Relationship between cross-equatorial flows over the Bay of Bengal and Australia in boreal summer: Role of tropical diabatic heating

Xuelei Jing Yunyun Guo Zhiping Wen b c

a Department of Atmospheric and Oceanic Sciences & Institute of Atmospheric Sciences, Fudan University, Shanghai, China

b Innovation Center of Ocean and Atmosphere System, Zhuhai Fudan Innovation Research Institute, Zhuhai, China

c Jiangsu Collaborative Innovation Center for Climate Change, Nanjing, China

Keywords: Cross-equatorial flow Equatorially asymmetric diabatic heating Linear baroclinic model

ABSTRACT The interannual variability of cross-equatorial flows (CEFs) over the Asian—Australian monsoon (AAM) region during boreal summer was analyzed by applying the empirical orthogonal function (EOF) method to the merid- ional wind at 925 hPa. The first mode (EOF1) exhibits an in-phase relationship among different CEF channels over the AAM region, which has received much attention owing to its tight linkage with ENSO. By contrast, the second mode (EOF2) possesses an out-of-phase relationship between the Bay of Bengal (BOB) CEF (90°E) and Aus- tralian CEF, among which the New Guinea CEF near 150°E shows the most significant opposite correlation with the BOB CEF. Observational and numerical model results suggest that the equatorially asymmetric heat source (sink) over the western (eastern) Maritime Continent, closely associated with the in-situ sea surface temperature anomaly, can induce cross-equatorial northerly (southerly) flow into the heating hemisphere, which dominates the out-of-phase relationship between the BOB and New Guinea CEFs. Furthermore, an equatorially symmetric heating over the central Pacific may indirectly change the CEFs by modulating the zonal atmospheric circulation near the Maritime Continent.

1. Introduction

Cross-equatorial flows (CEFs) serve as a bridge in the inter- hemispheric transportation of atmospheric mass, moisture, angular mo- mentum and energy, which significantly contributes to changes in tropi- cal precipitation, air temperature, typhoons, and so on. Due to their sig- nificant sensitivity to climate change, CEFs have attracted considerable scientific interest (e.g., Li and Lou, 1987 ; Kakade and Dugam, 2008 ).

As widely agreed, relatively predominant channels of CEFs lie over the Asian—Australian Monsoon (AAM) region in the lower troposphere (near 925 hPa) in summer, including the Somali CEF at around 45°E, Bay of Bengal (BOB) CEF at around 90°E, South China Sea (SCS) CEF at around 105°E, Celebes Sea CEF at around 125°E, and New Guinea CEF at around 150°E. All of these five CEF channels are named by their geo- graphic location of their climatological state, among which the Somali CEF exhibits the strongest magnitude but weakest interannual variance compared with the other four channels (e.g., Li and Lou, 1987 ; Peng and Jiang, 2003 ). Thus, the other four CEFs over the AAM region extending from 80°E to the tropical western Pacific are the main focus of discussion in the present work.

The relationships among different CEF channels have been investi- gated in recent decades. It has been found that the CEFs over the AAM region can be bracketed together since they are located in adjacent geographical situations and mainly modulated by the Australian high ( Peng and Jiang, 2003 ; Wang and Yang, 2008 ). By contrast, some stud- ies have suggested that the BOB CEF (90°E) changes independently of the others and further grouped the SCS (105°E), Celebes Sea (125°E), and New Guinea (150°E) CEFs into one category regarded as the Australian CEFs owing to the close correlation among them ( Zhu, 2012 ; Li and Li, 2014 ; Wang and Yang, 2014 ). However, these studies disagree about the relationships among different CEF channels over the AAM region,which therefore requires further study.

