Impact of tropical lower stratospheric cooling on deep convective activity : ( I ) Recent trends in tropical circulation

Large changes in tropical circulation, in particular those related to the summer monsoon and cooling of the sea 10 surface in the equatorial eastern Pacific, were noted from the mid to late 1990s. The cause of such recent decadal variations in the tropics was studied by making use of a meteorological reanalysis dataset. Cooling of the equatorial southeastern Pacific Ocean occurred in association with enhanced cross-equatorial southerlies, which resulted from a strengthening and poleward shift of the rising branch of the boreal summer Hadley circulation connected to the stratospheric Brewer‒Dobson circulation. From boreal summer to winter, the anomalous convective activity centre moves southward following the 15 seasonal march to the equatorial Indian Ocean‒Maritime Continent region, which strengthens the surface easterlies over the equatorial central Pacific. Accordingly, ocean surface cooling extends over the equatorial central Pacific. We suggest that the fundamental factor causing the recent decadal change in the tropical troposphere and the ocean is a poleward shift of the rising branch of the summertime Hadley cell, which can result from a strengthening of extreme deep convection penetrating into the tropical tropopause layer, in particular over the continents of Africa and Asia, and adjacent oceans. We conjecture 20 that this effect is produced by a combination of land surface warming due to increased CO2 and a reduction of static stability in the tropical tropopause layer due to tropical stratospheric cooling.


Introduction
Large changes in tropical circulation occurred from the mid to late 1990s.These include a slowdown, or hiatus, of global warming in association with a decrease in the tropical eastern Pacific sea surface temperature (SST) (Kosaka and Xie, 2013;25 England et al., 2014;Trenberth et al., 2014;Watanabe et al., 2014).Changes were also found in the advancement of the onset of the Asian summer monsoon (Kajikawa et al., 2012;Gautam and Regmi, 2013;Xiang and Wang, 2013;Yun et al., 2014) and an increase in precipitation in West Africa over the Sahel (Fontaine et al., 2011;Brandt et al., 2014;Maidment et al., 2015, Diawara et al., 2016).An increase in precipitation in southern Africa was also noted during the austral summer (Vizy and Cook, 2016).Besides these large-scale circulation changes, changes in mesoscale phenomena such as an increase 30 in the cyclone frequency and intensity over the Arabian Sea were also reported (Evan and Camargo, 2011).Wang  (2012) pointed out that these phenomena are, in fact, related to the abovementioned early onset of the Asian summer monsoon.A relationship between tropopause layer cooling and tropical cyclone activity in the Atlantic has also been suggested (Emanuel et al., 2013).Indeed, recent numerical model studies show that cooling of the tropopause impacts the intensity of tropical storms as well as SSTs (Ramsay, 2013;Wang et al., 2014).In this respect, the recent cooling of the tropical tropopause and lower stratosphere from around 2000 (Randel et al., 2006;Randel and Jensen, 2013) should be 5 investigated together with tropical tropospheric change.
The importance of the Pacific Decadal Oscillation (PDO) on decadal changes in global temperature and precipitation has been noted (Meehl et al., 2013;Dong and Dai, 2015;Trenberth, 2015).The most recent hiatus ended around 2013 followed by a large warming due to an El Niño event in 2015 (Hu and Fedorov, 2017;Liu and Zhou, 2017;Urabe et al., 2017;Xie and 10 Kosaka, 2017).However, the El Niño of 2015/16 was different from the previous large 1997/98 El Niño with less warming in the eastern Pacific (Paek et al., 2017), conforming to an increasing trend of the central Pacific-type El Niño (Kao and Yu, 2009;Johnson, 2013).In this sense, the anomalous tropical circulation from the 1990s did not terminate with the hiatus, but is persisting.Similarly, the northward shift of the convective zone in the boreal summer still continues, as shown below.

