Abstract

Changes of the Southern Hemisphere Hadley cell over the twentieth century are investigated using the Twentieth Century Reanalysis (20CR) and coupled model simulations from phase 5 of the Coupled Model Intercomparison Project (CMIP5). Trends computed on a 30-yr sliding window on the 20CR dataset reveal a statistically significant expansion of the Hadley cell from 1968 forced by an increasing surface global warming. This expansion is strongly associated with the intensification and poleward shift of the subtropical dry zone, which potentially explain the increasing trends of droughts in the subtropical regions such as southern Australia, South America, and Africa. Coupled models from the CMIP5 do not adequately simulate the observed amount of the Hadley expansion, only showing an average of one-fourth of the expansion as determined from the 20CR and only when simulations include greenhouse gas forcing as opposed to simulations including natural forcing only.

1. Introduction

The Hadley circulation dominates tropical and subtropical climate across Earth at both short-term and longer-term time scales. At its simplest, the energy transport from the low latitudes is achieved by the rising motion of warm air near the equator, a cooling and poleward motion in the upper reaches of the troposphere, broad subsidence in the subtropics, and a return flow toward the equator near Earth’s surface; these general circulation features form a closed cell when averaged across all longitudes. These Hadley cells characterize the persistence and extent of the trade winds and the subtropical high pressure belt that dominate tropical and subtropical climates (Webster 2004). The cells are variable, with seasonal and interannual fluctuations in their strength and location (e.g., Oort and Yienger 1996; Nguyen et al. 2013). There is also evidence of a gradual expansion of the Hadley cells (Fu et al. 2006; Seidel et al. 2008), with the location of the descending branch of the Hadley cells progressively moving poleward during recent decades (Lucas et al. 2014). The changes to the Hadley circulation and its implications for regions whose weather and climate are influenced by these features are significant for future resource management decisions. With the majority of arid regions across Earth located beneath the subsiding branches of the Hadley cells, such an expansion suggests a broadening of the subtropical dry zone leading to a drier subtropical climate.

In the Southern Hemisphere (SH), many studies (e.g., Perlwitz et al. 2008; Johanson and Fu 2009; McLandress et al. 2011; Polvani et al. 2011; Min and Son 2013) suggested that stratospheric ozone depletion is the main driver for the Hadley cell expansion during austral summer [December–February (DJF)], when following the annual peak of ozone loss (September–October), for the past few decades (since 1979) that include the period of maximum ozone depletion (1979–2000). Nguyen et al. (2013) showed that the Hadley cell expansion trends were most salient in summer and autumn. A recent modeling study (Allen et al. 2014) suggests that greenhouse gas and ozone forcings have equal effect on the annual mean Hadley cell expansion trend, but without evident linearity of the climate response, contradicting the results of Polvani et al. (2011) and Staten et al. (2012). Allen et al. (2014) also show that forcing from observed sea surface temperature (SST) is also significant, consistent with Staten et al. (2012). A synthesis of these results is that both factors (ozone depletion and increased greenhouse gas emission) largely contribute to tropical expansion, but quantifying the role of each individual forcing remains a challenge given the inability of climate models to estimate the correct amount of expansion and the large spread between models (Johanson and Fu 2009; Hu et al. 2013).

There is considerable evidence that changes in the Hadley circulation (HC) are behind an increasing trend of droughts in subtropical regions, such as southern Australia (CSIRO 2012; IOCI 2012), the Altiplano in South America (Morales et al. 2012), Africa, and Southeast Asia (Dai 2013). Fyfe et al. (2012) indicated that drying trends since 1957 in the Southern Hemisphere summer midlatitudes were driven mainly by greenhouse gas forcing. Ozone forcing was secondary, and largely balanced by aerosol forcing. Understanding the source of changes to the HC provides possible insight into its future behavior. If SH expansion is largely due to ozone depletion, then the future impacts will be reduced as the ozone hole is projected to recover by the mid-twenty-first century (WMO 2014). If greenhouse gas (GHG) forcing is significant, then impacts should be expected well into the future.

Here we examine the Southern Hemisphere Hadley cell expansion in context of global warming associated with increases in atmospheric greenhouse gas concentrations. We show that the Southern Hemisphere Hadley cell started expanding during the middle part of the twentieth century as the global mean surface temperature began to strongly increase. Using data from the latest generation of climate model simulations we show that such an expansion can occur when simulations include GHG forcing, whereas model simulations with natural forcing only show no expansion.

2. Data and methods

The Twentieth Century Reanalysis (20CR; Compo et al. 2011) was used to analyze trends and relationships between various indices, and then as reference to assess the model outputs. Despite the fact that 20CR includes only surface observations in its assimilation and has no upper-air or satellite observations, we are confident that the Hadley cell is reasonably represented (Nguyen et al. 2013). Indeed, Nguyen et al. (2013) showed that among the eight reanalysis products, 20CR evolves close to the ensemble mean and reproduces reasonable trends. Phase 5 of the Coupled Model Intercomparison Project (CMIP5) includes 10 models’ first realizations with three historical forcings: 1) including all forcings, which imposes changing conditions based on observations; 2) including GHG forcing only; and 3) including natural forcing only. The models used are listed in Table 1. Details on the experiment design can be found in Taylor et al. (2012). However in the GHG-only forcing simulations, half of the models also include ozone changes (Table 1) and only one model (CCSM4) has simulations including ozone forcing only. Therefore conclusions drawn concerning GHG response will be made with care taking the ozone factor into account when possible. Observed changes in surface temperature are obtained from three surface temperature products: HadCRUT3 (Brohan et al. 2006), NOAA (Smith and Reynolds 2005; Smith et al. 2008), and GISS (Hansen et al. 2010). The observed subtropical ridge (STR) is derived from the HadSLP2 data (Allan and Ansell 2006).

