GFS vs ECMWF Demand-Hub Outlook: 2026-05-04 00Z

Executive Summary

The 15-day outlook presents a highly uncertain and conflicting demand picture. The ECMWF model is materially colder, forecasting 110.5 demand-weighted HDDs, while the GFS is warmer at 96.9 weighted HDDs. Conversely, the GFS is hotter, projecting 38.3 weighted CDDs against the ECMWF's 33.1. This divergence is driven by a colder Midwest/East in the ECMWF and a much hotter South in the GFS. Confidence is low due to a poor convergence score (0.21) and significant, opposing model shifts from the prior cycle.

Hub-Level Weather

The national forecast is shaped by major disagreements in the highest-weighted demand hubs:

• Midwest (Chicago, 35% wt): This is the primary point of contention for heating demand. The ECMWF is significantly colder, with a cumulative 179.8 HDDs compared to the GFS's 151.3 HDDs. The ~28.5 HDD difference is most pronounced in the Day 7-12 period (May 10-15).
• East (New York City, 30% wt): Similar to the Midwest, the ECMWF projects a colder scenario with 105.1 HDDs versus 85.3 HDDs from the GFS. The models diverge most significantly after May 12th.
• South Central (Houston, 15% wt): For cooling demand, the models are flipped. The GFS is substantially hotter, forecasting 196.5 CDDs over the 15-day period, a full 43.3 CDDs more than the ECMWF's 153.2 CDDs. The GFS shows much stronger and more sustained heat in the latter half of the forecast.

Demand Implications (Natural Gas)

The competing model solutions imply different risks for natural gas consumption during this shoulder season month.

• Heating Demand: The colder ECMWF solution would support late-season residential and commercial (Res/Com) demand, particularly in the Midwest and East. This would be considered a bullish heating signal for early May.
• Cooling Demand: The hotter GFS outlook implies a stronger-than-expected start to cooling season demand in the South. This would drive higher gas-for-power burn to meet air conditioning load, a bullish signal for the power sector.
• Overall: The market is faced with conflicting signals. The net demand impact is ambiguous, hinging on whether late-heating or early-cooling materializes more strongly.

Model Agreement & Caveats

Confidence in the 15-day outlook should be low.

• Convergence: The model agreement is poor, with a convergence score of 0.21. The weighted average absolute difference between models is a notable 3.67 HDDs per day.
• Run-over-Run Changes: Forecast instability is a major concern. The GFS trended significantly colder versus its prior run (+11.7 weighted HDDs), while the ECMWF shifted much warmer (-19.5 weighted HDDs). These large, opposing trends indicate that the models are struggling to resolve the medium-range pattern, significantly reducing forecast reliability.
• Source: This analysis is based on publicly available GFS and ECMWF model data via Open-Meteo. Forecast accuracy naturally degrades beyond 7-10 days.