15
As a cause of recent tropical changes, multidecadal variations of the atmosphere-ocean coupled mode such as the Atlantic Multidecadal Oscillation have also been proposed (Wang et al., 2013;Kamae et al., 2017).The possible impact of SSTs in different oceanic basins on recent trends in monsoon precipitation was studied by using a coupled ocean model by Kamae et al. (2017).Their results show that the change in Atlantic SST can reproduce recent increasing trends in monsoon rainfall in the northern hemisphere (NH), except for the Asian monsoon.The Atlantic SSTs, however, have practically no effect on the 20 southern hemisphere (SH) such as the African and Australian monsoons.Another difference from observations is that the simulated increase in rainfall occurs mainly over the oceans in low latitudes between the equator and 15° N rather than over continents between around 10° N and 20° N as observed (see Fig. 3 of Kamae et al., 2017).Thus, it is difficult to attribute the recent global trends to a regional mode of decadal oceanic variation alone.In this paper, we first show that the fundamental factor causing the recent decadal trend in the tropics from the mid to late 1990s is not the PDO, but rather a 25 poleward shift of the summertime rising branch of the Hadley circulation connected to the upwelling branch of the stratospheric Brewer-Dobson circulation.
One aspect of the recent trends in tropical circulation that has been reported is the poleward expansion of the tropics (e.g., Davis and Rosenlof, 2012).The expansion of the tropics is rather related to a change in the sinking branch of the Hadley cell 30 driven by midlatitude transient eddies (Kang and Polvani, 2011).There are also other phenomena showing large trends in high latitudes and polar regions that are a part of global change and could be related to tropical change.However, in the interest of brevity, this paper focuses on changes in the tropics.The paper is organized as follows.After describing the data used for the study in Sect.2, the results of our analysis are presented in Sect.3. Finally, a summary and discussion of the causes of the recent tropical changes are presented in Sect. 4.

Data
We make use of meteorological reanalysis data produced by the Japan Meteorological Agency (JMA), JRA55 (Kobayashi et al., 2015).A large discontinuity was found at the end of the 1990s in the previous reanalysis product, JRA25, when the 5 TIROS Operational Vertical Sounder (TOVS) on board the National Oceanic and Atmospheric Administration (NOAA) satellite was switched to Advanced TOVS (ATOVS) (Li et al., 2000).This discontinuity has largely been reduced in the JRA55 reanalysis (Kobayashi et al., 2015).To investigate deficiencies in the general circulation model, we make use of a family of JRA55 product: AMIP-type Simulation (JRA55-AMIP), which uses the same model and boundary conditions as JRA55 reanalysis, but without assimilation of observational data in the atmosphere.10 Outgoing longwave radiation (OLR) data provided by NOAA are widely used for analysis of convective activity in the tropics.In the present study, we use monthly mean OLR data (1° × 1° latitude-longitude resolution) derived from the High Resolution Infrared Radiation Sounder (HIRS) (Lee et al., 2007), available at ftp://eclipse.ncdc.noaa.gov/cdr/hirsolr/monthly/. Analysis of the precipitation is performed by making use of Global Precipitation Climatology Project (GPCP) 15 monthly mean data Ver.2.3 (Adler et al., 2003).COBE monthly mean gridded SST data with 1° × 1° boxes compiled by the JMA (Ishii et al., 2005) are used for the study of oceanic surface change.

