The zonal structure of trends in the tropical tropopause layer during 1979–2014 is investigated by using reanalysis datasets and chemistry–climate model simulations. The analysis herein reveals that the tropical cold-point tropopause temperature (CPTT) trends during 1979–2014 are zonally asymmetric; that is, over the tropical central and eastern Pacific (CEP; 20°S–20°N, 160°E–100°W), the CPTT shows an increasing trend of 0.22 K decade−1, whereas over the rest of the tropical regions (non-CEP regions) the CPTT shows a decreasing trend of −0.08 K decade−1. Model simulations suggest that this zonal asymmetry in the tropical CPTT trends can be partly attributed to Walker circulation (WC) changes induced by zonally asymmetric changes of the sea surface temperatures (SSTs). The increasing (decreasing) SSTs over the western Pacific (CEP) result in a larger zonal gradient in sea level pressure over the tropical Pacific and intensified surface easterlies. The increased pressure gradient leads to enhanced convection over the Indo-Pacific warm pool and weakened convection over the CEP, facilitating a stronger WC. The downward branch of the intensified WC induces a dynamical warming over the CEP and the upward branch of the intensified WC induces a dynamical cooling over the non-CEP regions below 150 hPa. The significant warming in the upper troposphere and lower stratosphere (UTLS) caused by the WC descending and wave activity changes in the UTLS over the CEP shifts the cold-point tropopause height to a higher level, while the radiative effects of greenhouse gases, ozone, and water vapor changes in the UTLS make less important contributions to the trend of the tropical CPTT than SST changes.
The tropics are the main channel for tropospheric air entering the lower stratosphere (LS), as the tropical tropopause changes and related physical processes modulate the transport of the air mass between the troposphere and the stratosphere and affect the global climate. Because of its significant sensitivity to climate change, the tropical tropopause has attracted great scientific interest in recent years (e.g., Gettelman et al. 2009; Son et al. 2009; Fueglistaler et al. 2009; Feng et al. 2012).
Despite the large number of studies, there are still large uncertainties in the estimated tropical tropopause temperature. Some observational studies report that the tropical tropopause has been cooling during the past decades (e.g., Zhou et al. 2001; Randel et al. 2006; Rosenlof and Reid 2008). Randel et al. (2000) pointed out that the trend of zonal mean tropical tropopause temperature in the radiosonde data was decreasing during the period of 1979–97. Xie et al. (2014) also showed that the tropical-averaged cold-point tropopause temperature (CPTT) has exhibited a decreasing trend since the 1980s. Some other studies showed different tropical tropopause temperature changes from those in the abovementioned studies. Wilcox et al. (2012) found that the trends of zonal- and annual-mean tropical tropopause temperature are not significant. Wang et al. (2012) analyzed the radiosonde data and also found that the trends of tropical tropopause temperature reported in previous studies may have uncertainties. However, based on multimodel simulations, Gettelman et al. (2009) argued that the future tropical CPTT is projected to increase by 0.09 K decade−1. Recently, Wang et al. (2013) found that the tropical tropopause temperature during the last decade has increased by 0.8 K.
Most of the abovementioned studies are focused on zonal mean tropical tropopause temperature trends, but some other studies have documented that the tropical tropopause temperature has longitudinal variations (e.g., Sherwood and Dessler 2000; Gettelman and Forster 2002; Rosenlof and Reid 2008; Garfinkel et al. 2013). Rosenlof and Reid (2008) suggested that the tropopause temperature almost everywhere equatorward of 40° latitude has experienced a cooling since the mid-1990s using the NCEP dataset. More recently, Garfinkel et al. (2013) discussed the zonal structure of the tropical temperature over the upper troposphere and lower stratosphere (UTLS) during the period of 1980–2004. They pointed out that there exists a strong warming trend in the tropical upper troposphere (UT) and a cooling trend in the tropical LS near the Indo-Pacific warm pool, while the temperature changes in the western and central Pacific UTLS are much weaker. Questions arise here as to whether the trends of tropical tropopause temperature in different regions are different from the trend of zonal mean tropical tropopause temperature and what processes are mainly responsible for the longitudinal variations of the trends in the tropical tropopause temperature.
The long-term trend of the tropical tropopause layer is controlled by complex coupling of chemical, dynamical, and radiative processes (Fueglistaler et al. 2009; Grise and Thompson 2013). Changes in greenhouse gas (GHG) concentrations (e.g., Xie et al. 2008; Hu et al. 2015b), the strength of the Brewer–Dobson circulation (BDC; e.g., Wang et al. 2013; Fu et al. 2015), convection (e.g., Nishimoto and Shiotani 2012), and sea surface temperatures (SSTs; e.g., Hu et al. 2014) all have an impact on the tropical tropopause temperatures. Among these factors, the SST and GHG changes as well as convection activities play a key role in the long-term tropical CPTT changes (e.g., Rosenlof and Reid 2008; Xie et al. 2014). Based on model simulations, Wang et al. (2015) showed a statistically significant negative correlation (R = −0.5) between the SST and the tropical tropopause temperature. Rosenlof and Reid (2008) argued that the cooling of tropical tropopause is associated with convection anomalies caused by SST variations over the western Pacific. Furthermore, Xie et al. (2014) pointed out that the interannual variations of tropical CPTT could be modulated by El Niño Modoki activity through changing the intensity of the overshooting convection.