Table 1.

CMIP5 models used in the study. These models archived both GHG only and Natural forcing only. Possible forcing for CMIP5 models includes solar irradiance (Sl); volcanic aerosol (Vl); land-use change (LU); dust (Ds); anthropogenic sulfate aerosol, direct effects (SD); black carbon (BC); mineral dust (MD); organic carbon (OC); ozone, tropospheric and stratospheric (Oz); and anthropogenic aerosols (AA). (Expansions of acronyms are available online at http://www.ametsoc.org/PubsAcronymList.)

CMIP5 models used in the study. These models archived both GHG only and Natural forcing only. Possible forcing for CMIP5 models includes solar irradiance (Sl); volcanic aerosol (Vl); land-use change (LU); dust (Ds); anthropogenic sulfate aerosol, direct effects (SD); black carbon (BC); mineral dust (MD); organic carbon (OC); ozone, tropospheric and stratospheric (Oz); and anthropogenic aerosols (AA). (Expansions of acronyms are available online at http://www.ametsoc.org/PubsAcronymList.)
CMIP5 models used in the study. These models archived both GHG only and Natural forcing only. Possible forcing for CMIP5 models includes solar irradiance (Sl); volcanic aerosol (Vl); land-use change (LU); dust (Ds); anthropogenic sulfate aerosol, direct effects (SD); black carbon (BC); mineral dust (MD); organic carbon (OC); ozone, tropospheric and stratospheric (Oz); and anthropogenic aerosols (AA). (Expansions of acronyms are available online at http://www.ametsoc.org/PubsAcronymList.)

The Hadley cell is identified using the zonal mean meridional streamfunction Ψ, representing streamlines of mass transport in pressure coordinates. From Ψ, the Hadley cell edge (HCE) is defined as the location of the subtropical zero contour of Ψ averaged between 600 and 400 hPa, The Hadley cell intensity (HCI) is the average of the peak Ψ values between 900 and 200 hPa. These definitions are consistent with the metrics used in other studies (e.g., Lucas et al. 2014). Full details on the computation are described in Nguyen et al. (2013).

The zonal mean subtropical ridge or STR is characterized by its intensity (STRI) and position (STRP) and is here assumed to be associated with the absolute maximum zonal average mean sea level pressure (MSLP). Therefore STRI and STRP can be defined by identifying the amplitude and location of the MSLP peak. Given the coarse resolution of the MSLP data of 5°, a linear interpolation to a finer latitudinal resolution of 0.5° is applied to the zonal mean MSLP prior to the peak detection process to avoid artificial jumps in the location of the peak.

The linear least squares trend is applied to a running window containing 30-yr time series with a running step of one year. The result is time series of a 30-yr trend for a given index where the date coincides with the first year of the given window. The two-tailed probability and the 95% statistical significance level considering lag-1 autocorrelation to estimate the effective sample size were used to determine whether the trend is significantly different from a trend of zero. To examine how unusual the Hadley cell trend is in the historical context, we compute linear trends of the annual mean standardized anomalies to provide direct comparison of the trend size across the different indices using a yearly-running window of 30 yr in length. The study is focused on the Southern Hemisphere so all the indices above and results shown hereafter are exclusive to the Southern Hemisphere, unless indicated otherwise.

3. Expansion of the Southern Hemisphere Hadley cell

Here we focus on the more zonally homogenous Southern Hemisphere, where expansion is observed across a range of dataset and methodologies (e.g., Lucas et al. 2012; Choi et al. 2014; Lucas et al. 2014; Lucas and Nguyen 2015, and references therein). We show analyses with the 20CR because at the time of study this dataset was the only reanalysis product covering the entire twentieth century.

Time series of annual mean global surface temperature and the latitude of the Southern Hemisphere Hadley cell edge and intensity and the subtropical ridge position and intensity, together with their standardized anomalies derived from 20CR, are displayed in Fig. 1. The mean and standard deviation for these indices over this period are provided in Table 2. In Fig. 1, the observed global temperature, STRI, and STRP derived from observed mean sea level pressure are plotted for direct comparison with the 20CR. The global surface temperature in 20CR matches well with the observations only after 1950; 20CR is cooler than observed prior to 1950. The increase in assimilated ground-based observations from 1952 used by the model (Fan and van den Dool 2008) is likely to constrain the 20CR and improve the reanalyzed temperature in the model for the later period.

Fig. 1.

