Maize and wheat are two major staple foods that collectively contribute two-thirds of the world’s grain supply. The extensive use of nitrogen (N) fertilizers during the cultivation of both crops leads to significant losses of reactive nitrogen (Nr) into the environment. Here, using machine learning algorithms, we generate high-resolution maps of crop-specific soil Nr losses based on global field measurements. We estimate that global annual soil Nr losses from the use of synthetic N fertilizer in 2020, including direct emissions of nitrous oxide (N₂O), nitric oxide (NO), ammonia (NH₃), N leaching and run-off, amount to 0.18, 1.62, 0.09, 1.47 and 1.10 million tonnes N for maize, and 0.12, 1.33, 0.07, 1.21 and 0.95 million tonnes N for wheat, respectively. The annual indirect N₂O emissions induced by synthetic N fertilizer use from these soil Nr losses are estimated to be 45,000 and 37,000 tonnes for maize and wheat, respectively, with hydrologic pathways playing a predominant role. Enhancing N use efficiency up to 60% for regions below this value can achieve a total soil Nr loss mitigation potential of 4.00 million tonnes per year for the two crops, thereby reducing indirect N₂O emissions by 49%. Our results contribute to constrain global N budgets from the use of fertilizer in agriculture, which then can help to improve projections of nitrogen cycle–climate feedbacks using modelling approaches.

玉米和小麦是两种主要主粮，合计占世界粮食供应量的三分之二。在这两种作物的种植过程中大量使用氮肥会导致活性氮（Nr）大量流失到环境中。在这里，我们使用机器学习算法，根据全球现场测量生成特定作物土壤氮损失的高分辨率地图。我们估计，2020年全球因使用合成氮肥而导致的土壤氮年度损失，包括一氧化二氮（N ₂ O）、一氧化氮（NO）、氨（NH ₃ ）、氮淋失和径流的直接排放量，玉米氮用量分别为 0.18、1.62、0.09、1.47 和 110 万吨，小麦氮用量分别为 0.12、1.33、0.07、1.21 和 95 万吨。玉米和小麦因使用合成氮肥而造成的土壤 Nr 损失估计每年间接 N ₂ O 排放量分别为 45,000 吨和 37,000 吨，其中水文途径起着主导作用。将低于该值的地区的氮利用效率提高至60%，可以实现两种作物每年400万吨的土壤氮流失总量减缓潜力，从而减少49%的间接N ₂ O排放。我们的结果有助于限制农业中化肥使用的全球氮预算，从而有助于使用建模方法改进氮循环-气候反馈的预测。