Considering spatial heterogeneity of cultivation conditions can effectively improve the assessment of nitrogen use at the provincial scale in China

•China provincial annual nitrogen (N) use is assessed from 1985 to 2020.•An improved scheme for assessing the contribution of synthetic N and soil fertility to crop growth is proposed.•Crop N uptake is modeled directly with N input components by using Random Forest.•The response of crop N uptake and NUE to the regulation of N inputs is simulated. Improving nitrogen (N) management are crucial for food security and ecological protection in China and globally. Assessing the contribution and influence characteristics of N input components on crop growth is a key component. There are challenges in conducting such assessments at large regional scales, particularly in developing models that fit regional N inputs and crop N uptake. An improved scheme that fit regional N inputs and crop N uptake with consideration of spatial heterogeneity of cultivation conditions was proposed. Based on the division of homogeneous cultivation conditions zones, linear models and Random Forest models were developed to assess the contribution and influence characteristics of N input components on crop growth at the provincial scale in mainland China. And synthetic N contribution rate (SNCR) and soil fertility N contribution (SFNC) were innovatively proposed to represent the contribution of fertilizer and soil fertility to crop growth. The results showed that the overall N use efficiency ranged from 35 % to 55 % in 1985–2020, and higher NUE could be seen in major grain producing provinces. The SNCR generally declined, with a spatial pattern higher in the southwest, northwest, and northeast regions. The SFNC showed an increasing trend, with a spatial pattern higher in the northeast than southeast than west. For each N input component, Synthetic N input was critical in the northwest, southwest regions and north China, positively affecting crop N uptake. The improved scheme, which incorporated considerations for spatial heterogeneity, demonstrated superior accuracy compared to the model fitted by year, increasing from 0.27 (0.11) to 0.42 (0.47). Over the past 35 years, the crop N uptake and N surplus of mainland China had experienced a process of extensive-unsustainable-sustainable-conservative pattern changes. The response of crop N uptake and N use efficiency to the regulation of N inputs in typical regions was simulated, providing reference for provincial N inputs regulation in China. This study can provide support for the design of N management strategies in China to reduce N pollutio