Tropical SST changes can induce changes in various atmospheric processes (e.g., Guan et al. 2003; Vecchi et al. 2006; Zhang et al. 2010; Dong and Lu 2013; Hu et al. 2014; Guo et al. 2016; Zhang et al. 2016). Apart from convection anomalies, the tropical SST changes can also cause changes in large-scale circulation, such as the Walker circulation (WC; e.g., Vecchi et al. 2006; Meng et al. 2012; Dong and Lu 2013) and the BDC (e.g., Xie et al. 2008; Shu et al. 2011; Hu et al. 2014). However, the underlying mechanisms and extent of these large-scale circulation changes in modulating the tropical tropopause temperatures are still under debate, and the relative importance of radiative processes associated with radiatively active gases and dynamical processes associated with SST changes in influencing the CPTT trend is still unclear.
In this study, the zonal structure of the tropical CPTT and its trend are analyzed in detail using three reanalysis datasets, and the dynamical and radiative effects of SST and GHG changes on the long-term trend of the CPTT are diagnosed via a chemistry–climate model with fully interactive chemical–radiative–dynamical processes. The manuscript was organized as follows. Datasets, model, and numerical simulations are described in section 2. The trends of tropical CPTT from the reanalysis datasets and their relationship with SST changes are discussed in section 3. The effects of changing WC on the CPTT trends are discussed in section 4. The corresponding responses of the tropical tropopause height to the SST changes are presented in section 5. Conclusions and discussion are given in section 6.
2. Datasets and numerical experiments
The deviations from zonal mean CPTT—hereafter CPTT*, defined as CPTT* = CPTT − [CPTT], where the square brackets represent the zonal mean—were based on the annual-mean CPTT derived from the reanalysis datasets covering the period of 1979–2014. All trends and their statistical significance in this paper were estimated with the Sen median slope (Sen 1968) and the Mann–Kendall (Kendall 1975) method, respectively, since the nonparametric methods are less sensitive to outliers. All the trends discussed in the following are the annual-mean values.
First, we analyze the Modern-Era Retrospective Analysis for Research and Applications (MERRA) reanalysis dataset for 1979–2014 time period. The MERRA dataset has a horizontal resolution of 1.25° × 1.25° with 37 vertical layers from 1000 to 1 hPa (Rienecker et al. 2011). Second, we use the European Centre for Medium-Range Weather Forecasts interim reanalysis (ERA-Interim, hereinafter ERAI) dataset for 1979–2014 period. ERAI used here has a horizontal resolution of 1.5° × 1.5° on 37 vertical layers from 1000 to 1 hPa (Dee et al. 2011). Third, we analyze the National Centers for Environmental Prediction–U.S. Department of Energy Global Reanalysis 2 (NCEP2) dataset for the 1979–2014 period. The NCEP2 dataset has a horizontal resolution with 2.5° × 2.5° and 17 vertical pressure levels from 1000 to 10 hPa (Kanamitsu et al. 2002). We also analyzed the multimodel simulations from 16 chemistry–climate models (CCMs), which have been evaluated as part of the Stratospheric Processes and Their Role in Climate Chemistry–Climate Model Validation Activity, round 2. More details about these models can be found in Eyring et al. (2007) and Morgenstern et al. (2010). The long-term SST data analyzed in this study are the Extended Reconstruction SST, version 3b (ERSST.v3b; Smith et al. 2008) dataset, which covers the time period from 1979 to 2014.
The numerical model used in this study is the Whole Atmosphere Community Climate Model, version 4 (WACCM4). It has 66 vertical levels extending from the surface to about 150 km with a 1-km vertical resolution in the tropical tropopause layer and the LS. All the simulations in this study were performed at a horizontal resolution of 1.9° × 2.5° (latitude × longitude) with the chemistry processes switched on. More details of the model can be found in Garcia et al. (2007) and Marsh et al. (2013).
