Tracking aggregates of deep convective systems in the tropics

In the tropics, the spatio-temporal aggregation of deep convective systems (DCS) has strong implications for the regional and global energy balance, for cloud-circulation interactions, and for the production of heavy precipitation on various scales. Yet, no ubiquitous definition or tracking algorithm exists for identifying cloud "aggregates", or cloud "clusters" in an objective way. Current cloud tracking tools are either developed for detecting individual cloud entities, or for identifying cloud clusters faithful to the original definition of mesoscale convective systems. They thus exclude a wide variety of DCS ensembles that organize in space and time without always merging into contiguous structures.

This work introduces a methodology to define and track coherent aggregates of deep convective systems using infrared brightness temperature (Tb) retrievals from geostationary satellites intercalibrated across the tropical band. The novelty lies in the combination between a spatiotemporal clustering and a region-growing method: starting with the coldest and largest cloud cores, we gradually attribute to preexisting aggregates the new neighboring systems that appear when the Tb threshold is increased to warmer values, until reaching the warmer outer edge of anvil clouds. Using a few criteria to measure organizational properties of the segmentation, we tune the algorithm parameters to maximize the realism of final cloud aggregates, minimize split-and-merge issues, and ensure that each aggregate is not intertwined with its neighbors whenever possible.

We demonstrate the good performance of this algorithm through a variety of realistic modes for deep convective organisation. The example case studies chosen include mesoscale convective complexes and cyclones over tropical oceans, cloud clusters embedded in African Easterly Waves, and the upscale merging of DCS in the course of the diurnal cycle of convection over tropical continents. A few composite properties of aggregates are shown for the entire tropics, to illustrate the potential of this dataset in providing new insights on the variety of organized convective patterns and on their associated multiscale interactions. The dynamics of individual systems can now be explored as an integral component of larger convective aggregates, across the diversity of cluster morphologies that populate the current tropics.