{
 "profile": "tabular-data-package",
 "name": "meteosignal-verification-scores",
 "id": "https://meteosignal.com/scores/datapackage.json",
 "title": "MeteoSignal: verification scores of open weather forecasts over Europe",
 "description": "Independent verification of ECMWF's AI model AIFS against IFS, GFS and other open AI and physics-based models, at weather stations in Germany and Europe, at KNMI's Cabauw tower and against ERA5 over Europe. Read the usage rules before using the numbers.",
 "version": "20261116T194902Z",
 "created": "2026-11-16T19:49:02Z",
 "homepage": "https://meteosignal.com/scores/",
 "keywords": [
  "forecast verification",
  "numerical weather prediction",
  "AI weather models",
  "AIFS",
  "IFS",
  "GFS",
  "ERA5",
  "RMSE",
  "Europe"
 ],
 "licenses": [
  {
   "name": "CC-BY-4.0",
   "path": "https://creativecommons.org/licenses/by/4.0/",
   "title": "Creative Commons Attribution 4.0"
  }
 ],
 "contributors": [
  {
   "title": "Iman Soltanzadeh",
   "role": "author",
   "path": "https://meteosignal.com/method/#about"
  }
 ],
 "sources": [
  {
   "title": "ECMWF open data: AIFS and IFS forecasts",
   "path": "https://www.ecmwf.int/"
  },
  {
   "title": "NOAA/NCEP GFS and AIGFS, NOAA Open Data Dissemination on AWS",
   "path": "https://registry.opendata.aws/noaa-gfs-bdp-pds/"
  },
  {
   "title": "AI Weather Prediction Model Reforecasts, CIRA and NOAA GSL",
   "path": "https://registry.opendata.aws/aiwp/"
  },
  {
   "title": "Open-Meteo Previous Runs API (ICON, up to 7 November 2026)",
   "path": "https://open-meteo.com/"
  },
  {
   "title": "Deutscher Wetterdienst, 10-minute station observations",
   "path": "https://opendata.dwd.de/"
  },
  {
   "title": "GeoSphere Austria, DMI, KNMI, MeteoSwiss and the Finnish Meteorological Institute, station observations"
  },
  {
   "title": "NOAA NCEI Global Historical Climatology Network hourly (GHCNh)"
  },
  {
   "title": "KNMI CESAR Cabauw tower profiles"
  },
  {
   "title": "Copernicus Climate Change Service, ERA5 reanalysis",
   "path": "https://doi.org/10.1002/qj.3803"
  },
  {
   "title": "WeatherBench 2 climatology",
   "path": "https://doi.org/10.1029/2023MS004019"
  }
 ],
 "usage": [
  "Compare numbers only within one file and, inside it, one site or model_set, one era and one track. The checks (the station sets, the Cabauw tower and ERA5) have their own places, days and caveats, and their numbers are not comparable with each other.",
  "A difference between AIFS and another model is a finding only where verdict is ai-leads or other-leads. cannot-separate means the two cannot be told apart on this sample, which is not evidence that they are equal. not-judged means the number shows a sign only.",
  "Comparisons in one family share days and places, so they are not independent results. Do not add up verdicts across files or sets as if they were: the German stations are also part of the europe set.",
  "All results are provisional. No AIFS version has been scored over a full year, so each sample covers only the seasons between its first_date and last_date. The verdicts are tested again at every weekly update and are interim until a version covers a full year. AIFS versions (era) cover different seasons and must not be compared with each other.",
  "Lower RMSE, MAE and absolute bias are better. Higher skill and anomaly correlation are better. A negative rmse_difference_ai_minus_other means AIFS had the lower error.",
  "ERA5 is a model product and AIFS was trained mostly on it, which probably favours AIFS in the era5 files. Take the size of any short-range lead from the station files.",
  "Values are point estimates over the whole sample with 95% intervals; monthly values are not in these files. The site does not publish which model is better at any single station.",
  "When quoting a number, give the site or model_set, the era, the variable, the lead day, the number of days and the data_version."
 ],
 "provisional": true,
 "temporalCoverage": "2025-09-07/2026-11-12",
 "method": "https://meteosignal.com/method/",
 "citation": "Soltanzadeh, I. (2026). MeteoSignal: verification of open weather forecasts over Europe. Data version 16 November 2026, run 20261116T194902Z. https://meteosignal.com/",
 "attribution": "Credit the source data as listed on the method page; their licences travel with these derived scores.",
 "resources": [
  {
   "profile": "tabular-data-resource",
   "name": "stations-scores",
   "path": "meteosignal-stations-scores.csv",
   "title": "Scores at the stations and the tower",
   "description": "One row per station set, AIFS version, variable, track, lead day and series: RMSE, MAE, bias and skill against two baselines, with 95% intervals from a stationary block bootstrap over forecast dates.",
   "format": "csv",
   "mediatype": "text/csv",
   "encoding": "utf-8",
   "bytes": 377151,
   "hash": "sha256:27e9b8053c47a2783d46b09bce9f51fafedf29d77192a4758570823d5ca78e26",
   "rows": 1608,
   "dialect": {
    "delimiter": ",",
    "header": true,
    "doubleQuote": true
   },
   "schema": {
    "fields": [
     {
      "name": "site",
      "type": "string",
      "description": "The station set. Each set has its own days; compare numbers only within one set",
      "constraints": {
       "enum": [
        "germany",
        "europe",
        "noaa-stations",
        "cabauw-tower",
        "germany-with-frozen-icon",
        "germany-with-aigfs",
        "germany-with-other-ai-models"
       ]
      },
      "categories": [
       {
        "value": "germany",
        "label": "30 DWD stations in Germany. The main result."
       },
       {
        "value": "europe",
        "label": "The German stations and the stations of the national services of Austria, the Netherlands, Denmark, Switzerland and Finland, pooled into one set with its own tests."
       },
       {
        "value": "noaa-stations",
        "label": "Stations in further countries from NOAA's hourly reports (GHCNh), scored as a set of their own, without solar."
       },
       {
        "value": "cabauw-tower",
        "label": "KNMI's Cabauw tower in the Netherlands, one site. A case study, never pooled with the stations."
       },
       {
        "value": "germany-with-frozen-icon",
        "label": "The German stations on the days with ICON from the Open-Meteo archive, up to 7 November 2026. A closed set that no longer grows."
       },
       {
        "value": "germany-with-aigfs",
        "label": "The German stations on the days with NOAA's AIGFS, which starts from the GFS analysis."
       },
       {
        "value": "germany-with-other-ai-models",
        "label": "The German stations on the days with NOAA's re-runs of GraphCast, FourCastNet v2-small, Pangu-Weather and Aurora, started from IFS analyses. They are NOAA's re-runs, not the developers' own forecasts."
