Cloud System Resolving Modelling of the Tropical Atmosphere

The tropics are often described as the engine room of the Earth's climate system, powering the global circulations of the atmosphere and oceans. Absorption of sunlight heats the land and ocean surfaces strongly at low latitudes, producing convection that carries the energy from the surface into the atmosphere. Except over the deserts, the convection generates deep clouds that transport moisture evaporated from the ocean into the upper atmosphere. When these clouds rain, the release of 'latent heat' by condensing water produces further heating that drives the weather systems of the tropics and influences the winds all around the globe. But such deep convective clouds rarely exist in isolation; they are almost always organised into structures ranging from squall lines and cloud clusters to tropical storms, hurricanes and super-clusters; and convection varies on a wide range of timescales from that of an individual cloud element (hours), through the daily cycle to a plethora of waves with periods ranging up to the intra-seasonal oscillation, which can propagate around the world in 30-60 days. The tropical atmosphere thus organises itself on a huge range of space and timescales; the effect on the climate system is very different from that of random, or disorganised turbulence so that all these scales should be represented in a computer model if it is to reproduce the real world accurately. Until now, many of these effects have had to be represented in a highly simplified way, through a process known as parametrization. This tries to mimic the effect of the convection on the scales smaller than those represented explicitly in the model. The trouble is that across the cloud system from a few to a few hundreds of kilometres there is no preferred scale, and these scales span the range from those that are 'sub-grid' and therefore need parametrization to those which are resolved by the model. Even with the availability of modern supercomputers, this transition from parametrization to resolved motion takes place at around 100km in the latest climate models. All of the structures that occur in the real world below this scale are parametrized, so a completely artificial break occurs between what is resolved and what is parametrized. To compound the problem, the parametrization assumes that organization on scales which cannot be resolved is not important for determining the properties of the convection or its influence on the large-scale flow. But there is mounting evidence that this creates all sorts of problems in the models. Several decades have been invested in developing convection parametrizations and even after all this time and effort no fully satisfactory solution has been found. We therefore propose to take a radical but entirely logical approach to this problem. It is now possible to run models that do resolve convective systems explicitly (at least down to scales of around 1km) over very large domains that encompass all of the important scales mentioned above. Such cloud system resolving models provide a new tool for understanding how convection really works and organises itself, and how it should be parametrized in climate models. Our proposal links these models with new data from satellites and from the surface that will give us an unprecedented view of the evolution of clouds and rain-producing systems. We will bring this unique combination of modelling and observations to bear on what is regarded as one of the most fundamental problems in weather and climate. The results of the work will inform the development of a new generation of more accurate atmospheric models that will find employment in both climate prediction and weather forecasting.