(a) Annual mean and (b) standardized anomalies indices of (from top to bottom): global surface temperature (GT), Southern Hemisphere Hadley cell edge (HCE) and intensity (HCI), and subtropical ridge position (STRP) and intensity (STRI). The black lines are for indices from the 20CR and the red ones for indices from the NCEP–NCAR reanalysis. The gray lines represent time series of observed anomalies of GT (shaded), STRP (negative curve), and STRI (positive curve). The Hadley cell indices are derived from the zonal mean meridional streamfunction and the STR indices from zonal mean MSLP. Further definition of these indices is detailed in the methodology section.

Fig. 1.

(a) Annual mean and (b) standardized anomalies indices of (from top to bottom): global surface temperature (GT), Southern Hemisphere Hadley cell edge (HCE) and intensity (HCI), and subtropical ridge position (STRP) and intensity (STRI). The black lines are for indices from the 20CR and the red ones for indices from the NCEP–NCAR reanalysis. The gray lines represent time series of observed anomalies of GT (shaded), STRP (negative curve), and STRI (positive curve). The Hadley cell indices are derived from the zonal mean meridional streamfunction and the STR indices from zonal mean MSLP. Further definition of these indices is detailed in the methodology section.

Table 2.

Mean and standard deviations of annual mean indices: Global surface temperature (GT), Hadley cell extent (HCE) and intensity (HCI), and subtropical ridge position (STRP) and intensity (STRI) for the period 1950–2008 for observations and 20CR and 1950–2005 for the CMIP5 multimodel average.

Mean and standard deviations of annual mean indices: Global surface temperature (GT), Hadley cell extent (HCE) and intensity (HCI), and subtropical ridge position (STRP) and intensity (STRI) for the period 1950–2008 for observations and 20CR and 1950–2005 for the CMIP5 multimodel average.
Mean and standard deviations of annual mean indices: Global surface temperature (GT), Hadley cell extent (HCE) and intensity (HCI), and subtropical ridge position (STRP) and intensity (STRI) for the period 1950–2008 for observations and 20CR and 1950–2005 for the CMIP5 multimodel average.

To further investigate the discontinuity in the 20CR indices compared to observation, we apply the breakpoint analysis to the HCE time series following the Lund and Reeves (2002) method, which basically detects points where creating two linear fits explains statistical significantly (F test) more variance that a single fit across the whole dataset. This method enables to identify a marked breakpoint in July 1951, where the trend is +0.11 ± −0.07 and −0.26 ± 0.07 before and after this date, respectively (light blue lines in Fig. 2). This breakpoint is visually apparent and supports the hypothesis of data issues. Additional explanation for the breakpoint may also be the number of observation stations that reports monthly mean surface temperature from the Global Historical Climatology Network (GHCN) and Climate Anomaly Monitoring System (CAMS). Indeed, in their Fig. 1 Fan and van den Dool (2008) show a jump in the number of these stations in 1952. Note that there is another much bigger jump in 1981 for the CAMS stations but that is compensated by an important decrease in the GHCN stations in the early 1990s, which results in a total comparable to the late 1950s after the first shift in 1952. Repeating this breakpoint analysis for the 1952 subset suggests statistically significance breakpoints in July 1976 and around 1993–95, respectively. The first may be related to climate shift and/or the onset of the modern period of warming, whereas the second may happen because of Pinatubo recovery. Therefore in what follows we limit the study to 1952 onward.

Fig. 2.

20CR Hadley cell edge time series at monthly (black) and annual (blue) mean. Trends computed for the whole period (red) and each of the two periods at the breakpoint found in July 1951 (yellow) are superimposed.

Fig. 2.

20CR Hadley cell edge time series at monthly (black) and annual (blue) mean. Trends computed for the whole period (red) and each of the two periods at the breakpoint found in July 1951 (yellow) are superimposed.

Linear trends time series of the above indices from the observations and 20CR are shown in Fig. 3a, and the indicated year corresponds to the first year of the window used to compute the trend. The trends in 20CR global temperature are smaller than those from observations in all windows from 1958. Despite the weaker values, the 20CR trends follow the observed tendency; increasing in amplitude with time and remaining statistically significant at the 95% level throughout entire period, except for 1957.

Fig. 3.

(a) Linear trends over a 30-yr running window 20CR standardized anomalies indices of GT, HCE, HCI, STRP, and STRI (standardized unit per decade). The year on the x axis corresponds to the first year of the window. Values statistically significant at the 95% level are indicated by the square boxes. The gray plots are observed anomalies of GT (shaded), STRP (negative curve), and STRI (positive curve). (b) As in (a), but for NCEP.

Fig. 3.

(a) Linear trends over a 30-yr running window 20CR standardized anomalies indices of GT, HCE, HCI, STRP, and STRI (standardized unit per decade). The year on the x axis corresponds to the first year of the window. Values statistically significant at the 95% level are indicated by the square boxes. The gray plots are observed anomalies of GT (shaded), STRP (negative curve), and STRI (positive curve). (b) As in (a), but for NCEP.

