GFS vs ECMWF Demand-Hub Outlook: 2026-05-21 00Z

Executive Summary

The 15-day outlook shows a transition from light, lingering heating demand in the North to building cooling demand across most of the U.S. Overall, the GFS model projects slightly higher weather-driven gas demand (130.1 total weighted degree days) compared to the ECMWF (121.9 TDDs). However, the GFS made a significant warmer shift from its prior run, reducing confidence in its forecast, while the ECMWF remained more stable.

Hub-Level Weather

The national picture is primarily driven by the high-weighted Midwest and East hubs, which are also the source of the most significant model disagreement.

• Midwest (Chicago, 35% weight): This is the key battleground. While cumulative degree days are similar, the models diverge significantly in the 8-15 day period. For example, on June 3rd, the GFS forecasts cooling demand (8.6 CDD) while the ECMWF sees a return to heating demand (7.9 HDD).
• East (New York City, 30% weight): Both models show a cool start followed by a warm-up. The GFS is more aggressive with the late-period heat, forecasting 12.1 CDDs on May 29th versus only 3.1 CDDs from the ECMWF.
• South Central / South (Houston/Atlanta, 25% combined weight): The GFS is consistently hotter across the South, driving its higher national cooling demand total. The GFS projects 203.4 cumulative CDDs in Houston, a notable 28.7 CDDs more than the ECMWF's 174.8 CDD forecast.

Demand Implications (Natural Gas)

The forecast implies below-normal demand for late May as heating load quickly diminishes and is replaced by moderate, not-yet-extreme cooling demand.

• Total Demand: The GFS is forecasting 30.6 weighted HDDs and 99.5 weighted CDDs. The ECMWF is slightly less demanding, with 28.8 weighted HDDs and 93.1 weighted CDDs.
• Run-over-Run Shifts: The GFS shows significant run-to-run instability. It lost 5.8 weighted HDDs and 5.2 weighted CDDs compared to its previous cycle, a major bearish shift. The ECMWF was far more stable, with a minor change of -0.8 HDDs and +1.9 CDDs. This suggests the market may place more trust in the ECMWF solution this cycle.

Model Agreement & Caveats

Confidence in this forecast should be moderate, particularly beyond the first week, due to key model disagreements and GFS instability.

• Convergence: The model convergence score is 0.4761, indicating moderate agreement. This agreement has slightly worsened from the prior cycle (delta: +0.0399).
• Divergence: Disagreement increases materially after day 7, centered on the pattern evolution in the Midwest and East. The stability of the ECMWF provides a stronger anchor, but the GFS's large shift introduces forecast uncertainty.
• Source Limitations: This analysis is based on deterministic model runs for specific hub locations from Open-Meteo. It does not incorporate ensemble guidance, which would provide a better sense of probabilistic outcomes and forecast confidence.