       }
      ]
     },
     {
      "name": "era",
      "type": "string",
      "description": "AIFS version in service. Versions cover different seasons and are never pooled or compared",
      "constraints": {
       "enum": [
        "v1.1",
        "v2"
       ]
      },
      "categories": [
       {
        "value": "v1.1",
        "label": "AIFS 1.1, the AIFS version in service from 2025-08-27"
       },
       {
        "value": "v2",
        "label": "AIFS 2, the AIFS version in service from 2026-05-12"
       }
      ]
     },
     {
      "name": "variable",
      "type": "string",
      "description": "What is forecast",
      "constraints": {
       "enum": [
        "t2m",
        "wind10m",
        "solar",
        "wind100m",
        "z500",
        "t2m_era5",
        "wind10m_era5"
       ]
      },
      "categories": [
       {
        "value": "t2m",
        "label": "2 m air temperature at the station, in °C"
       },
       {
        "value": "wind10m",
        "label": "10 m wind speed at the station, in m/s"
       },
       {
        "value": "solar",
        "label": "Daily solar energy on a horizontal surface, in kWh/m² per day"
       },
       {
        "value": "wind100m",
        "label": "Wind speed at 100 m on the Cabauw tower, estimated from the 80 m and 140 m measurements, in m/s"
       },
       {
        "value": "z500",
        "label": "Geopotential height of the 500 hPa surface, against ERA5, in gpm"
       },
       {
        "value": "t2m_era5",
        "label": "2 m air temperature on the 0.25° grid, against ERA5, in K"
       },
       {
        "value": "wind10m_era5",
        "label": "10 m wind speed on the 0.25° grid, against ERA5, in m/s"
       }
      ]
     },
     {
      "name": "unit",
      "type": "string",
      "description": "Unit of every value in this row that is not a fraction or a probability",
      "constraints": {
       "enum": [
        "°C",
        "m/s",
        "kWh/m2/day",
        "gpm",
        "K"
       ]
      },
      "categories": [
       {
        "value": "°C",
        "label": "°C"
       },
       {
        "value": "m/s",
        "label": "m/s"
       },
       {
        "value": "kWh/m2/day",
        "label": "kWh/m2/day"
       },
       {
        "value": "gpm",
        "label": "gpm"
       },
       {
        "value": "K",
        "label": "K"
       }
      ]
     },
     {
      "name": "track",
      "type": "string",
      "description": "Raw or bias-corrected forecasts",
      "constraints": {
       "enum": [
        "raw",
        "corrected"
       ]
      },
      "categories": [
       {
        "value": "raw",
        "label": "Forecasts as issued"
       },
       {
        "value": "corrected",
        "label": "Forecasts after removing each series’ mean error over the 30 days before issue (stations only)"
       }
      ]
     },
     {
      "name": "lead_days",
      "type": "integer",
      "description": "How many days ahead the forecast was made",
      "constraints": {
       "enum": [
        "1",
        "3",
        "5",
        "7",
        "10"
       ]
      },
      "categories": [
       {
        "value": "1",
        "label": "1 day ahead"
       },
       {
        "value": "3",
        "label": "3 days ahead"
       },
       {
        "value": "5",
        "label": "5 days ahead"
       },
       {
        "value": "7",
        "label": "7 days ahead"
       },
       {
        "value": "10",
        "label": "10 days ahead"
       }
      ]
     },
     {
      "name": "days",
      "type": "integer",
      "description": "Number of forecast days in the sample. All models and climatology in one sample are scored on exactly the same pairs. At stations, persistence and the skill against it rest on persistence_days"
     },
     {
      "name": "first_date",
      "type": "string",
      "description": "First valid day of the sample, YYYY-MM-DD"
     },
     {
      "name": "last_date",
      "type": "string",
      "description": "Last valid day of the sample, YYYY-MM-DD"
     },
     {
      "name": "effective_days",
      "type": "number",
      "description": "About how many independent days the sample is worth, after allowing for day-to-day correlation"
     },
     {
      "name": "series",
      "type": "string",
      "description": "The model or baseline scored",
      "constraints": {
       "enum": [
        "AIFS",
        "IFS",
        "GFS",
        "ICON",
        "AIGFS",
        "GraphCast",
        "FourCastNet v2",
        "Pangu",
        "Aurora",
        "persistence",
        "climatology"
       ]
      },
      "categories": [
       {
        "value": "AIFS",
        "label": "AIFS Single 0.25°, AI model from ECMWF"
       },
       {
        "value": "IFS",
        "label": "IFS 0.25°, physics-based model from ECMWF"
       },
       {
        "value": "GFS",
        "label": "GFS 0.25°, physics-based model from NOAA"
       },
       {
        "value": "ICON",
        "label": "ICON Global, physics-based model from DWD, scored up to 2026-11-07 only (a closed set)"
       },
       {
        "value": "AIGFS",
        "label": "AIGFS 0.25°, AI model from NOAA"
       },
       {
        "value": "GraphCast",
        "label": "GraphCast Operational, AI model from NOAA AIWP"
       },
       {
        "value": "FourCastNet v2",
        "label": "FourCastNet v2-small, AI model from NOAA AIWP"
       },
       {
        "value": "Pangu",
        "label": "Pangu-Weather, AI model from NOAA AIWP"
       },
       {
        "value": "Aurora",
        "label": "Aurora, AI model from NOAA AIWP"
       },
       {
        "value": "persistence",
        "label": "Baseline: the measurement from the lead time earlier, carried forward (for solar, the energy of the day before that, the last day complete when the forecast was issued)"
       },
       {
        "value": "climatology",
        "label": "Baseline: the average for that date and time of day over earlier years"
       }
      ]
     },
     {
      "name": "kind",
      "type": "string",
      "description": "What kind of series it is",
      "constraints": {
       "enum": [
        "ai",
        "physics",
        "baseline"
       ]
      },
      "categories": [
       {
        "value": "ai",
        "label": "AI model"
       },
       {
        "value": "physics",
        "label": "Physics-based model"
       },
       {
        "value": "baseline",
        "label": "Reference forecast, not a model"
       }
      ]
     },
     {
      "name": "rmse",
      "type": "number",
      "description": "Root mean square error, in the row’s unit. Lower is better"
     },
     {
      "name": "rmse_low",
      "type": "number",
      "description": "Lower end of the 95% interval of rmse"
     },
     {
      "name": "rmse_high",
      "type": "number",
