Figure 2. (a) Correlation between DMI and surface chlorophyll anomalies. The black solid and dashed square indicates the southeastern IO (SEIO) and western IO (WIO) region respectively, used to define the DMI. (b) Correlation between SWI chlorophyll index and surface chlorophyll anomalies, and (c) correlation between WS chlorophyll index and surface chlorophyll anomalies. In (b) and (c) the purple solid and dashed square indicates the SWI and WS regions used to define the biological dipole index (BDI). In each plot significant regions at the 95% confidence interval are marked by contours.
To explore the characteristics of IOD and BDI, we further examined the amplitude and phase locking (Fig. 3). As shown in Fig. 3, both in the observations and GFDL-ESM4, IOD and BDI show a dominant seasonal dependency with a peak phase in autumn. In addition, consistent with previous studies IOD peaks in October in the observations (Li et al 2003, Saji et al 1999), while the BDI shows an early peak (September, Fig. 3a). On the other hand, although the BDI peak month in GFDL-ESM4 is consistent with observations, IOD in the GFDL-ESM4 shows an early peak compared to the observations (Fig. 3c). However, apart from the differences in the peak month, both observations and model show the strong phase locking of IOD and BDI in the boreal autumn.
We further examined the relationship between BDI and IOD during September to October (Fig. 3b). The months of September and October have been chosen as the peak months of the two indices. The results clearly show that there is a strong and significant correlation between BDI and IOD amplitudes during autumn. In agreement with the observations, GFDL-ESM4 also shows a strong and significant correlation (-0.85, Fig. 3d). We also examined the correlation between BDMI (defined according to Shi and Wang 2022) and IOD during autumn and found that the correlation (r = -0.62) is lower compared to that of BDI (Fig. S5). Also, the previous BDMI does not capture chlorophyll variability during extreme events because the region used to define the index does not reflect well the changes in chlorophyll in response to IOD. For example, the extreme negative IOD in 2016 is not apparent in the BDMI (BDMI=0.01 for September 2016, Fig. S5), but the event is well represented in the BDI (BDI=0.86 for September 2016, Fig. 3c). Therefore, compared to the BDMI in previous studies, the BDI can well represent the IOD-induced responses in chlorophyll. Thus, we demonstrate the presence of the biological dipole pattern in the tropical IO and propose a new definition (BDI) that can actively capture the variability of chlorophyll in response to IOD.