GFS vs ECMWF Demand-Hub Outlook: 2026-09-20 00Z

Executive Summary

The 15-day outlook presents significant model disagreement on weather-driven demand, particularly for heating load in the eastern U.S. The GFS model forecasts a colder and therefore more demand-supportive pattern for the East and a hotter South, while the ECMWF is notably colder in the Midwest. Overall, the GFS sees a higher national demand profile, but low model agreement warrants caution.

Hub-Level Weather

The national forecast is a tale of two distinct regional outlooks, driven by the highest-weighted demand hubs.

• Midwest (Chicago): The heaviest weighted hub (0.35) shows the ECMWF forecasting a colder pattern, with a cumulative 48.4 HDDs versus the GFS's 40.6 HDDs. The divergence grows significantly in the Day 8-12 period.
• East (New York City & Boston): In a sharp reversal of the Midwest pattern, the GFS is substantially colder across the East. In New York (0.30 weight), GFS projects 70.0 HDDs compared to just 44.9 HDDs from the ECMWF. This trend is amplified in Boston, where GFS sees 122.6 HDDs to the ECMWF's 82.7 HDDs.
• South (Houston & Atlanta): Both models forecast persistent cooling demand, but the GFS is consistently hotter. In Houston (0.15 weight), GFS forecasts 311.5 CDDs against 276.0 CDDs from the ECMWF, a material difference for power-sector gas burn.

Demand Implications (Natural Gas)

This shoulder-season forecast contains a mix of heating and cooling demand, with models painting different pictures for Natural Gas consumption.

• GFS Scenario (Higher Demand): The GFS is more bullish for NG prices. Its combination of significantly colder temperatures in the high-population East and hotter weather in the power-burn-heavy South results in a stronger national demand forecast (47.5 weighted HDDs, 65.1 weighted CDDs). This cycle's run trended colder on heating demand (+9.6 weighted HDDs), adding to its bullish tilt.
• ECMWF Scenario (Lower Demand): The ECMWF presents a less supportive outlook, with its milder solution for the East and South outweighing its colder Midwest forecast. The national total is a less impressive 38.6 weighted HDDs and 55.7 weighted CDDs. While this run added significant cooling demand (+10.0 weighted CDDs), it shed 4.0 weighted HDDs versus the prior cycle, diverging from the GFS.

Model Agreement & Caveats

Confidence in this forecast should be low.

• Convergence: The model convergence score is 0.31, indicating significant disagreement. This is primarily driven by the opposing solutions for the timing and intensity of cold air across the Midwest and Eastern U.S. in the 6-15 day period.
• Run-over-Run Shift: The models diverged materially in this cycle. GFS trended colder (more HDDs) while ECMWF trended warmer (fewer HDDs). This increases uncertainty and suggests the market should wait for further runs to confirm a direction.
• Source: This analysis is based on GFS and ECMWF model data from Open-Meteo, using a standard 65°F base for degree day calculations.