Deciphering summer precipitation patterns: A dual-step framework for classification in the Middle-Lower Yangtze River Basin

Precipitation spatial patterns significantly impact regional water resource management, agricultural production, and socioeconomic development. While conventional methods typically explain less than 80 % of total precipitation variability due to challenges in analyzing spatially heterogeneous rainfall data, this study introduces a novel dual-criterion framework integrating Agglomerative Hierarchical Clustering (AHC) with Rotated Empirical Orthogonal Function (REOF) to enhance pattern identification. Using meteorological station data (1951–2024) from the Middle-Lower Yangtze River Basin (MLYRB), we identified 11 homogeneous sub-regions and six dominant precipitation patterns explaining a notably high 85.1 % of total variance: basin-wide coherent (34.7 %), north-south dipole (18.6 %), western dominant (11.3 %), lower reaches dominant (8.0 %), and two tripole patterns (12.4 % combined). The basin-wide pattern, occurring most frequently (31 years), dominated historical catastrophic floods (1954, 1998, 2020). Circulation analysis revealed a distinctive tripole structure in the 500 hPa geopotential height field driving the basin-wide pattern, while anticyclonic circulation anomalies characterized the north-south dipole pattern. The positive phase of the basin-wide pattern features enhanced southwesterly moisture transport from tropical sources with peak convergence of −2.29 × 10 −6 kg/(m 2 ·s) over the central Yangtze region, mechanistically explaining basin-wide precipitation increases. Wavelet coherence analysis uncovered a critical temporal shift in ENSO's influence on basin-wide precipitation from short (2–4 years) to long (8–12 years) periods since the 1960s, fundamentally transforming teleconnection impacts on major flood events. Concurrently, PDO maintains a robust decadal-scale relationship with lower Yangtze precipitation patterns throughout the entire study period, offering potential for reliable long-term predictability. These findings provide a scientific foundation for improving seasonal precipitation forecasting and regional water resource management under changing climate conditions.