In this study, we apply the breeding method in a fully coupled atmosphere-ocean GCM model to identify the coupled perturbations related with ENSO variability. The ultimate goal is to use the results to improve ensemble forecasting and data assimilation in the operational coupled system of the NASA Seasonal and Interannual Predictability Project.
Previous work (refs, including AMS) has demonstrated that bred vectors can capture the fast growing modes associated with the instabilities of the evolving El Nino/La Nina background state. Experiments with independent breeding runs rescaled with their own SST rms norm in NINO3 region have shown that these ENSO related modes are robust. This is true for their atmospheric and oceanic fields within either tropical or global domain, using a one-month rescaling interval.
Several breeding runs will be examined for better understanding the optimal growth related with ENSO variability, including (1) a norm based on the amount of warm water volume above the thermocline (2) an oceanic energy norm related to Kelvin/Rossy waves. This can also help us explore which variables are the most reliable in dealing with ENSO predictability. Furthermore, in order to maintain the fast growing mode, we consider preparing another set for the breeding runs to keep the bred vectors young???.
One issue that needs to be addressed is that bred vectors are capable to capture the local instability evolving with the background flow without being overly sensitive to the chosen norm or optimal integration period as singular vectors do. This will be an advantage for describing the uncertainties in those areas sensitive to both dynamical and thermodynamical aspects, like the tropical eastern Pacific.
My plans include constructing the ensemble forecast members with the physical modes obtained by bred vectors and examine their potential benefits for ENSO predictability in this coupled system. If the results show positive impacts on forecast error, this information on the bred vectors should be also included in the data assimilation system for better model initialization to increase the forecast skill.