Earth System Models (ESMs), including the Energy Exascale Earth System Model (E3SM), generally struggle to represent low-level warm clouds realistically. Studies have shown considerable sensitivity of warm cloud properties to changes in microphysics scheme parameters affecting, for example, the rate of precipitation formation. This can alter key climate features such as aerosol-cloud interactions, precipitation characteristics, and climate sensitivity. Increasing resolution improves some aspects of cloud properties, but significant biases persist as resolution increases in E3SM. This suggests structural deficiencies in representing clouds in E3SM. Nonetheless, significant advances have been made recently in process-level understanding and simulation of cloud microphysics. A major advance has been the development of Lagrangian particle-based microphysics that tracks trajectories of individual “super-droplets” in the modeled flow representing a multitude of real cloud droplets (called the “super droplet” method, or SDM). SDM offers several key advantages compared to other detailed microphysical models including bin microphysics, particularly by eliminating numerical diffusion and improving representation of cloud-aerosol interactions. Though it is not feasible to use SDM directly in E3SM, it can be used to improve the bulk microphysics in E3SM. This idea is a key part of the proposed work: using SDM in high-resolution large eddy simulations (LES) of a variety of cloud regimes to generate an expansive dataset for improving the treatment of microphysics in E3SM.

A challenge is how to incorporate these advances in process-level knowledge in a computationally efficient and accurate way, which may require a re-thinking in the design of bulk microphysics schemes. To address this challenge, the team will replace the traditional approach of separate cloud and rain categories with a single liquid category (SLC) approach. Recent work has shown that the SLC approach can substantially improve the representation of rain generation through drop collision-coalescence compared to the traditional two-category approach. The SLC approach replaces separate processes for autoconversion, accretion, and self-collection in the two-category approach with a single collision-coalescence process, modeled by predicting at least four size distribution moments for the single category. The team will train SLC and obtain process rates from the SDM-LES dataset via machine learning (neural network emulation) and Bayesian inference. The SLC approach will be implemented into the P3 microphysics scheme, which is the scheme expected to be used in the E3SM version 3 and 4 releases. The team will test P3-SLC in single-column and global low-resolution tests using E3SM as well as the high-resolution doubly-periodic configuration called SCREAM. A focus through all stages of testing will be on how SLC alters the representation of cloud-aerosol interactions compared to the standard P3 scheme in E3SM. They will also pay particular attention to the coupling with the “cloud macrophysics” to account for subgrid-scale cloud variability, which is especially important for lower resolution applications. The proposed research directly addresses the ESMD topic “i) low level clouds” by using recent advances in detailed process-level modeling of cloud microphysics to improve the treatment of microphysics in E3SM, targeting low liquid clouds. Physically-based, process-level improvements to the microphysics are expected to bring greater fidelity to model simulations, addressing critical uncertainties in representing low clouds in E3SM including cloud-aerosol interactions and cloud-climate feedback.