In terms of understanding what controls the interannual variability of CEFs, numerous studies have suggested that it can be attributed to several factors, including the beta effect, land—sea contrast, baroclinicity in the boundary layer, diabatic heating, and so on ( Krishnamurti et al.,1979 ; Rodwell and Hoskins, 1995 ). El Ni?o—Sothern Oscillation (ENSO)and its associated diabatic heating have been demonstrated to be tightly linked with the Australian CEFs by modulating the Mascarene high and Australian high ( Zeng and Li, 2002 ; Zhu and Chen, 2002 ; Wu et al.,2018 ). Theoretical and numerical analyses have proven that diabatic heating anomalies asymmetric about the equator might induce a CEF into the heating hemisphere, which provides a mechanism responsible for the impact of ENSO on the interannual variability of CEFs ( Gill, 1980 ;Lee et al., 2009 ).

The present study aims to address the relationships among the main CEF channels over the AAM region on the interannual time scale by investigating the 925 hPa meridional winds during boreal summer. Furthermore, the causes of the CEFs’ variability are also examined via observational and numerical model data.

2. Data, methods, and model

The data used in this work are: (1) monthly variables (horizonal wind, vertical velocity, geopotential height, and air temperature obtained from ERA-Interim with a horizontal resolution of 1° × 1°( Dee et al., 2011 ); and (2) monthly mean sea surface temperature (SST)from the Hadley Center (HadISST) with a horizontal resolution of 1°× 1°( Rayner et al., 2003 ). The HadISST data have been verified to be an improvement upon the Global Sea Ice and SST dataset ( Rayner et al.,2003 ).

The study period covers 1979 to 2017 and the summer mean is taken as the average of June to August (JJA). To obtain the interannual components, all variables are detrended and the interdecadal components longer than 10 years are filtered out via the Fast Fourier transform technique. The vertically integrated atmospheric diabatic heating is estimated from the thermodynamic equation given in Yanai et al. (1973) .

The linear baroclinic model (LBM) used in the present work is based on a linearized atmospheric general circulation model (AGCM), which has been applied extensively to investigate how the atmosphere response to an idealized heat source (e.g., Lu and Lin, 2009 ; Guo et al.,2019 ). In this work, we employ the version with a horizonal resolution of T42 and 20 vertical layers based on a sigma coordinate system. Since a quasi-steady state is achieved after about two weeks, the numerical responses presented in Section 3.4 are approximated as the average of the last 15 days of a 40-day integration.

3. Two EOF modes of CEFs over the AAM region

3.1. In-phase mode of CEFs

Based on the leading mode (EOF1) of the CEFs over the AAM region applied onto the JJA-mean meridional wind anomaly averaged over 5°S—5°N at 925 hPa, a monopolar pattern is observed between 80°E and the date line ( Fig. 1 (a)). This suggests a coherent variation of four main CEF channels, including the BOB, SCS, Celebes Sea, and New Guinea CEFs. EOF1 can account for 61.22% of the total variance, whose principal component (PC1) shows more dominant cycles of CEFs on the interannual time scale ( Fig. 1 (c)).

Fig. 1 (e, g) exhibits the regression maps of horizontal wind at 925 hPa and the SST anomaly (SSTA) against PC1. Corresponding to a positive PC1, a significantly enhanced Mascarene high and Australian high are observed in the present case (figure not shown), north of which the anomalous southerly extends from 70°E to 160°E ( Fig. 1 (e)), inferring that the CEFs over the AAM region are significantly intensified.Accordingly, the change of the Somali CEF is insignificant owing to weak northerly anomalies near 45°E ( Fig. 1 (e)). The enhanced westerly anomaly over the BOB, together with more-active-than-normal precipitation over that region, suggests a relatively strong BOB summer monsoon. By contrast, the East Asian summer monsoon tends to be weakened owing to anomalous northerlies prevailing along 110°E and suppressed precipitation over eastern China (figure not shown).

