GFS vs ECMWF Demand-Hub Outlook: 2026-08-23T00:00Z

Executive Summary

The 15-day outlook shows sustained cooling demand, with the GFS forecasting a slightly higher national total than the ECMWF (159.4 vs. 155.5 weighted CDDs). The most significant development is a major bullish shift in the ECMWF, which added 27.5 TDDs cycle-over-cycle. While the models have a high national-level agreement score, they diverge significantly at the regional level, with the GFS much warmer in the South and the ECMWF showing stronger heat in the Midwest.

Hub-Level Weather

The national picture is driven by the two largest hubs, which show opposing model trends: * Midwest (Chicago, 35% weight): The ECMWF is materially warmer, forecasting 102.5 CDDs vs. the GFS's 90.7 CDDs. The divergence is most pronounced from August 29th to September 1st. * East (New York City, 30% weight): The GFS is slightly warmer, forecasting 137.8 CDDs against the ECMWF's 133.0 CDDs. * South (Houston/Atlanta, 25% combined weight): The GFS is significantly hotter across the South. In Atlanta, it forecasts 283.7 CDDs compared to just 230.2 CDDs from the ECMWF. * Northeast (Boston, 10% weight): The models show a major disagreement. The ECMWF is much warmer with 92.5 CDDs, while the GFS is cooler at 61.0 CDDs and even forecasts 2.6 HDDs in early September.

Demand Implications (Natural Gas)

Both models point to a period of above-normal cooling demand for late August and early September, which will support power burn for natural gas. The key takeaway from this cycle is the ECMWF's substantial warmer trend, adding +27.5 demand-weighted degree days. This represents a significant bullish shift versus its prior run and brings it closer to the GFS's total demand outlook. The GFS saw a minor cooling revision of -1.0 TDDs. This late-season heat provides underlying support for the front of the NG futures curve.

Model Agreement & Caveats

Model agreement is high at a national level, with a convergence score of 0.9828 (an improvement of +0.0115 from the prior cycle). However, this high-level score masks significant regional disagreements noted above. Confidence in the forecast should be tempered by these differences, particularly the GFS's heat in the South versus the ECMWF's in the Midwest. This forecast is based on deterministic runs from Open-Meteo and does not reflect the full range of outcomes present in ensemble model suites.