Exploring the usefulness of global warming levels for aligning agricultural productivity impact trajectories across GCMs

A key source of uncertainty in climate impact projections is the divergence of climatic variables across global climate models (GCMs) which are used to drive impact models. Here, we demonstrate the extent of this issue for agricultural impact modelling using the latest generation of GCM-driven global gridded crop models, and explore the usefulness of global warming levels (GWL) for aligning crop yield impact estimates across GCMs. To do this, we compare the spread in distributions of spatially aggregated yield change projections across GCMs using the GWL- and the commonly used fixed time window approaches. We find that at the global scale, the GWL approach is particularly effective in reducing GCM uncertainty in projections of interannual yield variability changes, and that this effect is robust across crops. In contrast, for changes in mean yields, the effectiveness of GWLs is strongly crop-dependent. These differences can be explained by different responses to increasing CO₂ concentrations across crops and yield metrics: a strong CO₂ fertilization effect on mean yields of the C3 crop wheat renders the GWL approach less effective, while the relative independence of both maize and wheat variability from CO₂ concentrations makes GWLs particularly effective in these cases. We find that in the agricultural modelling community, the GWL approach offers a means not only to align responses across GCMs but also to better understand impact drivers and components of uncertainty. The relevance of these findings also extends to the broader impact modelling community, particularly in settings where output from multiple climate models is used to drive impact models, such as studies based on the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) framework.