      "description": "Upper end of the 95% interval of rmse"
     },
     {
      "name": "rmse_raw_same_pairs",
      "type": "number",
      "description": "On the corrected track only: the RMSE of the raw forecasts on the same pairs, in the row’s unit"
     },
     {
      "name": "forecast_activity",
      "type": "number",
      "description": "Spread of the forecast’s departures from climatology over that of the truth, each station or grid cell centred on its own mean at each valid hour first, a fraction. 1 is as variable as the truth; below 1 is smoother"
     },
     {
      "name": "mae",
      "type": "number",
      "description": "Mean absolute error, in the row’s unit. Lower is better"
     },
     {
      "name": "mae_low",
      "type": "number",
      "description": "Lower end of the 95% interval of mae"
     },
     {
      "name": "mae_high",
      "type": "number",
      "description": "Upper end of the 95% interval of mae"
     },
     {
      "name": "bias",
      "type": "number",
      "description": "Mean error, forecast minus truth, in the row’s unit. 0 is unbiased"
     },
     {
      "name": "bias_low",
      "type": "number",
      "description": "Lower end of the 95% interval of bias"
     },
     {
      "name": "bias_high",
      "type": "number",
      "description": "Upper end of the 95% interval of bias"
     },
     {
      "name": "skill_vs_persistence",
      "type": "number",
      "description": "1 minus the RMSE over the RMSE of persistence, a fraction, on the days and pairs where persistence can be scored (persistence_days). 1 is perfect, 0 no better than persistence, below 0 worse"
     },
     {
      "name": "skill_vs_persistence_low",
      "type": "number",
      "description": "Lower end of the 95% interval of skill_vs_persistence"
     },
     {
      "name": "skill_vs_persistence_high",
      "type": "number",
      "description": "Upper end of the 95% interval of skill_vs_persistence"
     },
     {
      "name": "skill_vs_climatology",
      "type": "number",
      "description": "1 minus the RMSE over the RMSE of climatology, a fraction. 1 is perfect, 0 no better than climatology, below 0 worse"
     },
     {
      "name": "skill_vs_climatology_low",
      "type": "number",
      "description": "Lower end of the 95% interval of skill_vs_climatology"
     },
     {
      "name": "skill_vs_climatology_high",
      "type": "number",
      "description": "Upper end of the 95% interval of skill_vs_climatology"
     },
     {
      "name": "mse_skill_vs_persistence",
      "type": "number",
      "description": "1 minus the ratio of mean square errors against persistence, a fraction"
     },
     {
      "name": "mse_skill_vs_climatology",
      "type": "number",
      "description": "1 minus the ratio of mean square errors against climatology, a fraction"
     },
     {
      "name": "persistence_days",
      "type": "integer",
      "description": "Number of forecast days on which persistence could be scored at every lead as well as the models. The persistence series, skill_vs_persistence and mse_skill_vs_persistence rest on these days; every other value in the row rests on days. The two are equal where no earlier measurement is missing, and against ERA5"
     },
     {
      "name": "persistence_last_date",
      "type": "string",
      "description": "Last valid day of the days counted in persistence_days, YYYY-MM-DD"
     },
     {
      "name": "correction_change_in_rmse",
      "type": "number",
      "description": "On the corrected track: corrected RMSE minus raw RMSE on the same pairs, in the row’s unit. Negative: the correction lowered the error"
     },
     {
      "name": "correction_change_in_rmse_low",
      "type": "number",
      "description": "Lower end of the 95% interval of correction_change_in_rmse"
     },
     {
      "name": "correction_change_in_rmse_high",
      "type": "number",
      "description": "Upper end of the 95% interval of correction_change_in_rmse"
     },
     {
      "name": "data_version",
      "type": "string",
      "description": "The weekly run that wrote this row. Name it when quoting a number"
     }
    ],
    "primaryKey": [
     "site",
     "era",
     "variable",
     "track",
     "lead_days",
     "series"
    ],
    "missingValues": [
     ""
    ]
   }
  },
  {
   "profile": "tabular-data-resource",
   "name": "stations-aifs-vs-models",
   "path": "meteosignal-stations-aifs-vs-models.csv",
   "title": "AIFS against each other model at the stations and the tower",
   "description": "One row per station set, AIFS version, variable, track, lead day and other model: the paired RMSE difference of AIFS and that model on the same times and places, its interval, the test and the verdict.",
   "format": "csv",
   "mediatype": "text/csv",
   "encoding": "utf-8",
   "bytes": 116551,
   "hash": "sha256:ec4c551f4aed825c771a502219b71ee52c6c6458b6d8af45880b762f13c92606",
   "rows": 588,
   "dialect": {
    "delimiter": ",",
    "header": true,
    "doubleQuote": true
   },
   "schema": {
    "fields": [
     {
      "name": "site",
      "type": "string",
      "description": "The station set. Each set has its own days; compare numbers only within one set",
      "constraints": {
       "enum": [
        "germany",
        "europe",
        "noaa-stations",
        "cabauw-tower",
        "germany-with-frozen-icon",
        "germany-with-aigfs",
        "germany-with-other-ai-models"
       ]
      },
      "categories": [
       {
        "value": "germany",
        "label": "30 DWD stations in Germany. The main result."
       },
       {
        "value": "europe",
        "label": "The German stations and the stations of the national services of Austria, the Netherlands, Denmark, Switzerland and Finland, pooled into one set with its own tests."
       },
       {
        "value": "noaa-stations",
        "label": "Stations in further countries from NOAA's hourly reports (GHCNh), scored as a set of their own, without solar."
       },
       {
        "value": "cabauw-tower",
        "label": "KNMI's Cabauw tower in the Netherlands, one site. A case study, never pooled with the stations."
       },
       {
        "value": "germany-with-frozen-icon",
        "label": "The German stations on the days with ICON from the Open-Meteo archive, up to 7 November 2026. A closed set that no longer grows."
       },
       {
        "value": "germany-with-aigfs",
        "label": "The German stations on the days with NOAA's AIGFS, which starts from the GFS analysis."
       },
       {
        "value": "germany-with-other-ai-models",
        "label": "The German stations on the days with NOAA's re-runs of GraphCast, FourCastNet v2-small, Pangu-Weather and Aurora, started from IFS analyses. They are NOAA's re-runs, not the developers' own forecasts."