Furthermore, from the tropical perspective, anomalous descending motion is observed over the Maritime Continent (MC), where a negative SSTA lies ( Fig. 1 (g)), which may be associated with an interannual weakness of the Walker circulation. The SSTA shown in Fig. 1 (g) illustrates a canonical El Ni?o-like pattern with significant warming in the equatorial central-eastern Pacific and cooling to the west. Further analysis proves that EOF1 is relevant to the developing phase of such El Ni?o-like warming (figure not shown). The correlation between PC1 and Ni?o3.4 in the following winter (simultaneous summer) reaches 0.74(0.82). Aside from the BOB CEF, the modulation of the interannual variability of the SCS, Celebes Sea, and New Guinea CEFs by ENSO has been demonstrated in previous studies ( Zhu, 2012 ; Li and Li, 2014 ; Li et al.,2017 ). ENSO can change the Australian high through the large-scale air—sea interaction, and then the Australian high will help modulate the CEFs nearby ( Zeng and Li, 2002 ; Zhu and Chen, 2002 ). Furthermore,our result indicates that the effect of ENSO still works with respect to the BOB CEF.

3.2. Out-of-phase mode of CEFs

Different from EOF1, an out-of-phase relationship of the meridional flow between the MC and equatorial western Pacific is observed in the second EOF mode (EOF2) ( Fig. 1 (b)), which can explain 15.55% of the total variance. Note that EOF2 is characterized by a maximum and minimum near 90°E and 150°E, respectively, implying an opposite relationship between the BOB and New Guinea CEFs. Our result shows some disagreement with previous standpoints ( Wang and Yang, 2008 ; Zhu, 2012 ;Li and Li, 2014 ; Wang and Yang, 2014 ), which regarded the BOB, SCS,Celebes Sea, and New Guinea CEFs as the Australian CEF or grouped the latter three CEFs to one cluster because of their coherent interannual variability. The present result suggests that the CEFs spanning from 80°E to the date line may behave in an in-phase relationship in most years, as EOF1 suggests; however, these CEF channel may change in an opposite fashion in some years (such as 1987, 1991, 1993, 1994, 1999, 2010,and 2012), which is clearly detected by EOF2.

To further validate the robustness of EOF2, a CEF index (CEFI) is constructed by the difference in the 925-hPa meridional wind anomaly between 88°—95°E and 147°—153°E along the equator (5°S—5°N). Its correlation coefficient with PC2 reaches 0.88, suggesting a coherent variation between them. Furthermore, the composite results of the 925 hPa meridional wind anomaly by averaging the abnormal years, in which the normalized CEFI is greater (smaller) than 1.0 ( ? 1.0) standard deviation,bear a strong resemblance to the out-of-phase relationship between the BOB and New Guinea CEFs detected by EOF2 (figure not shown). Consequently, such an out-of-phase mode identified by the EOF technique is robust.

Corresponding to a positive principle component of EOF2 (PC2),there is a dipole pattern of meridional wind anomalies lying over the western and eastern MC (hereafter abbreviated as WMC and EMC, respectively) south to the equator. A significant southerly anomaly appears near 90°E and a northerly anomaly appears between 150°E and 160°E, which indicates an intensified BOB CEF and a weakened New Guinea CEF ( Fig. 1 (f)). Moreover, the meridional wind anomaly is extremely weak near 45°E, which may imply a neglected connection between the Somali CEF and EOF2. Identifying the characteristics of the Asian summer monsoon associated with this out-of-phase CEF mode is an interesting issue. It is found that the Mascarene high in the Southern Hemisphere weakens significantly and an anomalous westerly spans from the Arabian Sea to the tropical western Pacific along the zonal belt between 10°N and 20°N, while significantly enhanced precipitation is observed in northern India, the eastern BOB, SCS, and western North Pacific (figure not shown). This result indicates that EOF2 is plausibly related with an enhanced South Asian summer monsoon and SCS summer monsoon.

