multi_model_statistics assumes cubes have latitude and longitude #266
Description
The multi_model_statistics preprocessor still assumes that all cubes have latitude and longitude components.
This is not the case.
There are so many crucial metrics that do not have a spatial component. For instance, stuff like the Global Average Sea Level rise, the Atlantic Current.
We had previously fixed this, but there' been some kind of change recently where this bug was re-instated.
Here's a recipe:
# ESMValTool
# recipe_ocean_scarlar_fields.yml
---
documentation:
description: |
Debug recipe
authors:
- demora_lee
maintainer:
- demora_lee
references:
- demora2018gmd
projects:
- ukesm
preprocessors:
# --------------------------------------------------
# Time series preprocessors
# --------------------------------------------------
prep_timeseries_0D: # For 0D fields
multi_model_statistics:
span: full
statistics: [mean ]
diagnostics:
diag_timeseries_sealevel:
variables:
zostoga: # Global Average Thermosteric Sea Level Change
preprocessor: prep_timeseries_0D
additional_datasets:
- {dataset: ACCESS1-0, project: CMIP5, exp: historical, mip: Omon, ensemble: r1i1p1, start_year: 1850, end_year: 2005, }
- {dataset: ACCESS1-3, project: CMIP5, exp: historical, mip: Omon, ensemble: r1i1p1, start_year: 1850, end_year: 2005, }
scripts:
Regional_Ocean_Surface_mean_timeseries:
script: ocean/diagnostic_timeseries.py
This recipe fails with:
File "/users/modellers/ledm/workspace/ESMValTool/ESMValCore/esmvalcore/preprocessor/_multimodel.py", line 110, in _put_in_cube
lats = template_cube.coord('latitude')
As there's no latitude in the cube.
I've made a basic bugfix PR to this file which I'l submit shortly.