HTML
--> --> -->Many previous studies have focused on investigating the long-term trend of the zonally averaged HC (ZAHC) regarding its intensity and width. For the width of the ZAHC, although different studies have shown various magnitudes of the width trend, they have all agreed on an expansion of the ZAHC (e.g., Fu et al., 2006; Hu and Fu, 2007; Hu et al., 2011, 2018; Davis and Rosenlof, 2012; Lucas et al., 2014; Nguyen et al., 2015). For the strength of the ZAHC, however, considerable spread is observed across different datasets (Stachnik and Schumacher, 2011; Nguyen et al., 2013; Chen et al., 2014). For example, Nguyen et al. (2013) showed that the trends of the ZAHC intensity from various datasets are inconsistent, both within a given season and for the annual mean. Results regarding the intensity of the long-term ZAHC trend remain controversial.
In view of the uneven distribution of the land, oceans, and topography, the ZAHC could mask the regional difference, thereby overlooking its spatial variance at the regional scale (Chen et al., 2014; Schwendike et al., 2014; Nguyen et al., 2018). The ZAHC’s intensity is found to be stronger than normal with El Ni?o events (Oort and Yienger, 1996; Guo et al., 2016), while the accompanying regional HC intensities tend to be below-normal over the Indian Ocean and South America (Freitas and Ambrizzi, 2015; Freitas et al., 2017) and above-normal in the eastern Pacific (Wang, 2002; Zeng et al., 2011). Previous studies have also indicated that the regional HC could exert significant influences on regional climate variability (e.g., Wang, 2002; Zhang and Wang, 2013, 2015; Chen et al., 2014; Freitas and Ambrizzi, 2015; Freitas et al., 2017). For example, Zhang and Wang (2013, 2015) demonstrated that the interannual variability of the regional HCs over the Atlantic and the northeastern Pacific strongly modulates the tropical cyclone activities in these two regions. Huang et al. (2018a) found that the interannual variability of the regional HC over the western Pacific (WPHC) has pronounced influences on the precipitation and temperature anomalies over the Asian and Australian continents.
Chen et al. (2014) attempted to detect the longitudinally resolved HC intensity trend by separating the tropical zonal belt into six regions. They found that uncertainty remains in these regions regarding the long-term trend of the regional HC intensities among six reanalysis datasets. Notably, however, annual mean data were examined in their study, which makes it difficult to detail the various trends and explore the possible influences of the regional HC on climate extremes at shorter time scales. Freitas et al. (2017) further investigated the long-term trend of the regional HC intensity over the Indian Ocean during boreal summer and autumn, and found that there is a significant strengthening trend of the regional HC over there. Note, however, that in their study, they only used one reanalysis dataset, which raises a question about the robustness of the results. Overall, relatively few studies have been conducted on the seasonality of the regional HC intensity trend, particularly the trends in transition seasons, such as boreal spring.
Boreal spring (March–May, MAM) is a season with pronounced climate anomalies. The climatological mean HC tends to be equatorially symmetric during MAM, different from that in summer and winter (Oort and Rasmusson, 1970; Dima and Wallace, 2003). Feng et al. (2013) indicated that the robust warming trends of sea surface temperature (SST) over the Indo-Pacific region contribute to the long-term variation of the MAM HC. However, they mainly focused on the zonal mean HC. Guo and Tan (2018) investigated the variability of the regional HC intensity over the Indo-Pacific warm pool area during boreal spring, which has been found to significantly impact the genesis and intensity of the tropical cyclone activity over the western North Pacific in the following summer. Note, however, that their study focused on interannual time scales. Huang et al. (2018a) also demonstrated that the WPHC, with its rising branch located over the Indo-Pacific warm pool region, could exert significant influences on the climate anomalies over Asia and Australia. In addition, a predictability barrier of the tropical El Ni?o–Southern Oscillation (ENSO) exists in spring (Webster and Yang, 1992). Therefore, further investigation of the WPHC variability and trend during MAM is worthwhile, and has important implications in understanding Asian–Australian climate anomalies and change.
