We present crop model simulations results from the AgMIP GGCMI / ISIMIP Phase 3 multi-model ensemble across global gridded crop models, CMIP6 bias-adjusted climate models, SSP-RCPs, and time slices in order to understand mean changes and the volatility of regional yields over the 21^st century. In particular, we will focus on identifying:

• Linear and non-linear responses in comparison to global warming levels of important to policymakers considering greenhouse gas emissions. Non-linear characteristics will then be examined to explore connections between yields and non-linear hazards or the emergence of different or compound prevalent hazards over the course of the century.
• Broad patterns or tipping-point behaviors potentially related to biophysical hazard thresholds. These include limiting temperatures for a given crop species, such as temperatures exceeding 35℃, or periods with more extensive soil moisture drought.
• Regions where the coefficient of variation in annual yields emerges as a challenge to agricultural production even as mean yield changes are not dramatically different. These agricultural systems will have hidden costs on farmers and markets that are not typically captured in integrated assessment models or other applications utilizing only mean yield changes, and increase the potential for multi-breadbasket failures and disasters leading to systemic impacts.

Together, these results will shed light on the importance of the temporal evolution and consistency of crop yields beyond the most commonly examined mean yield changes for a given time slice and scenario.