Regional meridional cells governing the interannual variability of the Hadley circulation in boreal winter

Regional meridional cells governing the interannual variability of the Hadley circulation in... The Hadley circulation (HC) has conventionally been considered as thermally direct with uniform zonal distribution. How- ever, the meridional circulation in the tropics is far from uniform, including the thermally direct cells associated with the global monsoon heating and indirect cells in the absence of diabatic heating. This study aims at assessing the thermally direct and indirect cells in different regions, identifying the geographic sectors responsible for interannual variability of HC strength and boundaries, and unraveling the underlying mechanism for the interannual variability of HC edges and intensity. Results derived from ERA-Interim reanalysis and climate models show that the climatological HC in wintertime (December–January–February) obscures longitudinal diversity of regional meridional cells (RMC), including thermally direct RMCs over Eurasia and the Eastern Pacific, thermally direct southern limbs of RMCs over the Central Pacific and Western Atlantic with opposite circulation. For each of the regions, El Niño-South Oscillation and mid-latitude eddies are assessed in terms of their relative contributions to the interannual variability of HC intensity and extent. Their underneath physical mechanisms are thoroughly investigated. Keywords Hadley circulation · Global monsoon · Strength · Extent · Interannual variability · ENSO · Mid-latitude eddies · Jet · Reanalysis · CMIP5 model 1 Introduction and momentum. The intensity and extent of the HC have varied significantly over recent decades (Quan et al. 2004; The Hadley circulation (hereafter HC) is conventionally Hu and Fu 2007; Gastineau et al. 2011; Sun et al. 2013), depicted as the dominant large-scale atmospheric circula- with very pronounced interannual variability. tion in the Earth’s tropics and subtropics, as described by a An early attempt to understand the interannual vari- conceptual model with zonal average over the globe (Hadley ability of HC was reported in Oort and Yienger 1996 who 1735). It plays an important role in the transport and balance used radiosonde data and pinpointed a strong connection to of fundamental climate variables, such as moisture, energy, ENSO. Most recent studies have used reanalysis datasets to examine the seasonal cycle and interannual variability of the HC. For instance, a good agreement on the interan- * Yong Sun nual variability of the annual mean HC was reported among sunyong@mail.iap.ac.cn multi-reanalysis datasets (Nguyen et al. 2013). But HC has pronounced seasonality (Dima and Wallace 2003), annual State Key Laboratory of Numerical Modeling mean does not reveal its complete picture. A few other stud- for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy ies were reported to examine the interannual variability of of Sciences, Beijing 100029, China a specific season (Tanaka et al. 2004; Stachnik and Schu- Laboratoire des Sciences du Climat et de macher 2011). l’Environnement/IPSL, CEA-CNRS-UVSQ, Université Sun et al. (2013) showed that there is a mode of HC inter- Paris-Saclay, Gif-sur-Yvette, France annual variability that is driven by ENSO, and this mode Laboratoire de Météorologie Dynamique, CNRS, Sorbonne is thought to be linked with the interannual variability of Université, Paris, France HC strength associated with HC anomalies that are simi- Institute of Meteorology and Climate Research, Karlsruhe lar (opposite) to its mean state in the tropics (subtropics). Institute of Technology, Karlsruhe, Germany Vol.:(0123456789) 1 3 Y. Sun et al. Furthermore, Sun and Zhou (2014) investigated the behavior ENSO-induced meridional temperature gradient plays a key of HC in boreal summer. There are two prominent variability role in setting the edges of the HC. There is also an indirect modes for the developing and decaying phases of El Niño, effect of ENSO through ENSO-induced changes in mid-lat- respectively. The interannual variabilities of its northern itude eddies, as demonstrated by Guo and Li (2016) on the branch in both modes are largely under the control of the interannual variability of the northern HC edge. East Asian summer monsoon (EASM) variability, while that After an overview of numerous studies on global-scale of its southern branch results from the atmospheric response HC, it is necessary to emphasize that HC is decomposed to meridional shift of diabatic heating from the developing to of regional meridional cells (RMC) over a few geographic decaying phases of El Niño (Sun and Zhou 2014). A series sectors. They are collectively responsible for HC strength of recent studies have decomposed the HC variability into and extents, and there is a rich longitudinal diversity in the an equatorially asymmetric mode (ASM) and a symmetric tropics. We want to study the contributions