Skip to main content

Features

WSSI-US Features are the derived inputs that underpin both Signals and Events.

They provide clean, research-ready representations of weather-driven demand dynamics, forecast behaviour, and model disagreement across US regions.

Available Features​

hdd_cdd(region_id=..., region_type=...)​

Population-weighted Heating Degree Days (HDD) and Cooling Degree Days (CDD), in Fahrenheit, with p10/p50/p90 percentiles. Provides a standardised view of temperature-driven demand across US regions. Its as_of start is 2026-02-23 for all regions; the 2026-09-11 start applies only to the EIA-region rows (R31-R35) of hdd_observed and hdd_climatology.

forecast_shock()​

Run-to-run forecast shock features capturing how weather forecasts change between successive model runs. Useful for identifying periods where expectations are shifting rapidly.

model_disagreement()​

Cross-model disagreement features comparing GEFS and GFS forecast outputs. Elevated disagreement indicates greater uncertainty in near-term weather outcomes.

features()​

Composite WSSI-US feature matrix combining the underlying weather and demand metrics into a structured dataset suitable for modelling and quantitative research.

forecast_skill()​

Forecast skill of GEFS, GFS and ECMWF degree-day forecasts against NOAA CPC observed HDD/CDD, by model, forecast run (forecast_cycle: 0/6/12/18 UTC), lead time and region (CONUS plus seven regions). Columns are HDD/CDD MAE, RMSE and bias, hdd_mae_30d_rolling, skill_trend, and skill_score (100 - 5 x combined MAE, clipped to 0-100). Since model 4.0 (5 October 2026) each region is scored against CPC state values weighted by the population of each state inside the region's forecast box, and CONUS against CPC's own CONUS series. A row is knowable two days after the target date. It measures forecast accuracy, not storage-surprise skill; read bias by forecast_cycle.

Typical use cases​

  • Custom signal development using weather-driven demand features
  • Quantitative modelling of weather shock sensitivity
  • Point-in-time research on weather-driven storage pressure
  • Assessing forecast reliability across regions and seasons

WSSI-US Features are designed to integrate directly into quantitative research workflows.