Eastern Pacific cooling
The recent poleward shift of the convergence zone in the boreal summer is identified from the July-September mean 25 anomalous OLR in 1999-2016 from 30-year climatology (1981-2010) (Fig. 1a).A northward shift of the convective activity is seen around the summer monsoon regions over Africa, Asia, and Central America.A poleward shift is further evident in the zonal-mean OLR in Fig. 1b.Climatological OLR peaks around 10° N, while anomalous OLR of the recent period has a maximum around 15° N.There is a close relationship between the position of the tropical convergence zone and the development of cold tongues in the tropical oceans.Figure 2 shows the climatological mean annual cycle in zonal mean OLR, meridional wind at 925 hPa, and equatorial SSTs in the eastern Pacific−Atlantic sector.When convective activity, identified by low values of the OLR, shifts northward during boreal summer, cross-equatorial southerlies become stronger west of the American and African continents, which lead to a decrease in the SSTs along the coastal regions.Thus, the primary factor producing cold tongues 5 in the tropical SSTs is the shape of continents, air-sea interaction, and the position of the rising branch of the Hadley circulation, as described in Xie and Philander (1994) and Xie (2004).Therefore, changes in the rising branch of the Hadley circulation, such as can be seen in Fig. 1, can directly impact the equatorial eastern Pacific SSTs.
To investigate whether the northward shift of the convective zone is driven by the PDO, anomalous OLR during the two 10 periods of neutral and negative phases of the PDO is displayed in Figs.1c and 1d together with anomalous SSTs during those time periods (Figs.1e and 1f).A characteristic horseshoe pattern in the north Pacific SST is evident during the negative phase of the PDO.Anomalous OLR indicates that the convective activity is enhanced along 15° N-20° N latitude irrespective of the phase of the PDO, except for the sector under the direct influence of the PDO in the eastern Pacific, where cooling is larger during the negative phase.However, even during the neutral phase of the PDO, negative anomalies in the 15 SST exist in the tropics west of South America.This suggests that the SST cooling west of South America is not solely driven by the PDO, but is related to the strengthening and northward shift of the convergence zone.
The seasonal dependence of the recent trend in tropical convective activity is depicted in Fig. 3.The panels on the left show latitude-time sections of 3-monthly mean anomalous zonal-mean OLR from 1979 to 2016.To better illustrate the evolution 20 in latitudinal structure, the tropical mean (10° S-25° N) is further subtracted from the temporal anomaly in Fig. 3a.
Inspection reveals a change in zonal-mean OLR around 1999 (vertical lines) in all four seasons: the convergence zone in the NH tends to be located at higher latitudes in recent years compared to the previous period.The anomalous negative zone shifts northward after 1999 from boreal spring to summer and descends to lower latitudes from boreal summer to autumn following the seasonal march.25 The horizontal structure of the anomalous OLR in the recent period (1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) is depicted in the panels on the right for each season: April, May and June (AMJ), July, August and September (JAS), October November and December (OND), and January, February, and March (JFM).The anomalous negative centre of OLR is located around the Maritime Continent during boreal spring, which shifts northward by spreading longitudinally along 15° N, in particular along the African and 30 in the bottom panels.Because the response of the SST arises following atmospheric circulation, anomalous SSTs during the following month (i.e., August, September and October (ASO) and November, December and January (NDJ) ) are displayed in Figs.4c and 4d, respectively.In JAS, the anomalous cross-equatorial flow west of South America intensifies due to a poleward shift of a rising branch of the Hadley cell.The cross-equatorial flow changes from westward to eastward when it crosses the equator.This results in a strengthening of the climatological easterlies in the SH and enhances anomalous 10 convergence near New Guinea.In contrast, easterlies are weakened in the NH, which explains a warming in the north, but a cooling south of the equator.Such a meridional seesaw of the anomalous SSTs and cross-equatorial flow suggests an important role for wind−evaporation−SST (WES) feedback (Xie and Philander, 1994) in recent trends.The centre of anomalous negative OLR moves to the equatorial eastern Indian Ocean from boreal summer to autumn, which results in a strengthening of anomalous easterlies over the equatorial central Pacific and a westward extension of low SSTs over the 15 equator.
The impact of cooling of the eastern equatorial Pacific in the SH can also be seen as a structural change in El Niño/Southern Oscillation (ENSO) phenomena after 1999.Figure 5 shows a longitude-time section of the anomalous OLR over the equatorial SH (0°-10° S).Convective activity largely increases over the Pacific during El Niño events before 1999.20 However, after 1999, Pacific convective activity is suppressed; an increase in convective activity during El Niño is apparent only over the central Pacific.In contrast, convective activity west of 160° E over the Maritime Continent generally increases after 1999.

