Principal Component Analysis and Hierarchical Clustering of Heavy-Rainfall Events in the Baihetan Hydropower Station Dam Area

Rainstorm-related hazards frequently threaten the Baihetan Hydropower Station, a major water conservancy hub in southwestern China, and objective characterization of rainfall events is needed to support disaster prevention and mitigation in the dam area. Based on hourly precipitation observations from 11 meteorological stations in the dam area during 2012–2022, this study developed a classification framework for rainstorm feature analysis and classification that integrates Principal Component Analysis (PCA) and hierarchical clustering. Fifteen feature indicators were grouped into four dimensions, i.e., extremeness, intensity, and spatio-temporal distribution. PCA reduced them to six principal components explaining 87.49% of the total variance, and hierarchical clustering with the optimal number of clusters determined by the elbow method classified the 42 rainstorm events into three relatively robust types and one rare extreme case group: north-biased moderate-scale precipitation (69.05%), south-biased rainstorms (14.29%), north-biased long-duration local precipitation (11.90%), and anomalous station-concentrated bullseye rainfall (4.76%, n = 2, exploratory only). The latter is not interpreted as a statistically stable type because of its very small sample size. The four types are distinguishable in the principal component space and exhibit distinct spatial and temporal patterns; the southward-type heavy rain events occurred mainly in 2020–2021, whereas the two rare bullseye events occurred in 2015 and 2021, possibly linked to large-scale circulation anomalies. The proposed classification framework demonstrates good applicability and interpretability, providing a methodological reference for rainstorm risk assessment and early-warning indicator design in the Baihetan dam area.