GFS vs ECMWF Demand-Hub Outlook: 2026-05-09T00:00Z

Executive Summary

The 15-day outlook shows significant model disagreement on a regional basis, leading to an uncertain national demand picture. While total weighted HDDs are nearly identical between models, this masks a key divergence: the GFS is colder in the Midwest while the ECMWF is colder in the East. Both models trended substantially warmer than their prior runs, a bearish shift reducing overall heating demand expectations and increasing cooling load forecasts, particularly in the GFS solution for the southern U.S.

Hub-Level Weather

The national forecast is driven by conflicting signals from the primary demand hubs:

• Midwest (Chicago, 35% weight): Models diverge significantly. The GFS is colder, projecting 115.1 HDDs versus 101.2 HDDs from the ECMWF. The largest disagreements occur in the extended range, with the GFS seeing a cold shot around May 16-18 that the ECMWF misses, while the ECMWF is warmer in the final days of the forecast.
• East (New York City, 30% weight): The roles are reversed in the East, where the ECMWF is materially colder, forecasting 78.7 HDDs compared to the GFS's 60.4 HDDs. The ECMWF projects a cooler pattern from May 12-15 and is less warm than the GFS toward the end of the period.
• South Central (Houston, 15% weight): The GFS is significantly hotter, forecasting 223.1 CDDs against the ECMWF's 169.0 CDDs. This represents a major split in the outlook for power burn demand in Texas.
• South (Atlanta, 10% weight): Similar to Houston, the GFS is much warmer, projecting 127.9 CDDs versus the ECMWF's 72.1 CDDs.

Demand Implications (Natural Gas)

As this forecast covers a shoulder season, both heating and cooling demand are factors.

• Heating Demand: The national weighted HDD totals are deceptively similar (GFS: 71.4, ECMWF: 72.0). However, this high-level agreement is a result of offsetting regional forecasts. The market should be cautious, as a shift toward either model's solution would have a material impact. Critically, both models saw a significant bearish trend, with weighted HDDs falling by 21.6 (GFS) and 23.4 (ECMWF) compared to the prior cycle.
• Cooling Demand: The GFS is more bullish on cooling demand, with a national weighted total of 54.6 CDDs versus the ECMWF's 50.0 CDDs. This difference is driven entirely by the GFS's hotter outlook across the South. Both models trended warmer/more bullish for cooling demand versus the previous run (Deltas: +16.0 GFS, +17.8 ECMWF).

Model Agreement & Caveats

Confidence in this forecast is low due to poor model agreement, especially beyond the first week.

• Convergence: The convergence score of 0.27 indicates significant disagreement between the GFS and ECMWF solutions. This is further quantified by the weighted mean absolute HDD difference per day of 2.64.
• Caveats: The primary uncertainty is which regional solution (colder GFS Midwest vs. colder ECMWF East) will prove more accurate. The very close national HDD totals should be viewed with skepticism, as they mask these underlying and significant model differences.
• Source: This analysis is based on GFS Seamless (gfs_seamless) and ECMWF IFS 0.25° (ecmwf_ifs025) models via Open-Meteo. The data is suitable for directional analysis but is not an official source for trading or settlement.