In this study, the long-term trend of the WPHC intensity during boreal spring is investigated using six reanalysis datasets. The data employed are described in section 2. Section 3 presents the long-term trends of the spring WPHC intensity in the six reanalyses. We characterize the WPHC by examining the spatial structure of the meridional mass streamfunction, its central value, and the two-dimensional winds of the vertical section to obtain robust results of the WPHC trend. The factor responsible for the long-term strengthening of the spring WPHC during recent decades is also discussed. Section 4 provides conclusions and discussion.
| NCEP1 | NCEP2 | ERA-Interim | JRA-55 | 20CR | CFSR | |||
| Vertical pressure levels | 17 | 17 | 37 | 37 | 24 | 22 | ||
| Horizontal resolution | 2.5° × 2.5° | 2.5° × 2.5° | 2.5° × 2.5° | 1.25° × 1.25° | 2° × 2° | 0.5° × 0.5° | ||
| Data range | 1979–2016 | 1979–2016 | 1979–2016 | 1979–2016 | 1979–2014 | 1979–2012 | ||
| Reference | Kalnay et al. (1996) | Kanamitsu et al. (2002) | Dee et al. (2011) | Ebita et al. (2011) | Compo et al. (2011) | Saha et al. (2010) | ||
| WPHCI | NH | NCEP1 | 0.889 | 0.768 | 0.662 | 0.676 | 0.677 | |
| NCEP2 | 0.765 | 0.736 | 0.555 | 0.703 | ||||
| ERA-Interim | 0.896 | 0.587 | 0.813 | |||||
| JRA-55 | 0.422 | 0.846 | ||||||
| 20CR | 0.521 | |||||||
| SH | NCEP1 | 0.928 | 0.948 | 0.938 | 0.923 | 0.927 | ||
| NCEP2 | 0.898 | 0.931 | 0.896 | 0.915 | ||||
| ERA-Interim | 0.955 | 0.899 | 0.917 | |||||
| JRA-55 | 0.933 | 0.944 | ||||||
| 20CR | 0.911 | |||||||
| Notes: All correlation coefficients exceed the 95% confidence level. All reanalysis datasets are computed over the same interval (1979–2012) when calculating the correlations. | ||||||||
Table1. Reanalysis data used and correlation coefficients of the WPHCI in the NH and SH among these datasets.
where
Figure 1 displays the climatology (contours) and long-term trend (shading) of the WPHC during boreal spring. The MAM climatology of the WPHC is similar among the six reanalyses. It reveals a roughly symmetric and magnitude-comparable two-cell pattern, with the ascending branch located around the equator and descending branches around 30° latitude in each hemisphere, consistent with previous studies (e.g., Guo and Tan, 2018). The pattern also resembles the climatological structure of the boreal spring ZAHC (Dima and Wallace, 2003; Feng et al., 2013), which further verifies that the approach of the meridional mass streamfunction is also useful for defining and assessing the variations of the regionally averaged HC (e.g., Zhang and Wang, 2013, 2015; Schwendike et al., 2014; Huang et al., 2018b; Nguyen et al., 2018). For the long-term trend of the WPHC, all the reanalyses exhibit significant positive (negative) ψ trends in the Northern (Southern) Hemisphere extending throughout the troposphere. This suggests a strengthening of the WPHC during boreal spring in both hemispheres.
Figure1. Climatology (contours; units: 1010 kg s?1) and the linear trends [shading; units: 1010 kg s?1 (10 yr?1)] of the boreal spring regional meridional mass streamfunction over the western Pacific (110°–160°E) in (a) NCEP1, (b) NCEP2, (c) ERA-Interim, (d) JRA-55, (e) 20CR, and (f) CFSR. Positive (negative) values are indicated by solid (dashed) contours representing clockwise (counterclockwise) circulations. The contour intervals are 3.5 × 1010 kg s?1. Stippled regions denote trends that are significantly different from zero at the 90% confidence level.To further quantify the robust strengthening of the WPHC intensity, we define a regional circulation intensity index to capture the long-term trend of the WPHC intensity (WPHCI), following Oort and Yienger (1996) in defining the zonal mean HC. Specifically, the WPHCI in the Northern (Southern) Hemisphere is defined as the maximum (minimum) of the meridional mass streamfunction of the WPHC between 30°N (30°S) and the equator. Figure 2 shows the time series of the WPHCI in the two hemispheres derived from the six datasets over their respective time periods. It is clear from Fig. 2 that the interannual evolution of the WPHCI is similar among the reanalyses. However, a higher consistency of the interannual WPHCI variability among the datasets is seen in the Southern Hemisphere (SH) compared to that in the Northern Hemisphere (NH) (Fig. 2). Correlations of the WPHCI between individual reanalysis datasets are given in Table 1. The WPHCIs from the six reanalyses are highly correlated with each other—particularly in the SH, where the correlation coefficients exceed 0.89 (Table 1). Besides, the WPHCI in the SH exhibits evident variability on decadal time scales (Fig. 2b). An obvious strengthening is seen around 2000, which may be associated with the cold ENSO/Pacific Decadal Oscillation (PDO)-like SST pattern in recent decades (Fig. 2b), e.g., Wang and Liu, 2016. Furthermore, previous studies have suggested that the PDO variability contributes to the tropical expansion (e.g., Allen and Kovilakam, 2017; Grise et al., 2019). Therefore, whether the PDO is also a forcing factor for the decadal-scale variability of the WPHCI over the SH is an interesting issue and deserves further investigation in the future.