Three numerical experiments, which are denoted as AllForcing, SSTvar, and SSTfix, were performed by the WACCM4. The GHGs in these three simulations were adopted from the Intergovernmental Panel on Climate Change representative concentration pathway 4.5 (RCP4.5) scenarios. The specified SST and sea ice fields used in the model are taken from the Met Office Hadley Centre for Climate Prediction and Research (Marsh et al. 2013). In the reference run (AllForcing), all the forcings including GHGs and SSTs are time-varying. In run SSTvar, the SSTs are time-varying, but the GHGs are fixed for 1980 values. In run SSTfix, the SSTs are 12-month climatologies for 1979–2005 time period, but the other forcing fields are the same as those in the AllForcing run. All simulations were performed for 32 years, with the first 5 years being excluded for the model spinup, and the remaining 27-yr model output during the period of 1979–2005 used for analysis. The SSTvar run is used to isolate the effects of SST changes (with fixed radiative forcing) and the SSTfix run is used to isolate the effects of radiative forcing changes (with fixed SSTs).
3. Trends of tropical CPTT
Figures 1a–c show the spatial patterns of the CPTT* trends during 1979–2014 derived from the MERRA, ERAI, and NCEP2 datasets. The annual-mean CPTT* over the central and eastern Pacific (CEP; 20°S–20°N, 160°E–100°W) region has an increasing trend, while the CPTT* over the tropical non-CEP regions exhibits a decreasing trend during the past three decades in the three reanalysis datasets. Also noticeable is that cooling trends of tropical CPTT* over the western tropical Pacific warm pool resemble a horseshoe-shaped pattern, which has been extensively discussed in previous studies (e.g., Gill 1980; Yulaeva and Wallace 1994; Highwood and Hoskins 1998; Dima and Wallace 2007; Fueglistaler et al. 2009; Garfinkel et al. 2013). The CPTT* over the CEP region shows off-equatorial warming trends (Figs. 1a–c). The spatial variations of the CPTT* trend over the CEP resemble a stationary wave pattern known as the Matsuno–Gill pattern (Matsuno 1966; Gill 1980). Figure 1d gives the zonal variations of tropical-averaged (here, the tropics are defined as the zone between 20°S and 20°N) CPTT* trend over the period of 1979–2014 derived from the three reanalysis datasets. The results from the three reanalysis datasets all indicate that the tropical CPTT* over the CEP regions has a positive trend with the maximum of about 0.5 K decade−1 near 150°W, while the tropical CPTT* over the non-CEP regions has a negative trend with the maximum of about −0.2 K decade−1 near 120°E. Figure 1d also shows the zonal variations of tropical-averaged CPTT* trend over the period of 1979–2005 derived from multimodel simulations from 16 CCMs. It is apparent that the modeled CPTT also has a warming trend over the CEP and a cooling trend over the most of the non-CEP regions.
Figures 1e and 1f further give the time series of annual-mean anomalies of CPTT* averaged over the CEP and non-CEP regions derived from MERRA dataset. Note that both the time series with El Niño–Southern Oscillation (ENSO) and interdecadal Pacific oscillation (IPO) signals included and the time series with ENSO and IPO signals removed with the linear regression method are shown. The ENSO index is defined as the SST anomaly over the Niño-3.4 region (5°S–5°N, 120°–170°W) and the IPO index is developed by Henley et al. (2015). By comparing Figs. 1e and 1f, we can see that ENSO and IPO signals have no significant impact on CPTT* trends over the tropical CEP region, although slight differences in the magnitude of trends in the CPTT* exist between the CPTT* time series with and without ENSO and IPO signals. Overall, the results presented in Fig. 1 clearly show that the trends of tropical CPTT* during the past decades are zonally asymmetric. As discussed earlier, previous studies showed that the tropical zonal mean CPTT has had a decreasing trend during the past decades (e.g., Gettelman et al. 2009; Xie et al. 2014). Our results suggest that this decreasing trend of the tropical zonal mean CPTT* is mainly due to the negative trends of CPTT* over the non-CEP regions, and the increasing trends of CPTT* over the CEP partly offset the cooling trend of tropical zonal mean CPTT* during the past three decades.
As the SST changes have an important impact on the tropical CPTT changes (Rosenlof and Reid 2008; Xie et al. 2014), it is necessary here to examine the correlations between CPTT* and SST variations. Figure 2a gives the trends of SSTs during 1979–2014. It can be seen that the SSTs have a decreasing trend in the CEP and an increasing trend in the Indian Ocean and western Pacific regions during 1979–2014, in agreement with previous studies (e.g., Garfinkel et al. 2013; England et al. 2014). Note that the changes in the zonal structure of tropical CPTT* trends in Fig. 1 are also consistent with the changes in zonal structure of the tropical SST trends, further confirming that the zonal asymmetry in tropical CPTT* trends is closely related to the tropical SST changes. Xie et al. (2014) reported a positive correlation between the tropical zonal mean CPTT and the SST changes over the western Pacific, and a statistical insignificant negative correlation between the tropical zonal mean CPTT and the SST over the CEP regions. However, we find that the CEP-averaged CPTT* has a significant positive correlation with the SSTs over non-CEP regions and a significant negative correlation with the SSTs over CEP regions (Fig. 2b). The correlations between the non-CEP-averaged CPTT* and the tropical SSTs are reversed; that is, there is a significant negative correlation between the non-CEP-averaged CPTT* with the SSTs over non-CEP regions and a strongly positive correlation between the non-CEP-averaged CPTT* with the SSTs over CEP regions (not shown). The previous analysis suggests that the CPTT trends are connected with both the CEP cooling and the western Pacific warming during the past three decades.
