GFS vs ECMWF Demand-Hub Outlook: 2026-05-07 00Z

Executive Summary

The 15-day outlook shows significant model disagreement on the character of natural gas demand. While total demand is similar between models, the European (ECMWF) forecast is materially colder, projecting higher late-season heating demand, particularly in the East. Conversely, the American (GFS) model is much warmer, indicating a stronger and earlier start to cooling demand in the South. Confidence in the forecast is low, underscored by a poor convergence score (0.27) and significant, opposing run-over-run changes, with the GFS trending colder and the ECMWF trending warmer than the prior cycle.

Hub-Level Weather

The national forecast is driven by disagreement in the highest-weighted demand hubs:

• Midwest (Chicago, 35% weight): Models are in fair agreement for the first week. Divergence grows significantly in the Day 10-14 period (May 17-20), where the GFS warms sharply to generate 24.0 cumulative Cooling Degree Days (CDDs), while the ECMWF remains cooler, totaling only 18.9 CDDs and retaining some heating load.
• East (New York City, 30% weight): This is the area of greatest disagreement for heating demand. The ECMWF is substantially colder, forecasting 113.8 cumulative Heating Degree Days (HDDs) versus just 78.9 HDDs from the GFS. The 35 HDD difference is concentrated in the Day 10-15 period.
• South Central (Houston, 15% weight): The GFS is significantly hotter in key southern markets. It projects 213.3 cumulative CDDs for Houston, a remarkable 63.5 CDDs more than the ECMWF's 149.8 CDDs. This differential drives the GFS's higher national cooling demand forecast.

Demand Implications (Natural Gas)

The conflicting model solutions imply different outcomes for the composition of natural gas demand during this shoulder season period.

• ECMWF View: The 96.1 weighted HDDs suggest a lingering tail to the heating season, which would support residential/commercial demand, especially in the population-dense Northeast. This scenario would likely delay the seasonal ramp-up in storage injections.
• GFS View: The stronger cooling load (60.2 weighted CDDs vs. ECMWF's 45.1) points to an earlier and more robust start for power sector gas demand ("gas-for-power"), particularly in ERCOT and the Southeast. This would provide an earlier-than-expected source of summer demand support.

The primary uncertainty is the timing of the transition from heating to cooling demand, with the GFS showing a much faster shift.

Model Agreement & Caveats

Confidence in this forecast should be low due to material model divergence and recent instability.

• Model Agreement: The overall convergence score is 0.27, indicating poor agreement between the GFS and ECMWF solutions. A score closer to 1.0 would signal high confidence. The weighted mean absolute difference in daily HDDs is a notable 2.75.
• Run-over-Run Changes: The models shifted significantly and in opposite directions compared to their prior runs. The GFS trended colder, adding 7.5 weighted HDDs. The ECMWF trended substantially warmer, removing 13.6 weighted HDDs while adding 14.0 weighted CDDs. Such opposing trends suggest significant atmospheric pattern uncertainty and reduce the reliability of either model's extended outlook.
• Source Limitations: This analysis is based on raw, grid-point data from Open-Meteo's GFS and ECMWF-IFS model implementations. It does not account for proprietary model blending or downscaling, and the demand-weighting is a simplified representation of complex regional gas consumption. Forecast skill naturally decreases beyond the 7-10 day timeframe.