Vertical structure
A northward shift of convective activity is observed during boreal summer.Normally, Asian summer monsoons start from 25 the pre-Meiyu season around 27 May, and the convective activity jumps northward in midsummer around 18 July (Chiang et al., 2017).Differences in the northward shift of the convection between the recent decadal period (1999-2016 mean JAS anomalies) and the climatological seasonal march around the onset of midsummer (18 July) are depicted in Fig. 6.The horizontal structure of the OLR related to the recent trend (Fig. 6a) is similar to that related to the change around the onset of midsummer (Fig. 6b): On one hand, this pattern is characterized by increased convection around 15° N, in particular over 30 Africa and Asia and the western Pacific; on the other hand, convective activity over the equatorial region is generally suppressed, except for New Guinea in recent trends.The bottom panels compare meridional sections of zonal mean pressure vertical velocity during the same time periods as the top panels.Climatological mean pressure vertical velocity is also indicated by contours.In spite of the similarity of the horizontal structure of the OLR, the vertical structure of the pressure vertical velocity is largely different.In the case of the seasonal march (Fig. 6d), changes occur mainly horizontally as northward in the tropospheric.A northward shift in the recent trend in the lower troposphere is more or less similar to that in seasonal march (Fig. 6c).However, the vertical structure is quite different: In the upper troposphere, upwelling extends vertically to the tropical tropopause layer (TTL) along the rising branch of Hadley circulation indicated by climatological 5 upwelling region (contours in Fig. 6c).In the equatorial SH, intensification of upwelling is observed in the troposphere, but this does not extend into the stratosphere.
The recent increasing trend in the tropical lower stratospheric upwelling at 70 hPa, together with the increasing number of sudden stratospheric warming events, was identified by Abalos et al. (2015).It has been suggested that increased 10 stratospheric tropical upwelling tends to enhance deep convective activity, in particular deep penetrating clouds, through change in the static stability of the TTL (Eguchi et al., 2015;Kodera et al., 2015).A height-time section of the zonal mean temperature around the tropopause, together with the zonally averaged pressure vertical velocity at 70 hPa, shows year-toyear variation of the temperature associated with the quasi-biennial oscillation superimposed on a decadal cooling trend in the lower stratosphere.In the upper troposphere, warming pulses related to El Niño events are superimposed on a warming 15 trend in the troposphere (Fig. 7).A decrease in lower stratospheric temperature and an increase of upper tropospheric temperature lead to a decrease in the static stability of the TTL.
To investigate the coupling between the stratospheric and tropospheric variations in the tropics, a singular value decomposition (SVD) analysis (Kuroda, 1998) was conducted.First, we performed SVD analysis on the zonal mean pressure 20 vertical velocity around the TTL (30° S-30° N, 300-70 hPa) and the horizontal distribution of the OLR in the tropics (30° S-30° N) during boreal summer (JAS) based on the covariance matrix.Each value at the grid point was weighted by the cosine of the latitude and the layer thickness.The results are presented in Fig. 8.The left panels show the time coefficients.
The middle and right panels are heterogeneous correlations with OLR and the zonal mean vertical pressure velocity, respectively.To get a general view of the troposphere and stratosphere, the heterogeneous correlation of the vertical velocity 25 is extended to a height range of 1000 to 10 hPa.
The leading SVD mode (Fig. 8a) is related to the ENSO phenomenon.When El Niño occurs, convection over the equatorial central-eastern Pacific becomes active (negative OLR) and the time coefficients become large.Especially large amplitudes are found during the El Niño events of 1997/98 and 2015/16.The three types of El Niño identified by Paek et al. (2017) are 30 indicated above the time coefficients.The time coefficients of the second mode exhibit an increasing trend (Fig. 8b).The variation of the zonal mean pressure vertical velocity of the leading mode is confined essentially in the troposphere below 100 hPa, whereas that of the second mode extends into the lower stratosphere.Note that the upwelling of the second mode is also related to the variation of the middle stratosphere in the SH.This implies a strengthening of the Hadley circulation in connection with the enhancement of the Brewer-Dobson circulation.Tropospheric anomalous upwelling is disconnected from the stratosphere in the equatorial SH.A horizontal map of the OLR indicates that the upwelling region is located over the Maritime Continent.This increase in the regional convective activity is, in fact, coupled to a suppression of upwelling over surrounding oceans.

