| Numerical experiments with ocean general circulation models have been executed to study predictability of interannual variability of the Kuroshio transport. Using a model with simplified bottom topography, at first, we investigated basic processes of westward propagation of nondispersive Rossby waves excited by artificial windstress anomaly which is located around the center of the North Pacific (30N, 180E). In cases where oscillation period of the windstress forcing is interannual, barotropic Rossby waves can not cross ridges such as the Izu-Ogasawara Ridge and the Shatsky Rise, and they hardly have effect on volume transport of the western boundary current flowing south of Japan. On the other hand, first-mode baroclinic Rossby waves are transmitted largely across the ridges. The volume transport variability due to the first-mode waves reaches about half of the Sverdrup transport estimated theoretically. As for second-mode baroclinic waves, those generated by mode conversion from the first-mode waves on the ridges can arrive south of Japan, though they do not have much effect on the volume transport. Since it takes about 7-8 years for the first-mode waves to traverse the entire basin along 30N in the North Pacific, these results suggest that a two- or three-year forecast for the interannual variability of the Kuroshio transport shows high skill, when a past record of the windstress data over the North Pacific is available. To investigate this, then, we carried out a hindcast experiment covering the period 1970-2001 and 15 cases of four-year forecast ones, using a model with realistic bottom topography and windstress data from the NCEP/NCAR reanalysis. Each of the forecast runs, whose initial conditions are taken from the hindcast run at intervals of two years, is driven by the windstress in year preceding each forecast run. As for the interannual variability of the Kuroshio transport across the ASUKA-line south of Japan, the forecast runs reproduce the hindcast result well as far as within two years except for the cases in which remarkable windstress variability exists near Japan. The correlation coefficient of volume transport anomalies between the hindcast and forecast runs is about 0.7 in forecast years 1-2, and it falls to below 0.5 after 3 years, when the rms error becomes greater than the standard deviation of the transport anomaly. |
|
|