HCE and STRP trends tend to increase with time, as does the STRI trend although to a lesser extent. The HCE trend remains statistically significant from 1968 and reaches its maximum (0.7 standardized unit per decade) in the last window (1979–2008). Time series of the HCI trend, although showing intensification, retreated toward zero from 1976 and became statistically insignificant. The STRP trend time series varies in concert with HCE trends, although reaching statistical significance only between 1974 and 1977. Trends in the STRI increase from the late 1950s until the early 1970s, after which they begin to decrease. These trends are significant in windows between 1962 and 1975. Differences are numerous in the trends between the 20CR and observed STR metrics, partly due to data coarseness. However, discrepancies might also relate to the sensitivity of changes in STRI to global surface temperature changes in 20CR. While observations show global warming to be strengthening, in association with STRI trends, the slight decline in 20CR global warming in the last few windows is consistent with STRI trend weakening.

To identify the relationship between each of the variables and global warming, linear fits of the trends against global surface temperature and HCE and the variance explained (r2) are shown in Figs. 4a,b respectively). There is no strong relationship between global temperature increase and HCI intensification (r2 = 0.14) or STRI intensification (r2 = 0.24). In contrast, HCE widening and STRP poleward shift are aligned with global warming (r2 = 0.73 and 0.63, respectively). This suggests that the Hadley cell expansion and the STR poleward shift are both strongly related to global temperature increase. The relationship between HCE and STRP trends [r2 = 0.85, Fig. 4b)] confirms their covariance; this is lower with STRI (r2 = 0.41). No relation is obtained between HCE and HCI, as identified in Nguyen et al. (2013).

Fig. 4.

(a) Linear fits of the 20CR standardized trends against GT and HCE (red), HCI (green), STRP (purple), and STRI (blue). Values statistically significant at the 95% level are indicated with closed dots and the variance explained is shown. (b) As in (a), but the trends are linearly fitted against HCE. (c) and (d) As in (a) and (b) but for NCEP.

Fig. 4.

(a) Linear fits of the 20CR standardized trends against GT and HCE (red), HCI (green), STRP (purple), and STRI (blue). Values statistically significant at the 95% level are indicated with closed dots and the variance explained is shown. (b) As in (a), but the trends are linearly fitted against HCE. (c) and (d) As in (a) and (b) but for NCEP.

Noting that the NCEP–NCAR reanalysis product (hereafter simply NCEP; Kalnay et al. 1996) also cover the period of study, which spans from 1948 to present, we here compare the NCEP trends with 20CR in order to assess how consistent the two datasets are in terms of the different indices. We first compare the time series of the studied indices at annual mean and standardized anomalies for the period covering both NCEP and 20CR in Fig. 1. The mean surface temperature in NCEP has a systematic cool bias of about 1 K, also reported in Qian et al. (2006), who attribute it to errors from the mountainous areas such as elevations error of the data grids, snow cover error, large uncertainties in the land surface model (Bowen ratio), and airflow error. In addition, the NCEP surface temperature anomalies are marked by a slight cooling trend from 1948 to the 1970s and then a too strong warming afterward. This cooling bias is reported in Qian et al. (2006) as specific to the Northern Hemisphere while the warming may be caused by the forced constant concentration of atmospheric CO2 and GHG in the model.

The surface temperature bias in NCEP is also reflected in the Hadley cell characteristics with HCE and HCI time series showing marked differences with those from 20CR (Fig. 1). Annual mean and anomaly HCE from NCEP here support the direct relationship between global surface temperature and HC edge in the way that global cooling prior to the 1970s is associated with a slight shrinkage while global warming is associated with expansion. The HCE also expands too much in NCEP compared to 20CR, in response to the warming bias. Coincidentally with the 1-K cool bias in NCEP, annual mean HCI exhibits a systematic weak bias of about 20 × 109 kg s−1 and a hint of a weakening followed by a stronger intensification compared to 20CR, in association with changes in global surface temperature. In contrast, variation in the NCEP STR does not follow those in global surface temperature in terms of observed bias. Both annual mean and anomalies show fairly good agreement with 20CR in both position and intensity of the STR except the marked weaker STRP anomaly in NCEP prior to the 1970s which actually coincides with a cooler global surface temperature anomaly.

Time series of these indices 30-yr running window linear trends for NCEP shown in Fig. 3b confirm the behavior described above, namely the relative consistency in the direction of the trends between the NCEP and 20CR indices only from 1967, except for the STRI trends that evolve similarly for the entire period. Before 1967, all the other indices’ trends in NCEP contradict with those in 20CR as indicated in Fig. 1a. The scatterplots of each of the HC and STR indices with reference to global temperature and HC expansion shown in Figs. 4c,d respectively) for NCEP also tend to differ from 20CR. HCI and STRP trends are found in both directions with increased global temperature. Only HCE and STRI trends are on the increase associated with increase in global surface temperature. No coherent linear relationship was found between STRP and HCE trends in NCEP (Fig. 4d) as suggested in the time series in Fig. 3b. These trends together with the annual mean and anomalies above suggest that 20CR performs better than NCEP by reasonably reproducing the variability of the different indices and justify the choice of using the 20RC product in this study.

4. CMIP5 model assessment

To establish the influence of global warming on Hadley cell expansion, 10 CMIP5 climate models listed in Table 1 are used. Here we consider simulations with GHG forcing, natural forcing, or all forcings included. HC expansion inferred with natural forcing alone would rule out GHG influence as a cause of the expansion. Noting that half of the models in Table 1 include time-varying ozone in their GHG-forcing simulations, we separate them from the sample with only GHG forcing in order to assess how the inclusion of ozone affects HC expansion.

