HTML
-
It has been shown in Fig. 4 that strong variability in equatorial meridional winds appears in both the upper and lowest level of the troposphere over the Maritime Continent. A question therefore arises: Is there a linkage between upper- and lower-level CEFs on the interannual time scale? To address this question, we regress the meridional winds along the equator onto the low-level MC-CEF index (MC-LCEFI), which is defined as the standardized JJA-mean meridional wind anomalies at 925 hPa averaged over 102.5°−110°E, 122.5°−130°E, and 147.5°−152.5°E, following Li and Li (2014). The only difference between the present definition and theirs is that we use the equatorial anomalies in order to match with the other analyses in this study, while they used the averages over 2.5°S−2.5°N. However, our index is almost identical to theirs, as indicated by extremely high correlation coefficients between the two indices, which range from 0.9986 to 0.9998 among the three datasets.
The meridional wind anomalies regressed onto the MC-LCEFI are shown in Fig. 7. There are significant northerly anomalies in the upper troposphere and southerly anomalies in the lower levels of the troposphere over the Maritime Continent in all datasets. This indicates a strengthening of both the upper and lower branches of CEFs over the Maritime Continent. The northerly anomalies consistently appear between 90°E and 170°E around 200 hPa, with the strongest one at about 150°E. Although we define the MC-LCEFI using the three branches to the east of 100°E, there are also significant southerly anomalies between 90°E and 100°E in the lower troposphere. Weak but significant southerly anomalies appear in the mid troposphere over the Maritime Continent in both ERA-Interim and JRA-55, but these are absent in NCEP-2. In addition, to the west of 90°E, there are southerly anomalies in the upper troposphere and northerly anomalies in the lower levels. The upper-tropospheric southerly anomalies are very similar between ERA-Interim and JRA-55, but shift eastward in NCEP-2. The low-level northerly anomalies are significant, and this out-of-phase relationship between the Somali jet and the MC-LCEFs has been well documented by Li and Li (2014), Li et al. (2017a), and Li et al. (2017b). However, the anomalies associated with the Somali jet are much weaker, consistent with the weak interannual variance (Fig. 4), and they appear relatively eastward in NCEP-2.
Figure 7. Regression of the interannual component of the meridional winds along the equator onto the lower-level MC-CEF index based on (a) ERA-Interim, (b) JRA-55, and (c) NCEP-2. The contour interval is 0.2 m s−1, and the zero contours are omitted. The red (blue) shading denotes positive (negative) values, and dots represent regions significant at the 95% confidence level based on the Student’s t-test.
This negative relationship between the upper and lower troposphere over the Maritime Continent can also be confirmed from the viewpoint of upper-tropospheric CEFs. Figures 8a-c show the correlation coefficients of equatorial meridional winds between the reference point (200 hPa, 150°E) and all the grids. This reference point shows the strongest meridional wind anomaly associated with the low-level CEFs (Figs. 7a and b) and the greatest interannual variance (Figs. 4a and b) in ERA-Interim and JRA-55. The correlation coefficients over the Maritime Continent are characterized by positive values in the upper troposphere and negative ones in the lower levels in all the datasets. Strong negative correlations in the lower levels are centered in several branches, consistent with the large interannual variance (Figs. 4d-f). Besides, there are negative and positive correlations to the west of 90°E in the upper and lower troposphere, centered around 150 hPa and 850 hPa, respectively. This distribution is very similar to that shown in Fig. 7, including the negative correlations (Fig. 8) and southerly anomalies (Fig. 7) in the mid troposphere over the Maritime Continent, in ERA-Interim and JRA-55 but not in NCEP-2.
Figure 8. One-point correlation coefficient for the interannual component of the meridional winds along the equator based on (a) ERA-Interim, (b) JRA-55, and (c) NCEP-2. The reference point (200 hPa, 150°E) is marked by the white cross. The contour interval is 0.1, and values between −0.4 and 0.4 are omitted. The red (blue) shading denotes positive (negative) values, and dots represent regions significant at the 95% confidence level based on the Student’s t-test. As in (a−c) (d−f) are for the reference point (150 hPa, 150°E).
Considering that the strongest interannual variance and northerly anomaly in NCEP-2 appear at 150 hPa (Figs. 4c and 7c), we use (150 hPa, 150°E) as the reference point and repeat the analysis. The results (Figs. 8d-f) are similar, and the largest positive values appear at 200 hPa to the west of 135°E, even though the reference point is shifted to 150 hPa.
The above results demonstrate the significant negative relationship in the equatorial meridional winds between the upper and lowest level of the troposphere, over the Maritime Continent and Indian Ocean. The domains of significant correlation are closely consistent with those of large interannual variance. All these results imply that the negative relationship may play an important role in the interannual variability of the equatorial meridional winds over the Maritime Continent and Indian Ocean. To verify this, we perform an EOF analysis on the equatorial meridional winds in the domain of (30°E−160°W, 1000−70 hPa), which is the same as those for Figs. 4, 7 and 8. The first mode (EOF1) is separable from the other modes according to North et al. (1982), and accounts for 34.3%, 32.5% and 29.2% of the total interannual variance in ERA-Interim, JRA-55 and NCEP-2, respectively. Figure 9 shows the equatorial meridional wind anomalies regressed onto the standardized principal component of the first mode (PC1). EOF1 is characterized by northerly (southerly) anomalies in the upper (lower) troposphere over the Maritime Continent and southerly anomalies in the upper troposphere over the Indian Ocean in all the datasets. There are also significant, albeit weak, anomalies of the Somali jet. This distribution resembles very well the distributions shown in Figs. 7 and 8.
