Skip to content

Enrichment ​

Enrichment features compute derived data from the raw model forecasts. Each feature is independently toggleable and publishes to its own Home Assistant device per location.

Overview ​

FeatureDefaultDescription
ConsensusenabledCross-model aggregation with spread and agreement metrics
AlertsenabledWeather alerts derived from model data
DerivedenabledBeaufort scale, wind chill, comfort index, WMO descriptions
TrendsdisabledTemperature/wind/precipitation trend directions and timing
IndicesdisabledActivity scores, degree days, VPD
HistorydisabledForecast accuracy tracking and model weighting

Consensus ​

Aggregates forecasts across all models for a location into a single consensus value per parameter and horizon.

json
{
  "Enrichment": {
    "Consensus": {
      "Enabled": true,
      "Method": "Median",
      "TrimPercent": 0.1
    }
  }
}
OptionDefaultDescription
Method"Median"Aggregation method: "Median" or "TrimmedMean"
TrimPercent0.1Fraction to trim from each end when using TrimmedMean

Publishes to njord/{location}/consensus/h{N} with additional metadata: _spread (range), _agreement (ratio of models within threshold), _models_used (count), IQR, outlier detection, and confidence intervals.

Alerts ​

Generates weather alerts as binary sensors with severity levels (None, Yellow, Orange, Red) based on configurable thresholds.

json
{
  "Enrichment": {
    "Alerts": {
      "Enabled": true,
      "FrostThreshold": 0.0,
      "HeatThresholds": [30, 35, 40],
      "StormGustThreshold": 16.7,
      "HeavyRainHourlyThreshold": 10.0,
      "HeavyRainDailyThreshold": 25.0,
      "PressureDropThreshold": 5.0,
      "CapeThreshold": 1000.0,
      "ThunderstormPrecipThreshold": 5.0,
      "ThunderstormGustThreshold": 15.0
    }
  }
}

Alert types: Frost, Heat, Storm, HeavyRain, UV, Fog, Snow, PressureDrop, Thunderstorm

Each alert publishes as a binary_sensor with attributes:

  • severity: None, Yellow, Orange, or Red
  • confidence: how many models agree
  • Type-specific fields (e.g. expected_low, earliest_frost, models_agreeing)

TIP

Thresholds use metric units: temperatures in Celsius, wind speed in m/s, rain in mm, pressure in hPa, CAPE in J/kg.

Derived ​

Computes human-readable interpretations from raw forecast data.

json
{
  "Enrichment": {
    "Derived": {
      "Enabled": true
    }
  }
}

Per-horizon values:

  • beaufort: wind speed on the Beaufort scale (0 to 12)
  • wind_chill: perceived temperature in Celsius
  • dewpoint_comfort: comfort category based on dew point (e.g. "Comfortable", "Humid")
  • wmo_description: human-readable weather description from WMO code

Scalar values:

  • diurnal_amplitude: daily temperature range in Celsius
  • sunshine_pct: percentage of daylight hours with sunshine
  • inversion: temperature inversion detected (boolean)

Analyzes how forecast parameters change over time.

json
{
  "Enrichment": {
    "Trends": {
      "Enabled": false
    }
  }
}

Publishes to njord/{location}/trends:

  • Parameter trend directions for temperature, wind, precipitation, cloud cover
  • Weather change description (from WMO code transitions)
  • Precipitation timing: starts-in / ends-in hours
  • Temperature extrema timing: max-in / min-in hours
  • Consensus stability label and ratio
  • Predictability decay rate and reliable-hours estimate

Indices ​

Activity-oriented scores computed from forecast consensus data. Each score is 0–100 and aggregated over the next 24 hours. Scores include uncertainty envelopes (min/max/confidence) derived from model spread.

Basic configuration ​

json
{
  "Enrichment": {
    "Indices": {
      "Enabled": false,
      "Preferences": {
        "IdealOutdoorTemp": 22.0,
        "IndoorTemp": 22.0,
        "HeatSensitivity": 1.0,
        "HumiditySensitivity": 1.0,
        "WindSensitivity": 1.0,
        "RainSensitivity": 1.0
      }
    }
  }
}

Global preferences ​

These apply to all scores and all locations unless overridden.