Table 2 shows the mean and standard deviation values for the period 1950–2005 of the multimodel average indices. While the mean values in CMIP5 simulations are comparable to observations and the 20CR, standard deviations are weaker for all indices related to meridional mean circulation, and much weaker when compared with 20CR. However, standard deviation is markedly larger for global surface temperature if individual CMIP5 models are considered (not shown).

Figure 5a shows time series of annual mean indices from the multimodel average. As expected, the warming trend is reproduced only in anthropogenic GHG-induced (i.e., historical forcing and GHG forcing with or without ozone) time series. The main difference in these latter time series resides in the mean values, with the GHG being about 0.25 and 0.5 K warmer than the GHG with ozone and the all forcing, respectively. The standardized annual anomalies time series (Fig. 5b) of global temperature under natural forcing (NAT) is marked by three sharp cooling periods in 1963, 1982, and 1991 in response of major volcanic eruptions of Mount Agung, El Chichón, and Pinatubo, respectively. These cooling episodes are also seen in the all-forcing time series, but with reduced intensity. Both the GHG-induced time series exhibit a more continuous increase over the whole period.

Fig. 5.

(a) Annual mean and (b) standardized annual anomalies indices from 20CR (black) and CMIP5 multimodel average (HIS: orange, GHG only: dark green, NAT: blue, and GHG with O3 included: light green) from top to bottom: GT, HCE, HCI, STRP, and STRI.

Fig. 5.

(a) Annual mean and (b) standardized annual anomalies indices from 20CR (black) and CMIP5 multimodel average (HIS: orange, GHG only: dark green, NAT: blue, and GHG with O3 included: light green) from top to bottom: GT, HCE, HCI, STRP, and STRI.

The mean edge and intensity of the Hadley cell in the CMIP5 models ensemble mean are relatively similar to those from 20CR, except in the last 20 years for HCE, which tends to be smaller. A hint to HC expansion is then seen only in anthropogenic GHG-induced time series. As for 20CR, this expansion is concomitant with a tendency for a poleward shift of the STRP, although the mean values tend to be equatorward (~0.5° latitude) compared to 20CR. This is associated with more intense mean STRI in the CMIP5 models ensemble mean (~1 hPa). Note that with the GHG forcing without ozone annual mean global temperature time series being markedly the warmest, the associated time series of both HCE and STRP are markedly poleward from the other simulations, tending to evolve closest to the 20CR time series (Fig. 5a). However the variability in these modeled indices remains too small compared to the 20CR (Fig. 5b).

We now apply the same 30-yr-wide window running trend calculation to the standardized anomalies to assess the models ability to replicate the trends and the strong linear relationships between global warming and HC expansion. Strong linear relationships between the different HC indices obtained in CMIP5 are comparable to 20CR with the scatterplots in Fig. 6 of CMIP5 and 20CR. Remembering that half of the models here include ozone changes in their GHG forcing simulations (Table 1) and that ozone depletion had been reported in the literature to contribute to the Hadley cell expansion in DJF (Son et al. 2010; Polvani et al. 2011; Min and Son 2013) we separate these models from the one that do not include ozone in the GHG forcings in Fig. 6 to assess whether including ozone in the GHG-only simulations alter our conclusions. The first result from Fig. 6 is that there is no apparent difference between the GHG and GHG including ozone forcing simulations; both ensembles are clustered together. Second, the relationships established between the HC and STR indices and global surface warming and HC expansion are not altered whether ozone is included in the GHG forcing simulation or not. The linear fits with the GHG including ozone simulations show a slight improvement of the slope and the variance, except for the STRI and HCE trends scatterplot. Therefore for what follows, we will not make any distinction between GHG-only and GHG with ozone experiments.

Fig. 6.

Linear fits of the 20CR (black) and CMIP5 (HIS: orange, GHG only: dark green, NAT: blue, and GHG with O3 included: light green) standardized trends against GT of (a) HCE and (b) STRI and against HCE of (c) STRP and (d) STRI. Values statistically significant at the 95% level are indicated with closed dots and the variances explained are shown.

Fig. 6.

Linear fits of the 20CR (black) and CMIP5 (HIS: orange, GHG only: dark green, NAT: blue, and GHG with O3 included: light green) standardized trends against GT of (a) HCE and (b) STRI and against HCE of (c) STRP and (d) STRI. Values statistically significant at the 95% level are indicated with closed dots and the variances explained are shown.

The scatterplot between trends in global surface temperature and HCE (Fig. 6a) in CMIP5 indicates the clustering of the response according to the forcing applied, and a linear response of the HCE to global warming with r2 > 0.5. Global warming associated with Hadley cell expansion seen in the GHG-forcing simulations, either alone or combined with other forcing, contrast with the naturally induced simulations. However, the slope of the relations differs from 20CR; for a comparable amount of global warming, HC expansion is weaker in CMIP5 (by a factor of 4). This difference may point to the CMIP5 models underestimating HC expansion in response to global warming, compared to 20CR. This has potentially important implications for representation of precipitation changes for the subtropical regions. This underestimate may also be associated with the misrepresentation in the models of long-term changes in El Niño–Southern Oscillation, the Pacific decadal oscillation, and the southern annular mode that may also account for the observed tropical expansion (Allen et al. 2014; Lucas and Nguyen 2015).