To provide more evidence on the connection between the tropical CPTT* and the tropical SST, here we define a SST zonal gradient index, SSTG:
where the square brackets with a subscript represent the area-mean SSTs over the tropical western Pacific region (15°S–15°N, 90°–150°E) and over the tropical eastern Pacific region (15°S–15°N, 150°–90°W). Figure 2c shows the regressed CPTT* on the SSTG during the period of 1979–2014. The regressed CPTT* shows similar spatial patterns with the CPTT* trends shown in Fig. 1, with positive values over the CEP and negative values over the non-CEP regions. This further confirms that the zonal asymmetry of tropical CPTT trends over 1979–2014 is closely related to the changes of tropical SSTs.
The close relationship between tropical CPTT* and SSTs can be further elucidated by the corresponding results from the three numerical experiments. The modeled CPTT* trends in runs AllForcing, SSTvar, and SSTfix are shown in Figs. 3a–c, respectively. Figure 3d gives the zonal variations of tropical-averaged CPTT* trends in the three model simulations. Similar to Fig. 1, the modeled CPTT* has a decreasing trend over the non-CEP regions and an increasing trend over the CEP in the reference run AllForcing. Comparing Fig. 1d with Fig. 3d, we can see that the zonal asymmetry of tropical CPTT* trend is well simulated in the AllForcing run. It is apparent from Figs. 3a and 3b that the tropical CPTT* trends in runs SSTvar and AllForcing have similar spatial patterns although the magnitude of the changes of tropical CPTT* trends over the CEP region in SSTvar reduces by about 40% relative to that in AllForcing. Figures 3a and 3b together indicate that long-term variations of GHGs have relatively small impact on the CPTT* trends in the tropical ocean. It is also evident that the zonal asymmetry of the tropical CPTT* trends disappears in run SSTfix (Fig. 3c), further confirming that the zonal asymmetry of tropical CPTT* trends during the past three decades mainly resulted from the tropical SST variations. Note that there exist differences of CPTT near 180° longitude between AllForcing and SSTvar. The CPTT near 180° in SSTfix is warming (Figs. 3c,d), which partly offsets the differences of CPTT near 180° between AllForcing and SSTvar. But the difference of CPTT near 180° between AllForcing and SSTvar is not fully due to GHG trends. In SSTvar, although there are no GHG trends, the effects of increasing GHGs on climate are implied in SST increases. From a linear-response point of view, the differences of CPTT near 180° between AllForcing and SSTvar reflect the direct radiative effects of the increased GHGs. However, the radiative effects of increased GHGs cannot fully account for the CPTT difference near 180° between AllForcing and SSTvar. This difference results from complex radiative–dynamical feedbacks induced by GHG changes.
4. Relationship between the trends of CPTT and the WC
The effects of GHG and SST changes on the tropical CPTT involve both radiative and dynamical processes. The relative importance of radiative and dynamical processes on the tropical CPTT trends is highlighted in Figs. 4a–f, which show the trends of the dynamical heating rates and radiative heating rates (including the longwave and shortwave radiative heating rates) in the three runs. The dynamical heating rate is estimated from the transformed Eulerian-mean equation (e.g., Andrews et al. 1987; Hu et al. 2015b):
where is zonally averaged temperature, are the components of the residual meridional circulation, S is a stability parameter, are the eddy-forcing terms, z is the height, H is the scale height, and is the zonal mean diabatic heating. Primes denote zonal deviation and overbars denote zonal average.
In the AllForcing run, over the CEP regions there exists a positive trend in dynamical heating rate and a negative trend in the longwave radiative heating rate at about the CPT height (Figs. 4a and 4d). In contrast, a negative trend in dynamical heating rate and a positive trend in the longwave radiative heating rate can be noticed over the non-CEP regions, with the maximum dynamical cooling trend and radiative warming trend over the Indo-Pacific warm pool. Note that the trends of the shortwave radiative heating rate are rather small compared to the trends of the longwave radiative heating rate, but they are statistically significant. The trends of the gross heating rate (sum of the dynamical and radiative heating rates) are insignificant but are mostly positive at around the cold-point tropopause (CPT) height over the CEP (not shown), in accordance with the warming trends of the CPTT in this region. On the other hand, over the non-CEP regions, the gross heating rate trends are mostly negative, corresponding to the cooling trends of the CPTT over the non-CEP regions.
