GFS vs ECMWF Demand-Hub Outlook: 2026-06-04T00:00Z

Executive Summary

The 15-day outlook shows a significant divergence between models, with the ECMWF forecasting much stronger national cooling demand than the GFS. This split is driven almost entirely by a conflicting forecast for the Midwest in the 8-15 day period. Confidence in the overall demand signal is low, reflected in a poor convergence score (0.69), which has worsened since the prior model cycle.

Hub-Level Weather

The primary source of model disagreement is the Midwest (Chicago, 35% weight). While models are in loose agreement for the first week, they diverge sharply thereafter. The ECMWF projects sustained summer heat, accumulating 178.0 CDDs, whereas the GFS shows a significant cool-down that even generates minor heating load (11.0 HDDs) and accumulates only 74.9 CDDs. This represents a massive 103 CDD difference in a key demand region.

Elsewhere: * East (New York City, 30% weight): Models are in close agreement on the 15-day cumulative demand, with the GFS at 159.2 CDDs and the ECMWF at 158.6 CDDs. * South Central (Houston, 15% weight): The GFS is notably hotter than the ECMWF, forecasting 288.7 CDDs versus the ECMWF's 248.8 CDDs.

Demand Implications (Natural Gas)

The total demand-weighted outlook presents two very different scenarios for natural gas consumption. The ECMWF is significantly more bullish, calling for 172.1 weighted CDDs, implying robust and sustained power burn for cooling. The GFS is far more bearish at 145.6 weighted CDDs, with its forecast for a late-period Midwest cool-down suggesting a sharp drop-off in weather-driven demand.

The GFS run trended substantially cooler versus its prior cycle, shedding 27.0 weighted CDDs. The ECMWF was more stable, losing only 4.6 weighted CDDs. This bearish trend in the GFS, combined with the model split, injects significant uncertainty into the medium-range demand forecast.

Model Agreement & Caveats

Confidence in this forecast is low. The model convergence score is 0.69, indicating poor agreement, particularly in the critical 8-15 day range. This score has decreased by 0.13 since the prior run, signaling that the models are trending apart, not toward a consensus solution.

The significant disagreement in the high-weight Midwest hub is the primary reason for the low confidence. Traders should treat the medium-range portion of this forecast with extreme caution until the models begin to converge on a solution for the Midwest pattern.

Data Source: Open-Meteo, using a 65°F degree day base.