Fig. 2: Comparison of root mean square error (RMSE) between WenHai and a state-of-the-art numerical global ocean forecast system (GOFS) as a function of forecast lead time.
From: Forecasting the eddying ocean with a deep neural network
Globally averaged RMSE (the lower, the better) of the forecast temperature profile (a), salinity profile (b), sea surface temperature (SST) (c), sea level anomaly (SLA) (d), 15-m zonal current (e) and 15-m meridional current (f) as a function of forecast lead time. The zero-lead time represents the initial conditions. The blue, red, and grey lines correspond to WenHai, GLO12v4 and persistent forecast, respectively. For the temperature and salinity profile forecast, the RMSE is vertically averaged over the upper 643 m. The shading corresponds to the 50% confidence interval computed from a bootstrap method.
