Identifying spatiotemporal patterns and driving factors of nitrogen leaching in Henan Province by integrating a process-based model with machine learning

Understanding the spatiotemporal patterns and driving factors of nitrogen (N) leaching is crucial to mitigating groundwater nitrate contamination. This study integrated a process-based model with machine learning to quantify the spatiotemporal distribution of N leaching and identify its dominant drivers in Henan Province, a major wheat-maize rotation region in China. Thirty-year (1992–2022) simulations revealed a south-to-north decreasing gradient in annual N leaching, from 215 kg N ha⁻¹ in southern areas to 85 kg N ha⁻¹ in the north. Mean N leaching was substantially higher during the maize season (36–123 kg N ha⁻¹) than during the wheat season (42–100 kg N ha⁻¹), with the maize season contributing about 70% of annual N leaching in parts of Henan. Interannual variability was mainly driven by precipitation, with significantly greater maize-season leaching occurring in wet years. Water and N management were the dominant drivers across all temporal scales. Soil factors exerted stronger influence than meteorological conditions during the wheat season, but weaker influences during the maize season. Nitrogen application, precipitation, irrigation, relative humidity and field capacity (FC) collectively explained over 70% of leaching variability. Nitrogen application exhibited a softplus-type relationship with N leaching, whereas water-related variables displayed sigmoidal response patterns. Field capacity regulated the seasonal redistribution of N leaching. Relatively high FC promoted N accumulation during dry years, resulting in higher N leaching during wet years. These findings provide a scientific basis for developing targeted water and N management and soil amendment to mitigate N leaching risks.