Natural Gas Weather — Week of November 13, 2026
GFS vs ECMWF Demand-Hub Outlook: 2026-11-13 12:00 UTC
Executive Summary
Both models agree on a cold start to the 15-day window but split sharply from Nov 22–27, where GFS drives a much colder national pattern (weighted cumulative HDD 290.8 vs ECMWF 243.3, a ~47.6 weighted-HDD gap, ~20% more implied heating load on GFS). Convergence is low (0.136) and deteriorated vs the prior cycle, so the late-period signal deserves limited confidence. The near-term (days 1–6) ECMWF-colder gap of 12.6 weighted HDD is statistically typical for mid-November and not a new development.
Hub-Level Weather
| Hub (weight) | Cum. HDD GFS | Cum. HDD ECMWF | Mean abs daily HDD diff | Key disagreement |
|---|---|---|---|---|
| Midwest (Chicago, 0.35) | 358.8 | 291.6 | 7.44 | GFS ~28°F Nov 23–24 vs ECMWF 53–55°F Nov 22–23 |
| East (NYC, 0.30) | 297.6 | 276.1 | 5.49 | Nov 24: GFS 37.9°F vs ECMWF 53.4°F |
| Northeast (Boston, 0.10) | 319.3 | 294.6 | 5.16 | Same east-coast split, Nov 24–26 |
| South Central (Houston, 0.15) | 117.9 | 50.4 | 4.95 | Only CDD hub: GFS 18.4 vs ECMWF 29.0 CDD; Nov 25: GFS 40.1°F vs ECMWF 60.7°F |
| South (Atlanta, 0.10) | 263.8 | 213.6 | 8.53 (largest) | Nov 24: GFS 33.0°F (32.1 HDD) vs ECMWF 58.7°F (6.3 HDD) |
- Models track closely through ~Nov 19–21; the divergence is a days-10–15 problem, where GFS amplifies a cold trough into the Midwest, East, and Deep South while ECMWF holds a milder ridge.
- Chicago, carrying the largest demand weight, shows the biggest absolute HDD gap (+67.2 GFS-colder) and drives the national split.
- Atlanta shows the largest day-to-day disagreement (8.53 HDD), the classic Gulf-front timing battleground.
Demand Implications (Natural Gas)
- Weighted 15-day heating load: GFS 290.8 HDD vs ECMWF 243.3 — a 47.6 weighted-HDD spread, material for natural gas balance positioning into late November.
- Days 1–6 demand is the higher-confidence piece: weighted HDD 102.7 (GFS) vs 115.4 (ECMWF), a cold start for Chicago and the East; the 12.6 HDD gap here is the models' normal seasonal bias, not a demand signal.
- Cooling demand is negligible nationally (weighted CDD 2.8 GFS / 4.4 ECMWF); Houston is the only hub with meaningful CDD, and ECMWF keeps Gulf warmth (and lower heating load) through Nov 24.
- The bull case for NG heating demand rests on the GFS scenario verifying; the ECMWF scenario implies a much softer late-month load. See the weather model stream for updates.
Model Agreement & Caveats
- Convergence is weak: score 0.136, down 0.026 vs the prior cycle, with weighted mean absolute daily HDD divergence rising +1.19 to 6.36. Agreement is acceptable in days 1–8 and poor beyond, where forecast skill decays anyway — lean on the near-term signal, treat days 10–15 as scenario spread.
- Spread significance (status: ok): for days 1–6, today's ECMWF-minus-GFS spread is 12.64 weighted HDD vs a seasonal expectation of 13.28 — an anomaly of −0.63 (z = −0.13, 13th percentile), labeled typical. This is the models' usual October–February ECMWF cold bias, not news. Verification (n=97) shows ~40.6% of past days-1–6 anomalies were ultimately realized (5-day realized lag), statistically significant at p = 0.003 — but this describes historical windows and is not backtested as a forecast; it should not be converted into a corrected HDD number.
- Run-over-run: ECMWF's weighted cumulative HDD fell ~25.0 vs the prior cycle (prior ECMWF 00Z), i.e., ECMWF trended warmer even as model disagreement widened — the source of the "material change" report trigger (change 50.8). No prior GFS cycle was available for comparison.
- Data caveats: the feed flags the GFS 12Z run as not yet ingested, so GFS figures may reflect the earlier cycle; Open-Meteo's ECMWF historically runs colder than observed stations (Oct–Feb), which flatters the near-term ECMWF HDD. These figures are demand-weighted point forecasts at five hubs, not a full consumption model, and say nothing about Henry Hub price direction.