       }
      ]
     },
     {
      "name": "era",
      "type": "string",
      "description": "AIFS version in service. Versions cover different seasons and are never pooled or compared",
      "constraints": {
       "enum": [
        "v1.1",
        "v2"
       ]
      },
      "categories": [
       {
        "value": "v1.1",
        "label": "AIFS 1.1, the AIFS version in service from 2025-08-27"
       },
       {
        "value": "v2",
        "label": "AIFS 2, the AIFS version in service from 2026-05-12"
       }
      ]
     },
     {
      "name": "variable",
      "type": "string",
      "description": "What is forecast",
      "constraints": {
       "enum": [
        "t2m",
        "wind10m",
        "solar",
        "wind100m",
        "z500",
        "t2m_era5",
        "wind10m_era5"
       ]
      },
      "categories": [
       {
        "value": "t2m",
        "label": "2 m air temperature at the station, in °C"
       },
       {
        "value": "wind10m",
        "label": "10 m wind speed at the station, in m/s"
       },
       {
        "value": "solar",
        "label": "Daily solar energy on a horizontal surface, in kWh/m² per day"
       },
       {
        "value": "wind100m",
        "label": "Wind speed at 100 m on the Cabauw tower, estimated from the 80 m and 140 m measurements, in m/s"
       },
       {
        "value": "z500",
        "label": "Geopotential height of the 500 hPa surface, against ERA5, in gpm"
       },
       {
        "value": "t2m_era5",
        "label": "2 m air temperature on the 0.25° grid, against ERA5, in K"
       },
       {
        "value": "wind10m_era5",
        "label": "10 m wind speed on the 0.25° grid, against ERA5, in m/s"
       }
      ]
     },
     {
      "name": "unit",
      "type": "string",
      "description": "Unit of every value in this row that is not a fraction or a probability",
      "constraints": {
       "enum": [
        "°C",
        "m/s",
        "kWh/m2/day",
        "gpm",
        "K"
       ]
      },
      "categories": [
       {
        "value": "°C",
        "label": "°C"
       },
       {
        "value": "m/s",
        "label": "m/s"
       },
       {
        "value": "kWh/m2/day",
        "label": "kWh/m2/day"
       },
       {
        "value": "gpm",
        "label": "gpm"
       },
       {
        "value": "K",
        "label": "K"
       }
      ]
     },
     {
      "name": "track",
      "type": "string",
      "description": "Raw or bias-corrected forecasts",
      "constraints": {
       "enum": [
        "raw",
        "corrected"
       ]
      },
      "categories": [
       {
        "value": "raw",
        "label": "Forecasts as issued"
       },
       {
        "value": "corrected",
        "label": "Forecasts after removing each series’ mean error over the 30 days before issue (stations only)"
       }
      ]
     },
     {
      "name": "lead_days",
      "type": "integer",
      "description": "How many days ahead the forecast was made",
      "constraints": {
       "enum": [
        "1",
        "3",
        "5",
        "7",
        "10"
       ]
      },
      "categories": [
       {
        "value": "1",
        "label": "1 day ahead"
       },
       {
        "value": "3",
        "label": "3 days ahead"
       },
       {
        "value": "5",
        "label": "5 days ahead"
       },
       {
        "value": "7",
        "label": "7 days ahead"
       },
       {
        "value": "10",
        "label": "10 days ahead"
       }
      ]
     },
     {
      "name": "days",
      "type": "integer",
      "description": "Number of forecast days in the sample. All models and climatology in one sample are scored on exactly the same pairs. At stations, persistence and the skill against it rest on persistence_days"
     },
     {
      "name": "first_date",
      "type": "string",
      "description": "First valid day of the sample, YYYY-MM-DD"
     },
     {
      "name": "last_date",
      "type": "string",
      "description": "Last valid day of the sample, YYYY-MM-DD"
     },
     {
      "name": "effective_days",
      "type": "number",
      "description": "About how many independent days the sample is worth, after allowing for day-to-day correlation"
     },
     {
      "name": "compared_with",
      "type": "string",
      "description": "The model AIFS is compared with. Only comparisons of AIFS with another model are tested",
      "constraints": {
       "enum": [
        "IFS",
        "GFS",
        "ICON",
        "AIGFS",
        "GraphCast",
        "FourCastNet v2",
        "Pangu",
        "Aurora"
       ]
      },
      "categories": [
       {
        "value": "IFS",
        "label": "IFS 0.25°, physics-based model from ECMWF"
       },
       {
        "value": "GFS",
        "label": "GFS 0.25°, physics-based model from NOAA"
       },
       {
        "value": "ICON",
        "label": "ICON Global, physics-based model from DWD, scored up to 2026-11-07 only (a closed set)"
       },
       {
        "value": "AIGFS",
        "label": "AIGFS 0.25°, AI model from NOAA"
       },
       {
        "value": "GraphCast",
        "label": "GraphCast Operational, AI model from NOAA AIWP"
       },
       {
        "value": "FourCastNet v2",
        "label": "FourCastNet v2-small, AI model from NOAA AIWP"
       },
       {
        "value": "Pangu",
        "label": "Pangu-Weather, AI model from NOAA AIWP"
       },
       {
        "value": "Aurora",
        "label": "Aurora, AI model from NOAA AIWP"
       }
      ]
     },
     {
      "name": "rmse_difference_ai_minus_other",
      "type": "number",
      "description": "AIFS's RMSE minus the other model's, in the row’s unit. Negative: AIFS's error was lower (better)"
     },
     {
      "name": "rmse_difference_ai_minus_other_low",
      "type": "number",
      "description": "Lower end of the 95% interval of rmse_difference_ai_minus_other"
     },
     {
      "name": "rmse_difference_ai_minus_other_high",
      "type": "number",
      "description": "Upper end of the 95% interval of rmse_difference_ai_minus_other"
     },
     {
      "name": "rmse_difference_relative",
      "type": "number",
      "description": "The same difference as a fraction of the other model's RMSE. -0.10 means AIFS's RMSE was 10% lower"
     },
     {
      "name": "p_value",
      "type": "number",
      "description": "p-value of the paired test, before correction"
     },
     {
      "name": "p_value_holm",
      "type": "number",
      "description": "p-value after the Holm correction within its family of comparisons. A verdict needs it below 0.01, which is lower than the 0.05 of the intervals because the test is repeated at every weekly update"
     },
     {
      "name": "family_size",
      "type": "integer",
      "description": "Number of comparisons in the Holm family"
     },
     {
      "name": "verdict",
      "type": "string",
      "description": "The verdict on the RMSE difference. Only ai-leads and other-leads are findings",
      "constraints": {
       "enum": [
        "ai-leads",
        "other-leads",
        "cannot-separate",
        "depends-on-reading",
        "depends-on-height-correction",
        "not-judged"
       ]
      },
      "categories": [
       {
        "value": "ai-leads",
        "label": "Tested: AIFS's RMSE was lower, and the difference holds (95% interval excludes zero, Holm-adjusted p below undefined, enough blocks, same sign in both halves of the period)"
       },
       {
        "value": "other-leads",
        "label": "Tested: AIFS's RMSE was higher, and the difference holds"
       },
       {
        "value": "cannot-separate",
        "label": "The two models cannot be told apart on this sample. Not evidence that they are equal"
       },
       {
        "value": "depends-on-reading",
