Changes in crop yields and their variability at different levels of global warming

An assessment of climate change impacts at different levels of global warming is crucial to inform the political discussion about mitigation targets, as well as for the economic evaluation of climate change impacts e.g. in economic models such as Integrated Assessment Models (IAMs) that internally only use global mean temperature change as indicator of climate change. There is already a well-established framework for the scalability of regional temperature and precipitation changes with global mean temperature change (∆GMT). It is less clear to what extent more complex, biological or physiological impacts such as crop yield changes can also be described in terms of ∆GMT; even though such impacts may often be more directly relevant for human livelihoods than changes in the physical climate. Here we show that crop yield projections can indeed be described in terms of ∆GMT to a large extent, allowing for a fast interpolation of crop yield changes to emission scenarios not originally covered by climate and crop model projections. We use an ensemble of global gridded crop model simulations for the four major staple crops to show that the scenario dependence is a minor component of the overall variance of projected yield changes at different levels of ∆GMT. In contrast, the variance is dominated by the spread across crop models. Varying CO2 concentrations are shown to explain only a minor component of the remaining crop yield variability at different levels of global warming. In addition, we show that the variability of crop yields is expected to increase with increasing warming in many world regions. We provide, for each crop model and climate model, patterns of mean yield changes that allow for a simplified description of yield changes under arbitrary pathways of global mean temperature and CO2 changes, without the need for additional climate and crop model simulations.