GFS vs ECMWF Demand-Hub Outlook: 2026-08-17 00Z

Executive Summary

The 15-day outlook shows a material split between models, with the ECMWF forecasting stronger cooling demand (166.5 wCDD) than the GFS (157.8 wCDD). This national difference is driven almost entirely by a major disagreement in the Northeast, where the GFS sees a significant cool-down that the ECMWF does not. Both models trended cooler compared to their prior runs, reducing overall demand expectations.

Hub-Level Weather

The national demand picture is primarily driven by the Midwest (Chicago, 35% weight) and the East (New York City, 30% weight). While models are in general agreement in these key hubs, a significant divergence exists for the Northeast:

• Northeast (Boston): This is the main point of disagreement. The GFS forecasts a much cooler period, accumulating 8.8 HDDs and only 56.8 CDDs. In contrast, the ECMWF sees no heating demand and forecasts nearly double the cooling load at 110.4 CDDs. The GFS introduces cool periods around Aug 21-23 and again toward the end of the month.
• Midwest (Chicago): Models are in closer agreement, with the ECMWF slightly warmer (96.4 CDD) than the GFS (93.8 CDD).
• South (Houston & Atlanta): The ECMWF is warmer in the high-demand Houston hub (353.8 CDD vs. 341.6 CDD for GFS). Conversely, the GFS is slightly warmer in Atlanta.

Demand Implications (Natural Gas)

The late-summer forecast is dominated by cooling demand, which drives power burns for Natural Gas.

• The ECMWF's warmer outlook, particularly its rejection of a Northeast cool-down, implies a moderately more bullish scenario for NG futures compared to the GFS.
• The GFS solution would significantly reduce late-August power demand in the Northeast, a bearish catalyst if it verifies.
• Run-over-run, the demand picture has weakened. The GFS shed a significant 12.3 weighted CDDs from its prior forecast, while the ECMWF trimmed a more modest 4.3 wCDDs. This bearish trend in the models is a headwind for prices.

Model Agreement & Caveats

The overall model agreement is high, with a convergence score of 0.94. However, this top-line number masks the significant regional disagreement in the Northeast, which lowers forecast confidence.

• Run-over-Run Shift: Confidence has decreased from the prior cycle, with the convergence score falling by 0.04. This is due to the GFS trending much cooler than the ECMWF, widening the gap between the two solutions.
• Source: This analysis is based on GFS and ECMWF model data aggregated by Open-Meteo. The demand weighting is a simplified representation of U.S. natural gas consumption hubs.