Integrating Physical and Artificial Intelligence Models to Evaluate Urban Neonicotinoid Pollution: Rainfall Washoff Dynamics, Driving Factors, and Risk Implications in the Guangdong–Hong Kong–Macao Greater Bay Area

The extensive urban application of neonicotinoid insecticides (NNIs) has resulted in large-scale environmental contamination via rainfall runoff transport. However, their washoff dynamics, driving factors, and resulting ecological risks during rainfall events remain poorly understood. This study addresses this gap by developing a hybrid model that integrates a gated recurrent unit (GRU)-based module for drainage outlet rainfall-runoff prediction with a physical process-based module for NNIs pollution modeling in China’s Guangdong–Hong Kong–Macao Greater Bay Area (GBA). The model successfully simulated rainfall-runoff processes and NNI transport dynamics, demonstrating that an average of 2.49% of applied NNIs entered into waterways through drainage outlets, with thiamethoxam (THM) exhibiting the highest washoff ratio and acetamiprid (ACE) exhibiting the lowest. Spatially, NNI concentrations and washoff rates were highest in the GBA’s peripheral regions and decreased toward the Pearl River Estuary. Rainfall intensity, maximum rainfall, and maximum runoff drive NNIs washoff, while pesticide use and runoff volume determine spatial concentration distributions. High-risk proportions increase with rainfall intensity, while urban areas with mountainous as well as significant elevation changes have a high risk of NNIs. This research provides the first comprehensive framework for assessing NNI pollution at an urban agglomeration scale, offering critical insights for sustainable pesticide management in rapidly urbanizing watersheds.