Figure2. Time series of the boreal spring WPHC intensity in the (a) NH and (b) SH for the six reanalyses.Figure 3 shows the uncertainty of long-term WPHCI trends for the six reanalyses over their respective time periods. In the NH, all reanalyses show a statistically significant intensification of the WPHCI. In the SH, a pronounced intensified trend of the WPHCI appears in most datasets, except that in ERA-Interim. Nevertheless, the trend in ERA-Interim is still marginally significant at the 90% confidence level (Fig. 3). The statistical results remain the same when we compute the trends over the overlapping period (1979–2012) for the six datasets. Besides, in order to further quantify the intensification of the WPHC, Table 2 shows the relative changes of the intensity for the WPHC from 1979 to 2012 over the NH and SH in the six reanalysis datasets. Note that the relative changes of the intensity for the WPHC are calculated according to the following three steps. First, we obtain the linear fitting lines for the WPHCIs in the six reanalysis datasets based on the least-squares method. Second, we calculate the difference through using the value of the linear fitting line for the WPHCI in the last year minus that in the first year. Finally, the relative changes of the intensity for the WPHC are defined as the ratio of the difference derived from the second step to the value of the linear fitting line for the WPHCI in the first year. It is clear from Table 2 that all reanalyses show a positive relative change. Specifically, the multiple reanalysis ensemble mean (MEM) displays an intensification from 1979 to 2012 with a 61% relative change of the intensity for the WPHC in the NH, and a 36% relative change in the SH.
Figure3. Linear trends of the boreal spring WPHCI in the (a) NH and (b) SH for the six reanalyses. The error bars of the 90% confidence interval are estimated based on two-tailed Student’s t-tests. Units of the trends are 1010 kg s?1 (10 yr)?1. Positive (negative) trend values indicate strengthening of the NH (SH) WPHC.| NCEP1 | NCEP2 | ERA-Interim | JRA-55 | 20CR | CFSR | MEM | ||
| WPHCI | NH | 59.1% | 62.9% | 27.8% | 87.6% | 41.1% | 57.5% | 61% |
| SH | 46.4% | 29.9% | 26.1% | 53.8% | 36.3% | 16.3% | 36% | |
| Note: The MEM represents the multiple reanalysis ensemble mean. | ||||||||
Table2. Relative changes in the intensity of the WPHC from 1979 to 2012 over the NH and SH in the six reanalysis datasets.
To further confirm the reliability of the results derived from Figs. 1 and 3, which use the meridional mass streamfunction to describe the WPHC structure and define the WPHCI, Fig. 4 displays the long-term trend of the WPHC calculated from the two-dimensional wind fields of the vertical section during boreal spring. The WPHC trends show a largely similar spatial pattern compared to its climatology in all the reanalyses (compare Fig. 4 with Fig. 1), confirming that there is a significant tendency of WPHC intensification in all the datasets. In particular, there are pronounced upward trends over the tropics and downward trends over the subtropics, strengthening the climatological circulation. In addition, it is noted that there are slight differences in the NCEP1 and NCEP2 datasets, with the pattern of the WPHC trend shifted slightly southward in the NH, than those in the other four reanalyses, consistent with Figs. 1a and 1b. In summary, the results obtained from the meridional mass streamfunction and directly from the two-dimensional wind fields of the vertical section agree well, and suggest that all the reanalyses considered have pronounced intensification tendencies of the WPHC intensity in both hemispheres during boreal spring.
