GFS vs ECMWF Demand-Hub Outlook: 2026-08-30T00:00Z

Executive Summary

A major split defines the 15-day outlook, with the GFS forecasting significantly stronger late-season cooling demand than the ECMWF. The GFS model added a substantial 61.4 weighted CDDs run-over-run, projecting a total of 211.2 weighted CDDs, while the ECMWF remained stable and much cooler at 130.5 weighted CDDs. This significant divergence, particularly in the Day 8-15 period, lowers overall forecast confidence.

Hub-Level Weather

The national forecast divergence is driven primarily by the high-demand hubs in the Midwest and East, where the models show starkly different outcomes in the second week of the forecast.

• Midwest (Chicago, 35% weight): The GFS is radically hotter, forecasting 197.6 cumulative CDDs versus just 130.0 for the ECMWF. The split is most extreme from Sept 10-12, where the GFS shows temperatures in the high 80s to low 90s, while the ECMWF collapses into the 50s and 60s, even generating 17.3 cumulative HDDs.
• East (New York City, 30% weight): A similar pattern emerges, with the GFS projecting 183.9 cumulative CDDs against the ECMWF's 82.3. The GFS maintains temperatures in the 80s through Sept 13, whereas the ECMWF sees a sharp cooldown into the low 60s.
• South Central (Houston, 15% weight): While both models show consistent cooling demand, the GFS is materially hotter throughout the forecast, accumulating 323.0 CDDs to the ECMWF's 251.3.

Demand Implications (Natural Gas)

The two models imply vastly different scenarios for Natural Gas demand.

• The GFS outlook, with 211.2 weighted CDDs, points to robust power burn continuing well into September. This would be a bullish catalyst for NG futures, supporting prices by tightening the supply/demand balance and slowing the pace of storage injections.
• Conversely, the ECMWF's forecast of 130.5 weighted CDDs and 11.6 weighted HDDs suggests a much quicker end to summer demand. This scenario would be bearish relative to the GFS, allowing storage inventories to build more rapidly ahead of winter.

Given the magnitude of the GFS's bullish run-over-run shift (+61.4 wCDD), the market will likely treat it with skepticism until further model runs provide confirmation.

Model Agreement & Caveats

Forecast confidence is low due to poor model agreement, which has worsened since the prior cycle.

• Convergence: The national convergence score is 0.53, a low value indicating significant disagreement. This score deteriorated by 0.12 from the previous run, highlighting the growing uncertainty.
• Divergence Driver: The disagreement is almost entirely concentrated in the Day 8-15 period. While common, the difference in temperature regimes (sustained heat vs. sharp cooldown) is exceptionally large.
• Source: This analysis is based on deterministic GFS and ECMWF model runs from Open-Meteo. It does not incorporate ensemble guidance, which would provide a better view of the range of potential outcomes. Caution is warranted until the models come into better alignment.