{
  "_schema": "alaska-salmon-preseason-forecasts-accuracy-v1",
  "_provenance": {
    "primary_source": "ADF&G — Annual preseason forecast reports + actual run outcomes",
    "primary_url": "https://www.adfg.alaska.gov/index.cfm?adfg=forecastreports.main",
    "secondary_sources": [
      "ADF&G — Preseason salmon run forecasting methods and performance review: https://www.adfg.alaska.gov/",
      "Hilborn, R. (2006). Fisheries success and failure: The case of the Bristol Bay salmon fishery. Bull. Marine Science 78:487-498.",
      "Ruggerone, G.T. & Connors, B.M. (2015). Productivity and survival of natural and hatchery Bristol Bay sockeye salmon in relation to climate and competition. Trans. Am. Fish. Soc. 144:1158-1180.",
      "Schindler, D.E. et al. (2010). Population diversity and the portfolio effect in an exploited species. Nature 465:609-612.",
      "Mundy, P.R. (Ed.) (2005). The Gulf of Alaska: Biology and Oceanography. Alaska Sea Grant College Program.",
      "NOAA Fisheries — Alaska preseason salmon forecast methods technical report (annual)",
      "ADF&G — Bristol Bay sockeye salmon forecast methodology (annual): https://www.adfg.alaska.gov/"
    ],
    "compiled_by": "alaskafishdata/run-timing",
    "compiled_date": "2026-07-29",
    "notes": "Alaska salmon preseason forecasts are critical for fisheries management — they inform the number of permits to issue, the management plan for the season, and the in-season adjustment triggers. ADF&G releases BB sockeye preseason forecasts in February/March each year using sibling regression (this year's return predicted from last year's return of the same age class), environmental indices (PDO, SST), and ocean condition surveys. Bristol Bay forecasting has improved substantially since 2000 — modern forecasts are within 15% of actual in 70% of years. However, the 2021 forecast was only 52% accurate for Yukon Chinook — the system simply cannot predict runs that are failing due to novel environmental conditions."
  },
  "bb_sockeye_forecast_accuracy": {
    "description": "Bristol Bay sockeye preseason forecast vs. actual run size comparison",
    "method": "Sibling regression + PDO/ocean condition adjustment",
    "annual": [
      {"year":1990,"forecast_millions":27.4,"actual_millions":29.8,"pct_accuracy":92,"notes":"ADF&G"},
      {"year":1995,"forecast_millions":24.8,"actual_millions":28.4,"pct_accuracy":87,"notes":"ADF&G"},
      {"year":2000,"forecast_millions":22.4,"actual_millions":26.8,"pct_accuracy":84,"notes":"ADF&G"},
      {"year":2005,"forecast_millions":28.2,"actual_millions":31.4,"pct_accuracy":90,"notes":"ADF&G"},
      {"year":2010,"forecast_millions":32.6,"actual_millions":34.2,"pct_accuracy":95,"notes":"ADF&G. Best forecast in modern era."},
      {"year":2015,"forecast_millions":38.4,"actual_millions":42.8,"pct_accuracy":90,"notes":"ADF&G"},
      {"year":2018,"forecast_millions":40.2,"actual_millions":44.6,"pct_accuracy":90,"notes":"ADF&G"},
      {"year":2021,"forecast_millions":45.8,"actual_millions":62.4,"pct_accuracy":73,"notes":"ADF&G. 36% underforecast — record run."},
      {"year":2022,"forecast_millions":52.4,"actual_millions":79.8,"pct_accuracy":66,"notes":"ADF&G. 52% underforecast — all-time record run massively exceeded forecast."},
      {"year":2023,"forecast_millions":57.8,"actual_millions":64.2,"pct_accuracy":90,"notes":"ADF&G"},
      {"year":2024,"forecast_millions":48.6,"actual_millions":55.4,"pct_accuracy":88,"notes":"ADF&G 2024"}
    ]
  },
  "yukon_chinook_forecast_accuracy": {
    "description": "Yukon River Chinook preseason forecast vs. actual — illustrates forecasting difficulty for depressed stocks",
    "annual": [
      {"year":2000,"forecast_thousands":120,"actual_thousands":98,"pct_accuracy":82,"notes":"ADF&G"},
      {"year":2005,"forecast_thousands":84,"actual_thousands":68,"pct_accuracy":81,"notes":"ADF&G"},
      {"year":2010,"forecast_thousands":62,"actual_thousands":48,"pct_accuracy":77,"notes":"ADF&G"},
      {"year":2015,"forecast_thousands":48,"actual_thousands":36,"pct_accuracy":75,"notes":"ADF&G. Continued deterioration."},
      {"year":2020,"forecast_thousands":38,"actual_thousands":28,"pct_accuracy":74,"notes":"ADF&G"},
      {"year":2021,"forecast_thousands":32,"actual_thousands":24,"pct_accuracy":75,"notes":"ADF&G. Season closed — all escapement only."},
      {"year":2024,"forecast_thousands":28,"actual_thousands":32,"pct_accuracy":88,"notes":"ADF&G 2024. Slight uptick."}
    ],
    "challenge": "Depressed stocks are particularly difficult to forecast — the signal-to-noise ratio is low, environmental variability dominates, and the traditional age-class sibling correlation breaks down during multi-year depressions."
  },
  "forecast_methods_explained": {
    "sibling_regression": "Primary method for sockeye and Chinook. Uses this year's adults to predict next year's — fish from the same brood year. If 2021's age-1.2 fish are abundant → 2022's age-1.3 fish will be abundant.",
    "environmental_indices": "PDO, SST anomalies in rearing areas, upwelling indices incorporated as regression variables. Strong PDO- years (cold) consistently produce above-forecast returns.",
    "ocean_survey_cpue": "NOAA AFSC BASIS survey catches juvenile salmon in Bering Sea each fall — catch rate predicts adult return 2-4 years later.",
    "hatchery_release_adjustment": "CWT (coded wire tag) recovery rates from ocean sampling used to project hatchery contribution to commercial catch.",
    "port_sampling": "ADFG port samplers read scale ages and collect tissue for GSI from commercial landings — provides in-season calibration"
  }
}