OP-E8.1 SciDAC convergence project overview

                    

Poster TitleTowards numerically robust representation of sub-grid processes: improving solution accuracy and convergence for physics parameterizations
AuthorsHui WanCarol Woodward (Unlicensed)Michael BrunkeDavid Gardner (Unlicensed)Vince Larson, Huan Lei (Unlicensed), Jing Li (Unlicensed)Phil Rasch (pnl.gov)Balwinder SinghJeremy Sousa (Unlicensed)Panos Stinis (Unlicensed)Chris Vogl (Unlicensed), Xubin ZengShixuan Zhang 
First AuthorHui Wan
Session TypeE3SM session
Session IDE8
Submission TypePresentation
GroupAtmosphere
Experiment
Poster Link




Abstract

The development of atmospheric physics parameterizations traditionally focuses on the conceptualization of physical understanding and the handling of spatial scales. When it comes to the choices of time integration methods, the main considerations are usually the requirement of numerical stability and the limitations in computational resources. This presentation will introduce a SciDAC project that attempts to develop systematic methods to evaluate and improve the accuracy of time integration in E3SM's atmospheric physics parameterizations.

We will show that numerical artifacts are significant in E3SM's climate simulations. We will also demonstrate that the investigation into cases with poor time-step convergence can help identify code bugs, reveal numerical methods that are inconsistent with physical concepts of a parameterization, and indicate vulnerabilities in the model formulation. Issues that affect solution convergence can have substantial impact on features of the simulated climate. Theories and methods for solving deterministic partial differential equations and approaches for stochastic modeling are explored.

The ultimate goal of the efforts is to increase the numerical robustness of the climate simulations and to ensure that the numerical representation of the physical system is not contaminated by numerical artifacts. While our project focuses on E3SM, similar numerical issues are likely to exist in other models as well. Insights from this project are expected to be useful for the wider community of weather, climate, and Earth system modeling.