Fig. 3: Improved forecast skill in the ensemble mean as revealed by power spectral analysis of along-track SSHA.
From: Ensemble forecasting greatly expands the prediction horizon for ocean mesoscale variability
Scales constrained in the (a) 1 day forecast and (b) 10 day forecast in the high-resolution model (blue), lower resolution (red), and ensemble mean (black). R2, which is the ratio of the model error variance to the variance of the altimeter SSHA, values above 1 mean that variance of error in this frequency band is greater than variance of the altimeter signal and, hence, these scales are not constrained. R2 values below 1 indicate that variance of errors are lower than the variance of the signal and, hence, analysis and forecast have skill in this frequency band. The R2 values are spectrally smoothed and averaged over multiple forecasts (see details in the methods section). We excluded scales lower than 150 km as the power spectra density flattens off at this scale. The analysis is carried out for the globe (50°S to 50°N). The error bars of the R2 are computed at 95% confidence level (see methods for details).