from RMC of mode (SM) in terms of variability of annual mean (Feng different sectors to HC. We should point out that the lack et al. 2015), boreal winter (Ma and Li 2008) and boreal sum- of adequate metrics to gauge the regional characteristics of mer HC (Feng et al. 2011). The ASM mainly responds to HC increases the complexity and challenges to study the the SST warming over the Indo-Pacific warm pool and thus relationship between HC and RMC. explains the long-term variability of HC (Feng et al. 2013), The thermally direct summer monsoon cells in some while the SM linked to ENSO is responsible for the interan- geographic sectors (e.g. East Asia) are of inverse sense to nual variability of HC (Ma and Li 2008). the climatological HC in the summer hemisphere (North- By comparing the impacts of different meridional struc- ern Hemisphere) (Riehl et al. 1950). The interannual vari- ture of tropical SST on annual mean HC, ASM and SM ability of RMC over EASM sector plays a dominant role have been identified as the contrasting responses to equatori- in the interannual variability of the northern branch of the ally asymmetric and symmetric meridional SST structures HC in boreal summer (Sun and Zhou 2014). However, there (Feng et al. 2016). It is also true when the responses of the are no studies to assess the direct linkage between HC and HC to different meridional SST structures in the seasonal the global monsoon circulation. Additionally, it remains cycle are taken into account (Feng et al. 2017). A recent unknown whether the thermal direct circulation in associa- case study was reported in Feng and Li (2013) who investi- tion with the southern hemispheric monsoon regimes also gated different impacts of two ENSO types (canonical ENSO makes a reverse contribution to the southern Hadley cell in and ENSO Modoki) on the interannual variability of HC in austral summer. boreal spring. They revealed the important role of the merid- Most of these previous studies are based on the mass ional structure of sea surface temperature (SST) on the vari- stream function (MSF) that strictly satisfies the mass con- ation of HC. It has been reported that SST changes can also servation constraint. This is a powerful tool for diagnosing explain part of variability in HC strength and extent (Corvec the zonally-averaged HC. However, it neglects the regional and Fletcher 2017). The intensity of HC is proportional to diversity of the tropical atmospheric circulation. The verti- SST gradient between the tropics and subtropics (William- cal shear of the meridional wind between 200 and 850 hPa son et al. 2013), while the edges of HC are inversely pro- (V200–V850) is also a good indicator for HC and to some portional to meridional SST gradient (Adam et al. 2014). extent can remedy the shortcomings of MSF. It has been A rich literature can also be found in developing theories widely used as an alternative measure of boreal winter HC and mechanisms to understand the dynamics of HC. Under in earlier studies to depict the climatology of zonal HC (Oort the assumption of angular momentum conservation, Held and Yienger 1996; Quan et al. 2004), the interannual vari- and Hou (1980) provided a thermal model to explain suc- ation of HC related to ENSO (Oort and Yienger 1996; Sun cessfully the HC behavior. But this early theory neglected et al. 2013), and the strengthening trend of HC related to completely the role of mid-latitude eddies in the dynamics of the interdecadal intensification of ENSO amplitude (Quan HC. The effect of eddies has been highlighted in the subse- et al. 2004). A recent study has used V200-V850 to separate quent work (Held 2000), together with local Rossby number regions where the HC is reinforced (i.e., the features are R used to measure the relative importance of thermal forc- “Hadleywise”) from those where the HC is weakened (i.e., ing and eddy forcing on HC (Schneider 2006). they are “anti-Hadleywise”) and highlighted the dynamics ENSO and mid-latitude eddies both affect the HC strength of the HC and its inextricable linkage to subtropical drying (Oort and Yienger 1996; Walker and Schneider 2006), but a in a future scenario (Karnauskas and Ummenhofer 2014). In larger role is attributed to mid-latitude eddies for the interan- addition, a very recent study has attempted to investigate the nual variability of HC strength in boreal winter (Caballero regional variability of the extent of the HC in the Southern 2007; Caballero and Anderson 2009). ENSO can however Hemisphere by calculating the stream function via the verti- strongly affect the interannual variability of the extent and cal integral of the divergent meridional wind (Nguyen et al. boundaries of the HC in boreal winter, suggesting that the 2017). The strengthening trend of the HC in boreal winter 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… has also been validated from the perspective of the velocity one of the latest generation of