The rapid expansion of soybean-growing areas across Europe raises questions about the suitability of agroclimatic conditions for soybean production. Here, using data-driven relationships between climate and soybean yield derived from machine-learning, we made yield projections under current and future climate with moderate (Representative Concentration Pathway (RCP) 4.5) to intense (RCP 8.5) warming, up to the 2050s and 2090s time horizons. The selected model showed high R² (>0.9) and low root-mean-squared error (0.35 t ha⁻¹) between observed and predicted yields based on cross-validation. Our results suggest that a self-sufficiency level of 50% (100%) would be achievable in Europe under historical and future climate if 4–5% (9–11%) of the current European cropland were dedicated to soybean production. The findings could help farmers, extension services, policymakers and agribusiness to reorganize the production area distribution. The environmental benefits and side effects, and the impacts of soybean expansion on land-use change, would need further research.

欧洲大豆种植面积的迅速扩大引发了人们对农业气候条件是否适合大豆生产的质疑。在这里，我们利用机器学习得出的气候与大豆产量之间的数据驱动关系，对当前和未来气候下的产量进行了预测，气候从中等（代表性浓度路径（RCP）4.5）到强烈（RCP 8.5）变暖，直至 2050 年代和 2090 年代的时间范围。所选模型显示出高R ² (>0.9) 和低均方根误差 (0.35 t ha ⁻¹) 基于交叉验证观察到的和预测的产量之间。我们的研究结果表明，如果当前欧洲农田的 4-5%（9-11%）用于大豆生产，则在历史和未来气候下，欧洲可以实现 50%（100%）的自给自足水平。研究结果可以帮助农民、推广服务、决策者和农业企业重组生产区分布。环境效益和副作用，以及大豆扩张对土地利用变化的影响，都需要进一步研究。