Strong relationships can also be shown between global surface temperature and STRI in Fig. 6b. With the slope obtained from CMIP5 in agreement with the one from 20CR, albeit with a shift of ~0.2. This indicates that climate models underestimate the amount of STR intensification in relation to global warming. As with HC edge response to global warming, there is marked separation between natural and GHG clustering. Across all CMIP5 simulations, the strong relationship (r2 = 0.55) is larger than that for 20CR, due in part to the separated clustering based on the forcing used rather than an increased sensitivity to global warming in the CMIP5 models.

Internal to the meridional circulation, linear relationships match those in 20CR between the edge of the Hadley cell and the position of the STR (r2 > 0.8; Fig. 6c) with identical slopes. The models produce changes in the STRP which are tightly related to HC expansion with a slope close to 1 and crossing the zero point in the models inducing a similar amount of expansion of the Hadley cell as a poleward shift of the STR. Similarly, a strong linear relationship is also obtained from the multimodel average scatterplot between HCE and STRI, explaining two-thirds of the total variance (Fig. 6d). The linear best fit crosses the zero point in the models, with a slope typical of a strengthening of a standardized unit STR per standardized unit of HC expansion. These results confirm the coherence of the representation of the meridional circulation features in CMIP5 models with realistic changes in the surface climate properties of the STR in response to HC expansion.

These two internal relationships are not as tight in the 20CR, even presenting an offset from the origin suggesting that in that simulation, a shift equatorward of about 0.2 standardized units and an intensification of 0.2 standardized units coincide with no changes in the HC edge. These results point to the uncertainty in the exact amount of tropical expansion that can be inferred from 20CR as well as was discussed in the literature (Lucas et al. 2014; Lucas and Nguyen 2015), highlighting the limits with using the reanalysis dataset.

5. Concluding remarks

The 20CR data show a Hadley cell expansion in Southern Hemisphere since the 1950s with statistical significance from 1968. This Hadley cell expansion is consistent with increasing global warming and intensification of the zonal mean subtropical ridge. Examination of these trends in 10 CMIP5 historical simulations using natural and GHG forcings confirms that GHG forcing is a mechanism that can explain at least part the Hadley cell expansion. However, while the expansion was underestimated in the models for a given amount of global warming (by a factor of 4), the CMIP5 models still capture the internal characteristics of changes in the mean meridional circulation, correctly simulating the poleward shift of the subtropical ridge associated with Hadley cell expansion.

Stratospheric ozone depletion (e.g., Lu et al. 2009; Son et al. 2010; Polvani et al. 2011; Min and Son 2013) has been suggested to be the main driver for Southern Hemisphere Hadley cell expansion but exclusively in the austral summer from 1979. However, previous studies (e.g., Staten et al. 2012; Nguyen et al. 2013) showed statistically significant expansion not only in DJF but also in March–May. In addition, ozone depletion only started in the early 1980s (WMO 2014) and we showed here that the Southern Hemisphere annual mean Hadley cell started expanding in the 1950s, consistent with Allen et al. (2014). Allen et al. (2014) also suggested that anthropogenic greenhouse gases (GHGs) were an equally important driver, showing equivalent amounts of expansion inferred from any of the three anthropogenic forcings (all, GHG, or ozone forcing), contrasting with natural or anthropogenic aerosol forcing only showing no to little contraction trends. However, from their results the HC expansion inferred from the combined all-forcing simulations is not larger than those inferred from each single forcing, possibly because different models were used to generate the above trends. Staten et al. (2012) showed the linearity of the HC response to GHG increase, ozone depletion, and SST warming, which all contribute to the expansion. They further show that sea surface temperature warming was the dominant effect of the expansion and that the ozone effect was limited to austral spring and summer due to their seasonal cycle. This might explain why the expansion trends inferred from the AMIP simulations in Allen et al. (2014) exceed those inferred from the coupled all forcing ones and why previous studies found strong Southern Hemisphere summer HC expansion response associated with ozone forcing. Here we further show that GHG forcing whether with or without ozone included is a necessary condition for HC expansion and that the expansion occurs under global warming. Given the large underestimate of HC expansion from the models, the statistical significance of the attribution is likely to be weak because of the difference of HC changes per single forcing and per model. We are currently exploring a case study where if one model simulates reasonable HC expansion in the historical context, then attribution analysis may be conducted. In summary, stratospheric ozone depletion causing stratospheric cooling in late spring might be driving Hadley cell expansion in summer by affecting the tropospheric jet. Here, the ongoing global surface temperature increase attributable to the increase in GHG concentration and its effect on atmospheric radiative forcing, which has a direct effect on midlatitude westerly winds to strengthen and shift poleward (Deser and Phillips 2009), is demonstrated to be behind annual mean Hadley cell expansion in the Southern Hemisphere.

