This is not a cloud model: A new take on data-driven, stochastic modeling of organized cumulus fields using GOES-16 high-resolution data

Abstract:

The emergence of organized patterns resulting from convection is ubiquitous, observed throughout different cloud types around the world, across a wide range of scales. The reproduction of such patterns by cloud-resolving models (CRMs) or by large eddy simulation (LES) models remains a grand challenge. The new advances in data-driven modeling techniques over the last decade opens up prodigious perspectives for cloud modeling that are yet in their very infancy.

This talk will address such a data-driven modeling problem in the context of organized, mesoscale continental shallow cumulus, from high-resolution satellite datasets collected from GOES-16, in the course of the day. By relying on the state-of-the art of data-driven stochastic modeling techniques from partial observations of multi-scale complex systems, low-dimensional models of such datasets will be presented. The key to success relies on the efficient learning of hidden stochastic variables able to emulate the lacking, unobserved variables which are central in the organization of the observed larger scale patterns. As will be discussed, the approach allows not only for reproducing the cloud field’s key organizational features, such as cloud streets and convective inertia gravity waves, but also allow for simulating a wide range of out-of-sample days, presenting their own variability while respecting the patterns’ physical characteristics. This work is partially supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program [Grant Agreement No. 810370].