        "label": "The verdict did not hold when the models were read at the nearest land grid point instead of by interpolation, so none is given"
       },
       {
        "value": "depends-on-height-correction",
        "label": "The temperature verdict did not hold under other height corrections, so none is given"
       },
       {
        "value": "not-judged",
        "label": "No verdict by rule: against ERA5 only 500 hPa height after day 1 is judged. The number is shown for its sign only"
       }
      ]
     },
     {
      "name": "borderline",
      "type": "boolean",
      "description": "Always false since a verdict needs a Holm-adjusted p-value below 0.01: it marked verdicts given at 0.02 or above when the level was 0.05"
     },
     {
      "name": "reason",
      "type": "string",
      "description": "Why no verdict was given where the test alone might have given one. Empty when there is nothing to explain",
      "constraints": {
       "enum": [
        "too-few-blocks",
        "unstable-halves",
        "depends-on-reading",
        "depends-on-height-correction",
        "analysis-confounded",
        "reference-confounded"
       ]
      },
      "categories": [
       {
        "value": "too-few-blocks",
        "label": "The sample holds fewer blocks than a verdict needs"
       },
       {
        "value": "unstable-halves",
        "label": "The test would pass, but the sign differs between the two halves of the period"
       },
       {
        "value": "depends-on-reading",
        "label": "See verdict depends-on-reading"
       },
       {
        "value": "depends-on-height-correction",
        "label": "See verdict depends-on-height-correction"
       },
       {
        "value": "analysis-confounded",
        "label": "Against ERA5 at day 1: the models start from different analyses and ERA5 is an ECMWF product"
       },
       {
        "value": "reference-confounded",
        "label": "Against ERA5 for temperature and wind: each model's temperature refers to its own terrain, and AIFS was trained on ERA5"
       }
      ]
     },
     {
      "name": "mae_difference_ai_minus_other",
      "type": "number",
      "description": "The same comparison on MAE, in the row’s unit"
     },
     {
      "name": "mae_difference_ai_minus_other_low",
      "type": "number",
      "description": "Lower end of the 95% interval of mae_difference_ai_minus_other"
     },
     {
      "name": "mae_difference_ai_minus_other_high",
      "type": "number",
      "description": "Upper end of the 95% interval of mae_difference_ai_minus_other"
     },
     {
      "name": "mae_verdict",
      "type": "string",
      "description": "The verdict on the MAE difference",
      "constraints": {
       "enum": [
        "ai-leads",
        "other-leads",
        "cannot-separate",
        "depends-on-reading",
        "depends-on-height-correction",
        "not-judged"
       ]
      },
      "categories": [
       {
        "value": "ai-leads",
        "label": "Tested: AIFS's RMSE was lower, and the difference holds (95% interval excludes zero, Holm-adjusted p below undefined, enough blocks, same sign in both halves of the period)"
       },
       {
        "value": "other-leads",
        "label": "Tested: AIFS's RMSE was higher, and the difference holds"
       },
       {
        "value": "cannot-separate",
        "label": "The two models cannot be told apart on this sample. Not evidence that they are equal"
       },
       {
        "value": "depends-on-reading",
        "label": "The verdict did not hold when the models were read at the nearest land grid point instead of by interpolation, so none is given"
       },
       {
        "value": "depends-on-height-correction",
        "label": "The temperature verdict did not hold under other height corrections, so none is given"
       },
       {
        "value": "not-judged",
        "label": "No verdict by rule: against ERA5 only 500 hPa height after day 1 is judged. The number is shown for its sign only"
       }
      ]
     },
     {
      "name": "data_version",
      "type": "string",
      "description": "The weekly run that wrote this row. Name it when quoting a number"
     }
    ],
    "primaryKey": [
     "site",
     "era",
     "variable",
     "track",
     "lead_days",
     "compared_with"
    ],
    "missingValues": [
     ""
    ]
   }
  },
  {
   "profile": "tabular-data-resource",
   "name": "era5-scores",
   "path": "meteosignal-era5-scores.csv",
   "title": "Scores against ERA5 over Europe",
   "description": "One row per model set, AIFS version, variable, lead day and series, scored against ERA5 on a 0.25° grid from 35 to 72°N and 12.5°W to 40°E, weighted by latitude. ERA5 is a model product and AIFS was trained mostly on it: a second view, not a measurement.",
   "format": "csv",
   "mediatype": "text/csv",
   "encoding": "utf-8",
   "bytes": 114809,
   "hash": "sha256:1275b3846deab061c0bbd761afc5362a5eb1632c5afb225cc4cdc38e422d6d86",
   "rows": 450,
   "dialect": {
    "delimiter": ",",
    "header": true,
    "doubleQuote": true
   },
   "schema": {
    "fields": [
     {
      "name": "model_set",
      "type": "string",
      "description": "The group of models scored together on the same days. Compare numbers only within one set",
      "constraints": {
       "enum": [
        "all",
        "operational",
        "aigfs"
       ]
      },
      "categories": [
       {
        "value": "all",
        "label": "The models scored together on the same days: AIFS, IFS, GFS, GraphCast, FourCastNet v2, Pangu, Aurora, AIGFS (the members differ by AIFS version, as available)."
       },
       {
        "value": "operational",
        "label": "The models scored together on the same days: AIFS, IFS, GFS (the members differ by AIFS version, as available)."
       },
       {
        "value": "aigfs",
        "label": "The models scored together on the same days: AIFS, IFS, GFS, AIGFS (the members differ by AIFS version, as available)."
       }
      ]
     },
     {
      "name": "era",
      "type": "string",
      "description": "AIFS version in service. Versions cover different seasons and are never pooled or compared",
      "constraints": {
       "enum": [
        "v1.1",
        "v2"
       ]
      },
      "categories": [
       {
        "value": "v1.1",
        "label": "AIFS 1.1, the AIFS version in service from 2025-08-27"
       },
       {
        "value": "v2",
        "label": "AIFS 2, the AIFS version in service from 2026-05-12"
       }
      ]
     },
     {
      "name": "variable",
      "type": "string",
      "description": "What is forecast",
      "constraints": {
       "enum": [
        "t2m",
        "wind10m",
        "solar",
        "wind100m",
        "z500",
        "t2m_era5",
        "wind10m_era5"
       ]
      },
      "categories": [
       {
        "value": "t2m",
        "label": "2 m air temperature at the station, in °C"
       },
       {
        "value": "wind10m",
        "label": "10 m wind speed at the station, in m/s"
       },
       {
        "value": "solar",
        "label": "Daily solar energy on a horizontal surface, in kWh/m² per day"
       },
       {
        "value": "wind100m",
        "label": "Wind speed at 100 m on the Cabauw tower, estimated from the 80 m and 140 m measurements, in m/s"
       },
       {
        "value": "z500",
        "label": "Geopotential height of the 500 hPa surface, against ERA5, in gpm"
       },
       {
        "value": "t2m_era5",
        "label": "2 m air temperature on the 0.25° grid, against ERA5, in K"
       },
       {
        "value": "wind10m_era5",
        "label": "10 m wind speed on the 0.25° grid, against ERA5, in m/s"
       }
      ]
     },
     {
      "name": "unit",
      "type": "string",
      "description": "Unit of every value in this row that is not a fraction or a probability",