Figure4. Linear trends of the boreal spring WPHC from 1979 to 2016 in (a) NCEP1, (b) NCEP2, (c) ERA-Interim and (d) JRA-55; from 1979 to 2014 in (e) 20CR; and from 1979 to 2012 in (f) CFSR. Vertical velocities have been multiplied by ?100 in the figure, with positive (negative) values indicating ascending (descending) motion. Units for the trends of the divergent meridional wind are m s?1 (10 yr)?1 and those for vertical velocity are 10?2 Pa s?1 (10 yr)?1. Shading indicates the trends that are significantly different from zero at the 90% confidence level.A question then needs to be addressed is why the WPHC intensity during boreal spring has been strengthening during recent decades? Given that the HC is a thermally driven meridional circulation, the variation of the HC is closely linked to the underlying thermal structure in the tropics (e.g., Allen and Kovilakam, 2017; Feng et al., 2018a, b; Grise et al., 2019). For example, Feng et al. (2018a, b) suggested that it is the meridional SST structures that play an important role in impacting the response of the HC to SST. In addition, previous studies have shown that the recent HC widening is driven by the tropical SST (e.g., Allen and Kovilakam, 2017; Grise et al., 2019). Therefore, regarding the strengthening of the MAM WPHC intensity, one potential reason may be the significant warming of the SST over the tropics—especially the SST warming in the Indo-western Pacific region (Quan et al., 2004; Feng et al., 2013). Figure 5 shows long-term trend of SST during boreal spring derived from the four SST datasets. It is clear from Fig. 5 that the long-term trends of SST are similar among the four SST datasets, with pronounced warming in the tropical Indian Ocean, tropical Atlantic, and tropical western Pacific. The warming trends are also found in the subtropics of the central North and South Pacific (Fig. 5), consistent with the findings of previous studies (e.g., Deser et al., 2010; Sun et al., 2017).
Figure5. Linear trends of the boreal spring SST from 1979 to 2016 in (a) ERSST.v3, (b) HadISST, (c) KaplanSST, and from 1982 to 2016 in (d) OISSTv2. Units for the trends of SST are °C (10 yr)?1. Stippling denotes trends significantly different from zero at the 95% confidence level. Regions A, B, and C, indicated by the black boxes and selected to further investigate the trend of SST in some specific regions, are (5°S–15°N, 120°–160°E), (20°–35°N, 180°–150°W), and (15°S–30°S, 180°–150°W), respectively.To further investigate the relationship between the SST trends and the long-term trends of the WPHC intensity, we also display the spatial distributions of the correlations between the WPHCI of both hemispheres and the boreal spring SST (Fig. 6). From Fig. 6, it is apparent that there are similar spatial patterns among different datasets (e.g., compare Figs. 6a–d with Figs. 6e–h). Furthermore, the spatial distributions of the correlation coefficients between the boreal spring WPHCI in both hemispheres and the boreal spring SST also show a similar pattern, with significant positive correlations over the tropical western Pacific and the subtropics of the central North and South Pacific (Fig. 6). These areas of significant positive correlation are coincident with the significant warming areas of the SST trends, indicating strengthening of the boreal spring WPHCI is connected to the warming of the SST over those regions (Fig. 5 and Fig. 6). Furthermore, in a previous study, Zeng et al. (2011) found that an anomalous enhanced WPHC over the NH is accompanied by an enhanced midlatitude zonal cell characterized by air parcels rising in the central North Pacific and descending in the western North Pacific. That is, the ascent around the tropical western Pacific and the central North Pacific makes a positive contribution to the strengthening of the WPHC over the NH (Zeng et al., 2011). Therefore, according to Fig. 5 and Fig. 6 and the aforementioned previous study, we select several regions for further investigation. Regions A, B, and C represent the tropical western Pacific, the subtropical central North Pacific, and the subtropical central South Pacific, respectively (Fig. 5 and Fig. 6). Note that the robustness of the results reported below is not sensitive to a reasonable change of the regions selected (figure not shown).
