Hi-
I'm tasked with creating a training dataset by running 54-hour forecasts starting 24 hours apart through the summer months of many years. I have discovered I can use the SRW config.sh params DATE_FIRST_CYCL and DATE_LAST_CYCLE to run multiple cycles with one workflow. Will this create output identical to that from running a workflow for each day of my period of interest? I have been using this latter approach (one cycle and one workflow per day) and it's been causing resource problems with the cron host on cheyenne. So I am looking for an alternative.
For example, I want to run 54-hour forecasts for each day of July, 2022. I have a script that creates a config.sh for each of the 31 days in July and runs ./generate_FV3LAM_workflow from regional_workflow/ush. In each of these configs, DATE_FIRST_CYCL is the same as DATE_LAST_CYCLE, so the workflow runs one cycle. My question is whether I could accomplish exactly the same output with a single workflow configured with:
DATE_FIRST_CYCL="20220701"
DATE_LAST_CYCL="20220731"
I see that when I use this multiple-cycle mechanism output is generated in a separate, dated, directory for each cycle under the expt_dir for my workflow. But does anyone have a sense if there is a functional difference between these two ways of running the model? Is initialization done differently, or anything else under the hood that I should be aware of?
Thank you in advance, and please let me know if you need more information about what I am trying to do,
-Paddy McCarthy
- 10 views