GFS vs ECMWF Demand-Hub Outlook: 2026-05-17 00z

Executive Summary

The 15-day outlook presents a highly uncertain demand picture, with both models showing similar total cooling demand but for vastly different reasons. The GFS forecasts significantly stronger cooling demand in the South, offset by a late-period cold shot in the Midwest and Northeast that drives heating demand. The ECMWF, in contrast, shows a milder and more consistent pattern. Overall confidence is very low, reflected in a convergence score of 0.40, which deteriorated from the prior cycle as models moved in opposite directions.

Hub-Level Weather

The primary driver of the forecast disagreement is the Midwest hub (Chicago, 35% weight), where the models diverge sharply in the second week. The GFS is materially colder, forecasting 65.6 cumulative HDDs versus just 27.4 HDDs from the ECMWF. This difference is most pronounced from May 28-30, where the GFS shows a significant cold intrusion.

A secondary, but also significant, divergence exists across the South. The GFS is substantially hotter than the ECMWF in both the South Central (Houston) and Southeast (Atlanta) hubs. Over the 15-day period, the GFS forecasts 54 more CDDs in Houston and 86 more CDDs in Atlanta compared to the ECMWF, indicating a much stronger and more persistent heat pattern.

Demand Implications (Natural Gas)

The competing weather patterns imply different natural gas demand profiles. * GFS: The combination of strong heat in the South and late-season cold in the Midwest/Northeast results in a more supportive demand outlook. It suggests robust power burn for cooling in southern markets concurrently with a late surge in residential/commercial heating demand in northern markets. The GFS projects 33.0 weighted HDDs and 104.3 weighted CDDs. * ECMWF: This model points to a more benign demand scenario, with consistent cooling load but an almost complete lack of heating demand. The ECMWF forecasts only 20.9 weighted HDDs against 102.8 weighted CDDs.

While total cooling demand is nearly identical between models, the GFS's additional ~12 weighted HDDs make it the more bullish forecast for total weather-driven natural gas consumption.

Model Agreement & Caveats

Forecast confidence is very low. The convergence score of 0.40 indicates significant disagreement between the models, particularly in the Day 8-15 period.

Analysis of the run-over-run deltas shows the models are diverging further: * The GFS trended colder, adding +2.3 weighted HDDs since its prior run. * The ECMWF trended warmer, removing -2.1 weighted HDDs and adding +3.4 weighted CDDs. * This divergence resulted in a -0.07 point drop in the convergence score, reinforcing the low-confidence outlook.

Caveats: This analysis is based on a discrete set of demand-weighted hubs. The underlying data is from Open-Meteo's gfs_seamless and ecmwf_ifs025 models; results may differ from other data providers or model versions.