References
whitepaperNovember 15, 2026· Ilika Geospatial

Calibrated probabilistic weather intelligence for New Jersey energy affordability

Technical whitepaper: how the 51-member ECMWF ensemble is turned into calibrated, station-anchored probability distributions of daily temperature for New Jersey's utility zones, the out-of-sample accuracy of those distributions against the raw ensemble and climatology, and how a calibrated distribution acts on procurement margins, hedging, peak-day operations and household load shifting.

New Jersey's electricity bills carry the cost of uncertainty about the weather: the margin a load-serving entity holds against peak demand, the hedge it buys against heating and cooling load, and the operations it runs on peak days are all sized against a temperature forecast whose error is rarely stated and rarely verified.

This whitepaper describes a platform that turns the 51-member ECMWF ensemble into calibrated probability distributions of daily temperature for New Jersey's utility zones, on the station basis that indices settle on, verified out of sample against the raw ensemble and against climatology, and served every day. Three layers do the work: ensemble calibration per grid cell and month, station anchoring per settlement sensor and month, and daily delivery of fifty-one calibrated members in the frame a settlement calculator consumes.

Across 59 stations and 1.19 million forecast and observation pairs, the station-anchored distribution lowers the continuous ranked probability score from 2.50 degrees Celsius for the raw ensemble to 1.73, and its 80 percent band covers 77 percent of outcomes against 52 percent for the raw ensemble. At Newark the band covers 77 percent against 32 to 58 percent. The paper reports the results at the New Jersey zones, explains the method, sets out the cost-chain mechanism, and states the limits.

The full whitepaper is available as a PDF. This work was supported by the NJ AI Innovation Challenge (New Jersey Economic Development Authority, administered by Plug and Play NJ).

Want the record behind this note?

Per-station and per-lead breakdowns, verification data, and protocols are shared with prospective clients on request.