Figure6. Spatial distribution of the correlation coefficients between the boreal spring WPHCI over the NH obtained from the multiple reanalysis ensemble mean and the boreal spring SST anomalies in (a) ERSST.v3, (b) HadISST, (c) KaplanSST, and (d) OISSTv2. Stippling denotes correlation coefficients significantly different from zero at the 95% confidence level. (e–h) As in (a–d), but for anomalies correlated upon the minus one WPHCI over the SH. Definitions of Regions A, B, and C are the same as in Fig. 5.Figure 7 displays time series and the long-term trends of the regionally averaged SST over Regions A, B, and C derived from the four SST datasets. It is clear that the interannual evolution of the SST over these regions is similar among the four SST datasets (Fig. 7). Besides, all datasets show a statistically significant warming trend over Regions A, B, and C (Fig. 7), corresponding well to the findings obtained in Fig. 5. Therefore, the warming of the SST over these regions contributes positively to the strengthening of the WPHC intensity during boreal spring (Figs. 5, 6, and 7). Particularly, the WPHCI over the NH is highly correlated with that over the SH, with correlations of about ?0.56, significant at the 99% confidence level. Hence, the warming of the tropical western Pacific plays a key role in the pronounced intensification tendencies of the WPHC through influencing the mutual ascending branch of the WPHC in both hemispheres (Figs. 5, 6, and 7). In addition, the warming in the subtropics of the central North and South Pacific (Figs. 5, 6, and 7) also makes some positive contributions to the strengthening of the WPHC, with an enhanced rising branch over there, flowing westward aloft, and sinking in the subtropical western Pacific, similar to previous studies (e.g., Zeng et al., 2011). By contrast, the warming in the tropical Indian Ocean makes a negative contribution to the long-term trend of the WPHC intensity, as the WPHCI has a negative correlation with the SST in the Indian Ocean (Fig. 5 and Fig. 6). The warming SST produces increased ascending motion over the Indian Ocean, which flows eastward aloft through the tropical Walker circulation and then sinks in the tropical western Pacific, weakening the intensity of the WPHC.
Figure7. Normalized time series of the boreal spring SST (units: °C) averaged over (a) Region A, (b) Region B, and (c) Region C, obtained from four SST datasets. (d) Linear trends (units: 10 yr?1) of normalized time series in (a–c). The error bars of the 95% confidence intervals are also shown in (d). A positive trend value indicates warming of the SST. Definitions of Regions A, B, and C are the same as in Fig. 5.It is noted that some other factors may also play a role in contributing to the long-term trend of the WPHC intensity during boreal spring, such as changes in the global SST warming pattern (Feng et al., 2013), the meridional temperature gradient (Seo et al., 2014), and the gross static stability (Ma et al., 2012). In addition, the increasing concentrations of greenhouse gases stemming from human activities may also be contributing to the long-term trend of the WPHC (Tanaka et al., 2004). The relative contributions of the internal climate variability and external forcing to the long-term trend of the WPHC remain to be explored.
This study shows that there is good agreement among the six reanalyses in capturing the boreal spring WPHC intensity trends, which is in sharp contrast to previous studies that concentrated on the zonal-mean HC and showed large discrepancies among reanalyses. Thus, the result obtained from this study may improve understanding of the HC variability at the regional scale. Note that this study focuses on the WPHC during boreal spring; the long-term trends of the WPHC intensity during other seasons is still unclear, which is worthy of further discussion. Figure 8 displays the time series and long-term trends of the boreal summer WPHCI in the two hemispheres derived from the six reanalysis datasets over their respective time periods. It is clear from Fig. 8 that, although the interannual WPHCI variability during boreal summer is similar among different reanalyses, discrepancies remain regarding the long-term trends in the boreal summer WPHC intensity (Fig. 8). Similarly, the boreal autumn interannual variation of the WPHCI is consistently captured by all the reanalyses (Fig. 9), but the trends display inconsistency among different datasets (Fig. 9). In addition, a previous study showed that uncertainty remains concerning the long-term trends in the boreal winter WPHC intensity (Huang et al., 2018b). Overall, unlike boreal spring (Fig. 2 and Fig. 3), the WPHC intensity during other seasons (boreal summer, autumn, and winter) to a certain extent shows no significant discernible long-term trend, although some datasets display pronounced tendencies (Fig. 8, Fig. 9; Huang et al., 2018b), which highlights a need to investigate the long-term trend of the WPHC intensity at the seasonal scale. Besides, the forcing factor responsible for the trends of the WPHC intensity during other seasons is also an interesting issue deserving of further investigation in the future. In addition, aside from the WPHC, several previous studies (e.g., Chen et al., 2014; Freitas et al., 2017) have focused on the long-term trends of the regional HC intensity over other regions. This is also important and remains to be further explored in the future.
Figure8. Time series of the boreal summer WPHC intensity in the (a) NH and (b) SH for the six reanalyses. (c, d) As in (a, b), but for the linear trends of the boreal summer WPHCI for the six reanalyses. The error bars of the 90% confidence interval are estimated based on two-tailed Student’s t-tests. Units of the time series and the trends are 1010 kg s?1 and 1010 kg s?1 (10 yr)?1, respectively. Positive (negative) trend values indicate strengthening of the NH (SH) WPHC.