Ultimately these changes in the intensity of the subtropical ridge impact the rainfall in the subtropics: a rainfall decline is expected in association with intensification in the subtropical ridge. The underestimation of intensification of subtropical ridge relates to the underestimation of the Hadley cell expansion by the models, and opens up the possibility that future projections of rainfall declines in the subtropical regions are underestimated. The likelihood of this is worth investigating given the importance of potential consequences defined from climate change projections (Kang et al. 2013), as the IPCC Fifth Assessment Report (AR5) reported that the Hadley circulation is likely to slow down and widen at the end of the twenty-first century under the representative concentration pathway 8.5 (RCP8.5) scenario (IPCC 2013).

Acknowledgments

This project is supported by the Victorian Climate Initiative (VicCI). We warmly thank Scott Power, Eun-Pa Lim, and Andrew Dowdy for their insightful comments on the manuscript prior to submission. Special thanks also go to the anonymous reviewers for their contributions to improve the manuscript.

REFERENCES

REFERENCES
Allan
,
R.
, and
T. A.
Ansell
,
2006
:
New globally complete monthly historical gridded mean sea level pressure dataset (HadSLP2): 1850–2004
.
J. Climate
,
19
,
5816
5842
, doi:.
Allen
,
R. J.
,
J. R.
Norris
, and
M.
Kovilakim
,
2014
:
Influence of anthropogenic aerosols and the Pacific decadal oscillation on tropical belt width
.
Nat. Geosci.
,
7
,
270
274
, doi:.
Brohan
,
P.
,
J. J.
Kennedy
,
I.
Harris
,
S. F. B.
Tett
, and
P. D.
Jones
,
2006
:
Uncertainty estimates in regional and global observed temperature changes: A new data set from 1850
.
J. Geophys. Res.
,
111
,
D12106
, doi:.
Choi
,
J.
,
S.-W.
Son
,
J.
Lu
, and
S.-K.
Min
,
2014
:
Further observational evidence of Hadley cell widening in the Southern Hemisphere
.
Geophys. Res. Lett.
,
41
,
2590
2597
, doi:.
Compo
,
G. P.
, and Coauthors
,
2011
:
The Twentieth Century Reanalysis Project
.
Quart. J. Roy. Meteor. Soc.
,
137
,
1
28
, doi:.
CSIRO
,
2012
: Climate and water availability in south-eastern Australia: A synthesis of findings from Phase 2 of the South Eastern Australian Climate Initiative (SEACI). Commonwealth Scientific and Industrial Research Organisation, 41 pp.
Dai
,
A.
,
2013
:
Increasing drought under global warming in observations and models
.
Nat. Climate Change
,
3
,
52
58
, doi:.
Deser
,
C.
, and
A. S.
Phillips
,
2009
:
Atmospheric circulation trends, 1950–2000: The relative role of sea surface temperature forcing and direct atmospheric radiative forcing
.
J. Climate
,
22
,
396
413
, doi:.
Fan
,
Y.
, and
H.
van den Dool
,
2008
:
A global monthly land surface air temperature analysis for 1948–present
.
J. Geophys. Res.
,
113
,
D01103
, doi:.
Fu
,
Q.
,
C. M.
Johanson
,
J. M.
Wallace
, and
T.
Reichler
,
2006
:
Enhanced mid-latitude tropospheric warming in satellite measurements
.
Science
,
312
,
1179
, doi:.
Fyfe
,
J. C.
,
N. P.
Gillett
, and
G. J.
Marshall
,
2012
:
Human influence on extratropical Southern Hemisphere summer precipitation
.
Geophys. Res. Lett.
,
39
,
L23711
, doi:.
Hansen
,
J.
,
R.
Ruedy
,
M.
Sato
, and
K.
Lo
,
2010
:
Global surface temperature change
.
Rev. Geophys.
,
48
,
RG4004
, doi:.
Hu
,
Y. Y.
,
L. J.
Tao
, and
J. P.
Liu
,
2013
:
Poleward expansion of the Hadley circulation in CMIP5 simulations
.
Adv. Atmos. Sci.
,
30
,
790
795
, doi:.
IOCI
,
2012
: Western Australia’s weather and climate: A synthesis of Indian Ocean Climate Initiative (IOCI) stage 3 research. CSIRO and BoM, 119 pp.
IPCC
,
2013
: IPCC Climate Change 2013: The Physical Science Basis. T. F. Stocker et al., Eds., Cambridge University Press, 1535 pp.
Johanson
,
C. M.
, and
Q.
Fu
,
2009
:
Hadley cell widening: Model simulations versus observations
.
J. Climate
,
22
,
2713
2725
, doi:.
Kalnay
,
E.
, and Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project
.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
, doi:.
Kang
,
S. M.
,
S. M.
Deser
, and
L. M.
Polvani
,
2013
:
