Generate data for multiple CIMH domains and cases

fossell commented 2 years ago

This is a sub-issue of #40

Describe the Sub-Issue

Generate data for multiple CIMH domains and cases to be archived for later evaluation and verification. Cheyenne resources available to use before Oct 1.

Experiments to run:

4 domains
5th domain that'a union of all 4 (if time/resources allow)
For each domain, 1-2 high impact events and 1 week (or more, if time/resources allow) period of 00z initializations

@michelleharrold will initiate the task to get things set up and ready to run. Will assign specific tasks at future meeting.

Time Estimate

1 month. Resources expire Sep 30, 2022

Define the Metadata

Assignee

[ ] Select engineer(s) or no engineer required
[x] Select scientist(s) or no scientist required

Labels

[ ] Select component(s)
[x] Select priority

Projects and Milestone

[x] Select Project
[x] Select Milestone as the next official version or Backlog of Development Ideas

fossell commented 2 years ago

Current plan is to run 00Z and 12Z initializations for August 2021 - early/mid-September 2021 to capture a longer period and several hurricane cases within that time period.

fossell commented 2 years ago

Forecast length? As long as possible with given resources? Perhaps run a forecast for each domain to get some time/resource estimates and we can go from there.

michelleharrold commented 2 years ago

After some benchmarking, to run 120-h (5-day) forecasts for the 12-km full Caribbean and 4-km W. Caribbean, S. Caribbean, and E. Caribbean, it would take a little under 200K core hours for 70 inits (00 and 12 UTC from Aug. 1 - Sept. 4, 2021). I also ran 72-h (3-day) and 48-h (2-day) -- those would be a little over 100K and 70K, respectively.

NCAR / wrfcloud