Figure 9. Regression of the interannual component of the meridional winds along the equator onto the normalized PC1, with the EOF analysis performed for the domain identical to that shown in the figure, based on (a) ERA-Interim, (b) JRA-55, and (c) NCEP-2. The contour interval is 0.2 m s−1, and the zero contours are omitted. The red (blue) shading denotes positive (negative) values, and dots represent regions significant at the 95% confidence level based on the Student’s t-test.
To compare the first mode and these significant meridional wind anomalies over the Maritime Continent and Indian Ocean, we define several indexes to depict the meridional wind anomalies. First, considering the distribution of the anomalies over the Maritime Continent shown in Figs. 7 and 8 and the large variances shown in Fig. 4, we define MC-HCEFI as the standardized JJA-mean equatorial meridional winds at 200 hPa averaged over 110°−170°E. Second, we define IO-HCEFI at 150 hPa over 45°−75°E along the equator based on a similar consideration. Third, as mentioned before, we define MC-LCEFI, following Li and Li (2014), but only the equatorial winds are used here. We also follow Li and Li (2014) and define the Somali jet index (Somali-I) by averaging the equatorial meridional winds at 850 hPa over 37.5°−62.5°E.
Figure 10 shows the interannual and decadal variations of these indexes. High consistency exists among the three datasets in depicting the interannual variability in the upper troposphere and low-level branches over the Maritime Continent (Figs. 10a and b), consistent with the results shown in the preceding section. For instance, the correlation coefficient between ERA-Interim and JRA-55 is 0.98 for MC-HCEFI and 0.99 for MC-LCEFI. The similarity among different datasets over the Indian Ocean on the interannual time scale, albeit weaker than that over the Maritime Continent, can also be found from the series of IO-HCEFI and Somali-I (Figs. 10c and d), with the correlation coefficient being 0.88 for IO-HCEFI and 0.92 for Somali-I. In contrast, the decadal indexes show large differences (Figs. 10e-h). ERA-Interim and JRA-55 tend to show consistent decadal variations of the low-level CEFs over the Maritime Continent (Fig. 10f), but they show quite different variations of the high-level MC-CEF and the CEFs over the Indian Ocean (Figs. 10e, g and h). The large differences shown by the decadal indexes indicate that there is great uncertainty in the decadal variations of CEFs over the Maritime Continent and Indian Ocean in the current reanalysis data.
Figure 10. Time series of (a) MC-HCEFI, (b) MC-LCEFI, (c) IO-HCEFI, and (d) Somali-I calculated by the interannual component based on ERA-Interim, JRA-55 and NCEP-2. As in (a−d) (e−h) are for the interdecadal component.
We calculate the correlation coefficients between PC1 and these indexes and show them in Table 1. PC1 is highly correlated with MC-HCEFI and MC-LCEFI, suggesting that the first mode can explain the majority of the interannual variance of CEFs in both the upper and lowest level of the troposphere over the Maritime Continent. In addition, PC1 is also significantly correlated with IO-HCEFI and Somali-I. These correlation coefficients confirm the close relationship between the first mode and CEFs over the Maritime Continent and Indian Ocean. Or, in other words, the relationship between the CEFs over the Maritime Continent and Indian Ocean, particularly the seesaw pattern between the upper and lower troposphere over the Maritime Continent, contributes significantly to the first mode.
Index MC-HCEFI MC-LCEFI IO-HCEFI Somali-I PC1 MC-LCEFI −0.86 − − − − IO-HCEFI −0.61 0.72 − − − Somali-I 0.48 −0.68 −0.55 − − PC1 −0.98 0.93 0.68 −0.59 − Niño3.4 −0.67 0.83 0.68 −0.62 0.75 Table 1. Correlation coefficients between the indexes used in this study based on ERA-Interim. Results based on JRA-55 and NCEP-2 are very similar with ERA-Interim and are therefore not shown here.
In addition, a strong relationship between CEFs and ENSO, especially for MC-CEF, can be found in Fig. 10. For example, MC-HCEFI reaches a minimum (maximum) in 1997 (1998), which was a developing summer for a strong El Niño (La Niña). This relationship can be verified by the correlation coefficients with the Niño3.4 index (Table 1), which is defined as the standardized JJA-mean SST anomalies averaged over (5°S−5°N, 120°−170°W), using the monthly mean SST data provided by ERSST.v5. Actually, the Niño3.4 index is also highly correlated with PC1 (0.75; Table 1). All these strong correlation coefficients suggest that ENSO may contribute much to the leading mode. To confirm this, we regress the equatorial meridional wind anomalies onto the JJA-mean Niño3.4 index, and show the results in Fig. 11. Over the Maritime Continent there are northerly anomalies in the upper troposphere and southerly anomalies in the low-level branches. There are also southerly anomalies in the upper troposphere over the Indian Ocean and weak northerly anomalies in the lower troposphere around 50°E. This distribution is closely consistent with the leading mode shown in Fig. 9. All these results indicate that ENSO contributes much to the leading mode of equatorial meridional winds and the linkage between CEFs in the upper and lower troposphere over the Maritime Continent and Indian Ocean. The ENSO-related meridional wind anomalies tend to appear over a relatively narrow scope around 200 hPa in the vertical direction over the Maritime Continent, in comparison with those associated with PC1 (Fig. 9), indicating that ENSO is most strongly correlated with CEFs at 200 hPa and thus providing an extra advantage of defining CEFs by using 200-hPa meridional winds.
Figure 11. Regression of the interannual component of the meridional winds along the equator onto the Niño3.4 index in JJA based on (a) ERA-Interim, (b) JRA-55, and (c) NCEP-2. The contour interval is 0.2 m s−1, and the zero contours are omitted. The red (blue) shading denotes positive (negative) values, and dots represent regions significant at the 95% confidence level based on the Student’s t-test.