GFS vs ECMWF Demand-Hub Outlook: 2026-05-23T00:00Z

Executive Summary

The 15-day outlook shows similar total demand between models, but with a significant divergence in composition. The GFS forecasts higher heating demand (37.5 weighted HDDs) driven by a cool Midwest, while the ECMWF is warmer overall, forecasting higher cooling demand (92.6 weighted CDDs). Both models trended significantly warmer compared to their prior runs, adding confidence to a period of above-normal cooling load, particularly in the South.

Hub-Level Weather

The primary driver of the national forecast is a major disagreement in the most heavily weighted hub, the Midwest (Chicago, 35% weight).

• Midwest (Chicago): The models are in stark opposition. The GFS forecasts a cool period requiring significant heating (48.9 cumulative HDDs). In contrast, the ECMWF shows a much warmer, cooling-dominant pattern, forecasting almost no heating (5.2 HDDs) and substantial cooling demand (100.0 CDDs).
• East (New York City): Models are in better agreement on total demand but differ on timing. ECMWF is slightly more demand-heavy overall (44.8 HDDs, 63.4 CDDs) compared to GFS (39.7 HDDs, 58.5 CDDs).
• South Central (Houston): Both models show strong cooling demand, as expected for late May. However, the GFS is notably hotter, calling for 219.2 cumulative CDDs versus 178.6 CDDs from the ECMWF.

Demand Implications (Natural Gas)

The conflicting model solutions present different risks for natural gas demand during this shoulder season period.

• GFS Scenario: The cooler Midwest pattern would support stronger residential/commercial (res/com) heating demand than typical for this time of year, offsetting some seasonal demand weakness. This is coupled with strong power burn demand from persistent heat in the South.
• ECMWF Scenario: This forecast implies weaker res/com demand nationally but stronger power burn for cooling, particularly in the high-demand Midwest and East regions.
• Run-over-Run Trend: Both models shifted significantly warmer versus the prior cycle. The GFS added +23.0 weighted CDDs while shedding -8.7 weighted HDDs. The ECMWF saw a similar, substantial trend (+15.3 CDDs, -4.9 HDDs). This consistent trend increases the probability that cooling demand will outperform prior expectations.

Model Agreement & Caveats

Confidence in this forecast is low due to significant model disagreement in key regions.

• Convergence: The national convergence_score of 0.32 indicates poor agreement between the GFS and ECMWF solutions. This is driven primarily by the large forecast spread in the Midwest.
• Caveats: The forecast for the next 15 days is highly dependent on which solution for the Midwest proves correct. Traders should be cautious until subsequent model runs show better agreement. The provided hubs are proxies for regional demand and may not capture all local weather variations. Data is sourced from Open-Meteo.