The faithful representation of tropical convection in weather and climate models remains one of most difficult tasks in atmospheric science. This is so, because the complex interactions of small, meso and large-scale processes that occur in convection need to be parametrised in those models. This requires the behaviour of convective, and hence unresolved, scales to be inferred from the evolution of the larger, resolved, model scales.
Since the 1990s, the operational and research radar network around Darwin has been instrumental in significantly improving our understanding of the scale interactions in tropical convection. The use of the data produced by these systems has accelerated over the past decade and has led to the generation of several unique long-term data sets for atmospheric convection research. In this lecture, we will summarise some key findings of recent research and demonstrate their utility in inspiring entirely new approaches to the parametrisation of convection in climate models.
We first show how innovative retrieval and analysis techniques have opened new avenues to understand the behaviour of tropical convection. We show that convective heating in an area is entirely dominated by the fraction of the area that experiences convection. Following in the footsteps of field studies in the 70’s and 80’s we identify archetypal convective states and study the relationship to their environment. We show that the most intense convection is associated with large clouds that occur in a dry and often descending atmosphere, while the largest area-average rain results from a moderate number of moderate size clouds embedded in a humid and ascending atmosphere. Combining cloud-structure information with estimates of vertical velocity, enables us for the first time to estimate convective mass-fluxes over long periods of time and large areas.
We use the inspiration drawn from the observations to design a new approach to cumulus parametrisation. We implement this approach in two climate models and demonstrate improvements in the convective behaviour in both. In doing so we connect detailed process research based on observations with the models used in weather climate prediction, providing strong evidence that only large, sustained and collaborative research efforts can tackle the hard and important problems remaining in weather and climate science.