GFS vs ECMWF Demand-Hub Outlook: 2026-08-29T00:00Z

Executive Summary

The 15-day outlook shows a significant model split, with the ECMWF forecasting notably stronger national cooling demand than the GFS. While the models are in fair agreement for the first week, they diverge sharply in the 8-15 day period, particularly across the high-demand Midwest and East Coast hubs. Run-over-run trends exacerbate this split, with the GFS turning significantly cooler while the ECMWF trended slightly warmer, reducing overall forecast confidence.

Hub-Level Weather

The national picture is primarily driven by major disagreements in the most heavily weighted demand centers:

• Midwest (Chicago, 35% weight): This is the largest point of contention. The ECMWF is substantially hotter, projecting 227.1 CDDs over the 15-day period, a stark contrast to the GFS's much milder forecast of 165.6 CDDs. The divergence begins around September 4th and widens dramatically thereafter.
• East (New York City, 30% weight): A similar pattern unfolds here, with the ECMWF calling for a much hotter second week. It accumulates 175.6 CDDs versus only 136.2 CDDs from the GFS.
• South Central (Houston, 15% weight): In a reversal of the national trend, the GFS is the hotter model in the South, forecasting 354.6 CDDs compared to the ECMWF's 284.8 CDDs.
• Northeast (Boston, 10% weight): The models show an extreme split. The GFS turns cool enough late in the period to generate 10.5 HDDs, while the ECMWF stays warm, accumulating 128.7 CDDs and zero heating demand.

Demand Implications (Natural Gas)

The conflicting temperature forecasts present two distinct scenarios for natural gas demand.

• The ECMWF outlook, with 210.5 total weighted CDDs, implies a robust extension of summer power burn demand well into September.
• The GFS outlook, with only 187.1 weighted CDDs, suggests a more rapid seasonal decline in cooling demand. The GFS run trended -11.5 CDDs cooler than its prior cycle, signaling a potentially bearish shift for NG futures if it verifies.
• In contrast, the ECMWF trended slightly warmer run-over-run (+1.7 CDDs), reinforcing its stronger demand outlook. The presence of GFS-forecasted HDDs in the Northeast, however small, is a notable psychological signal of seasonal change.

Model Agreement & Caveats

Confidence in the 8-15 day forecast is low due to significant model disagreement.

• The headline convergence score is 0.93, which typically indicates good agreement. However, this is misleading as it masks the major divergence in the second week of the forecast.
• Forecast confidence has decreased since the previous model cycle, with the convergence score falling by -0.06. This is driven by the GFS trending significantly cooler while the ECMWF held its ground.
• Traders should be cautious of the growing split between models. The starkly different outcomes for late-season cooling demand create considerable uncertainty for near-term gas balances.
• This analysis is based on deterministic model runs sourced from Open-Meteo. For a more robust view of probabilities, especially beyond day 7, consulting ensemble model guidance is recommended.