Assessing and Improving the Numerical Solution of Atmospheric Physics in E3SM

The project addresses a crucial but largely overlooked source of error in Energy Exascale Earth System Model (E3SM) and similar models, namely the time-discretization and process-integration errors associated with parameterized subgrid-scale atmospheric physics. The objectives are to (1) understand the cause of recently revealed strong time-step sensitivity and poor time-step convergence in the E3SM atmosphere model, (2) develop new time-discretization and process-integration methods to improve convergence and accuracy in the physics parameterizations related to boundary-layer clouds, and (3) provide an assessment of the feasibility of using a stochastic perspective to develop time-integration methods for complex atmospheric processes.

An efficient testing method and a hierarchy of simplified model configurations are used to identify physical processes, numerical algorithms, and code pieces associated with unsatisfactory numerical properties. Theories of deterministic and stochastic partial differential equations are used to verify the hypothesis that the poor time-step convergence and strong time-step sensitivity in simulations from the E3SM atmosphere model are caused by insufficient representations of fast processes resulting from the combined use of long time-step sizes and crude numerical algorithms. New time-integration methods are developed and evaluated first in simplified model configurations and then in the full E3SM atmosphere model.

The key outcome of this project is a version of the E3SM atmosphere model with components that show substantially improved time-step convergence and reduced time-step sensitivity compared to E3SM version 1. Numerically robust and accurate solutions of the model equations can help to improve the model’s fidelity and trustworthiness. Proper convergence can provide a necessary basis for the development of accurate and efficient time-integration methods for E3SM at different spatial resolutions. The stochastic perspective might lead to a new paradigm in the treatment of fast processes in atmospheric models. 

Project Term: 
2017 to 2020
Project Type: 
Laboratory Funded Research

Publications:

None Available

Research Highlights:

None Available