1521 km2.