Cropland nitrogen use efficiency in China: key predictors, threshold responses, and management prioritization under...

Improving cropland nitrogen use efficiency (NUE) is crucial for advancing agricultural green transformation and mitigating water eutrophication. However, the spatiotemporal coupling, regional heterogeneity, and underlying associations linking agricultural modernization level (AML) and NUE remain insufficiently understood. Here, we analyzed spatiotemporal changes in NUE and AML across 30 Chinese provinces from 2010 to 2024 using a random forest model combined with interpretable machine learning. Nationally, NUE rose from 27.9% to 43.1% and AML from 0.238 to 0.308. High-NUE areas concentrated in Northeast China, high-AML in southeastern coastal provinces, and western regions showed low but rapidly growing AML. Three AML indicators-per capita livestock product output (PCLP), per capita cropland area of farmers (PCCF), and per capita income of farmers (PCI)-exhibited the strongest positive predictive associations with NUE, whereas pesticide application rate (PEST) showed a negative predictive association with NUE. Model-derived response ranges associated with relatively high and stable predicted NUE were identified: PCLP, 0.10-0.31 t per capita; PCI, 17,038-29,309 CNY; PEST, 1.78-19.38 kg ha^-1; and PCCF, 0.86-1.38 ha per capita. Provincial diagnosis further showed that most provinces remained below the PCI and PCCF ranges, while southeastern provinces such as Fujian and Hainan exceeded the PEST range. These results provide quantitative reference points for province-level NUE management prioritization and highlight differentiated modernization bottlenecks across regions.