Fig. 1. Spatial pattern of the (a, b) first two EOF modes and (c, d) corresponding principal components of the 925-hPa meridional wind anomaly along the equator(5°S—5°N) over the AAM region during 1979—2017. (e—h) Regression maps of (e, f) 925 hPa meridional wind anomalies (shading; units: m s —1 ) and horizontal wind(vectors; units: m s —1 ), and (g, h) SSTAs (shading; unit: K) and vertical velocity anomalies (contours; units: Pa s —1 , dashed contours represent negative values), onto PC1 (left) and PC2 (right). The boxes denote the area of interest for CEFs (10°S—10°N, 80°E—180°). Stippling indicates significance of the meridional wind in (e, f)and the SSTA in (g, h) at the 90% confidence level. Black vectors represent significance of the horizontal wind at the 90% confidence level.

The regressed SSTA against PC2 shows a similar dipole pattern firmly located near the WMC and EMC, with centers lying south to the equator ( Fig. 1 (h)). Furthermore, a large-scale cooling SSTA and its associated descending anomaly are observed in the equatorial central-eastern Pacific, jointly merging a tripole pattern of both the SSTA and vertical velocity anomaly. Upon examining the SSTA in every abnormal year when PC2 is larger (smaller) than 1.0 (—1.0) standard deviation,only 2 out of 5 (2 out of 6) years show a large-scale cooling (warming) in the central-eastern Pacific (figure not shown). It is hypothesized that the local forcing of the SSTA in the WMC and EMC may play more essential roles in inducing this out-of-phase CEF mode, compared with the SSTA in the central-eastern Pacific, since EOF2 could still exist with the absence of the SST forcing in the central-eastern Pacific.

3.3. Role of diabatic heating in the out-of-phase mode of CEFs

It has been demonstrated that atmospheric diabatic heating, which is tightly connected with SST change, serves as a direct driver of atmospheric circulation. Both theoretical and observational analyses demonstrate that tropical heating asymmetric about the equator feeds a CEF into the heating hemisphere, which may influence the CEFs ( Gill, 1980 ;Zhang and Krishnamurti, 1996 ; Lee et al., 2009 ).

Upon further examination of the regressed diabatic heating anomaly against PC2, it is found that the warm (cool) SSTA in the EMC (WMC)is associated with an anomalous heating source (sink) center lying at 5°S and 150°E (7°S and 95°E), which is asymmetric about the equator( Fig. 2 (a)). Note that there is a significant heating sink located over the equatorial central Pacific (CP), closely related to the local cooling SSTA( Fig. 1 (h)).

Fig. 2. Regression maps of the (a) vertically integrated diabatic heating anomaly (shading; units: K d —1 ) onto PC2 and (b) 925 hPa meridional wind(shading; units: m s —1 ), horizontal wind (vectors; unit: m s —1 ), and 500 hPa vertical velocity anomaly (contours; units: Pa s —1 ) onto the tripole index (TI). In (a),only values significant at the 90% confidence level are shown. In (b), stippled areas (black vectors) represent statistical significance of the meridional wind(horizontal wind) at the 90% confidence level.

A tripole index (TI) is constructed for a better understanding of the relationship between the tropical heating and the out-of-phase mode of CEFs. The TI is calculated by the following formula:

where EMCQ1is the vertically integrated diabatic heating anomaly areaaveraged over the EMC (20°S—0°, 140°—170°E), WMCQ1is the average over the WMC (20°S—0°, 85°—110°E), and CPQ1represents the areaaverage over the CP (15°S—10°N, 175°E—120°W).

The correlation coefficient between TI and PC2 is 0.90, with statistical significance at the 99% confidence level ( Table 1 ), and that between TI and PC1 is near zero. This suggests a very close linkage between the tripole pattern of tropical diabatic heating and the out-of-phase CEF mode, which is further supported by the regression map of the meridional wind anomaly against the TI ( Fig. 2 (b)). Thus, this tripole pattern of tropical diabatic heating may be one of the important factors that results in the out-of-phase relationship between the BOB and New Guinea CEFs.

