Attributing Long-Term Trends in Marine Low Cloud Morphologies to Aerosols and Large-Scale Meteorology With Deep Learning

The response of marine low-cloud mesoscale morphologies to climate change and emission reductions remains poorly understood. Here, we link long-term trends in six cloud morphologies to variations in large-scale meteorology and aerosols. The trends show strong spatial heterogeneity, with closed and disorganized mesoscale cellular convection decreasing in the Northeast Pacific and Southeast Atlantic. We develop a deep learning model (UMorNet) to predict instantaneous cloud morphologies from meteorology and cloud droplet number concentration (N_d), a proxy for aerosols. UMorNet achieves an average test accuracy of 0.55 and captures spatial patterns of climatology and long-term trends. Out-of-sample test with a marine heatwave event further demonstrates the model's performance. Sensitivity experiments identify N_d, marine cold-air outbreak index, sea surface temperature, and inversion strength as key drivers. Different responses of clustered Cu and suppressed Cu to N_d was identified. These findings highlight the potential role of aerosols in shaping cloud morphological changes.