Multi-scale modeling of Arctic clouds

Description

The presence and properties of clouds are critically important to the radiative budget in the Arctic, but clouds are notoriously difficult to represent in global climate models (GCMs). The challenge stems partly from a disconnect in the scales at which these models are formulated and the scale of the physical processes important to the formation of clouds (e.g., convection and turbulence). Because of this, these processes are parameterized in large-scale models. Over the past decades, new approaches have been explored in which a cloud system resolving model (CSRM), or in the extreme a large eddy simulation (LES) is embedded into each grid-cell of a traditional model to explicitly simulate more of these important processes. This approach is attractive in that it allows for more explicit simulation of small-scale processes while also allowing for interaction between the small and large-scale processes. This framework has been implemented into the new Energy Exascale Earth System Model (E3SM) using the System for Atmospheric Modeling (SAM) as the embedded CSRM.​The goal of this study is to quantify the performance of this framework in simulating Arctic clouds relative to a traditional global model, and to explore the limitations of such a framework using stand-alone LES and CSRM simulations with the embedded model. Simulations from the global model are compared ​both with satellite retrievals of cloud fraction partitioned by cloud phase from CALIPSO, and with point observations taken from the ARM Barrow and Oliktok Point measurement facilities. Limited-area CSRM and LES simulations are performed and compared with ground-based and tethered-balloon measurements from the ARM Barrow and Oliktok Point measurement facilities to better understand the limitations of the embedded high-resolution model in simulating Arctic clouds under ideal conditions. A suite of simulations from CSRM to LES scales are performed with various resolutions and configurations to understand how the resolution and configuration of the embedded model affects simulation fidelity and to separate model errors that arise from setup of the embedded model from those that arise from coupling between the global and embedded models.