
Summer 2025-26 Featured Climate Extremes Across Australia
03/04/2026, 8:19 am ESTClimate Impact Company Dynamic/AI Models Verification Report
Issued: Friday, March 6, 2026
Highlight: AIFS ENS and ECM ENS remain top North America medium-range forecast models with Graph Cast and (surprisingly) CMC ENS competitive.

Fig. 1: North America 2-meter temperature anomaly correlation forecast skill scores from the past 30 and 7 days for the 6-10-day period.
Discussion: A check on skill scores for North America medium range forecasts as meteorological winter closes. Using anomaly correlation provided by CWG/SVMW, the 6-10-day skill scores indicate AIFS ENS remains (routinely) the top model (Fig. 1). ECM ENS typically ranks second to AIFS ENS and collectively, using both 30-day and 7-day skill scores, that ranking remains. Climate Impact Company routinely uses AIFS ENS and/or ECM ENS for daily medium range forecasts (for the U.S.). Graph Cast is also competitive with AIFS ENS and ECM ENS. Surprising is the success of the Canadian Ensemble (CMC ENS) emerging during the winter 2025-26 season. The AIFS skill scores are also competitive with top models for the past 30 and 7 days for North America 2-meter temperature outlooks in the 6-10-day period. The lowest skill scores are from WeatherNext2, 4CastNet GFS ENS, CMC, and GFS.
In the 11-15-day period, skill scores are lower by approximately 20-40% forecasting 2-meter temperature across North America. Once again, AIFS ENS, Graph Cast, and ECM ENS rank highest with GFS ENS and CMC ENS competitive (Fig. 2). Collectively, from both the past 30 and 7 days, WeatherNext2, 4CastNet V2 GFS ENS, ECM, and GFS rank last.

Fig. 2: North America 2-meter temperature anomaly correlation forecast skill scores from the past 30 and 7 days for the 11-15-day period.

