GFS vs ECMWF Demand-Hub Outlook: 2026-05-15T00:00Z

Executive Summary

The latest 15-day outlook shows a significant divergence between models, driven by a major bullish shift in the GFS. The GFS model added substantial cooling demand (+30.4 weighted CDDs) versus its prior run, forecasting significant late-period heat in the East and South. The ECMWF remains largely unchanged and presents a more moderate, lower-demand scenario. Confidence in the forecast is low (convergence score: 0.43), warranting caution against overweighting the GFS's warmer trend.

Hub-Level Weather

The national forecast divergence is primarily driven by the two largest demand hubs, the Midwest and the East, particularly in the 8-15 day period.

• Midwest (Chicago, 35% weight): Models show a volatile and conflicting picture. The GFS forecasts both more heating demand (56.2 cumulative HDDs vs. 30.5 for ECMWF) due to late-period cold shots, while the ECMWF forecasts more consistent warmth and higher cooling demand (58.9 cumulative CDDs vs. 37.4 for GFS).
• East (New York City, 30% weight): This is a key driver of the GFS's warm shift. The GFS forecasts significantly more cooling demand than the ECMWF (97.7 cumulative CDDs vs. 58.4), projecting a strong heat event from May 25th onward that is absent in the ECMWF.
• South Central (Houston, 15% weight): Both models show consistent cooling demand, but the GFS is materially hotter throughout the period, totaling 238.0 cumulative CDDs versus 187.8 for the ECMWF.

Demand Implications (Natural Gas)

The competing forecasts present two distinct scenarios for natural gas power burn during this shoulder season.

• GFS Scenario (Higher Demand): The substantial increase in forecasted CDDs, especially in the East and South, implies a significant upside risk for power burn to meet air conditioning load. If the GFS verifies, it would be a bullish catalyst for natural gas demand, far outweighing the minor heating demand it projects for the Midwest.
• ECMWF Scenario (Moderate Demand): The ECMWF suggests a more balanced and seasonally normal demand profile. Its forecast implies moderate cooling demand that is less likely to stress the grid or drive significant incremental gas consumption compared to the GFS outlook.

The GFS's +30.4 run-over-run increase in weighted CDDs is the most significant feature of this cycle. The ECMWF's stability provides a less volatile anchor, suggesting the GFS may be an outlier.

Model Agreement & Caveats

Forecast confidence is low. The multi-hub convergence score is 0.43, indicating poor agreement between the GFS and ECMWF solutions.

• Divergence Point: The models are in reasonable agreement for the first 5-7 days before diverging significantly, a common issue in long-range forecasting.
• Run-over-Run Instability: The primary reason for low confidence is the GFS model's volatility. It shed over 30 weighted HDDs and added over 30 weighted CDDs in a single cycle. In contrast, the ECMWF was exceptionally stable (delta HDD +0.3, delta CDD -0.1). This suggests the GFS is struggling to resolve the long-range pattern, and its solution should be viewed with skepticism until subsequent runs provide confirmation.
• Source Limitations: This analysis is based on deterministic model runs for specific hub locations. It does not incorporate ensemble model data, which would provide a better sense of the range of potential outcomes and overall forecast uncertainty.