Significant trend in vertical wind velocity variability revealed by machine learning

Vertical air motion plays a critical role in cloud formation, mixing, and transport processes. Due to their limited spatial resolution, global atmospheric models cannot directly capture the sub-kilometer scale components of this motion. Instead, these fine-scale motions are represented by the standard deviation of vertical wind velocity within a grid cell, denoted as Sw. Using a novel deep learning model constrained by observational data applied to long-term climate reanalyses, we have identified significant trends in Sw from 1980 to 2023. These trends, which show an increase of up to 1% per year in low and mid-level oceanic regions, point to a more turbulent atmosphere and enhanced cloud formation over the past few decades. Using explainable machine learning techniques we attribute these changes to global shifts in water vapor, temperature, and convection patterns, indicating a linkage between increased warming, turbulence, and cloud microphysics. Our analysis suggests that over the industrial era, the enhanced cloud formation resulting from increased Sw contributes to a radiative forcing of approximately −0.1 ± 0.21 W m−2, which slightly offsets the effects of greenhouse warming.