Changes in high frequency internal variability alter extreme heat stress projections

Heat stress will increase as the global temperature rises, causing a wide range of harmful effects from reduced labor capacity to increases in morbidity and mortality from heat-related illnesses. By necessity, projections of future heat stress are generally based only on the change in the background state, implicitly assuming fixed high frequency internal variability (IV). Whether IV changes under anthropogenic climate forcing, and how those changes alter projections of heat stress or other extreme events, has yet to be explicitly tested.

We characterize heat stress using daily maximum Wet Bulb Globe Temperature (WBGT) output from the 100-member CESM2 Large Ensemble simulation (SSP3-7.0), calculated from 3-hourly model output and bias corrected using quantile delta mapping. The large ensemble allows assessing heat stress via a "snapshot" approach, counting days per year above WBGT thresholds, with statistics calculated in the ensemble dimension. The snapshot approach provides a simple and robust measure of the forced changes in climate variability (ensemble mean), and the remaining IV. We assess the effect of changing IV by comparing the full ensemble heat stress to the counterfactual case assuming fixed IV at the current warming level.

Future IV perturbations cause regionally coherent changes in the projections of heat stress. In areas where heat stress is "emerging", up to 100% of events are caused (or prevented) by changes in IV. Within regions experiencing stronger heat stress, up to ±50% of total heat stress is due to IV perturbations, equivalent to a shift in the background global warming by 0.5°C. Most often the regional IV changes are equivalent to 0.2-0.4°C, and the impact of changing IV is usually consistent across a range of warming levels. In general, the lower the probability of a heat stress event occurring, the more that changes in IV alter the projected number of events. Changes in IV are thus non-negligible when projecting heat stress, and are critical for extreme (i.e. very rare) event projections. In practice, this means an ensemble of climate models may be required to obtain reliable results.