It is evident from Fig. 4a that the warming trends of the tropical CPTT over the CEP in AllForcing is mainly caused by the dynamical heating. A comparison between Figs. 4a and 4d and Figs. 4b and 4e reveals that the trends of the dynamical heating rates and longwave radiative heating rates have little to do with the trends of GHGs, although the increased GHGs lead to a small negative trend in the dynamical and longwave radiative heating rates over the CEP (Figs. 4c and 4f, respectively). It appears that the dynamical processes lead to a heating (cooling) in the UTLS over the CEP (non-CEP) region, which is further damped by greater (less) infrared emission and thus increased (decreased) longwave radiative cooling to space. However, the trends in dynamical and longwave radiative heating rates at around CPT level over the tropical central Pacific (around 180° longitude) in SSTvar are slightly different from that in AllForcing. Without the effect of increased GHGs, the trends of the dynamical heating rate at 100 hPa over the tropical central Pacific become negative while the trends of the longwave radiative heating rate become positive. Also note that both the longwave and shortwave radiative heating rate changes in SSTfix are rather small (Fig. 4f), in agreement with the relevant results in previous studies (e.g., Xie et al. 2008; Shu et al. 2011; Hu et al. 2014). Nevertheless, the results shown in Figs. 4a–f confirm that the effects of the SST and GHG changes on the CPTT trends over the tropical ocean are intimately connected via some complex feedbacks between dynamical and radiative processes. Generally, the balance between the dynamical heating rate changes and the radiative heating rate changes due to anthropogenic GHG changes together with water vapor and ozone changes in the UTLS region determines the CPTT changes. In tropical UTLS, the modeled ozone and water vapor have a decreasing trend over the Indo-Pacific warm pool region and an increasing trend over the CEP region (not shown). However, the positive CPTT trend over the CEP and negative CPTT trend over the non-CEP region seem to be mainly caused by dynamical processes as suggested by Fig. 4.
A key question arises here as what dynamical processes associated with SST changes are responsible for the zonal asymmetry of the tropical CPTT trends. The spatial pattern of SST trends in Fig. 2 implies an intensification of the zonal gradient of the SST over the Pacific, and hence a strengthened WC (e.g., Tokinaga et al. 2012). By developing the concept of effective wind for water vapor transport from satellite products and adopting the wind values from reanalysis datasets, Sohn and Park (2010) also found that the WC during the period of 1979–2007 has strengthened. Figure 5 further shows the longitude–height distributions of the trends in the dynamical heating rates, and the composite of vertical velocity trends and zonal wind trends averaged over 20°S–20°N in three runs. The rising motion over the Indo-Pacific and the sinking motion over the eastern Pacific can be clearly seen from the climatological distribution of the vertical winds in three runs. In AllForcing, the strongest rising (sinking) branch of the vertical velocity climatologies is accompanied with a clockwise (anticlockwise) rotation indicated by the composite vectors of vertical and zonal winds trends, implying that the WC over the Pacific is strengthening. Also, the maximum trend center of vertical and zonal winds is shifted westward compared to the climatological center of vertical winds (Fig. 5a). The strengthened rising motion over the Indo-Pacific warm pool and the sinking motion over the central and eastern Pacific can also be well simulated in run SSTvar, implying that the strengthened WC is mainly caused by the SST changes. Slight differences between AllForcing and SSTvar near 160°E, where a regional sinking branch of the composite vectors exists in run AllForcing, are possibly related to the effects of GHG changes (Fig. 5c). The trends of dynamical heating rate show a quite consistent spatial pattern with the composite vectors of vertical wind trends and zonal wind trends; that is, the ascending (descending) motion is accompanied with the decreasing (increasing) dynamical heating rates, particularly in the troposphere below 150 hPa. However, the descending trends over the CEP above 150 hPa are rather weak and the dynamical warming over the CEP above 150 hPa mainly results from the upward flux of planetary-scale wave activity from below, as also suggested by Dima and Wallace (2007).
