The general workflow for running an experiment is as follows: 0. Bias correct model data using FUDGE 1. Create config file (pick suitable cshell_script) 2. submit cshell_script to queue from analysis node 3. Rscript steps a. Interpret config file b. Copy input files to local temp directory c. Copy/create output files with appropriate metadata in temp directory d. Sub-sample (eventually also spatially restrict) input files e. Establish and start local R-based cluster f. Construct analogues g. Stop local R-based cluster h. Copy outputs to final storage (/archive?) i. Clean up local temp directory

Steps are discussed in more detail below.

1. Create a config file We're currently writing our config files in yaml. Right now, there is a recent config file under sample_config.yaml; we will try to update it as changes to the workflow warrant.

2. cshell_script There was a step that used to make (and optionally submits) a c-shell script suitable for running in the batch workflow. It also set up a scripts directory for keeping track of the script and the logfiles from batch submission. You invoked it like so:

    python make_cshell_workflow.py -i sample_config.yaml
   

   Now batch submission is more generic and should be handled concurrently with creating the config.

3. Constructed Analogue R code CA_runner.R details the overall steps of the R code in a function run_experiment()