      "constraints": {
       "enum": [
        "°C",
        "m/s",
        "kWh/m2/day",
        "gpm",
        "K"
       ]
      },
      "categories": [
       {
        "value": "°C",
        "label": "°C"
       },
       {
        "value": "m/s",
        "label": "m/s"
       },
       {
        "value": "kWh/m2/day",
        "label": "kWh/m2/day"
       },
       {
        "value": "gpm",
        "label": "gpm"
       },
       {
        "value": "K",
        "label": "K"
       }
      ]
     },
     {
      "name": "track",
      "type": "string",
      "description": "Raw or bias-corrected forecasts",
      "constraints": {
       "enum": [
        "raw",
        "corrected"
       ]
      },
      "categories": [
       {
        "value": "raw",
        "label": "Forecasts as issued"
       },
       {
        "value": "corrected",
        "label": "Forecasts after removing each series’ mean error over the 30 days before issue (stations only)"
       }
      ]
     },
     {
      "name": "lead_days",
      "type": "integer",
      "description": "How many days ahead the forecast was made",
      "constraints": {
       "enum": [
        "1",
        "3",
        "5",
        "7",
        "10"
       ]
      },
      "categories": [
       {
        "value": "1",
        "label": "1 day ahead"
       },
       {
        "value": "3",
        "label": "3 days ahead"
       },
       {
        "value": "5",
        "label": "5 days ahead"
       },
       {
        "value": "7",
        "label": "7 days ahead"
       },
       {
        "value": "10",
        "label": "10 days ahead"
       }
      ]
     },
     {
      "name": "days",
      "type": "integer",
      "description": "Number of forecast days in the sample. All models and climatology in one sample are scored on exactly the same pairs. At stations, persistence and the skill against it rest on persistence_days"
     },
     {
      "name": "first_date",
      "type": "string",
      "description": "First valid day of the sample, YYYY-MM-DD"
     },
     {
      "name": "last_date",
      "type": "string",
      "description": "Last valid day of the sample, YYYY-MM-DD"
     },
     {
      "name": "effective_days",
      "type": "number",
      "description": "About how many independent days the sample is worth, after allowing for day-to-day correlation"
     },
     {
      "name": "series",
      "type": "string",
      "description": "The model or baseline scored",
      "constraints": {
       "enum": [
        "AIFS",
        "IFS",
        "GFS",
        "ICON",
        "AIGFS",
        "GraphCast",
        "FourCastNet v2",
        "Pangu",
        "Aurora",
        "persistence",
        "climatology"
       ]
      },
      "categories": [
       {
        "value": "AIFS",
        "label": "AIFS Single 0.25°, AI model from ECMWF"
       },
       {
        "value": "IFS",
        "label": "IFS 0.25°, physics-based model from ECMWF"
       },
       {
        "value": "GFS",
        "label": "GFS 0.25°, physics-based model from NOAA"
       },
       {
        "value": "ICON",
        "label": "ICON Global, physics-based model from DWD, scored up to 2026-11-07 only (a closed set)"
       },
       {
        "value": "AIGFS",
        "label": "AIGFS 0.25°, AI model from NOAA"
       },
       {
        "value": "GraphCast",
        "label": "GraphCast Operational, AI model from NOAA AIWP"
       },
       {
        "value": "FourCastNet v2",
        "label": "FourCastNet v2-small, AI model from NOAA AIWP"
       },
       {
        "value": "Pangu",
        "label": "Pangu-Weather, AI model from NOAA AIWP"
       },
       {
        "value": "Aurora",
        "label": "Aurora, AI model from NOAA AIWP"
       },
       {
        "value": "persistence",
        "label": "Baseline: the measurement from the lead time earlier, carried forward (for solar, the energy of the day before that, the last day complete when the forecast was issued)"
       },
       {
        "value": "climatology",
        "label": "Baseline: the average for that date and time of day over earlier years"
       }
      ]
     },
     {
      "name": "kind",
      "type": "string",
      "description": "What kind of series it is",
      "constraints": {
       "enum": [
        "ai",
        "physics",
        "baseline"
       ]
      },
      "categories": [
       {
        "value": "ai",
        "label": "AI model"
       },
       {
        "value": "physics",
        "label": "Physics-based model"
       },
       {
        "value": "baseline",
        "label": "Reference forecast, not a model"
       }
      ]
     },
     {
      "name": "rmse",
      "type": "number",
      "description": "Root mean square error, in the row’s unit. Lower is better"
     },
     {
      "name": "rmse_low",
      "type": "number",
      "description": "Lower end of the 95% interval of rmse"
     },
     {
      "name": "rmse_high",
      "type": "number",
      "description": "Upper end of the 95% interval of rmse"
     },
     {
      "name": "rmse_raw_same_pairs",
      "type": "number",
      "description": "On the corrected track only: the RMSE of the raw forecasts on the same pairs, in the row’s unit"
     },
     {
      "name": "forecast_activity",
      "type": "number",
      "description": "Spread of the forecast’s departures from climatology over that of the truth, each station or grid cell centred on its own mean at each valid hour first, a fraction. 1 is as variable as the truth; below 1 is smoother"
     },
     {
      "name": "mae",
      "type": "number",
      "description": "Mean absolute error, in the row’s unit. Lower is better"
     },
     {
      "name": "mae_low",
      "type": "number",
      "description": "Lower end of the 95% interval of mae"
     },
     {
      "name": "mae_high",
      "type": "number",
      "description": "Upper end of the 95% interval of mae"
     },
     {
      "name": "bias",
      "type": "number",
      "description": "Mean error, forecast minus truth, in the row’s unit. 0 is unbiased"
     },
     {
      "name": "bias_low",
      "type": "number",
      "description": "Lower end of the 95% interval of bias"
     },
     {
      "name": "bias_high",
      "type": "number",
      "description": "Upper end of the 95% interval of bias"
     },
     {
      "name": "skill_vs_persistence",
      "type": "number",
      "description": "1 minus the RMSE over the RMSE of persistence, a fraction, on the days and pairs where persistence can be scored (persistence_days). 1 is perfect, 0 no better than persistence, below 0 worse"
     },
     {
      "name": "skill_vs_persistence_low",
      "type": "number",
      "description": "Lower end of the 95% interval of skill_vs_persistence"
     },
     {
      "name": "skill_vs_persistence_high",
      "type": "number",
      "description": "Upper end of the 95% interval of skill_vs_persistence"
     },
     {
      "name": "skill_vs_climatology",
      "type": "number",
      "description": "1 minus the RMSE over the RMSE of climatology, a fraction. 1 is perfect, 0 no better than climatology, below 0 worse"
     },
     {
      "name": "skill_vs_climatology_low",
      "type": "number",
      "description": "Lower end of the 95% interval of skill_vs_climatology"
     },
     {
      "name": "skill_vs_climatology_high",
      "type": "number",
      "description": "Upper end of the 95% interval of skill_vs_climatology"
     },
     {
      "name": "mse_skill_vs_persistence",
      "type": "number",
      "description": "1 minus the ratio of mean square errors against persistence, a fraction"
     },
     {
      "name": "mse_skill_vs_climatology",
      "type": "number",
      "description": "1 minus the ratio of mean square errors against climatology, a fraction"
     },
     {
      "name": "persistence_days",
      "type": "integer",