Uncertainty in climate change projections of the Hadley circulation: The role of internal variability
.
J. Climate
,
26
,
7541
7554
, doi:.
Lu
,
J.
,
C.
Deser
, and
T.
Reichler
,
2009
:
Cause of the widening of the tropical belt since 1958
.
Geophys. Res. Lett.
,
36
,
L03803
, doi:.
Lucas
,
C.
, and
H.
Nguyen
,
2015
:
Regional characteristics of tropical expansion and the role of climate variability
.
J. Geophys. Res. Atmos.
, doi:, in press.
Lucas
,
C.
,
H.
Nguyen
, and
B.
Timbal
,
2012
:
An observational analysis of Southern Hemisphere tropical expansion
.
J. Geophys. Res.
,
117
,
D17112
, doi:.
Lucas
,
C.
,
B.
Timbal
, and
H.
Nguyen
,
2014
:
The expanding tropics: A critical assessment of the observational and modelling studies
.
Wiley Interdiscip. Rev.: Climate Change
,
5
,
89
112
, doi:.
Lund
,
R. B.
, and
J.
Reeves
,
2002
:
Detection of undocumented changepoints: A revision of the two-phase regression model
.
J. Climate
,
15
,
2547
2554
, doi:.
McLandress
,
C.
,
T. G.
Shepherd
,
J. F.
Scinocca
,
D. A.
Plummer
,
M.
Sigmond
,
A. I.
Jonsson
, and
M. C.
Reader
,
2011
:
Separating the dynamical effects of climate change and ozone depletion. Part II: Southern Hemisphere troposphere
.
J. Climate
,
24
,
1850
1868
, doi:.
Min
,
S.-K.
, and
S.-W.
Son
,
2013
:
Multimodel attribution of the Southern Hemisphere Hadley cell widening: Major role of ozone depletion
.
J. Geophys. Res. Atmos.
,
118
,
3007
3015
, doi:.
Morales
,
M. S.
, and Coauthors
,
2012
:
Precipitation changes in the South American Altiplano since 1300 AD reconstructed by tree-rings
.
Climate Past
,
8
,
653
666
, doi:.
Nguyen
,
H.
,
A.
Evans
,
C.
Lucas
,
I.
Smith
, and
B.
Timbal
,
2013
:
The Hadley circulation in reanalyses: Climatology, variability, and expansion
.
J. Climate
,
26
,
3357
3376
, doi:.
Oort
,
A. H.
, and
J. J.
Yienger
,
1996
:
Observed interannual variability in the Hadley circulation and its connection to ENSO
.
J. Climate
,
9
,
2751
2767
, doi:.
Perlwitz
,
J.
,
S.
Pawson
,
R. L.
Fogt
,
J. E.
Nielsen
, and
W. D.
Neff
,
2008
:
Impact of stratospheric ozone hole recovery on Antarctic climate
.
Geophys. Res. Lett.
,
35
,
L08714
, doi:.
Polvani
,
L. M.
,
D. W.
Waugh
,
G. J. P.
Correa
, and
S.-W.
Son
,
2011
:
Stratospheric ozone depletion: The main driver of twentieth-century atmospheric circulation changes in the Southern Hemisphere
.
J. Climate
,
24
,
795
812
, doi:.
Qian
,
T.
,
A.
Dai
,
K. E.
Trenberth
, and
K. W.
Oleson
,
2006
:
Simulation of global land surface conditions from 1948 to 2004. Part I: Forcing data and evaluations
.
J. Hydrometeor.
,
7
,
953
975
, doi:.
Seidel
,
D. J.
,
Q.
Fu
,
W. J.
Randel
, and
T. J.
Reichler
,
2008
:
Widening of the tropical belt in a changing climate
.
Nat. Geosci.
,
1
,
21
24
, doi:.
Smith
,
T. M.
, and
R. W.
Reynolds
,
2005
:
A global merged land–air–sea surface temperature reconstruction based on historical observations (1880–1997)
.
J. Climate
,
18
,
2021
2036
, doi:.
Smith
,
T. M.
,
R. W.
Reynolds
,
T. C.
Peterson
, and
J.
Lawrimore
,
2008
:
Improvements to NOAA’s historical merged land–ocean surface temperature analysis (1880–2006)
.
J. Climate
,
21
,
2283
2296
, doi:.
Son
,
S.-W.
, and Coauthors
,
2010
:
Impact of stratospheric ozone on Southern Hemisphere circulation change: A multimodel assessment
.
J. Geophys. Res.
,
115
,
D00M07
, doi:.
Staten
,
P. W.
,
J. J.
Rutz
,
T.
Reichler
, and
J.
Lu
,
2012
:
Breaking down the tropospheric circulation response by forcing
.
Climate Dyn.
,
39
,
2361
2375
, doi:.
Taylor
,
K. E.
,
R. J.
Stouffer
, and
G. A.
Meehl
,
2012
:
An overview of CMIP5 and the experiment design
.
Bull. Amer. Meteor. Soc.
,
93
,
485
498
, doi:.
Webster
,
P. J.
,
2004
:
The elementary Hadley circulation
. The Hadley Circulation: Past, Present and Future. H. Diaz and R. Bradley, Eds., Cambridge University Press, 9–60.
WMO
,
2014
: Assessment for decision-makers: Scientific assessment of ozone depletion: 2014. World Meteorological Organization, Global Ozone Research and Monitoring Project Rep. 56, 92 pp. [Available online at https://www.wmo.int/pages/prog/arep/gaw/ozone_2014/ozone_asst_report.html.]