Meteorological conditions associated with the shallow mesoscale clouds in the southern Atlantic

Shallow oceanic clouds strongly influence the global energy balance by cooling the planet and mediating exchanges of heat and moisture between the ocean and the atmosphere. A key challenge arises from the fact that these shallow clouds frequently organize into a range of spatial structures that are unresolved in global climate models and are heavily parameterized based on the meteorological conditions. Understanding the coupling of these clouds to meteorology is therefore essential to improve their representation in the models and for reducing uncertainties in future climate projections related to their feedbacks.

Using one year of SEVIRI/MSG data at 15-min temporal resolution, this study first explores the potential of deep machine learning (ML) to detect and classify mesoscale low-level cloud patterns. Using a supervised convolutional neural network, the shallow clouds are then classified into the dominant spatial patterns. The associated cloud properties and underlying meteorological conditions are further analysed using joint histograms based on the CM SAF CLAAS3 cloud climate data record and ERA5 reanalysis datasets to investigate if such information can be useful to evaluate and to better represent these clouds in global climate models.