GFS vs ECMWF Demand-Hub Outlook: 2026-05-19T00:00Z

Executive Summary

The 15-day outlook presents a highly divergent view between models, lowering forecast confidence. The GFS is materially colder, forecasting more national heating demand (41.2 weighted HDDs) than the ECMWF (30.3 HDDs), driven by a much cooler outlook in the Midwest. This is partially offset by the GFS forecasting significantly more heat and cooling demand in the South. Notably, the GFS has trended sharply cooler since its prior run, shedding 35.8 weighted CDDs and introducing significant forecast uncertainty.

Hub-Level Weather

The national forecast is primarily shaped by two key hubs with significant model disagreement:

• Midwest (Chicago, 35% weight): This is the main point of contention. The GFS forecasts a cool period with 57.8 cumulative HDDs and only 11.2 CDDs. In stark contrast, the ECMWF expects a much warmer pattern, with only 42.0 HDDs and a substantial 47.4 CDDs, implying a stronger start to cooling season.
• South Central (Houston, 15% weight): Both models forecast strong cooling demand, but the GFS is significantly hotter. It projects 234.0 cumulative CDDs, a full 56.2 CDDs more than the ECMWF's 177.8 CDDs.
• East (New York City, 30% weight): The models are in closer agreement here. The GFS is slightly cooler (39.8 HDDs) than the ECMWF (26.1 HDDs), while cumulative cooling demand is nearly identical (70.9 GFS CDDs vs. 68.1 ECMWF CDDs).

Demand Implications (Natural Gas)

The conflicting model solutions imply different trajectories for natural gas demand.

• The GFS solution, with its persistent Midwest chill, suggests lingering residential/commercial heating demand that would be unseasonably strong for late May.
• The ECMWF points to a more typical seasonal transition, with heating demand fading in favor of growing power burn for air conditioning, particularly in the Midwest.
• The significant bearish revision in the GFS's cooling demand forecast (-35.8 weighted CDDs run-over-run) signals a major reduction in expected power generation load compared to 12 hours prior, a key development for the market.

Model Agreement & Caveats

Confidence in this 15-day forecast is low.

• Convergence: The model convergence score is 0.43, indicating poor agreement and a high degree of uncertainty. The primary drivers of this divergence are conflicting temperature patterns in the Midwest and South Central demand hubs.
• Run-over-Run Shift: The GFS model experienced a substantial shift colder/less hot versus its prior cycle (delta_weighted_cumulative_cdd_gfs: -35.8). The ECMWF also trended cooler (delta_weighted_cumulative_hdd_ecmwf: -4.1). Large revisions like this reduce forecast reliability and suggest the potential for further significant changes in subsequent runs.
• Source: This analysis is based on GFS and ECMWF model data from Open-Meteo, using a 65°F degree day base.