Abstract
Large multi-lab projects on climate model development involves a collaboration of technical staff with very different backgrounds and expertise. For climate science, we are interested in the accuracy of dynamics and physics in the atmosphere; however, most of the automated model validation used during model development only covers the averaged climate behavior presented in diagnostic packages.
During the post-analysis of a version of CESM output from “high-res” project, we found the energetics and the atmospheric variability, including the wave activity and some key atmospheric phenomena, are wrong. After an in depth examination of the model output and the model itself, we found it to be affiliated with incorrect subcycling (time-stepping bug) of the dynamics in the atmosphere. It indicates a serious issue in the atmosphere coupler, which is responsible for the energy conversion between dynamical and physical processes. This issue, due to its nonlinearities, contaminated most of the dynamical processes in the model’s atmosphere.
We highlight this case here to emphasize that the atmospheric dynamics, and thus the associated shorter timescale variability, are the most fundamental process to the climate model. However, because these simulations are expensive, these aspects are often not not be analyzed until most of the allocation is spent for a given run. Therefore, establishing more significant and quantitative early testing of the high frequency atmospheric variability is important, especially during the developing stage of the model.