PATC 2023 SUNSET hands-on

Hi @aho

This is a potential recipe for the 2023 PATC tutorial. On WS, this data takes about 2 minutes to load, and the whole workflow took around 7 minutes to run. We could make it shorter by reducing the number of forecast time steps (for example, running 2 or 3 instead of 6).

I chose the EU region because we already have the parameters to adjust the plot size thanks to Núria's work.

Let me know your thoughts on the data size and running time.

Description:
  Author: V. Agudetse
  Description: Analysis of MF System 7 with temperature
Analysis:
  Horizon: Seasonal
  Variables:
    name: tas
    freq: monthly_mean
  Datasets:
    System:
      name: Meteo-France-System7
    Multimodel: False
    Reference:
      name: ERA5
  Time:
    sdate: '1101'
    fcst_year: '2020'
    hcst_start: '1993'
    hcst_end: '2016'
    ftime_min: 1
    ftime_max: 6
  Region:
    name: "EU"
    latmin: 20
    latmax: 80
    lonmin: -20
    lonmax: 40
  Regrid:
    method: bilinear
    type: to_system
  Workflow:
    Anomalies:
      compute: yes # yes/no, default yes
      cross_validation: yes # yes/no, default yes
      save: 'none' # 'all'/'none'/'exp_only'/'fcst_only'
    Calibration:
      method: evmos
      save: 'none' # 'all'/'none'/'exp_only'/'fcst_only'
    Skill:
      metric: RPSS BSS10 BSS90
      cross_validation: yes
      save: 'all' # 'all'/'none'
    Probabilities:
      percentiles: [[1/3, 2/3], [1/10, 9/10]]
      save: 'none' # 'all'/'none'/'bins_only'/'percentiles_only'
    Visualization:
      plots: skill_metrics, forecast_ensemble_mean, most_likely_terciles
      multi_panel: no
      projection: cylindrical_equidistant
    Indicators:
      index: no
  ncores: 10
  remove_NAs: yes
  Output_format: S2S4E
Run:
  Loglevel: INFO
  Terminal: yes
  output_dir: /esarchive/scratch/vagudets/auto-s2s-outputs/
  code_dir: /esarchive/scratch/vagudets/repos/auto-s2s/