2229 scores rmse vertical profile pressure levels add ensemble capability#2252
Conversation
…otting of vertical profiles without sequence_coordinate for 1D cubes.
…tial field and preserve time as sequence coordinate to slide over. This required an addition to scoreswrappers.py that maps the xarray time dimensions with iris auxilliary time dimension requirements. The scoreswrapper was developed with help of Github copilot.
…or rmse_scores_profile when aggregating over an entire case study's time coordinate resulting in a single rmse profile.
…teps in a case study and in additional rmse profile for evevry single time step.
… (sequence) and vertical profile RMSE over whole case. In addition identifying base_model at start of loader.
Sylvia Bohnenstengel (Sylviabohnenstengel)
linked an issue
…rity for basis of RMSE calculation for profiles.
…cipes_no_variables - IndexError: list index out of range. Solution suggested by Github copilot.
…ation info into title.
…emble mean. Adding information about ensemble mean into title when producing rmse from ensemble mean. Used GitHub Copilot to code.
1a8d04e to
b443918
Compare
…evels_add_ensemble_capability
|
adding ensemble capability breaks the pytests. Specifically it gives It seams that collapsing the ensemble down to a mean causes the problem. I end up with 2 cubes marked as cset_comparison_base=1 as an attribute as the collapsed cube that gives the ensemble mean also has the cet_comparison_base=1 . It looks like _sort_cubes_for_verification.py in scoreswrappers.py requires modification to allow handling of collapsed cubes alongside the original cubes? James Ross (@james-a-f-ross) James Frost (@Fraetor) Do any of you have an idea to solve this as the indicator for base vs other model is still required here. |
|
James Frost (@jfrost-mo) ping for you to look at in work hours. |
|
how will this be used outside of the tests? can we not just sneak in a |
3 participants
expanding on PR #2229 to include ensemble capability for vertical profiles of RMSE.
For the calculation the ensemble mean of the variable is calculated first for all ensembles and then the scores package is used to
As the CRSP is more useful for analysing ensembles we only use the RMSE based on the ensemble mean here at the moment,.
Contribution checklist
Aim to have all relevant checks ticked off before merging. See the developer's guide for more detail.