5
In the above analysis, we made use of the covariance matrix to calculate the SVD.In this case, a mode of variability having large amplitude such as the ENSO is extracted as the leading mode.To extract a variation of smaller amplitude but spreading over a large area, the use of normalized covariance, or the correlation matrix, may be appropriate.In such an analysis, the second mode in the previous analysis becomes the leading mode (Fig. 8c).An increasing trend related to the stratospheric circulation and the northward shift of convection over the Asian-African sector becomes further obvious.The ENSO related 10 variation becomes the second mode (Fig. 8d).The spatial structure of the OLR of this mode is slightly modified: negative OLR extends less eastward over the equatorial central Pacific.Accordingly, the time coefficients of the east Pacific type El Niño become smaller compared to those in Fig. 8a, and a slight increasing trend appears in the time series.
To focus on the stratosphere-related variation during the boreal summer, the JAS mean vertical pressure velocity (ω) at 30 15 hPa averaged over tropical SH (0°-25° S) is chosen as the index of stratospheric mean meridional circulation (Iω) (Fig. 9a).
The correlation coefficient between Iω and zonal mean ω at each grid point (Fig. 9b) shows a correlation pattern very similar to the leading mode in Fig. 8c.To indicate the relationship between the climatological residual mean-meridional circulation, the mass-weighted stream function diagnosed by the method of Iwasaki (1992) is displayed by contours in Figs.9b and 9c.
We can see that the variation in stratospheric upwelling in the Brewer-Dobson circulation is connected to the Hadley 20 circulation in the TTL around 15° N.
The same correlation, but with the zonal-mean temperature at each grid point from the 90°S to 90°N, is shown in Fig. 9c.
The tropical upwelling is not only related to a cooling in the tropics and the summer hemisphere, but it is also related to a warming in the downwelling region around the winter polar stratosphere (Fig. 9c).This illustrates the dynamical nature of 25 the recent tropical stratospheric cooling.Stratospheric upwelling is also connected with convective activity along 15° N−20° N (Fig. 9d), as already discussed above.The correlation coefficients between Iω and 925 hPa zonal and meridional winds at each grid point are shown by arrows in Fig. 9e.An increase in the cross-equatorial winds in the eastern Pacific and Atlantic is observed.The impact of near surface wind variation on the SST can be seen in the lagged correlation with the SST in Fig.Here we have simply shown possible relationships among the variables.However, a discussion of the statistical significance of the relationships between variables that exhibit large trends in a short data record is practically impossible.Thus, we will address this issue in a companion paper focusing on transient events (Kodera et al., in preparation).

Convection over continents and ocean
Two different modes of variability extracted by the SVD analysis in Fig. 8 are particularly pronounced, one over continents 5 and the other over ocean sectors.Here we examine the characteristics of the variations in two sectors: the African continent (10° W-40° E) and the Pacific Ocean (170° W-120° W).The Pacific sector corresponds to the longitudes of Niño 3.4.To identify the different characteristics of convective activity over continents and ocean, the climatological annual cycle in the zonal mean pressure vertical velocity at 300 hPa is depicted in Figs.10a and 10d.A region of enhanced convective activity migrates north and south over the continent following the seasonal variation of solar heating.It should be noted that the 10 convective zone over Africa shows a jump during the summer monsoon season (Hagos and Cook, 2007).In the case of the Pacific Ocean, the convective zone shows only a small latitudinal excursion and is located in the NH around 5° N throughout the year.
Latitude-time sections of the 3-monthly mean anomalous (departures from climatology) 300 hPa pressure vertical velocity 15 are shown in Figs.10b and 10e from February 1979 to November 2016.Over the African sector (Fig. 10b), the vertical velocity increases from the mid-1990s in both hemispheres around 10°−20° in latitude corresponding to the rising branch of the summertime Hadley circulation in the TTL.Accordingly, the annual mean precipitation over Africa increases in the recent period (1999−2016) in both hemispheres over the Sahel and Namibia.Over the Pacific Ocean sector (Fig. 10e), strong upward motion appears over the equator when El Niño events occur.The anomalous convective region, however, tends to 20 stay north of the equator after 1999.Accordingly, the annual mean anomalous precipitation during the recent period exhibits a large increase around 5° N−10° N, but decreases over the equator and the SH (Fig. 10f).Three-monthly mean meridional wind at 925 hPa averaged over the equator (5° S−5° N) and a latitude−time section of 3-monthly anomalous SST of the Niño 3.4 sector are shown in Figs.10g and 10h, respectively.Before 1999, an increase in equatorial SST was accompanied by a strengthening of convection over the equator and the anomalous southward wind.After 1999, although SST increased over 25 the equator especially during El Niño, the anomalous northward wind was stronger and the convective activity tended to remain in the NH.

