Chronic and rainfall-associated ammonia nitrogen export varies across land-use and river-connectivity contexts in Chinese river networks.

Extreme rainfall can alter riverine nitrogen export, but its effects may depend on land use and river-network connectivity and may differ between chronic pollution pressure and event-related amplification. We integrated ammonia nitrogen (NH₃-N) monitoring data, reach-scale discharge, climate, land cover and the Connectivity Status Index (CSI) to examine monthly NH₃-N export across Chinese river reaches during 2021-2024. Monthly Yield quantified area-normalised absolute export, whereas SurgeRate quantified export relative to each reach's mean Yield during months without extreme-rainfall days. Linear mixed-effects models were used to evaluate conditional interactions, and histogram-based gradient boosting regression with spatial cross-validation and SHAP analysis was used to characterise nonlinear predictive structures. The final datasets contained 52,656 reach-month observations from 1097 reaches for Yield and 49,164 observations from 1095 reaches for SurgeRate. The association between extreme-rainfall days and both responses became progressively more positive with increasing impervious and cropland cover. Connectivity did not uniformly amplify the rainfall response; its association with SurgeRate varied with cropland context, although the higher-order interaction was comparatively weak and sensitive to baseline specifications. Total monthly precipitation was the leading predictor of both responses, while Yield retained stronger landscape and geomorphic structure and SurgeRate showed greater temporal and seasonal dependence. Combining background Yield with event-month SurgeRate separated 1110 reaches into low/stable (39.2%), chronic high-output (43.0%), event-sensitive (10.8%) and compound high-risk (7.0%) types. Conditional future experiments for 2030-2098 indicated comparatively stable but climate-model-dependent national predicted ln(Yield + 1) and declining SurgeRate, with the most consistent decline under SSP5-8.5. Compound high-Yield and high-SurgeRate states remained rare, although their estimated prevalence was threshold-sensitive. These findings indicate that absolute export and relative rainfall-associated sensitivity represent distinct dimensions of riverine NH3-N risk and support differentiated source-control and rainfall-targeted monitoring strategies.