Applications of Conditional Nonlinear Optimal Perturbation to the Studies of Predictability Problems
条件非线性最优扰动在可预报性问题研究中的应用

This paper summarizes the applications of conditional nonlinear optimal perturbation (CNOP) in recent predictability studies. These include four main contributions. First, the CNOP approach was extended to consider not only initial perturbations but also model parametric perturbations. The extended CNOP approach can study the predictability problems induced by either initial errors or model errors as well as those induced by both initial errors and model errors. Second, the extended CNOP approach was applied to the predictability studies of ENSO events and Kuroshio path anomalies. The effect of initial errors and model parametric errors on the predictability of these events was demonstrated and it was shown that the initial errors play a dominant role in the predictability of ENSO and the Kuroshio path anomalies. The CNOP approach was also applied to investigate the optimal precursors (OPRs) of the onset of blocking events and optimally growing initial errors (OGRs). The results demonstrated that OPRs and OGRs are often concentrated at a localized region; furthermore, their patterns are very similar. Finally, the CNOP approach was used to study adaptive observations of typhoons. With the CNOP, the sensitive areas of some typhoon cases were determined and with the data from the Observing-Systems Simulation Experiments (OSSEs) and/or Observing-Systems Experiments (OSEs) the validity of the sensitive area was demonstrated. Specifically, the OGRs of typhoon cases often concentrate in a particular region. Increasing the number of the observations in this region may significantly improve the forecasting skill for typhoons. The region identified by OGRs may represent the sensitive area of typhoon forecasting. The OGRs of El Ni o events, Kuroshio path anomalies, and blocking events are also localized in a particular region. Based on the approach of the typhoon adaptive observation, the sensitive areas associated with these events may be identified as the localized regions of the OGRs.