Internal variability, multiple equilibria, and convection-SST coupling in a cloud-resolving model with an interactive ocean

Future changes in tropical convection will be closely tied to changes in the underlying sea surface temperature (SST) pattern. To understand the convective response to warming in a coupled atmosphere-ocean system, we perform a series of idealized, 20-year radiative-convective equilibrium experiments with a 2D cloud-resolving model coupled to a 25-m slab ocean. The domain length is that of the tropical Pacific basin, and different climates are achieved by varying the parameterized ocean heat transport (q-flux). The simulations are characterized by two distinct regimes of  convection-SST coupling: an oscillatory regime that occurs when the mean SST is near that of the present-day tropical Pacific (27-30 °C), and a non-oscillatory regime at warmer temperatures (>36 °C).

The oscillatory regime is defined by internal, 3°C oscillations in mean SST driven by variations in low cloudiness. During the warming phase of the cycle, SSTs are homogeneous, deep convection occurs in two regions, and low clouds are sparse. During the cooling phase, there are well-defined warm and cold pools, deep convection aggregates into a single region, and expansive low cloud decks act to decrease the mean SST.  

In the warmer, non-oscillating regime, distinct warm and cold pools still form, but convection is no longer limited to the warmest SSTs. Rather, convection develops over cooler SSTs and is then advected to the warm pool by the mean flow. The expansion of deep convection to cooler SSTs impedes low cloud formation over the cold pool and inhibits the low cloud-driven oscillations in mean SST. Changes in sub-cloud buoyancy explain the expansion of the convectively unstable region.

Both regimes (oscillatory and non-oscillatory) can be achieved for the same q-flux depending on initial conditions. Intermediate SSTs (30-36 °C) are unstable on long timescales and eventually revert to one regime or the other. While certain aspects of this behavior are likely sensitive to simulation design, our broader set of experiments suggests potential shifts in convection-SST coupling as the climate warms.