Based on clustering analysis, the NWP TCs are grouped into three different clusters (C1, C2 and C3) (Fig. 1). The C1 cluster is characterized by TCs that generated south of 15°N and east of SCS over the NWP and have recurved tracks while most of TCs in C2 generated over the subtropical region, north of 15°N, and move northward. The C3 cluster is characterized by TCs that generated west of 140°E, half of them in the SCS, and mostly have belong to straight trajectories moving westward. In this study we focus on studying the TCs in C1 and C2 and found that both of them experienced interannual and interdecadal variabilities.
As shown in Fig. 2a and 2b, the red lines stand for the IPO index and the blue ones for C1 (C2) time series. The correlation coefficient between IPO (ENSO) and C1 index is 0.66 (0.55), while -0.6 (-0.5) between IPO (ENSO) and C2 time series. The correlation coefficient between C1 and C2 is -0.43, above the 95% confidence level based on Student’s T test. Clearly, there is a seesaw relationship between C1 and C2 TC frequency, which is contributed by the dipole effect of ENSO. Using the moving T test with a 5-year window, we found that the C1 TCs encounters a significant interdecadal shift after 1997 (Fig. 2c), which is consistent with the IPO phase change. Moreover, we used the Fast Fourier Transform (FFT) to filter signals more than 11 years and found that both the IPO (red line) and C1 time series (blue line) changed from positive to negative values around 1997/1998 (Fig. 2d). So we separate the whole time series into two periods: the prior period (1980-1997) and the post period (1998-2015).
We calculated the correlation relationship between the IPO index and C1 (C2) time series for two periods and found that the correlation coefficients are 0.5 (-0.72) from 1980-1997) while 0.74 (-0.5) from 1998-2015, all of them above 95% confidence level based on Student’s T test. We found that the Cluster 1 (C1) TC frequency over the tropical NWP, south of 15°N and east of South China Sea (SCS), is significantly decreased after 1998 (Fig. 2a and Fig. 3b) when the IPO turned from its positive phase to negative phase. The TCs frequency in C1 abruptly decreased from 123 in prior period to 47 in post period (Fig. 3b). Meanwhile, the tracks of Cluster 2 (C2) TCs north of 15°N, as well as C1 tracks, over the NWP experienced a westward shift after 1998 (Fig. 3a and 3c). During the negative IPO phase, the La Niña-like SST pattern strengthened the Walker circulation, intensified equatorial easterly, increased the vertical zonal wind shear, weakened the Monsoon Trough (MT) and finally suppressed the TC genesis over the eastern tropical NWP (Fig. 4). The IPO in recent negative phase also strengthened NWP Subtropical High (NWPSH), leading to anomalous westward steering flow and thus a westward shift. The results from model simulation further confirmed our analysis that the IPO are responsible for the interdecadal change of NWP TCs.