Figure9. As in Fig. 8 but for boreal autumn.The investigations noted above focused on the long-term trends of the WPHC and its intensity. However, the variability of the width of the WPHC is still unclear. Some recent studies have suggested that poleward expansion of the HC has distinct regional signatures, implying widening of the HC does not occur at all longitudes (e.g., Chen et al., 2014; Grise et al., 2018; Nguyen et al., 2018; Staten et al., 2019). Thus, it is interesting to further investigate whether the boreal spring WPHC is widening. According to previous studies (e.g., Nguyen et al., 2018) we know that there are some potential difficulties in defining the poleward edges of the regional HC, because it is not clearly bound in the subtropics by Ferrel cells from a regional view (Fig. 1). Therefore, although the edge of the zonal-mean HC is commonly characterized as the poleward extent of the sinking branch and defined as the latitude where the streamfunction falls to zero (e.g., Oort and Yienger, 1996), the streamfunction for the regional HC does not necessarily equal zero at its poleward edge (e.g., Nguyen et al., 2018). Hence, following the approach of Nguyen et al. (2018), we define the edge of the WPHC as where the streamfunction falls to a specified percentage of the maximum streamfunction value in the tropics. After multiple attempts, we find that it is hard to detect the NH boundary of the WPHC (e.g., the NH cell in Fig. 1), indicating that quantifying the width of the HC regionally remains an ongoing research question. Therefore, we only show the long-term trend of the SH boundary of the WPHC in the following.
Figures 10a and c display the long-term trends of the WPHC edge in the SH for the six reanalyses over their respective time periods. Figures 10b and d are as in Figs. 10a and c, but for different thresholds. Figures 10a and c show the edge of the WPHC in the SH is defined as the latitude where the streamfunction values first reach 20% of the peak value at 500 hPa, but a 30% threshold is used in Figs. 10b and d. Note that different thresholds (e.g., 10%, 15%, 20%, 25%, 30%, and 35%) were tested and the results were not sensitive to the choice of threshold. In addition, the expansion trends of the WPHC over the SH are also not sensitive to the selection of the layers (e.g., 400 hPa, 500 hPa, 600 hPa, and 700 hPa) (figure not shown). Since a higher value ensures a more robust detection of the edge of the WPHC in the SH without impacting its long-term trend, the 20% and 30% thresholds are used (Fig. 10). From Fig. 10, we can see that the year-to-year variability in the extent of the WPHC over the SH is similar among different reanalyses. All reanalyses show a poleward expansion of the SH WPHC, but there is no significant discernible trend over their respective periods in the extent of the WPHC over the SH (Fig. 10). Note that, although different reanalyses have different temporal coverage, the results do not change when the trends in all datasets are computed over the same interval (1979–2012) (figure not shown). The results in Fig. 10 suggest that, although the zonal-mean HC has clearly been expanding (e.g., Fu et al., 2006; Hu and Fu, 2007; Hu et al., 2011; Lucas et al., 2014; Nguyen et al., 2015), the long-term trends of the regional HC poleward edges could have regional variability, similar to conclusions reached in previous studies (e.g., Chen et al., 2014; Grise et al., 2018; Nguyen et al., 2018; Staten et al., 2019). This highlights the importance to understand tropical belt changes regionally.
Figure10. (a) Time series of the latitudes where the streamfunction values first reach 20% of the peak value at 500 hPa in the southern subtropics for the SH WPHC during boreal spring obtained from six reanalyses. (c) As in (a) but for the linear trends for six reanalysis datasets over their respective time periods. The error bars of the 90% confidence intervals according to a two-tailed Student’s t-test are also shown in (c). Units for (a, c): degrees of latitude and degrees of latitude per decade, respectively. Negative (positive) trend values indicate a poleward (an equatorward) expansion (contraction) of the SH WPHC. (b, d) As in (a, c) but with a 30% threshold used, indicating that the edge of the SH WPHC is defined as the latitudes where the streamfunction values first reach 30% of the peak value at 500 hPa in the southern subtropics.Acknowledgments. We thank the two anonymous reviewers and the Editor for their constructive suggestions, which helped to improve the paper. This study was jointly supported by the National Key Research and Development Program of China (Grant No. 2016YFA0600604), the National Natural Science Foundation of China (Grant Nos. 41605050, 41721004, and 41530425), the Chinese Academy of Sciences Key Research Program of Frontier Sciences (Grant No. QYZDY-SSW-DQC024), and the Guangdong Province Science and Technology Project (Grant No. 2017B020244002). The NCEP1, NCEP2 and 20CR data were derived from
