FORECAST GAS WEIGHTED DEGREE DAYS · SEVEN MODELS · 3,108 COUNTIES BRIEF DASHBOARD MAP EXPLORER OBSERVED STORAGE SUPPLY WEEKS 3–4 MODELS VERIFICATION METHOD

STORAGE

The EIA weekly storage report against the 5-year average, and a weather-driven estimate of the next three reports from observed and forecast degree days.

Loading
–
 

Loading…

The next three reports

Estimated net change in Lower 48 working gas for each EIA report week (Saturday to Friday), from gas-heating-weighted HDD and power-burn-weighted CDD. Days already observed use NOAA nClimGrid; the rest use the newest model run covering them. The range is ± the model's typical error.

Weather scenarios

ECMWF ENS members grouped by their North American 500-hPa pattern, each group priced in gas-weighted degree days and Bcf.

Degree days by report week

Every model's total for each EIA report week, observed days included, against the 5-year average, last year and the 1991–2020 normal.

Gas-heating HDD

Weighted to where homes and businesses heat with gas: the demand side of the storage balance.

Power-burn CDD

Weighted to where gas-fired plants burn: the summer demand side.

Inventory

Lower 48 working gas in storage this year against last year and the 5-year range.

Working gas, Lower 48

Bcf at the end of each report week. Shaded: the 5-year minimum to maximum for the same week.

By region

The latest report for each EIA storage region.

How good is the estimate?

Reported vs estimated weekly change

Each of the last 52 weeks: the change EIA reported, and what the model estimated from that week's observed degree days, fitted only on earlier weeks.

How the estimate works

Weekly net change in Lower 48 storage is regressed on that week's gas-heating-weighted HDD and power-burn-weighted CDD, with an annual cycle and a trend for the growth in production and LNG exports, over report weeks since 2021. The last eight weeks' mean residual is added back, so a persistent supply shift (a pipeline outage, an LNG plant in maintenance) is carried forward rather than ignored.

Its accuracy is scored honestly: each of the last 104 weeks is estimated from a model fitted only on the weeks before it. That is the error you should expect, and it is the range shown on each estimate. Weather explains most of the week-to-week swing, not all of it: salt-cavern cycling, holidays, reclassifications and revisions are outside any weather model. Storage data: U.S. Energy Information Administration, Weekly Natural Gas Storage Report.