      "description": "Number of forecast days on which persistence could be scored at every lead as well as the models. The persistence series, skill_vs_persistence and mse_skill_vs_persistence rest on these days; every other value in the row rests on days. The two are equal where no earlier measurement is missing, and against ERA5"
     },
     {
      "name": "persistence_last_date",
      "type": "string",
      "description": "Last valid day of the days counted in persistence_days, YYYY-MM-DD"
     },
     {
      "name": "anomaly_correlation",
      "type": "number",
      "description": "Uncentred anomaly correlation with ERA5 against the WeatherBench 2 climatology of 1990 to 2019, a fraction. 1 is perfect"
     },
     {
      "name": "anomaly_correlation_low",
      "type": "number",
      "description": "Lower end of the 95% interval of anomaly_correlation"
     },
     {
      "name": "anomaly_correlation_high",
      "type": "number",
      "description": "Upper end of the 95% interval of anomaly_correlation"
     },
     {
      "name": "anomaly_correlation_centred",
      "type": "number",
      "description": "The same with the area-mean anomaly of each field removed first, a fraction"
     },
     {
      "name": "anomaly_correlation_centred_low",
      "type": "number",
      "description": "Lower end of the 95% interval of anomaly_correlation_centred"
     },
     {
      "name": "anomaly_correlation_centred_high",
      "type": "number",
      "description": "Upper end of the 95% interval of anomaly_correlation_centred"
     },
     {
      "name": "data_version",
      "type": "string",
      "description": "The weekly run that wrote this row. Name it when quoting a number"
     }
    ],
    "primaryKey": [
     "model_set",
     "era",
     "variable",
     "track",
     "lead_days",
     "series"
    ],
    "missingValues": [
     ""
    ]
   }
  },
  {
   "profile": "tabular-data-resource",
   "name": "era5-aifs-vs-models",
   "path": "meteosignal-era5-aifs-vs-models.csv",
   "title": "AIFS against each other model, against ERA5 over Europe",
   "description": "One row per model set, AIFS version, variable, lead day and other model: the paired RMSE difference against ERA5, its interval, the test and the verdict. Only 500 hPa height after day 1 is judged.",
   "format": "csv",
   "mediatype": "text/csv",
   "encoding": "utf-8",
   "bytes": 52794,
   "hash": "sha256:b30ac86c99cbc642b45c9efc31611502cd85371587b8db98ca84629103218343",
   "rows": 270,
   "dialect": {
    "delimiter": ",",
    "header": true,
    "doubleQuote": true
   },
   "schema": {
    "fields": [
     {
      "name": "model_set",
      "type": "string",
      "description": "The group of models scored together on the same days. Compare numbers only within one set",
      "constraints": {
       "enum": [
        "all",
        "operational",
        "aigfs"
       ]
      },
      "categories": [
       {
        "value": "all",
        "label": "The models scored together on the same days: AIFS, IFS, GFS, GraphCast, FourCastNet v2, Pangu, Aurora, AIGFS (the members differ by AIFS version, as available)."
       },
       {
        "value": "operational",
        "label": "The models scored together on the same days: AIFS, IFS, GFS (the members differ by AIFS version, as available)."
       },
       {
        "value": "aigfs",
        "label": "The models scored together on the same days: AIFS, IFS, GFS, AIGFS (the members differ by AIFS version, as available)."
       }
      ]
     },
     {
      "name": "era",
      "type": "string",
      "description": "AIFS version in service. Versions cover different seasons and are never pooled or compared",
      "constraints": {
       "enum": [
        "v1.1",
        "v2"
       ]
      },
      "categories": [
       {
        "value": "v1.1",
        "label": "AIFS 1.1, the AIFS version in service from 2025-08-27"
       },
       {
        "value": "v2",
        "label": "AIFS 2, the AIFS version in service from 2026-05-12"
       }
      ]
     },
     {
      "name": "variable",
      "type": "string",
      "description": "What is forecast",
      "constraints": {
       "enum": [
        "t2m",
        "wind10m",
        "solar",
        "wind100m",
        "z500",
        "t2m_era5",
        "wind10m_era5"
       ]
      },
      "categories": [
       {
        "value": "t2m",
        "label": "2 m air temperature at the station, in °C"
       },
       {
        "value": "wind10m",
        "label": "10 m wind speed at the station, in m/s"
       },
       {
        "value": "solar",
        "label": "Daily solar energy on a horizontal surface, in kWh/m² per day"
       },
       {
        "value": "wind100m",
        "label": "Wind speed at 100 m on the Cabauw tower, estimated from the 80 m and 140 m measurements, in m/s"
       },
       {
        "value": "z500",
        "label": "Geopotential height of the 500 hPa surface, against ERA5, in gpm"
       },
       {
        "value": "t2m_era5",
        "label": "2 m air temperature on the 0.25° grid, against ERA5, in K"
       },
       {
        "value": "wind10m_era5",
        "label": "10 m wind speed on the 0.25° grid, against ERA5, in m/s"
       }
      ]
     },
     {
      "name": "unit",
      "type": "string",
      "description": "Unit of every value in this row that is not a fraction or a probability",
      "constraints": {
       "enum": [
        "°C",
        "m/s",
        "kWh/m2/day",
        "gpm",
        "K"
       ]
      },
      "categories": [
       {
        "value": "°C",
        "label": "°C"
       },
       {
        "value": "m/s",
        "label": "m/s"
       },
       {
        "value": "kWh/m2/day",
        "label": "kWh/m2/day"
       },
       {
        "value": "gpm",
        "label": "gpm"
       },
       {
        "value": "K",
        "label": "K"
       }
      ]
     },
     {
      "name": "track",
      "type": "string",
      "description": "Raw or bias-corrected forecasts",
      "constraints": {
       "enum": [
        "raw",
        "corrected"
       ]
      },
      "categories": [
       {
        "value": "raw",
        "label": "Forecasts as issued"
       },
       {
        "value": "corrected",
        "label": "Forecasts after removing each series’ mean error over the 30 days before issue (stations only)"
       }
      ]
     },
     {
      "name": "lead_days",
      "type": "integer",
      "description": "How many days ahead the forecast was made",
      "constraints": {
       "enum": [
        "1",
        "3",
        "5",
        "7",
        "10"
       ]
      },
      "categories": [
       {
        "value": "1",
        "label": "1 day ahead"
       },
       {
        "value": "3",
        "label": "3 days ahead"
       },
       {
        "value": "5",
        "label": "5 days ahead"
       },
       {
        "value": "7",
        "label": "7 days ahead"
       },
       {
        "value": "10",
        "label": "10 days ahead"
       }
      ]
     },
     {
      "name": "days",
      "type": "integer",
      "description": "Number of forecast days in the sample. All models and climatology in one sample are scored on exactly the same pairs. At stations, persistence and the skill against it rest on persistence_days"
     },
     {
      "name": "first_date",
      "type": "string",
      "description": "First valid day of the sample, YYYY-MM-DD"
     },
     {
      "name": "last_date",
      "type": "string",
      "description": "Last valid day of the sample, YYYY-MM-DD"
     },
     {
      "name": "effective_days",
      "type": "number",
      "description": "About how many independent days the sample is worth, after allowing for day-to-day correlation"
     },
     {
      "name": "compared_with",
      "type": "string",