Figure 6 gives the trends of the sea level pressure (SLP), surface wind, outgoing longwave radiation (OLR), and vertical velocity at 500 hPa within the tropics in the three runs. In AllForcing, the SLP trends over the CEP are positive while those over the non-CEP regions are negative. The different SLP trends over the CEP and non-CEP regions imply that the zonal pressure gradient over the tropical Pacific is increased, hence resulting in an intensification of the surface easterlies over the tropical Pacific (Fig. 6a). England et al. (2014) also found that the Pacific trade winds over the past two decades have experienced a pronounced strengthening. The vertical velocity at 500 hPa has an increasing trend and the deep convective activities represented by the OLR are enhanced over the tropical western Pacific (i.e., the OLR has a decreasing trend over the tropical western Pacific). In contrast, the 500-hPa vertical velocity has a decreasing trend and the OLR has an increasing trend over the CEP; that is, the deep convective activities are depressed over the CEP in run AllForcing (Fig. 6e). It is noted that the depressed deep convective activities near 10°N over the CEP are stronger than those near the equator. The abovementioned features can also be seen in the reanalysis datasets (not shown) and these results suggest that the WC has been strengthening during the past three decades. We further used an index ΔSLP (defined as the tropical Indo-Pacific SLP gradient) to serve as a proxy of the intensity of the Pacific WC (e.g., Vecchi et al. 2006). The annual anomalies of ΔSLP changes during the past three decades from the three reanalysis datasets and three WACCM4 experiments are shown in Figs. 5d and 5h, respectively. The ΔSLP derived from the three reanalysis datasets exhibits positive trends, confirming that the WC has been strengthening during the past three decades. The result here is in accordance with L’Heureux et al. (2013), who showed that the SLP over the Indonesian region has a decreasing trend whereas the SLP over the eastern Pacific has had an increasing trend since the 1950s. It is apparent from the previous analysis that the intensification of WC, as has been reported in previous studies, is robust and still persistent in recent years, and this strengthening of the WC is closely related to the zonally asymmetric SST changes in the tropical ocean that result in a positive trend of ΔSLP between the eastern and western Pacific.
As expected, the spatial patterns of the trends in the SLP, surface winds, OLR, and vertical velocity at 500 hPa in SSTvar are similar to those in AllForcing. Comparing the results in SSTfix with those in AllForcing, the spatial variations of the trends in the SLP and surface winds are quite different. It is evident that the model experiment (SSTfix) forced with the climatological SSTs cannot simulate the strengthening of the WC seen in the reanalysis datasets and the radiative effects of GHG variations have rather small contributions to the trend of WC. This result can also be seen in Fig. 6h, which shows the changes of ΔSLP in three runs. The ΔSLP in both AllForcing and SSTvar has an increasing trend and is in agreement with that in the reanalysis datasets, while the ΔSLP run SSTfix has no significant trends.
The intensified WC over the tropical Pacific Ocean corresponds to enhanced upward motions and stronger convection over its ascending branch (i.e., over the western Pacific) and strengthened downward motion and weakened convective activity over its descending branch (i.e., over the CEP). The enhanced upward motions over the western Pacific lead to an increased dynamical cooling in the UTLS region, while the strengthened descending over the CEP results in a dynamical warming in the UTLS region as shown in Fig. 4. This analysis clearly shows that the WC has had a strengthening trend during the past three decades, and the negative trends in the dynamical heating rate over the western Pacific and the positive trends in dynamical heating rate over the CEP as depicted in Fig. 4 are associated with the WC changes in the UT and the changes of planetary wave activity in the LS.
To provide more insights into the tropical CPTT trends, Fig. 7 gives the height–longitudinal variations of the temperature trend averaged over 20°S–20°N derived from three numerical experiments and three reanalysis datasets. It is apparent that the temperature trends in the tropical UTLS region presented in Fig. 7 are consistent with CPTT trends in Fig. 1. The zonal structure of temperature trends over the Pacific UTLS regions in the run AllForcing and in the three reanalysis datasets are similar (i.e., the temperature in the upper troposphere has a positive trend over the Indo-Pacific warm pool region and a negative trend over the CEP region). The vertical structure of the tropical temperature deviations from zonal mean in the UTLS region resembles a quadrupole, which is consistent with the equatorial planetary wave structure documented by Dima and Wallace (2007) and Fueglistaler et al. (2009). Note that the temperature trends at around and above the tropopause level over the CEP and non-CEP regions are reversed compared with the corresponding temperature trends in the UT. This is consistent with the results in Garfinkel et al. (2013). They proposed that the warming in the UT over the Indo-Pacific warm pool is driven by the warming SSTs over the Indian Ocean, which enhances the moist heating in the UT, and in turn to a Gill-like response extending into the LS and hence a cooling there.
Over the CEP, the descending motion of the WC leads to an adiabatic heating on one hand. On the other hand, the dynamical warming due to wave activities near the tropopause is also important. The wave-induced dynamical heating in the tropical UTLS has been reported in the previous literature (e.g., Dima and Wallace 2007; Nishimoto and Shiotani 2012). Figure 1 indicates that the spatial structure of CPTT* resembles the Matsuno–Gill-wave-response pattern (Matsuno 1966; Gill 1980). The trends of the streamfunction at 200 hPa in three runs shown in Fig. 8 further support this Matsuno–Gill response. In the AllForcing and SSTvar runs, there exists a pair of cyclones in the UT and a Kelvin wave pattern presented over the CEP region, while a pair of anticyclones in the UT and an equatorially trapped Rossby wave pattern exist over the Indian Ocean. The spatial variations of streamfunction trends in the UT over the CEP resemble the results in Gill (1980), Highwood and Hoskins (1998), and Garfinkel et al. (2013). On the other hand, the radiative heating associated with ozone and water vapor changes in the tropical UTLS has no significant contribution to the warming of the UTLS over the CEP.
