Bred vectors and forecast error in the NASA coupled general circulation model

Breeding experiments are performed in the operational CGCM of the NASA seasonal to interannual prediction project (NSIPP). The system uses the optimal interpolation to update the oceanic initial state with the real temperature observations. In this study, we examined the characteristics and relationships between bred vectors and forecast error in order to explore the potential application to use the bred vector as initial coupled ensemble perturbations for seasonal-to-interannual ensemble forecasting. Our results indicate that the one-month forecast error and the bred vector share many similar characteristics in the SST and subsurface temperature in both space and time. Also, such features are very sensitive to the background anomalous variations, i.e. ENSO variability. Our results indicate that the one-month forecast error in NASA CGCM is dominated by dynamical errors whose shape can be captured by bred vectors and the agreement between them is particularly good when the bred vector growth rate is large. The results suggest that the bred vector can be used to detect the structure of the forecast error. This can benefit the ensemble forecasting in providing the shape of the dominant uncertainties in the seasonal-to-interannual related variabilities. We will also discuss how those characteristics from bred vectors can be applied to augment the background error covariance in order to adjust its time-independent and Gaussian structure to have a locally seasonal-to-interannual dependent feature