Add Hovmöller time vs. lat/lon plot to diagnostic monitor/multi_datsets.py

[lukruh]

See https://de.wikipedia.org/wiki/Hovm%C3%B6ller-Diagramm.

Aim: add new plot type hovmoeller_time_vs_lat_or_lon to monitor/multi_datasets.py.

Large parts of the necessary code can be copy-pasted from here:

ESMValTool/esmvaltool/diag_scripts/monitor/multi_datasets.py

Lines 1418 to 1494 in 0d2b68d

	def create_zonal_mean_profile_plot(self, datasets, short_name):
	"""Create zonal mean profile plot."""
	plot_type = 'zonal_mean_profile'
	if plot_type not in self.plots:
	return
	if not datasets:
	raise ValueError(f"No input data to plot '{plot_type}' given")
	# Get reference dataset if possible
	ref_dataset = self._get_reference_dataset(datasets, short_name)
	if ref_dataset is None:
	logger.info("Plotting %s without reference dataset", plot_type)
	else:
	logger.info("Plotting %s with reference dataset '%s'", plot_type,
	self._get_label(ref_dataset))
	# Get plot function
	plot_func = self._get_plot_func(plot_type)
	# Create a single plot for each dataset (incl. reference dataset if
	# given)
	for dataset in datasets:
	if dataset == ref_dataset:
	continue
	ancestors = [dataset['filename']]
	if ref_dataset is None:
	(plot_path, netcdf_paths) = (
	self._plot_zonal_mean_profile_without_ref(plot_func,
	dataset)
	)
	caption = (
	f"Zonal mean profile of {dataset['long_name']} of dataset "
	f"{dataset['dataset']} (project {dataset['project']}) "
	f"from {dataset['start_year']} to {dataset['end_year']}."
	)
	else:
	(plot_path, netcdf_paths) = (
	self._plot_zonal_mean_profile_with_ref(plot_func, dataset,
	ref_dataset)
	)
	caption = (
	f"Zonal mean profile of {dataset['long_name']} of dataset "
	f"{dataset['dataset']} (project {dataset['project']}) "
	f"including bias relative to {ref_dataset['dataset']} "
	f"(project {ref_dataset['project']}) from "
	f"{dataset['start_year']} to {dataset['end_year']}."
	)
	ancestors.append(ref_dataset['filename'])
	# If statistics are shown add a brief description to the caption
	if self.plots[plot_type]['show_stats']:
	caption += (
	" The number in the top left corner corresponds to the "
	"spatial mean (weighted by grid cell areas).")
	# Save plot
	plt.savefig(plot_path, **self.cfg['savefig_kwargs'])
	logger.info("Wrote %s", plot_path)
	plt.close()
	# Save netCDFs
	for (netcdf_path, cube) in netcdf_paths.items():
	io.iris_save(cube, netcdf_path)
	# Provenance tracking
	provenance_record = {
	'ancestors': ancestors,
	'authors': ['schlund_manuel'],
	'caption': caption,
	'plot_types': ['vert'],
	'long_names': [dataset['long_name']],
	}
	with ProvenanceLogger(self.cfg) as provenance_logger:
	provenance_logger.log(plot_path, provenance_record)
	for netcdf_path in netcdf_paths:
	provenance_logger.log(netcdf_path, provenance_record)

Use /work/bd1179/esmvaltool_hackathon/recipes/recipe_hovmoeller_time_vs_lat_or_lon.yml to get started. You find more detailed instructions in this file.

TODO:

• Discuss which variable/which regions/which observations we want to show.
• adjust labels and order depending on lat or lon
• documentation in diagnostic file
• test meridional and zonal pp
• single plot without reference
• multi panel with reference and bias
• consistent order in docs, methods, plot types etc..
• try to enable/change all recipe options
• • rasterize does not work
• • remove show stats
• • replace minor tick labels with minor ticks
• update captions and provenance
• discuss defaults and options with others working on hovmoellers
• example recipes in monitor with some example options
• example plots in docs doc/sphinx/source/recipes/recipe_monitor.rst add plot image to the bottom (compressed <100k)
• flake, codespell, isort multi_datasets.py
• add authors to recipe (and diag?)

[lukruh]

Example for Hovmoeller using iris:
https://scitools-iris.readthedocs.io/en/stable/generated/gallery/meteorology/plot_hovmoller.html

[lukruh]

we should test _add_stats() again.