GFS vs ECMWF Demand-Hub Outlook: 2026-05-22 00z

Executive Summary

The 15-day outlook shows significant disagreement between the GFS and ECMWF models, resulting in a low confidence forecast. The GFS is materially colder and slightly less hot on a national basis (35.5 weighted HDDs / 78.3 weighted CDDs) compared to the ECMWF (23.3 weighted HDDs / 87.8 weighted CDDs). This divergence is driven almost entirely by a split forecast for the Midwest, where the GFS sees late-season heating demand and the ECMWF projects a much warmer, cooling-demand-driven pattern.

Hub-Level Weather

The primary driver of the national forecast divergence is the Midwest (Chicago), our most heavily weighted hub (0.35): * GFS: 47.3 HDDs and 23.9 CDDs. * ECMWF: 11.6 HDDs and 88.5 CDDs. The GFS indicates periods of below-normal temperatures, while the ECMWF is significantly warmer, particularly in the 8-15 day period.

In the South, the GFS is consistently hotter, projecting much stronger cooling demand: * South Central (Houston): GFS forecasts 251.0 CDDs vs. 175.2 CDDs for the ECMWF. * South (Atlanta): GFS forecasts 163.4 CDDs vs. 91.4 CDDs for the ECMWF.

The models show better agreement in the East (New York City), with cumulative HDDs (GFS: 39.2, ECMWF: 42.3) and CDDs (GFS: 52.1, ECMWF: 59.0) remaining relatively close.

Demand Implications (Natural Gas)

The competing model solutions present different narratives for shoulder-season gas demand. * The GFS view implies a broader support for demand, combining lingering heating load in the Midwest and Northeast with robust cooling demand across the South. * The ECMWF view points to a more rapid transition to a summer demand profile, with minimal heating load and a stronger concentration of cooling demand building in the Midwest and East.

Compared to the previous cycle, both models trended slightly colder (GFS: +1.1 weighted HDDs, ECMWF: +1.6 weighted HDDs). However, the GFS also trended significantly hotter in cooling regions (+4.2 weighted CDDs), while the ECMWF trended slightly less hot (-1.2 weighted CDDs), widening the overall model disagreement.

Model Agreement & Caveats

Forecast confidence is low. The convergence score of 0.37 indicates significant model disagreement, a decrease in agreement from the prior run (delta: -0.14). The large forecast spread in the high-impact Midwest hub suggests caution is warranted, and the market will likely await further model runs for a clearer signal.

This analysis is based on Open-Meteo's implementation of the gfs_seamless and ecmwf_ifs025 models. The demand-weighting is based on a simplified five-hub system and may not capture all regional nuances. All degree-day calculations use a standard 65°F base.