Spatial coherence of climate extremes across Brazil’s river basins: a clustering approach

Understanding how extreme climate events organize spatially is essential for robust climate regionalization. This study proposes a spatial clustering framework that integrates Moran’s I with the Weighted Total Sum of Squares (WTSS) through a convex combination, balancing intra-cluster compactness and spatial coherence. Using 45 monthly extreme climate indices, the Principal Component Analysis (PCA) was applied to reduce dimensionality and stabilize k-means clustering under strong spatial dependence. Two streamlined experiments are conducted: (i) clustering each of Brazil’s twelve hydrographic basins to provide a natural spatial cross-validation of the method, and (ii) clustering the entire country to identify broad-scale patterns. The basin-level results converge to similar optimal configurations, while national clustering reveals coherent regions consistent with known climatic divisions. The proposed approach enhances spatial cluster validation by explicitly incorporating autocorrelation, producing interpretable and geographically meaningful representations of climate extremes.