Timescale-dependent impacts of extreme precipitation on watershed nutrient removal: Insights from five decades (1970–2022)

Intensifying extreme precipitation (EP) due to global climate change poses severe challenges to controlling watershed non-point source pollution. However, the performance of best management practices (BMPs) under EP and its associated timescale-dependent mechanisms are poorly understood. We develop a multiscale, long-term assessment framework to quantify BMP effectiveness and resilience under EP at annual, seasonal, and monthly scales, and apply it to a tributary watershed of the Pearl River, China’s third-largest river, using 1970–2022 data. Our findings reveal that EP significantly amplifies performance variability, with the mean coefficient of variation (CV) for total nitrogen (TN) reduction surging by 110 %. Application of the resilience index (RI) indicates substantial divergence in BMP resilience: vegetated filter strips (VFS) prove highly resilient (RI < 0.15), with TN removal enhanced under EP (increasing from 30.83 % to 35.95 %), whereas fertilizer reduction is vulnerable, with its TN reduction declining from 8.18 % to 5.45 %. Crucially, BMP responses are strongly scale-dependent. For instance, conservation tillage shows improved annual TN removal but degrades performance during the rainy season, demonstrating that annual-level assessments can mask critical seasonal vulnerabilities. This study underscores the necessity of multiscale analysis to develop climate-adaptive watershed management. It provides decision-relevant evidence for designing resilient BMP portfolios, such as prioritizing stable measures like VFS, to ensure long-term pollution control in an era of increasing climate extremes.