Global gridded soybean yield dataset for future warming scenarios using emergent constraints

Reliable projections of future soybean yields under global warming are essential for food security and climate adaptation planning. However, current yield projections from multiple climate-crop models are subject to substantial uncertainty (i.e., inter-model spread), thereby limiting the reliability of subsequent adaptation decisions. Here, we developed a global gridded soybean yield dataset with reduced uncertainty under 1.5℃ and 2.0℃ warming scenarios. The raw multi-model yield simulations from different emissions scenarios were first aligned to consistent global warming levels. The emergent constraint approach was then applied to further constrain the multi-model projections using observed historical yields, leveraging robust emergent relationships between simulated historical and future yields across models. The resulting constrained projections exhibit a marked reduction in uncertainty, with a median reduction exceeding 65% relative to the raw multi-model simulations. The constrained soybean yields also show improved agreement with both subnational statistics and independent gridded yield datasets. This dataset can serve as a more reliable input for integrated assessments of climate change impacts on global soybean market.