Seminar by Jan Haerter

Convective self-aggregation is a heavily-studied emergent phenomenon in the atmosphere over the tropical ocean. There, the lower boundary condition (sea surface temperature) can approximately be considered constant and oscillations in solar irradiation can be neglected. Over extended periods (weeks), buoyancy-driven heat transport from the lower to the upper boundary can hence reach a stationary state. Under such conditions, clouds have been found to organize in numerical models, often leading to one or several large patches (~1000 km in diameter) of clouds (each cloud is only few km in diameter). In the past years, the imbalance of radiative effects in the cloudy vs. cloud-free regions has been made responsible for this "aggregation". 

A typical feature of precipitating clouds, namely the evaporation (and subsequent cooling) of a fraction of their precipitation on the way to the ground, has been found to hamper such aggregation. With evaporation of rain, clouds are "shut down" because they remove their own buoyancy. We show that, from a very simple model, the formation of such "cold pools" under clouds can indeed be reconciled with clustering, or even aggregation.

The model is very simple: a unit domain is seeded by any number of circle origins at random, which then grow in radius. Where two circles overlap they are deactivated. When three overlap, they form a new circle, and so forth. The model gives non-trivial effects of clustering, intermittancy, and highlights the aspect of clouds coming in generations.

We propose that variants of the model could potentially explain aspects of large scale clustering over the Indian and Pacific Ocean, such as the Madden Julian Oscillation.