Representing coastal tropical convection in global climate models using stochastic models

Precipitation is a challenging meteorological variable in general circulation models, especially in coastal areas such as the Maritime Continent. Compared with observations modeled rainfall is usually too weak and too frequent. In coastal regions, the spatial pattern and timing of convection and rainfall becomes worse. Although advances in convection parameterization and model resolution have been made to tackle the problem of rainfall timing and intensity the main issues in simulating the diurnal precipitation cycle in the coastal tropics remain unsolved.

Classical cumulus convection parameterizations consists of a set of rules that translate the large-scale behavior into rainfall. Commonly these rules are either the same globally, or contain a land-ocean differences. Parametrizing coastal convection is complicated due to the complexity of land-sea interactions and the inability of most global models to resolve the coastlines. Therefore to our knowledge, there is no existing cumulus parametrization that allows for a different behavior near coastlines.

To tackle the problem of parameterizing coastal convection we propose a stochastic multicloud modelling framework rather than a deterministic scheme. The model operates on a much higher resolution of up to 1 km and to incorporates an newly developed nearest-neighborhood interaction scheme to represent local coastally driven interaction, such as land-sea-breeze circulations systems. When coupled to a simple convection scheme the stochastic model is able to reproduce the broad spatial and temporal organization that is well known for convection in tropical coastal regions.

Therefore we conjecture that the proposed stochastic multicloud model framework has potential to contribute to an improvement of the representation of coastal convection within numerical weather and climate models.