Table 1 Correlation coefficients among different diabatic heating indices and PC2 during the period 1979—2017. The WMCQ1 and CPQ1 are multiplied by ? 1 for a better comparison with TI. The values in brackets represent the partial correlation coefficients. An asterisk denotes that the value is significant at the 99% confidence level.

To further unravel the individual effect of the tropical diabatic heating over three key regions, a partial correlation technique is applied as in Ashok et al. (2007) . For a better comparison with Fig. 2 (b), the WMCQ1and CPQ1are multiplied by ? 1 when performing the partial correlation analysis. After excluding the impact of EMCQ1and CPQ1, there are anomalous southerlies dominating over the WMC in correspondence to a local heat sink anomaly ( Fig. 3 (a)), suggesting a local effect of the WMC heating on the BOB CEF near 90°E and SCS CEF near 105°E. Consistently, the partial correlation between WMCQ1multiplied by ? 1 and PC2 increases to 0.79 ( Table 1 ), partly contributing to the out-of-phase mode of the CEFs.

Fig. 3. Partial correlation of the 925 hPa meridional wind anomaly (shading;units: m s —1 ) and horizontal wind anomaly (vectors; unit: m s —1 ) with the (a)WMCQ 1 multiplied by ? 1, (b) EMCQ 1 and (c) CPQ 1 multiplied by ? 1. The stippled areas (black vectors) represent statistical significance of the meridional wind (horizontal wind) at the 90% confidence level.

Similarly, the heating source over the EMC is related to the southerly anomaly over Indonesia and northerly anomaly east of Indonesia with a maximum correlation of ? 0.5 near 160°E, which indicates a significant influence of EMC heating on the New Guinea CEF ( Fig. 3 (b)). Furthermore, when the focus is on the individual impact of CPQ1, negative correlations appear over the tropical MC between 110°E and 160°E, accompanied by positive correlations lying on its western and eastern flank( Fig. 3 (c)). Although such a pattern shows opposite signs of correlation near 100°E and 160°E, it somehow differs from the out-of-phase mode of the CEFs ( Fig. 1 (f)), especially the negative correlations between 110°E and 145°E. Plausibly, CPQ1can help weaken the CEF channels east of 110°E, but only slightly influence the BOB CEF around 90°E. As mentioned in Section 3.2 , CPQ1cannot always take effect and might only promote an out-of-phase variation of the BOB and New Guinea CEFs in some specific years.

3.4. Numerical results

Four simulation experiments were designed based on the LBM to confirm the effect of diabatic heating in each key domain on the outof-phase CEF mode. The results are shown in Fig. 4 . According to the spatial distributions of the observational tripole pattern of the diabatic heating over the tropical Pacific ( Fig. 2 (b)), the heating sink (source)is placed onto the WMC/CP (EMC) with a heating rate at the center of? 2.4 K d?1/ ? 2.1 K d?1(2.4 K d?1) at the sigma level of 0.45. Since the LBM is linear, the responses of the joint impact simulation are actually the results adding the first three numerical experiments together.

As expected, the response of meridional wind to the heat sink over the WMC shows a significant southerly anomaly in-situ ( Fig. 4 (a)), consistent with the Gill-type response of a CEF into the heating hemisphere to the heating asymmetric about the equator. Furthermore, the result confirms the strengthened effect of the WMC heat sink on the CEFs between 90°E and 110°E, as presented in the observational analysis( Fig. 3 (a)).

Fig. 4. Numerical responses of the 950 hPa meridional wind anomaly (shading; units: m s —1 ) and horizontal wind (vectors; units: m s —1 ) to the heat forcing (contours;units: K d —1 ) over the (a) WMC, (b) EMC, (c) CP, and (d) three key regions imposed into the LBM at the sigma level of 0.45.