Recall that the CPTT* over the CEP regions shown in Fig. 1 resembles a stationary wave pattern known as the Matsuno–Gill pattern (Matsuno 1966; Gill 1980). Nishimoto and Shiotani (2012) pointed out that the Matsuno–Gill-response pattern of the tropical tropopause temperature is related to convective heating. Our previous analysis indicates that the CPTT* trends over the CEP are mainly affected by the dynamical warming near the tropopause in response to zonally asymmetric SST changes, which is closely related to the dynamical heating associated with planetary wave changes near the tropical tropopause.
5. Trends of CPT height
Figure 9 further shows the trends of annual-mean CPT pressure (CPTP) averaged over the regions of CEP and non-CEP derived from the three reanalysis datasets. Consistent with the previous studies (e.g., Santer et al. 2003; Seidel and Randel 2006; Wilcox et al. 2012; Xie et al. 2014), Fig. 9 indicates that the tropical CPTP over both the CEP and the non-CEP regions has a decreasing trend in all the three reanalysis datasets (i.e., the tropical tropopause height is lifted during the past decades). In general, a lifted (lowered) tropopause height corresponds to a cooler (warmer) tropopause temperature via an adiabatic cooling (warming) process, and this is the case over the non-CEP regions where the decreased CPTP corresponds to a decrease in the tropopause temperature. It is interesting that the CPTP over the CEP regions is also decreasing while the CPTT is warming. Previous studies reported that the zonal mean tropopause temperature and height variations have an anticorrelation (e.g., Seidel and Randel 2006; Gettelman et al. 2009; Xie et al. 2014); however, our results indicate that the regional variations in the CPT temperature and height over the CEP in recent years are positively correlated. The warming CPT results from the significant warming in the UTLS, which is possibly related to dynamical heating associated with the WC in the UT and wave activity in the LS.
It is worthwhile, at this stage, to look into the responses of the potential temperature lapse rate minimum (LRM; the height where the dθ/dt is minimum) temperature and height to the WC changes. Following the definition of Gettelman and Forster (2002), the LRM level locates the minimum in dθ/dz, which marks the level where radiation begins to influence the temperature and it departs from a saturated moist adiabat. As the LRM is the level where the static stability begins to change and the maximum convective outflow locates, it has a close relationship with convective activities. Figure 10 shows the LRM pressure (LRMP) and LRM temperature (LRMT) trends averaged over the CEP and non-CEP regions derived from three reanalysis datasets. We can see that the trends of tropical LRM during the past decades are also zonally asymmetric. Over the CEP, the LRMP and LRMT both have an increasing trend whereas over the non-CEP regions the LRMP and LRMT trends are opposite; that is, the LRM height and the LRMT are anticorrelated in the three reanalysis datasets. Over the CEP, the intensified WC over the tropical Pacific Ocean shown in Fig. 6 leads to strengthened downward motion and weakened convective activities; therefore, the LRM height has a decreasing trend accompanied by an increasing trend of LRMT there. However, over the non-CEP regions the stronger convection over the ascending branch of WC corresponds to a higher LRM, which leads to the adiabatic cooling of LRMT.
It is evident from the previous analysis that the trends in the LRMT and LRMP are predominately influenced by the convection changes in response to the strengthened WC. Over the CEP, the strengthened descending of WC leads to weakened convective activities, so the LRM height has a decreasing trend accompanied by an increasing trend of LRMT there. However, the CPT height is not solely controlled by convection but rather is dependent on the subtle balance between radiative and dynamical heating in influencing the temperature profiles in the tropical UTLS, as is evident from that depicted in Fig. 11c. The warming of the UTLS changes the temperature profile in a sense that the level of the coldest temperature is moved up and a small part of the lowermost stratospheric air becomes the tropospheric air. The model results from WACCM4 suggest that the warming over the CEP near the tropopause is mainly caused by zonally asymmetric changes of the SSTs via dynamical processes. This dynamical warming is possibly influenced by wave activity changes, as is suggested in Dima and Wallace (2007) and Nishimoto and Shiotani (2012), while over the Indo-Pacific warm pool region the stronger convective activities and significant cooling of the UTLS give rise to a cooler and higher CPT (Fig. 11b).
