GFS vs ECMWF Demand-Hub Outlook: 2026-09-09T00:00Z

Executive Summary

The 15-day outlook shows national cooling demand totals that are nearly identical between models, with the GFS forecasting 132.1 weighted CDDs and the ECMWF 132.7. However, this top-line agreement masks significant regional disagreements, particularly a much warmer ECMWF in the Midwest versus a hotter GFS across the South. The GFS model has trended significantly cooler cycle-over-cycle (-31.5 CDDs), introducing early-season heating demand that the more stable ECMWF does not yet see.

Hub-Level Weather

The two most heavily weighted hubs show major forecast divergence, driving the national picture:

• Midwest (Chicago, 35% weight): This is the primary point of contention. The ECMWF is substantially warmer, projecting 105.5 cumulative CDDs versus just 63.7 from the GFS. The GFS introduces a late-period cooldown, generating 2.8 HDDs where the ECMWF sees none.
• East (New York City, 30% weight): The models show better agreement here, though the GFS projects slightly more cooling demand (118.0 CDDs) compared to the ECMWF (104.1 CDDs).
• South Central (Houston, 15% weight): The GFS is materially hotter, forecasting 307.5 CDDs against the ECMWF's 256.1. This warmer GFS solution in the South helps offset its cooler Midwest forecast in the national weighted average.
• Northeast (Boston, 10% weight): The GFS is significantly colder in the back half of the forecast, calling for 32.8 HDDs versus only 19.8 from the ECMWF.

Demand Implications (Natural Gas)

The forecast for mid-to-late September remains supportive of late-season cooling demand for Natural Gas power burn, though below peak summer levels. The key takeaway is the significant shift in the GFS model compared to its prior run.

• Run-over-Run Shift: The GFS saw a major trend cooler, shedding 31.5 weighted CDDs and adding 4.3 weighted HDDs from its previous cycle. The ECMWF was far more stable, with a minor cooling adjustment of -4.8 CDDs.
• Demand Profile: While total CDDs are similar, the GFS solution implies a sharper end to summer demand in the Midwest and Northeast, replaced by the first hints of heating demand. The ECMWF maintains stronger cooling demand in the Midwest for longer. This divergence presents conflicting signals for the NG contract balance as the market transitions from cooling to heating season.

Model Agreement & Caveats

Confidence in this forecast is moderate and has decreased since the last model cycle.

• Convergence: The model convergence score is 0.78, reflecting the significant regional disagreements noted above. This score has decreased by 0.08 from the prior run, indicating the models are moving further apart, which reduces overall forecast confidence.
• Caveats: The primary divergence occurs in the 8-15 day period, which carries inherently lower forecast accuracy. The significant cycle-over-cycle shift in the GFS suggests instability in its long-range solution. Traders should be cautious of the GFS's cooler trend until the ECMWF begins to show similar verification. This analysis is based on raw Open-Meteo model data.