Connecting tropical cloud dynamics to global climate sensitivity

The spread in model-estimated equilibrium climate sensitivity arises largely from differences in the feedback from low clouds, for reasons not yet understood. Here we show that differences in the simulated strength of convective mixing between the lower and middle tropical troposphere (here denoted LT-mixing) explain about half of the variance in climate sensitivity estimated by 43 climate models. The apparent mechanism is that LT-mixing dehydrates the low-cloud layer at a rate that increases as climate warms, and this rate of increase depends on the initial mixing strength, linking LT-mixing to cloud feedback in a mechanism we dub the MILC (Mixing-Induced Low Cloud) feedback. This feedback mechanism explains both the spread of low-cloud feedback and its tendency to be positive in nearly all models.

The mixing behind this feedback occurs both within the convective schemes in models and explicitly; most of the explicit mixing is associated with steady ascent and shallow divergence in the Pacific and Atlantic ITCZ, and transient midlevel outflows in extratropical cyclones. We investigate, using observations and cloud-resolving models, the conditions that permit persistent LT-mixing, and explore the reasons why in GCMs this mixing fails to slow down in a warmer and moister atmosphere. This somewhat puzzling persistence causes LT-mixing to transport significantly more water vapor, and exert the MILC feedback.