Stratosphere−troposphere connection
The connection between tropospheric and stratospheric circulations around the tropopause region is shown in Figs. 8 and 9. are displayed in Fig. 11.Horizontal distributions of the anomalous pressure vertical velocity at 150 hPa and the anomalous surface precipitation are also displayed in the lower panels.Anomalous upwelling, extending from the upper troposphere to the lower stratosphere, is connected with the surface precipitation over continents and their vicinity.The vertical structure of the present decadal change in both hemispheres has a similar structure: upwelling tends to be enhanced over the continents, but suppressed over the oceans.Contrast between the continent and ocean sector is clearly seen in the stratosphere of the SH 5 where the distribution of land is simpler.Note also that the variation of surface precipitation is larger over the ocean than over land, but the associated height of the upwelling is lower and has a smaller impact on the stratospheric circulation.

Summary and discussion
Characteristics of the recent tropical circulation change from the middle to the end of the 1990s can be summarized as follows.Cooling of the ocean in the equatorial SH occurred in association with a strengthening of cross-equatorial 10 southerlies near the surface, which was induced by a northward shift of the convective zone connected to the rising branch of the Hadley circulation and the stratospheric Brewer−Dobson circulation.The transport of water vapour by the enhanced cross-equatorial southerlies further amplified the convective activity in the NH.An increase in the vertical velocity was most apparent in the TTL.This could be related to a decrease in the static stability in the TTL due to the combined effects of lower stratospheric cooling and upper tropospheric warming.Thus, we conclude that the recent tropical circulation change 15 originates primarily from a strengthening of deep convective activity over the continents and their vicinity in the summer hemisphere, in particular over the African-Asian sector (30° W−150° E).Aumann and Ruzmaikin (2013) reported that deep convection over land in the tropics shows an increasing trend, while that over tropical oceans shows an opposite decreasing trend based on 10 years of Atmospheric Infrared Sounder (AIRS) data.