OptionDefaultDescription
IdealOutdoorTemp22.0Peak of the outdoor comfort curve (°C). A 26 makes the outdoor score favour warmer days.
IndoorTemp22.0Reference indoor temperature for the ventilation score (°C)
HeatSensitivity1.0Multiplier for temperature penalties across all scores (0.0–5.0). Higher = more sensitive to heat/cold.
HumiditySensitivity1.0Multiplier for humidity penalties (0.0–5.0). Increase if you find humid days uncomfortable.
WindSensitivity1.0Multiplier for wind penalties (0.0–5.0). Increase if wind bothers you.
RainSensitivity1.0Multiplier for rain penalties (0.0–5.0). Increase if you want stricter rain avoidance.
RunningIdealTempLow5.0Lower bound of ideal running/cycling temperature range (°C)
RunningIdealTempHigh20.0Upper bound of ideal running/cycling temperature range (°C)
BbqMinTemp10.0Minimum temperature for the BBQ score (°C). Below this, BBQ scores drop sharply.
BbqIdealWindLow1.0Lower bound of ideal BBQ wind range (m/s)
BbqIdealWindHigh3.0Upper bound of ideal BBQ wind range (m/s)

Per-score overrides ​

Override any preference for a specific score. Useful when you want different sensitivity for different activities.

json
{
  "Enrichment": {
    "Indices": {
      "Preferences": { "HeatSensitivity": 1.0 },
      "ScoreOverrides": {
        "Running": { "HeatSensitivity": 0.7 },
        "Bbq": { "RainSensitivity": 2.0, "BbqMinTemp": 15.0 }
      }
    }
  }
}

Valid score names: Laundry, Outdoor, Running, Cycling, Bbq, Irrigation, Solar, Ventilation.

Per-location overrides ​

Override preferences for a specific location. Each location can have its own global preferences and score overrides.

json
{
  "Enrichment": {
    "Indices": {
      "Preferences": { "IdealOutdoorTemp": 22.0 },
      "LocationOverrides": [
        {
          "Location": "Lucerne",
          "Preferences": { "IdealOutdoorTemp": 24.0 },
          "ScoreOverrides": {
            "Outdoor": { "WindSensitivity": 0.8 }
          }
        }
      ]
    }
  }
}

Location names are matched case-insensitively against the configured locations.

Resolution cascade ​

For any (location, score, property), the first non-null value wins:

  1. LocationOverrides[location].ScoreOverrides[score].{Property}
  2. LocationOverrides[location].Preferences.{Property}
  3. ScoreOverrides[score].{Property}
  4. Preferences.{Property}
  5. Hardcoded default

Published values ​

  • Activity scores (0–100): laundry, outdoor, running, cycling, BBQ, irrigation, solar, ventilation
  • Score envelopes: min, max, confidence for each activity score
  • Frost: hours-until-frost, confidence
  • VPD: vapour pressure deficit value (kPa) and category

History ​

Tracks forecast accuracy over time by comparing past predictions against observations. Requires persistence to be configured.

json
{
  "Enrichment": {
    "History": {
      "Enabled": false,
      "RetentionDays": 30,
      "MinSampleSize": 48,
      "SnapshotInterval": 100
    }
  }
}
OptionDefaultDescription
RetentionDays30How many days of history to keep
MinSampleSize48Minimum data points before publishing accuracy metrics
SnapshotInterval100Events between persistence snapshots

Published values:

  • Per-model: MAE (7-day and 30-day), weight, drift
  • Seasonal best model
  • Anomaly detection (boolean + deviation in sigma)
  • Accuracy-weighted temperature

WARNING

History enrichment depends on Akka.Persistence. Make sure a persistence provider is configured and the data volume is mounted.

Open-Meteo data is licensed under CC BY 4.0.