Episode 4 Running a recipe (First example):

[ledm]

Internal name: first_recipe

The outline document says:

• Questions:

  • What is a recipe?
  • How can I do the same preprocessing on many different datasets?
  • What are the files and directories after running a recipe?
  • What happens when I run a recipe?

• Objectives:

  • Run an ESMValTool recipe
  • Understand the purpose of different settings in the recipe
  • Inspect the output directories
  • Examine the log information

• Contents:

  • How to run ESMValTool
  • esmvaltool -c user-config.yml recipe.yml
  • A basic recipe that takes a simple dataset and produces a simple plot.
  • We’ve found that the best example here is a scalar (time series) field, like the thetaoga (global average ocean temperature). This data is small and light, (hundreds of kb per file).

• Discussion:

  • Inspect the example recipe
  • What do you think it will produce?
  • Exercise: Run the example recipe to produce a simple time series of ~20 years of data.

• Discussion: Inspect the output:

  • Plots
  • Logs
  • Preprocessed netcdf
  • Provenance and citation information
  • Settings.yml and metadata/yml files

• Exercise:

  • edit the file to include other models or a different variable and re-run.

• Exercise:

  • edit the recipe to run:
  • Atmosphere: Temperature at 2m
  • Land surface: Average land surface tempureate in your local region.
  • Ocean: Surface ocean temperature

• Callout: Common issues & tips

  • ESMValTool can’t locate the data (user didn’t correctly edit user-config.yml)
  • Esmvaltool not found (user didn’t active or correctly install conda)
  • Diagnostic path problems (explain ESMValTool’s methods to determine diagnostic path, and how it should appear in the recipe)
  • FX files not found.
  • The preprocessor works but the diagnostic fails. How to tell them appart and how to re-run a failed diagnostic (but not the preprocessor)
  • Your recipe’s name/project/reference isn’t recognised by ESMValTool.

• Key points:

  • You can’t break A a recipe does not break by fiddling with it
  • Log information is useful to How to interpret the first warnings/errors
  • The dataset section in the recipe relates to Understanding the directory structure of CMIP data on your server/disk and knowing how to use data from different experiments such as CMIP/ScenarioMIP.

[ledm]

We're working in a google doc. When we're happy with that content, we'll copy it into the respository.