20
The impact of the northward shift of summertime Hadley circulation in the Pacific sector is amplified through wind−evaporation−SST (WES) feedback (Fig. 5).During boreal autumn and winter, anomalously active convective centre moves southward to the equatorial Indian Ocean-Maritime Continent region, which induces stronger easterlies over the equatorial central Pacific.As a result, sea surface cooling extends from the equatorial eastern Pacific to the central Pacific.
This could explain the recent trend in the suppression of warming in the eastern Pacific including the aborted El Niño of 25 2014 (Maeda et al., 2016;Wu et al., 2017) and a decrease in the occurrence of east Pacific-type El Niños.It should also be noted that the El Niño of 2015/16 was one of the largest events; however, it was a mixed type with smaller warming in the eastern Pacific (Paek et al., 2017), as can also be seen in the OLR field in Fig. 5.
Results of the present study indicate some similarities between the recent trends in the two summer hemispheres, which can 30 be qualified as a strengthening of the rising branch of the Hadley circulation connected to the stratospheric circulation (Fig. 11).The increasing trend in the Earth's surface temperature is generally attributed to the increase in greenhouse gases like Sahelian sector, whereas it suppresses upwelling over the oceanic sector in the Pacific (Fig. 8 of Gaetani et al., 2016).Initial changes in convective activity over land and ocean can be amplified through associated circulation changes: warming of the 5 ocean surface is reduced by increased evaporation, and the water vapour transported to the continents further enhances convective activity over land.In fact, synchrony of the recent increasing trends in Sahelian rainfall with global oceanic evaporation has been documented by Diawara et al. (2016).
The tropospheric response to change in the radiative forcing described above is a fundamental process that should be easily 10 reproduced in global models.However, model simulations using realistic forcings, such as the Coupled Model Intercomparison Project Phase 5 (CMIP5) (Taylor et al., 2012), fail to reproduce the recent cooling tendency in the equatorial Pacific in the model ensemble mean.Therefore the equatorial Pacific cooling is often attributed to the internal variation of the atmosphere-ocean coupled system (Meehl et al., 2014) though it could also be due to model deficiencies.
Here we briefly investigate whether the effect of extreme deep penetrating convection on large scale circulation is well 15 simulated in a model.For convenience, we compare the climatological mean vertical velocities from the JRA55 reanalysis and the JRA55-AMIP run.The latter uses the same atmospheric general circulation model and boundary conditions (SST and sea ice) as JRA55, but without the assimilation of observed data (Kobayashi and Iwasaki, 2016).The results show that upwelling in the summer tropics and downwelling in the winter polar stratosphere are weaker in JRA55-AMIP (Fig. 12a).
The horizontal distribution of the difference in pressure vertical velocity at 150 hPa (lower panels in Fig. 12a) show that 20 model upwelling is weaker in summer over land and adjacent oceans around 10°-20° in latitude, where deep penetrative convection is frequent (Liu and Zipser, 2005).Regions of weak upwelling in the model coincide with the regions of increasing recent decadal trends in upwelling (see Fig. 9).
The above results based on a general circulation model used for JRA55 reanalysis are common biases found in other models: 25 the NH intertropical convergence zone is weaker and the eastern Pacific SST is warmer compared to observations owing to weaker cross-equatorial winds (Li and Xie, 2014;Richter, 2015).It is also noted that equatorial eastern Pacific SST is very sensitive to the model convective cloud scheme (Braconnot et al., 2007).Because the observed change is the opposite of these model biases, it is possible that the failure of the model to simulate the recent cooling in the tropical eastern Pacific originates from these model deficiencies.The recent decadal trend in the tropical troposphere is connected to the change in 30 the stratosphere, as seen in Figs. 8 and 11.Note that the extreme deep convection penetrating into the TTL is not only influenced by the underlying SST, but is also sensitive to the stratospheric circulation (Wang et al., 2014;Eguchi et al., 2015: Kodera et al., 2015).Therefore, the unsuccessful simulation could be the result of unrealistically low sensitivity to a cooling in the lower stratosphere in the models.An increase in CO2 raises the Earth's surface temperature, but also decreases stratospheric temperatures.Such changes in the vertical structure of the atmosphere may create favourable conditions for the development of extreme deep convection penetrating the TTL over and around the continents.At the same time, warming of the upper troposphere over remote oceanic regions where cloud tops are generally lower leads to a suppression of convective activity over the ocean owing to 5 increased static stability.The difference in convective activity over land vs. ocean is further amplified through the transport of water vapour from the ocean to the land.The present working hypothesis (Fig. 13) applies well during boreal summer in particular and where a large land mass is located off the equator.Although an increase in the concentration of CO2 produces a cooling effect in the stratosphere, recent cooling in the lower stratosphere-tropopause region is largely due to a dynamical effect (Abalos et al., 2015).The sudden decrease in lower stratospheric water vapour around the year 2000 (Randel et al., 10 2006;Hasebe and Noguchi, 2016) also supports the dynamical forcing of the change.Further investigation is needed to determine whether the stratosphere is merely passively responding to the tropospheric warming or playing an active role in the tropospheric circulation change.
Studies of long-term trends face problems that often arise from changes in the observation systems, stations or the drift of the 15 orbit of the satellites; even reanalysis datasets making use of the same model and the same method of analysis are subject to these artifacts.For instance, a large erroneous decrease in TTL temperatures was found in the NCEP-1 reanalysis around the end of the 1990s (Vecchi et al., 2013), which coincides with the change in the observing system from TOVS to ATOVS at NOAA (Li et al., 2000).Therefore, to avoid these problems and to better understand the stratosphere-troposphere coupling processes, we conducted additional case studies on the northward shift of convective activity similar to that associated with 20 the recent trend, but associated with a sudden cooling of the tropical stratosphere during a boreal summer monsoon (Kodera et al., 2018, paper in preparation).In the present study, we made use of daily mean OLR data to investigate the evolution of convective activity due to its long data record.However, OLR is not the most suitable variable for investigating extreme deep convection that plays a critical role in stratospheric and tropospheric coupling (Kodera et al., 2015).Therefore, a companion paper will investigate the coupling processes by using convective overshooting and cloud top data derived from 25 recent satellite observations.
Asian sectors during boreal summer.Next, the negative OLR region shifts and clusters over the Indian Ocean region during boreal autumn.At the same time, positive OLR anomalies develop over the equatorial eastern Pacific, suggesting a suppression of convective activity over the Pacific Ocean.During boreal winter, the anomalous convective active region Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-128Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 16 February 2018 c Author(s) 2018.CC BY 4.0 License.shifts eastward over the Maritime Continent, and suppression of convective activity over the eastern Pacific is sustained during the cold season.The impact of the recent decadal variation of convective activity on the SST is depicted in Fig. 4. The top panels show the 1999-2016 mean anomalous OLR in (a) JAS, (b) OND.The anomalous horizontal winds at 925 hPa and SSTs are presented 5 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-128Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 16 February 2018 c Author(s) 2018.CC BY 4.0 License.
Continuity in a zonally averaged field, however, does not necessarily indicate actual continuity at each location.To confirm 30 continuity within the rising branch of the Hadley circulation from the upper troposphere to the stratosphere, longitude−height sections of the normalized anomalous pressure vertical velocities averaged over 10°−20° latitudes in the summer hemisphere Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-128Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 16 February 2018 c Author(s) 2018.CC BY 4.0 License.
Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-128Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 16 February 2018 c Author(s) 2018.CC BY 4.0 License.CO2 (IPCC, 2013).Such a change in the radiative forcing may explain the global characteristics of the recent change.The effect of increased CO2 can be divided into a direct radiative effect and an indirect effect through changes in the SST.Model experiments have shown that the direct radiative effect of CO2 increases tropical upward motion particularly over the Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-128Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 16 February 2018 c Author(s) 2018.CC BY 4.0 License.

