Can We Trust Model Projections of Changes in Climate Extremes over the Tropical Americas?

An outstanding issue in climate change research is whether regional precipitation changes associated with global warming are governed mainly by a general increase of atmospheric humidity (“thermodynamic control”) or by changes in the statistics of large-scale tropospheric vertical velocity (“dynamic control”). Thermodynamic control is stronger and more robust than dynamic control in current climate models. Our investigation of observed precipitation changes over the last 70 years, however, tells a different story. In most regions around the globe, including the tropical Americas, the observed changes in both the mean and extreme daily precipitation are much more consistent with similar changes in vertical velocity, i.e. with dynamic control. Most tellingly, and with important implications for changes in drought, the observed changes in the probability of dry days are remarkably consistent with changes in the probability of days of tropospheric descent. This aspect of the precipitation statistics is consistent with dynamic control, but is totally beyond thermodynamic control.

We find that uncoupled atmospheric GCM simulations of the last 70 years with prescribed observed time-varying SSTs and radiative forcing are reasonably able to capture the observed changes in precipitation and vertical velocity, and hence the dominance of dynamic control. On the other hand, fully coupled climate models, which predict instead of prescribing the SSTs, do not correctly capture these changes, suggesting that the problem originates in their misrepresentation of the SSTs, especially tropical SSTs. Specifically, the current generation of climate models underestimate the changes in tropical SST gradients, and hence the magnitude of changes in the mean atmospheric circulation and storminess, leading to an underestimation of the dynamic control of precipitation that in reality is actually stronger than the thermodynamic control. This underestimation of the circulation and storminess changes also compromises climate model projections of changes in hot temperature extremes, which in reality have been stronger over the tropical Americas than implied merely by the mean shift of the temperature probability distribution towards warmer temperatures.