Fig. 4: Comparison of the CRPS (first row), the globally averaged latitude-weighted Spread (second row) and SSR (third row) of the ECMWF ensemble (light purple lines) and the FuXi ensemble (light red lines) for 2 upper-air variables, including Z500 (first column) and T850 (second column), and 2 surface variables, such as MSL (third column) and T2M (fourth column), in 15-day forecasts using testing data from 2018. | npj Climate and Atmospheric Science

Fig. 4: Comparison of the CRPS (first row), the globally averaged latitude-weighted Spread (second row) and SSR (third row) of the ECMWF ensemble (light purple lines) and the FuXi ensemble (light red lines) for 2 upper-air variables, including Z500 (first column) and T850 (second column), and 2 surface variables, such as MSL (third column) and T2M (fourth column), in 15-day forecasts using testing data from 2018.

From: FuXi: a cascade machine learning forecasting system for 15-day global weather forecast

Fig. 4

The FuXi ensemble is evaluated against the ERA5 reanalysis dataset, and the ECMWF ensemble is evaluated against ENS-fc0.

Back to article page