Both climate and impact models are essential for understanding and quantifying the impact of climate change on agricultural productivity. Multi-model ensembles have highlighted considerable uncertainties in these assessments, yet a systematic approach to quantify these uncertainties is lacking. We propose a standardized approach to attribute uncertainties in multi-model ensemble studies, based on insights from the Agricultural Model Intercomparison and Improvement Project. We find that crop model processes are the primary source of uncertainty in agricultural projections (over 50%), excluding unquantified hidden uncertainty that is not explicitly measured within the analyses. We propose multidimensional pathways to reduce uncertainty in climate change impact assessments.

气候和影响模型对于理解和量化气候变化对农业生产力的影响至关重要。多模型集成凸显了这些评估中相当大的不确定性，但缺乏量化这些不确定性的系统方法。基于农业模型比较和改进项目的见解，我们提出了一种标准化方法来归因多模型集成研究中的不确定性。我们发现作物模型过程是农业预测不确定性的主要来源（超过 50%），不包括分析中未明确测量的未量化的隐藏不确定性。我们提出了减少气候变化影响评估不确定性的多维途径。