By contrast, as a response to the heat source over the EMC, strong northerly anomalies appear between 145°E and 165°E south of the equator ( Fig. 4 (b)), implying a significant weakening of the New Guinea CEF.However, some discrepancies are observed between the observational and simulation results —in particular, the meridional wind anomalies between 100°E and 140°E ( Figs. 3 (b) and 4 (b)). The model response suggests a more significant in-situ meridional flow to the heat source over the EMC.

The meridional wind response to the CP heat sink, with an amplitude smaller than 0.2 m s?1, is quite weak over the AAM region ( Fig. 4 (c)).A cross-equator response disappears to be replaced by a Kelvin wave response with an equatorial easterly to the east of the heat sink, together with a weak Rossby wave response with a pair of anticyclones to the northwest and southwest of the heat sink. The weak impact of the heat sink over the CP may be attributable to the heat forcing, which is symmetric about the equator. This cannot induce the CEF in a simplified AGCM model, on which the LBM is constructed. The discrepancies between the observation and model results lead us to speculate that the heating over the CP may exert an indirect impact on the CEFs via modulation of the zonal atmospheric circulation near the MC. Such an indirect pathway may to some extent result in the significant correlation between CPQ1and the out-of-phase mode of CEFs.

By linear superposition, the atmospheric response to the tripole pattern of diabatic heating is shown as Fig. 4 (d). This indicates that the tropical heating can contribute significantly to the out-of-phase relationship between the BOB and New Guinea CEFs. By superposing the responses of the WMC and EMC cases (figure not shown), it is found that the equatorially asymmetric heating over these two domains may hold the key to the out-of-phase mode of the CEFs, whereas the equatorially symmetric heating over the CP has a limited impact via directly inducing cross-equator wind.

4. Summary and discussion

Based on EOF analysis, the second mode of the 925 hPa meridional wind averaged between 5°S and 5°N presents an out-of-phase pattern with opposite meridional wind anomalies over the WMC and EMC.

It is found that EOF2 is characterized by an enhanced BOB CEF and a weakened New Guinea CEF when its corresponding component is positive. There is a significant warming anomaly in the EMC and cooling in the WMC, together with a large-scale cooling lying in the tropical central-eastern Pacific. Correspondingly, a tripole pattern of tropical heating is observed, which is right above the SSTA centers. Moreover,the heating anomalies east/west of the MC are both asymmetric about the equator, which might contribute to the out-of-phase relationship between the BOB and New Guinea CEFs via inducing the CEF according to the Gill-type response.

With respect to the heating over the CP, some disagreements between the observation and simulation are apparent. The TI possesses a correlation coefficient of 0.90 with PC2; however, the correlation decreases to 0.73 if another heating index is constructed by calculating the difference between WMCQ1and EMCQ1after excluding the variability of CPQ1. Thus, the diabatic heating over the CP is believed to partly contribute to the out-of-phase mode of the CEFs. Nevertheless,the heat-induced meridional wind is quite weak in the LBM experiment.Consequently, we hypothesize that the CP heating anomaly may indirectly affect the CEFs by modulating the atmospheric circulation near the MC. Note that the heat forcing in the CP cannot always take effect,and hence its counterparts near the MC are most essential in leading to such an out-of-phase CEF mode in summer.

SST forcing is regarded as one of the crucial factors causing change in tropical diabatic heating. Moreover, previous studies suggest that the Indonesian Throughflow plays an important role in connecting the Pacific and Indian Ocean ( Meng and Wu, 2000 ; Wu et al., 2010 ), which may affect the out-of-phase relationship between the BOB and Australian CEFs.This issue remains unclear and requires further study.

Funding

This research was jointly supported by the National Key Research and Development Program of China [grant number 2016YFA0600601]and the National Natural Science Foundation of China [grant numbers 42030601 and 41875087 ].

Acknowledgments

The authors deeply appreciate the editor’s and two reviewers’ comments and suggestions, which helped to improve this manuscript. Special thanks go to Dr. Yanke Tan, Dr. Ruifen Zhan, and Dr. Jiacan Yuan for inspiring discussion in our group seminar.