[iluise]

Description

New design for the plotting folder to accommodate for the parallelization:

Every plot is classified by what it is drawn from:

• Field plots: drawn from the loaded prediction / target arrays (ReaderOutput). They are drawn where the data is loaded: by each worker under MPI, by _process_stream in a serial run. A worker draws only the samples × forecast steps of its own chunk.
• Verification plots: drawn from the scores (MetricResults: .score, .spatial, .by_init_hour). Scores are averages over samples, so these plots are drawn once, after the per-chunk results are merged. That happens on the master under MPI, and at the end of evaluate_from_config in a serial run.

The master never needs the loaded data. The workers never need another chunk's results. This is the same split as scoring: score_orchestration computes MetricResults per chunk, and the master merges them.

plot_utils.py holds the common helpers; field_plots.py and verification_plots.py one side each, with the score maps in their own score_maps.py and the videos of both sides in animations.py; plot_orchestration.py holds what the run loop calls.

Module	Content	Called from
field_plots.py	plot_data (orchestration of maps and histograms: config, availability, fstep filtering, colour ranges, task building, dispatch); _plot_single_sample, _plot_histograms_across_samples, _select_ensemble_data, _dispatch_timeseries_plots	plot_orchestration.py
verification_plots.py	plot_scores (orchestration of the score_plots entries over metric × region); the eight drivers lead_time_plot_metric_region, ratio_plot_metric_region, heat_maps_metric_region, score_card_metric_region, bar_plot_metric_region, quantile_plot_metric_region, psd_plot_metric_region, init_hour_metric_region; _flatten_or_average, _extract_psd_attrs	plot_orchestration.py
score_maps.py	plot_score_maps (orchestration of the score maps, including the per-(metric, region) colour ranges); _plot_score_maps_per_stream, _plot_score_map. Verification side: drawn from the merged MetricResult.spatial	run_evaluation.py for now then plot_orchestration.py from one of the next PRs
animations.py	the videos of both sides, built from the frames on disk: _build_single_animation, _dispatch_field_animations, _dispatch_score_map_animations	field_plots.py, score_maps.py
plot_utils.py	the generic helpers	all
plot_orchestration.py	only the two functions the run loop calls	run_evaluation.py

NO CODE CHANGE. Functions are moved as they are, except for the following renames:

Today	After	Why
run_score_map_pipeline	plot_score_maps	it plots; "pipeline" says nothing
_scatter_plot_single	_plot_score_map	it draws only score maps; the old name reads like the per-sample map path (Plotter.create_maps_per_sample)
plot_summary + plot_init_hour_summary	plot_scores	one entry point for all score_plots entries; no "summary" naming
body of plot_init_hour_summary	init_hour_metric_region	a driver like the others
plot_metric_region	lead_time_plot_metric_region	every driver plots "a metric for a region"; this one is the score vs lead time
_plot_all_samples	_plot_histograms_across_samples	it draws only the across-sample histograms
_dispatch_animations	_dispatch_field_animations	symmetric with _dispatch_score_map_animations; it builds only map and histogram videos

Issue Number

Closes #2993

Is this PR a draft? Mark it as draft.

Checklist before asking for review

• I have performed a self-review of my code
• My changes comply with basic sanity checks:
  • I have fixed formatting issues with ./scripts/actions.sh lint
  • I have run unit tests with ./scripts/actions.sh unit-test
  • I have documented my code and I have updated the docstrings.
  • I have added unit tests, if relevant
• I have tried my changes with data and code:
  • I have run the integration tests with ./scripts/actions.sh integration-test
  • (bigger changes) I have run a full training and I have written in the comment the run_id(s): launch-slurm.py --time 60
  • (bigger changes and experiments) I have shared a hegdedoc in the github issue with all the configurations and runs for this experiments
• I have informed and aligned with people impacted by my change:
  • for config changes: the MatterMost channels and/or a design doc
  • for changes of dependencies: the MatterMost software development channel

FastEvaluation

• I have updated the public documentation if necessary

[iluise]

Tested on ERA5, NPPATMS, SurfaceCombined with the following config:

global_plotting_options:
 image_format : "png" #options: "png", "pdf", "svg", "eps", "jpg" ..

evaluation:
  metrics:
    - fbi:
        thresh: 0.5
    - rmse
    - froct
    - troct
 
  regions: ["global"] 
  score_plots: [lead_time, init_hour, score_map, score_animation]
  summary_dir: "./plots_paper/"
  plot_ensemble: "members" #supported: false, "std", "minmax", "members"
  print_summary: false #print out score values on screen. it can be verbose
  
  log_scale: false
  add_grid: false
  
run_ids :
  test_multi_stream_1de37:
  streams:
      ERA5:
        channels: ["z_500", "t_850"] #,"u_850", "v_850", "q_850"] #"2t", "10u", "10v", 
        regions: ["global"] #, "nhem", "tropics"] 
        evaluation:
          forecast_step: "1-5" #"all" #[1,2,3,4,5,6,7] #"all" #[1, 2, 3] #,3, 2]
          sample: "0-4" #[0]
          ensemble: "all"
      NPPATMS:
        channels: ["obsvalue_rawbt_1", "obsvalue_rawbt_5"]
        regions: ["global"] #, "nhem", "tropics"]
        evaluation:
          forecast_step: "1-5" #"all" #[1,2,3,4,5,6,7] #"all" #[1, 2, 3] #,3, 2]
          sample: "0-2" #[0]
          ensemble: "all" 
         plotting:
           sample: [0] 
           forecast_step: [1, 2] 
           ensemble: "all"
           data_plots:
             - maps:
                 predictions: [image, video]
                 target: [image]
                 bias: [image, video]
             - histograms_per_sample        # or histograms_per_sample: [image, video]
             - timeseries
      SurfaceCombined:
        regions: ["global"] #, "nhem", "tropics"]
        evaluation:
          forecast_step: "1-5" #"all" #[1,2,3,4,5,6,7] #"all" #[1, 2, 3]
          sample: "0-2" #[0]
          ensemble: "all"
         plotting:
           sample: [0] 
           forecast_step: [1, 2] 
           ensemble: "all"
           data_plots:
             - maps:
                 predictions: [image, video]
                 target: [image]
                 bias: [image, video]
             - histograms_per_sample        # or histograms_per_sample: [image, video]
             - timeseries