GFS vs ECMWF Demand-Hub Outlook: 2026-05-06 00z

Executive Summary

The 15-day outlook shows a major divergence between the GFS and ECMWF models, resulting in very low forecast confidence. The GFS is significantly warmer, forecasting much lower heating demand and higher cooling demand than the ECMWF. This model disagreement has widened considerably since the prior cycle, with the GFS trending warmer and the ECMWF trending substantially colder, particularly for the Midwest and East.

Hub-Level Weather

The primary driver of the national forecast divergence is the Midwest (Chicago), the largest demand hub (35% weight). Over the 15-day period, the GFS forecasts 131.2 HDDs for Chicago, while the ECMWF calls for a much colder 189.4 HDDs. The disagreement is most pronounced in the Day 8-15 period; for example, on May 18th, the GFS forecasts 7.4 CDDs while the ECMWF forecasts 20.4 HDDs.

A similar, significant split exists in the East (NYC, 30% weight), where the GFS sees 62.7 total HDDs versus 129.2 HDDs from the ECMWF. In contrast, the models show a large split on cooling demand in the South Central region (Houston), with the GFS forecasting 190.7 CDDs compared to the ECMWF's 127.0 CDDs.

Demand Implications (Natural Gas)

The conflicting model solutions present opposing implications for natural gas demand. * ECMWF: The colder European model, with 124.2 total weighted HDDs, suggests a sustained period of late-season heating demand that would be supportive for natural gas consumption. This represents a significant bullish shift from its prior run, adding 36.3 weighted HDDs. * GFS: The American model is far more bearish for heating demand, totaling only 77.5 weighted HDDs—a loss of 10.6 HDDs from the last cycle. The GFS solution transitions more quickly to summer, with 51.1 weighted CDDs versus only 24.4 in the ECMWF.

The net difference between models is substantial, with the ECMWF carrying 46.7 more weighted HDDs than the GFS over the forecast horizon.

Model Agreement & Caveats

Forecast confidence is extremely low. The convergence_score is 0.19, indicating significant disagreement on the sensible weather outlook, especially in the Day 8-15 period.

The models have trended in opposite directions since the last cycle, further eroding confidence. The GFS became warmer (less demand), while the ECMWF became significantly colder (more demand). This major divergence suggests instability in the pattern and warrants caution against placing high confidence in either model's solution beyond the first week.

This analysis is based on Open-Meteo API data, which provides point forecasts for city hubs and may not fully represent demand across broader regions.