Bug in cloud scatterplot diagnostic

[bouweandela]

Reported by @lewisjared

  File "/home/jared/code/climate-ref/packages/climate-ref-core/src/climate_ref_core/diagnostics.py", line 555, in run
    self.execute(definition)
    │    │       └ ExecutionDefinition(diagnostic=CloudScatterplotCltSwcre(name='Scatterplots of two cloud-relevant variables (clt vs swcre)'), ...
    │    └ <function CommandLineDiagnostic.execute at 0x7f533730e8e0>
    └ CloudScatterplotCltSwcre(name='Scatterplots of two cloud-relevant variables (clt vs swcre)')

  File "/home/jared/code/climate-ref/packages/climate-ref-core/src/climate_ref_core/diagnostics.py", line 598, in execute
    cmd = self.build_cmd(definition)
          │    │         └ ExecutionDefinition(diagnostic=CloudScatterplotCltSwcre(name='Scatterplots of two cloud-relevant variables (clt vs swcre)'), ...
          │    └ <function ESMValToolDiagnostic.build_cmd at 0x7f533407e520>
          └ CloudScatterplotCltSwcre(name='Scatterplots of two cloud-relevant variables (clt vs swcre)')

  File "/home/jared/code/climate-ref/packages/climate-ref-esmvaltool/src/climate_ref_esmvaltool/diagnostics/base.py", line 116, in build_cmd
    recipe_path = self.write_recipe(definition)
                  │    │            └ ExecutionDefinition(diagnostic=CloudScatterplotCltSwcre(name='Scatterplots of two cloud-relevant variables (clt vs swcre)'), ...
                  │    └ <function ESMValToolDiagnostic.write_recipe at 0x7f533407e480>
                  └ CloudScatterplotCltSwcre(name='Scatterplots of two cloud-relevant variables (clt vs swcre)')

  File "/home/jared/code/climate-ref/packages/climate-ref-esmvaltool/src/climate_ref_esmvaltool/diagnostics/base.py", line 95, in write_recipe
    self.update_recipe(recipe, input_files)
    │    │             │       └ {<SourceDatasetType.CMIP6: 'cmip6'>:                start_time             end_time                                          ...
    │    │             └ {'documentation': {'title': 'Model evaluation with focus on clouds.', 'description': 'Plot 2D histograms using the Seaborn di...
    │    └ functools.partial(<function update_recipe at 0x7f533407f1a0>, var_x='clt', var_y='swcre')
    └ CloudScatterplotCltSwcre(name='Scatterplots of two cloud-relevant variables (clt vs swcre)')

  File "/home/jared/code/climate-ref/packages/climate-ref-esmvaltool/src/climate_ref_esmvaltool/diagnostics/cloud_scatterplots.py", line 58, in update_recipe
    suptitle = "CMIP6 {dataset} {ensemble} {grid} {timerange}".format(**datasets[0])
                                                                        └ [{'project': 'CMIP6', 'activity': 'CMIP', 'dataset': 'CNRM-ESM2-1', 'ensemble': 'r1i1p1f2', 'institute': 'CNRM-CERFACS', 'exp...

KeyError: 'timerange'

[bouweandela]

@lewisjared From looking at the code, I think this could happen if start_time or end_time is not available in the database:

climate-ref/packages/climate-ref-esmvaltool/src/climate_ref_esmvaltool/recipe.py

Lines 51 to 98 in 93e5e4a

	def as_timerange(group: pd.DataFrame) -> str | None:
	"""Format the timeranges from a dataframe as an ESMValTool timerange.
	Parameters
	----------
	group
	The dataframe describing a single dataset.
	Returns
	-------
	A timerange.
	"""
	# TODO: apply some rounding to avoid problems?
	# https://github.com/ESMValGroup/ESMValCore/issues/2048
	start_times = group.start_time.dropna()
	if start_times.empty:
	return None
	end_times = group.end_time.dropna()
	if end_times.empty:
	return None # pragma: no cover
	return f"{as_isodate(start_times.min())}/{as_isodate(end_times.max())}"
	def as_facets(
	group: pd.DataFrame,
	) -> dict[str, Any]:
	"""Convert a group from the datasets dataframe to ESMValTool facets.
	Parameters
	----------
	group:
	A group of datasets representing a single instance_id.
	Returns
	-------
	A :obj:`dict` containing facet-value pairs.
	"""
	facets = {}
	first_row = group.iloc[0]
	project = first_row.instance_id.split(".", 2)[0]
	facets["project"] = project
	for esmvaltool_name, ref_name in FACETS[project].items():
	facets[esmvaltool_name] = getattr(first_row, ref_name)
	timerange = as_timerange(group)
	if timerange is not None:
	facets["timerange"] = timerange
	return facets

Would it be possible that this is the case?

[lewisjared]

For https://staging.climate-ref.org/executions/3473?tab=datasets, the datasets of interest have bounds that appear correct.

SELECT d.slug, df.*
FROM dataset_file df
     JOIN dataset d ON df.dataset_id = d.id
WHERE d.slug in (
     'CMIP6.CMIP.CNRM-CERFACS.CNRM-ESM2-1.historical.r1i1p1f2.Amon.clwvi.gr.v20181206',
     'CMIP6.CMIP.CNRM-CERFACS.CNRM-ESM2-1.historical.r1i1p1f2.Amon.pr.gr.v20181206'
);

slug	id	dataset_id	end_time	start_time	path
CMIP6.CMIP.CNRM-CERFACS.CNRM-ESM2-1.historical.r1i1p1f2.Amon.clwvi.gr.v20181206	64341	2760	2014-12-16 12:00:00.000000	1850-01-16 12:00:00.000000	/mnt/data/intake/CMIP6/CMIP/CNRM-CERFACS/CNRM-ESM2-1/historical/r1i1p1f2/Amon/clwvi/gr/v20181206/clwvi_Amon_CNRM-ESM2-1_historical_r1i1p1f2_gr_185001-201412.nc
CMIP6.CMIP.CNRM-CERFACS.CNRM-ESM2-1.historical.r1i1p1f2.Amon.pr.gr.v20181206	8971	525	2014-12-16 12:00:00.000000	1850-01-16 12:00:00.000000	/mnt/data/intake/CMIP6/CMIP/CNRM-CERFACS/CNRM-ESM2-1/historical/r1i1p1f2/Amon/pr/gr/v20181206/pr_Amon_CNRM-ESM2-1_historical_r1i1p1f2_gr_185001-201412.nc

[bouweandela]

Then I'm out of ideas. The changes in #399 will probably fix the issue.

[bouweandela]

@lewisjared I did't see this one in the latest runs anymore, has it been solved?

[lewisjared]

Yes this exectution is now available after rerunning with the followup-fix in #426

https://staging.climate-ref.org/executions/868?tab=executions&detect_outliers=iqr&include_unverified=false