UCLALES-based cloud base updraft emulator for global models

Climate models cannot afford the computational cost of the meter-scale resolution needed to accurately resolve turbulence and convection in the boundary layer. Machine learning based Gaussian process emulators (GPEs) have been recently presented as an alternative to close the gap between meter-scale and kilometer-scale resolutions (Ahola et al., 2022, https://doi.org/10.5194/acp-22-4523-2022). An emulator offers an improved alternative for climate models to include turbulence effects in the boundary layer on the formation of stratocumulus clouds. Here we have trained a GPE using vertical winds from the large-eddy model UCLALES following the approach of Ahola et al. (2022). The training data of our updraft emulator includes a wide range of stratocumulus conditions both over land and sea. The predicted standard deviation of cloud base vertical wind can be used directly in the activation calculation of global models. We have additionally implemented our emulator to the OpenIFS global climate model. In this study, we present a comparison of different parametrizations for updraft velocities, including our emulator, and how these affect cloud droplet number concentration and aerosol radiative forcing in the global scale.

This project has received funding from Horizon Europe programme under Grant Agreement No 101137680 via project CERTAINTY (Cloud-aERosol inTeractions & their impActs IN The earth sYstem).