Spatial Structure and Scale Interactions of Tropical Atmospheric Convection

In the tropical atmosphere, convection drives motion from the turbulent cloud–scale to the global overturning circulations. None of these scales of motion occur in isolation, instead their coexistence results in complex scale–interactions from which additional emergent dynamics and phenomena arise. Due to spatial resolution constraints, weather and climate models have historically struggled to represent convection and the associated scale–interactions, leading to errors in tropical forecasting and uncertainty in climate projections. This thesis aims to diagnose scale–interactions in modern, high resolution models, and to assess the impact of the representation of convection in each case, over a range of scales and modelling domains.

Spatio-temporal relationships are identified between convection and later convection at sub-daily scales, supporting a growing body of evidence that scale–interactions provide predictability on short timescales. The aggregated effect of some of the same mesoscale dynamics, convective cold–pools, then alter regional scale circulation. Model representations of convection, and domain size, also affect measures of rainfall and moisture variability. The observed correlation between convective clustering and net radiation is examined at the regional scale over Africa, and the mechanism for the correlation is shown to be more realistic in a km–scale model. Finally, using concepts from statistical mechanics, model errors in distributions of water vapour are explained in terms of statistical contributions to moisture tendency. This final analysis sheds light on the processes leading to the observed bimodal distribution of column vapour over the tropical oceans, and offers an explanation for why it is often poorly represented in models. All these results contribute novel understanding to the field of tropical scale-interactions, provide new diagnostics for future model evaluation, and support investment in both km–scale simulation capacity and continued parameterization development.