GFS vs ECMWF Demand-Hub Outlook: 2026-10-05 00:00 UTC

Executive Summary

The 15-day outlook reflects a classic shoulder-season pattern with national demand-weighted cumulative heating degree days (HDDs) showing near-identical totals between models (55.56 HDDs for GFS vs. 55.70 HDDs for ECMWF). However, underlying regional divergence is substantial, as GFS carries a much warmer bias in the South and East, yielding 43.32 cumulative cooling degree days (CDDs) compared to ECMWF’s 33.24 CDDs. Run-over-run adjustments show GFS shedding 12.68 weighted HDDs from its prior cycle, while ECMWF held relatively stable (+1.68 HDDs).

Hub-Level Weather

Material timing and spatial disagreements emerge across key demand centers:

• Midwest (Chicago, 35% weight): GFS projects higher total heating load at 73.80 HDDs vs. ECMWF’s 65.71 HDDs, with a daily mean absolute HDD difference of 2.71. GFS brings a sharper cold shot in the 11–15 day window (10.53 HDDs on Oct 18 and 11.25 HDDs on Oct 19 vs. ECMWF's 5.90 and 6.10 HDDs, respectively).
• East (New York City, 30% weight): ECMWF forecasts colder conditions overall with 69.65 HDDs vs. GFS’s 55.88 HDDs. The daily mean absolute divergence is elevated at 3.62 HDDs, driven by ECMWF projecting earlier and more sustained mid-range heating demand (e.g., 8.08 HDDs on Oct 15 vs. 0.00 HDDs on GFS).
• Northeast (Boston, 10% weight): GFS leads heating demand with 129.13 cumulative HDDs vs. 111.68 HDDs on ECMWF. GFS projects sharp cold on Oct 19 (17.48 HDDs; 47.52°F mean) versus ECMWF's milder 6.29 HDDs (58.71°F mean).
• South Central (Houston, 15% weight) & South (Atlanta, 10% weight): GFS projects persistent late-season warmth in Houston (214.28 CDDs vs. ECMWF’s 168.05 CDDs) and Atlanta (68.40 CDDs vs. 37.19 CDDs). ECMWF introduces an early cooldown in Houston by Oct 17–19 (averaging ~3.7 to 4.9 CDDs daily vs. GFS's 13.8 to 14.1 CDDs).

+---------------------------+--------+---------------+-----------------+---------------+-----------------+
| Hub                       | Weight | GFS Cum. HDDs | ECMWF Cum. HDDs | GFS Cum. CDDs | ECMWF Cum. CDDs |
+---------------------------+--------+---------------+-----------------+---------------+-----------------+
| Midwest (Chicago)         |  0.35  |     73.80     |      65.71      |      8.32     |      12.32      |
| East (New York City)      |  0.30  |     55.88     |      69.65      |      4.76     |       0.00      |
| Northeast (Boston)        |  0.10  |    129.13     |     111.68      |      0.00     |       0.00      |
| South Central (Houston)   |  0.15  |      0.00     |       0.00      |    214.28     |     168.05      |
| South (Atlanta)           |  0.10  |      0.51     |       6.44      |     68.40     |      37.19      |
| Weighted National Total   |  1.00  |     55.56     |      55.70      |     43.32     |      33.24      |
+---------------------------+--------+---------------+-----------------+---------------+-----------------+

Demand Implications (Natural Gas)

Aggregate weather-driven load for Natural Gas (NG) remains subdued due to the October shoulder transition. While total weighted HDDs appear matched, the geographic breakdown presents divergent drivers: - Residential/Commercial Space Heating: If ECMWF verifies, earlier East Coast heating demand will firm regional cash markets, whereas GFS shifts heating support deeper into the Midwest/Northeast late in the period. - Power Burn: GFS maintains residual cooling demand across the ERCOT and Southeast footprints, adding roughly 10.08 weighted CDDs over ECMWF. Run deltas show GFS losing substantial heating intensity (-12.68 HDDs), reducing implied early-winter residential burn expectations.

Model Agreement & Caveats

• Convergence: The overall convergence score sits at a low 0.2915, accompanied by a weighted daily mean absolute HDD difference of 2.43. Although cumulative HDDs match at ~55.6, this is a mathematical coincidence masking significant regional and timing disagreements (particularly across Days 8–15 in New York and Chicago).
• Forecast Confidence: Low. Market participants should avoid leaning heavily on aggregate degree-day totals given wide intraday temperature spreads in the extended forecast.
• Data Limitations: Surface inputs derive from Open-Meteo operational model points (GFS Seamless vs. ECMWF IFS 0.25°), which exclude ensemble probability distributions and local urban heat island adjustments.