      "description": "The model AIFS is compared with. Only comparisons of AIFS with another model are tested",
      "constraints": {
       "enum": [
        "IFS",
        "GFS",
        "ICON",
        "AIGFS",
        "GraphCast",
        "FourCastNet v2",
        "Pangu",
        "Aurora"
       ]
      },
      "categories": [
       {
        "value": "IFS",
        "label": "IFS 0.25°, physics-based model from ECMWF"
       },
       {
        "value": "GFS",
        "label": "GFS 0.25°, physics-based model from NOAA"
       },
       {
        "value": "ICON",
        "label": "ICON Global, physics-based model from DWD, scored up to 2026-11-07 only (a closed set)"
       },
       {
        "value": "AIGFS",
        "label": "AIGFS 0.25°, AI model from NOAA"
       },
       {
        "value": "GraphCast",
        "label": "GraphCast Operational, AI model from NOAA AIWP"
       },
       {
        "value": "FourCastNet v2",
        "label": "FourCastNet v2-small, AI model from NOAA AIWP"
       },
       {
        "value": "Pangu",
        "label": "Pangu-Weather, AI model from NOAA AIWP"
       },
       {
        "value": "Aurora",
        "label": "Aurora, AI model from NOAA AIWP"
       }
      ]
     },
     {
      "name": "rmse_difference_ai_minus_other",
      "type": "number",
      "description": "AIFS's RMSE minus the other model's, in the row’s unit. Negative: AIFS's error was lower (better)"
     },
     {
      "name": "rmse_difference_ai_minus_other_low",
      "type": "number",
      "description": "Lower end of the 95% interval of rmse_difference_ai_minus_other"
     },
     {
      "name": "rmse_difference_ai_minus_other_high",
      "type": "number",
      "description": "Upper end of the 95% interval of rmse_difference_ai_minus_other"
     },
     {
      "name": "rmse_difference_relative",
      "type": "number",
      "description": "The same difference as a fraction of the other model's RMSE. -0.10 means AIFS's RMSE was 10% lower"
     },
     {
      "name": "p_value",
      "type": "number",
      "description": "p-value of the paired test, before correction"
     },
     {
      "name": "p_value_holm",
      "type": "number",
      "description": "p-value after the Holm correction within its family of comparisons. A verdict needs it below 0.01, which is lower than the 0.05 of the intervals because the test is repeated at every weekly update"
     },
     {
      "name": "family_size",
      "type": "integer",
      "description": "Number of comparisons in the Holm family"
     },
     {
      "name": "verdict",
      "type": "string",
      "description": "The verdict on the RMSE difference. Only ai-leads and other-leads are findings",
      "constraints": {
       "enum": [
        "ai-leads",
        "other-leads",
        "cannot-separate",
        "depends-on-reading",
        "depends-on-height-correction",
        "not-judged"
       ]
      },
      "categories": [
       {
        "value": "ai-leads",
        "label": "Tested: AIFS's RMSE was lower, and the difference holds (95% interval excludes zero, Holm-adjusted p below undefined, enough blocks, same sign in both halves of the period)"
       },
       {
        "value": "other-leads",
        "label": "Tested: AIFS's RMSE was higher, and the difference holds"
       },
       {
        "value": "cannot-separate",
        "label": "The two models cannot be told apart on this sample. Not evidence that they are equal"
       },
       {
        "value": "depends-on-reading",
        "label": "The verdict did not hold when the models were read at the nearest land grid point instead of by interpolation, so none is given"
       },
       {
        "value": "depends-on-height-correction",
        "label": "The temperature verdict did not hold under other height corrections, so none is given"
       },
       {
        "value": "not-judged",
        "label": "No verdict by rule: against ERA5 only 500 hPa height after day 1 is judged. The number is shown for its sign only"
       }
      ]
     },
     {
      "name": "borderline",
      "type": "boolean",
      "description": "Always false since a verdict needs a Holm-adjusted p-value below 0.01: it marked verdicts given at 0.02 or above when the level was 0.05"
     },
     {
      "name": "reason",
      "type": "string",
      "description": "Why no verdict was given where the test alone might have given one. Empty when there is nothing to explain",
      "constraints": {
       "enum": [
        "too-few-blocks",
        "unstable-halves",
        "depends-on-reading",
        "depends-on-height-correction",
        "analysis-confounded",
        "reference-confounded"
       ]
      },
      "categories": [
       {
        "value": "too-few-blocks",
        "label": "The sample holds fewer blocks than a verdict needs"
       },
       {
        "value": "unstable-halves",
        "label": "The test would pass, but the sign differs between the two halves of the period"
       },
       {
        "value": "depends-on-reading",
        "label": "See verdict depends-on-reading"
       },
       {
        "value": "depends-on-height-correction",
        "label": "See verdict depends-on-height-correction"
       },
       {
        "value": "analysis-confounded",
        "label": "Against ERA5 at day 1: the models start from different analyses and ERA5 is an ECMWF product"
       },
       {
        "value": "reference-confounded",
        "label": "Against ERA5 for temperature and wind: each model's temperature refers to its own terrain, and AIFS was trained on ERA5"
       }
      ]
     },
     {
      "name": "mae_difference_ai_minus_other",
      "type": "number",
      "description": "The same comparison on MAE, in the row’s unit"
     },
     {
      "name": "mae_difference_ai_minus_other_low",
      "type": "number",
      "description": "Lower end of the 95% interval of mae_difference_ai_minus_other"
     },
     {
      "name": "mae_difference_ai_minus_other_high",
      "type": "number",
      "description": "Upper end of the 95% interval of mae_difference_ai_minus_other"
     },
     {
      "name": "mae_verdict",
      "type": "string",
      "description": "The verdict on the MAE difference",
      "constraints": {
       "enum": [
        "ai-leads",
        "other-leads",
        "cannot-separate",
        "depends-on-reading",
        "depends-on-height-correction",
        "not-judged"
       ]
      },
      "categories": [
       {
        "value": "ai-leads",
        "label": "Tested: AIFS's RMSE was lower, and the difference holds (95% interval excludes zero, Holm-adjusted p below undefined, enough blocks, same sign in both halves of the period)"
       },
       {
        "value": "other-leads",
        "label": "Tested: AIFS's RMSE was higher, and the difference holds"
       },
       {
        "value": "cannot-separate",
        "label": "The two models cannot be told apart on this sample. Not evidence that they are equal"
       },
       {
        "value": "depends-on-reading",
        "label": "The verdict did not hold when the models were read at the nearest land grid point instead of by interpolation, so none is given"
       },
       {
        "value": "depends-on-height-correction",
        "label": "The temperature verdict did not hold under other height corrections, so none is given"
       },
       {
        "value": "not-judged",
        "label": "No verdict by rule: against ERA5 only 500 hPa height after day 1 is judged. The number is shown for its sign only"
       }
      ]
     },
     {
      "name": "data_version",
      "type": "string",
      "description": "The weekly run that wrote this row. Name it when quoting a number"
     }
    ],
    "primaryKey": [
     "model_set",
     "era",
     "variable",
     "track",
     "lead_days",
     "compared_with"
    ],
    "missingValues": [
     ""
    ]
   }
  }
 ]
}
