Estimating the impact of 2D urban landscape patterns on extreme precipitation based on non-stationary models in the Guangdong-Hong Kong-Macao Greater Bay Area, China.

Abstract

Rapid urbanization has significantly altered surface landscape configurations, leading to complex urban climates. While much attention has been focused on impervious surfaces' impact on extreme precipitation, a critical gap remains in understanding how various 2D urban landscape components influence extreme precipitation across different durations. Through an analysis of the non-stationarity and spatiotemporal variations in extreme precipitation across the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) from 1990 to 2020, we constructed the non-stationary Generalized Additive Models for Location Scale and Shape (GAMLSS) model by introducing six urban landscape structural metrics as explanatory variables for each of the 27 meteorological stations in the GBA. Additionally. we assessed the frequency of these metrics in the best-fitting models and predicted design values across different interannual periods. Our findings reveal that aggregation metrics (patch density: PD) and diversity metrics (Shannon's Diversity Index: SHDI) appeared more frequently in the best-fitting models than other metrics within all extreme precipitation indices. For short-duration extreme precipitation indices (≤ 3 h), the area matrix (Impervious Surface Percentage: ISP), PD, and SHDI were selected more often than other metrics, whereas for long-duration (> 3 h), PD and SHDI had a higher relative frequency as ISP's impact decreased. Design values peaked in the 2010s across all return periods (100, 50, and 20 years), highlighting the importance of integrating urban landscape features into non-stationary models of extreme precipitation. This research provides valuable insights for improving the management of urbanization-induced heavy precipitation and flood risks.