Figure 8 :
Figure 8: SVD analysis of the zonal mean anomalous pressure velocity in the upper troposphere-lower stratosphere (300-70 hPa: gray lines) and anomalous OLR (0°-360° E) in the tropics (30° S-30° N) during JAS from 1979 to 2016.(a) Leading mode and (b) second mode based on the covariance matrix: (from left to right) time coefficients, heterogeneous correlation of OLRs, and 5

Figure 9 :
Figure 9: (a) Time series of JAS mean vertical pressure velocity (ω) at 30 hPa averaged over 0°-25° S as an index for the tropical stratospheric vertical velocity (Iω).Correlation coefficients between Iω and (b) zonal mean ω, (c) zonal mean temperature T at each grid, (d) OLR, and (e) horizontal winds at 925 hPa (arrows).Lagged correlation with DJF mean SST is also presented as colored 5

Figure 10 :
Figure 10: (a) Latitude-time section of the climatological zonal-mean pressure vertical velocity at 300 hPa averaged over the 5

Figure 11 :
Figure 11: (a) (top panel) Height-longitude section of the standardized (with respect to the interannual variation) anomalous pressure vertical velocity averaged over 10° N-20° N, during boreal summer (JAS) 1999-2016.(middle panel) Standardized pressure vertical velocity at 150 hPa during boreal summer (JAS) 1999-2016.(bottom panel) As in the middle panel, but for 5

Figure 12 :
Figure 12: (a) Difference in the climatological JAS (top) and DJF (bottom) mean residual vertical velocity between JRA55-AMIP and JRA55 reanalyses (color shading).Contours indicate the mass stream function of the mean residual circulation in JRA55-AMIP.Streamlines connected to the winter polar stratosphere from the tropics are highlighted using thick lines.Solid and dashed lines represent clockwise and counterclockwise circulations, respectively.Statistically significant differences are indicated in color.

5 (Figure 13 :
Figure 13: Schematic of the recent changes in the tropics from the late 1990s.Earth's surface warms while the tropical stratosphere cools in response to enhanced stratospheric upwelling.This tends to enhance extreme deep convective activity penetrating the TTL in boreal summer over land, but reduces deep convective activity over equatorial oceans.Differences in convective activity are further amplified through the transport of water vapour in the lower troposphere.(See text for details.)