E3SM Leadership Meeting Focused on Current and Future Issues
E3SM Leadership Team (E3SM Council, Group Leads and their deputies, NGD Leads, and the Program Manager) held a very productive, virtual three-day E3SM Leadership Meeting* on April 28-30, 2021 to discuss the current status of the project and future plans with a particular focus on project issues. Each group gave an update on their current status and future plans, noting the work towards E3SM version 2 (v2) code and plans for v2 simulations, as well as the issues and lessons learned in phase 2 and possible solutions and changes. The Water Cycle Group (as the team charged with deciding on and validating the composition of features in the final v2 code, as well as tuning the v2 model configuration) discussed the issues they encountered, such as new features that were not tested enough in the coupled system or had performance issues or were incomplete. Another problem discussed was that feature developers seem to wait for other features to be developed, wanting their feature to be the last one added so their feature would be tested in the code that has all the other features in it, but this creates a positive feedback loop that causes delays. Another issue mentioned was a lack of visibility in team member’s availability leading to difficulties in re-assigning work, which points to the need for a critical look at the phase 2 project’s structure and the possible need to restructure the project in phase 3.
The Leadership Team devoted substantial time at the meeting to discuss these issues, such as revisiting the code review process. The team decided on the scope of a Code Review Deep Dive* that will convene in the near future to dissect problems encountered, to look into possible solutions, and present a recommendation for changes to the current process. To help with the standardization of the processes and documentation, the team discussed proposed standards for the v2 simulation campaign*, including naming conventions, how to run the model, postprocessing, and documentation conventions. For the v2 simulation campaign, the project will enforce a procedure using the simplified run_e3sm script, utilizing the short-term archiving feature, post-processing with a new tool called zppy, implementing long-term archiving with zstash, and documenting each run on a standardized Confluence page. The policies for the model output and their publication to ESGF were also discussed. Taking into account and supporting the requests from outside projects funded by the BER Earth System Model Development (ESMD) program area, the team decided to publish to ESGF the most popular variables in standard CMIP format as single-variable time series on the standard grid for all target v2 simulations, in addition to the publication of all output in a native file format on the native component grid.
The team reviewed the vision for the v3/v4 science goals, model configurations, the detailed v3 and phase 3 Timeline*, and a timeline and action items for the phase 3 proposal development.
Other subjects discussed at the meeting included coordination with external projects, specifically their contributions to the E3SM code, new model configurations and use cases, infrastructure and performance needs regarding large ensembles, a carbon mass conservation deep dive, the INCITE proposal for 2022 and beyond, and possible Exascale Computing Project (ECP) funding opportunities.
All agreed that the meeting was very fruitful and that we should have such meetings more frequently.
—
* Links are internal