Abstract

Marine atmospheric boundary layer (MABL) clouds cover vast areas over the ocean and have important radiative effects on the Earth's climate system. These radiative effects are known to be sensitive to the local organization, or structure, of the mesoscale cellular convection (MCC). A convolutional neural network model is used to identify the two idealized classes of MCC clouds, namely open and closed, over the Southern Ocean (SO) and Northwest Pacific (NP) from high-frequency geostationary Himawari-8 satellite observations. The results of the climatology show that MCC clouds are evenly distributed over the mid-latitude storm tracks for both hemispheres, with peaks poleward of the 40latitude. Open-MCC clouds are more prevalent than closed MCC in both regions. An examination of the presumed meteorological forcing associated with open- and closed-MCC clouds is conducted to illustrate the influence of large-scale meteorological conditions. We establish the importance of the Kuroshio western boundary current in the spatial coverage of open and closed MCC across the NP, presumably through the supply of strong heat and moisture fluxes during marine cold-air outbreaks events. In regions where static stability is higher, we observe a more frequent occurrence of closed MCCs. This behavior contrasts markedly with that of open MCCs, whose formation and persistence are significantly influenced by the difference in temperature between the air and the sea surface. The occurrence frequency of closed MCC over the SO exhibits a significant diurnal cycle, while the diurnal cycle of closed MCC over the NP is less noticeable.