6. Conclusions and discussion
The longitudinal variations of the tropical CPTT trends during the period of 1979–2014 are investigated using three reanalysis datasets and a chemistry–climate model. It is found that the trends of CPTT* during the period of 1979–2014 are zonally asymmetric; that is, the annual-mean CPTT* over the non-CEP regions has a decreasing trend, while the CPTT* over the CEP regions has an increasing trend both in the reanalysis datasets and in the model simulations. Our results reveal that the decreasing trends in the zonal mean CPTT during the past decades reported in previous studies are mainly contributed by the decreasing CPTT over the non-CEP regions, but are partly offset by the increasing CPTT over the CEP. The magnitude of the increasing trend in the CPTT* over the CEP region during 1979–2014 is larger than its decreasing trend over the non-CEP regions in our results. It is speculated that the warming in the CPTT* over the CEP will exceed the cooling in the CPTT* over the non-CEP regions. Therefore, the zonal mean tropical CPTT is likely to increase in the future, consistent with the results in Gettelman et al. (2009), which showed that the tropical CPTT in multimodel simulations is projected to increase in the future climate.
The zonally asymmetric tropical CPTT* trends during the past decades are mainly caused by zonally asymmetric changes in SSTs. A schematic of the relationship between the SST changes and the tropical tropopause variations and the relevant processes is shown in Fig. 11. The changes in the zonal distribution of tropical SSTs (warming over the non-CEP regions, cooling over the CEP regions) lead to a stronger WC, accompanied by the strengthened SLP gradient over the tropical Pacific (the positive pressure trends over the CEP and the negative pressure trends over the non-CEP regions) and an intensification of the surface easterlies over the tropical Pacific. This leads to strengthened convective activities over the tropical Indo-Pacific warm pool (i.e., over the ascending branch of WC) and weakened convective activities over the CEP regions (i.e., over the descending branch of WC). The strengthened WC causes a dynamical warming (cooling) over the CEP (non-CEP regions), particularly below 150 hPa, while changes in the upward flux of planetary wave activity from below cause a dynamical warming (cooling) over the CEP (non-CEP regions) above 150 hPa. Hence the tropical CPTT and LRMT are increasing over the CEP regions and decreasing over the non-CEP regions.
Because of the enhanced convective activities over the tropical Indo-Pacific warm pool and the corresponding dynamical cooling in the UTLS, the CPT exhibits a lifted and cooling trend over this region. Over the CEP, the depressed convective activities and the significant warming in the UTLS lead to a lifted but warmer CPT. Stronger ascending motion over the non-CEP regions and descending motion over the CEP are among the factors responsible for the cooling of the UTLS over the non-CEP regions and the warming of the UTLS over the CEP. However, the dynamical effects of Kelvin and Rossby waves, which make major contributions to the horseshoe pattern of tropical tropopause temperature anomalies (Highwood and Hoskins 1998; Nishimoto and Shiotani 2012), are also responsible for temperature changes in the UTLS region. The trends in LRMT and LRM height over the CEP are dominantly induced by the strengthened WC. And the CPTT and CPT height trends over the CEP are dominantly affected by the dynamical heating changes induced by wave activities in the tropical UTLS, while the significant warming of the tropical UTLS over the CEP tends to gives rise to a lifted but warming tropopause.
Finally, GHG, ozone, and water vapor changes in the UTLS region may also have impacts on the CPTT; however, our analysis suggests that their direct radiative effects on the long-term trend of the tropical CPTT over the CEP are less important than the effects of SST variations. On the other hand, some previous studies revealed that there is less water vapor entering the stratosphere during the past decades due to the increases of SST in the Indo-Pacific warm pool (e.g., Garfinkel et al. 2013; Hegglin et al. 2014). However, some other studies showed that the water vapor in the tropical LS during the past decades is increasing (e.g., Oltmans et al. 2000; Rosenlof et al. 2001; Tian and Chipperfield 2006; Hu et al. 2015a). Our analysis suggests that zonal asymmetric SST changes in the past three decades may allow more water vapor entering into the LS from the CEP regions due to the warming CPTT there, although the water vapor in the LS over the tropical Indo-Pacific warm pool has a slight decreasing trend.
We greatly appreciate Dr. Brian Soden and three anonymous reviewers for their constructive comments and suggestions. We are also grateful to the groups and agencies for providing the datasets used in this study. The MERRA data were provided by NASA (available online at http://disc.sci.gsfc.nasa.gov/daac-bin/FTPSubset.pl). The NCEP2 data used here were accessed online from http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis2.pressure.html. The ECMWF data were obtained from the European Centre for Medium-Range Weather Forecasts (available online at http://apps.ecmwf.int/datasets/data/interim-full-moda/levtype=pl/), and the HadISST data, obtained from the Met Office Hadley Centre, are available online at http://www.metoffice.gov.uk/hadobs/hadisst/data/download.html. This work was supported jointly by the National Natural Science Foundation of China (41330425 and 41225018), the Postdoctoral Science Foundation of China (2016M591882, the Natural Science Foundation of Jiangsu Province (BK20160949), and the General Program of Natural Science Foundation of Jiangsu Higher Education Institutions (16KJB170006).