An enhanced method for rainfall pattern recognition and its asymmetrical change in Guangdong-Hong Kong-Macao Greater Bay Area (GBA)

Study region Guangdong-Hong Kong-Macao Greater Bay Area (GBA). Study focus This study investigated whether long-term changes in heavy rainfall differ among rainfall patterns in the GBA. Using hourly rainfall observations from 23 stations during 1971–2020, independent rainfall events were identified and classified with a duration-adaptive rainfall pattern recognition framework to reduce mismatch between the number of identification points and event duration. Rainfall amount, duration, peak and precipitation concentration index (PCI) were analyzed for quantile-specific temporal trends and spatial variation. New hydrological insights for the region Unimodal rainfall patterns dominated the GBA, with Mode I accounting for 36.12% of all classified events. Rainfall pattern recognition was sensitive to mismatch between the number of identification points and event duration, and instability increased markedly when the former was smaller. Rainfall amount, duration and PCI showed clear quantile-dependent asymmetry, with lower quantiles tending to show negative or near-zero trends and upper quantiles tending to show positive trends, while rainfall peak increased across most of the distribution. This asymmetry was strongest for Mode I and much weaker for the bimodal patterns. Spatially, rainfall amount and peak showed broader centroid movement at upper quantiles, whereas duration and PCI were more variable at lower quantiles. These findings indicate that long-term rainfall change in the GBA is not uniform across events, but is strongly conditioned by rainfall pattern and quantile level.