atmospheric reanalyses pro- potential (Zhou et al. 2016). However, the stream function duced by the European Centre for Medium-Range Weather and velocity potential are obtained from the incompressible Forecasts (ECMWF), with a higher horizontal resolution of and irrotational components of air flow, respectively. Obvi- 1.25° × 1.25° (Dee et al. 2011), and (4) the Japanese 25-year ously, neither includes full details of the horizontal wind Reanalysis (JRA25) obtained from the Japan Meteorological and may not give a complete description of the regional Agency (JMA) (Onogi et al. 2007). diversity of the HC (Sun et al. 2017). Moreover, there are In general, the four reanalyses are consistent. Throughout increasing attempts to depict features of monsoon circulation the paper, we use the ERA-Interim reanalysis to demonstrate using regional MSF via a direct integral of zonally averaged the diversity of HC climatology and the implications for HC meridional wind over the specific monsoon sector (Bordoni intensity and edges, except where multi-reanalyses are used and Schneider 2008; Toma and Webster 2010; Jayakumar to confirm the differences in the HC strength and extent in et al. 2013; Walker et al. 2015). Out of consideration above, each of the four geographic sectors (Table 2). in the present paper we attempt to study the longitudinal We also examine model performance on simulating HC diversity of RMC within global monsoon domains by calcu- behavior using the Atmospheric Model Intercomparison lating the MSF over specific sectors from the original wind Project (AMIP) simulations involved in the Coupled Model fields and scaled by multiplying the coefficient (Δθ is Intercomparison Project phase 5 (CMIP5) (Table 1). The longitudinal range of each sector) (Zhang and Wang 2015). multi-model ensemble (MME) is used to evaluate model Given the general context and state-of-the-art for HC, performance, despite the fact that MME tends to reduce the firstly we would like in the present study to focalize on the variance and overestimate correlation, compared with each longitudinal diversity of RMCs within different monsoon single model (Knutti et al. 2009). domains. We are not only interested in long-term mean cli- matology, but also the interannual variability of HC (inten-2.2 Methodology sity and its southern and northern edges) and the associated RMC in each sector. Secondly, we would like to qualitatively 2.2.1 Mass stream function assess the relative contribution of ENSO and mid-latitude eddies in each sector on HC edges and intensity at interan- The meridional–vertical MSF is an important metric of nual time scale. Finally, our aim is also to assess the perfor- the HC (Oort and Yienger 1996). The MSF can be easily mance of models forced by realistic SST in reproducing the obtained via a mass-weighted vertical integral of the zonal interannual variability of HC and its connection to that of mean meridional velocity. In spherical coordinates, the MSF, RMC within each global monsoon sector. Ψ, at each pressure level, p, and latitude,  , can be expressed Due to the expected strong influence of monsoon on HC, as: transition seasons (spring and autumn) are not our ideal choice. But the boreal winter provides a good case. Moreo- 2๐œ‹ a cos ๐œƒ Ψ(๐œƒ , p)= V(๐œƒ , p) ⋅ dp, (1) ver, ENSO usually reaches its strongest amplitude during g boreal winter (December–January–February, DJF). Finally selecting boreal winter also offers us the possibility of com- where a is the Earth’s radius, g the gravitational accelera- parison with previous work (Caballero 2007; Ma and Li tion, and V the zonal mean meridional velocity. For mass 2008; Caballero and Anderson 2009). conservation, the zonal mean meridional velocity and verti- This paper is organized as follows. Descriptions of the cal motion ( ) must satisfy: data and methodology are presented in Sect. 2. Section 3 g g ๐œ• Ψ ๐œ• Ψ gives the main results and a summary and conclusions are V = and =− . (2) 2๐œ‹ a cos ๐œƒ ๐œ• p 2๐œ‹ a cos ๐œƒ ๐œ•๐œƒ provided in Sect. 4. We should keep in mind that V ( ) has the same (opposite) Ψ Ψ sign as . By examining the climatology of the MSF 2 Data and methodology (Fig. 1), it is clear that the MSF can depict the thermodynamic 2.1 Reanalysis datasets and model simulation circulation, characterized by ascending motion near the equa- ๐œ• Ψ tor ∼− < 0 and descending motion in the subtropics ๐œ•๐œƒ We first evaluate the consistency of four reanalysis datasets ๐œ• Ψ ∼− > 0 as well as poleward flow in depicting the climatology of HC and its interannual vari- ๐œ•๐œƒ ability. They are (1) the NCEP/NCAR Reanalysis (NCEP1) ๐œ• Ψ ๐œ• Ψ ฬ„ ฬ„ southerliesโˆถ V ∼ > 0 or northerliesโˆถ V ∼ < 0 at ๐œ• p ๐œ• p (Kalnay et al. 1996), (2) the NCEP–DOE AMIP-II Reanal- upper levels and equatorward flow ysis (NCEP2) (Kanamitsu et al. 2002), (3) ERA-Interim, 1 3 ฬ„๐œ” ฬ„๐œ” ฬ„๐œ” ฬ„๐œ” ฬ„๐œ” Y. Sun et al. Table 1 List of CMIP5 models used in this study Model Affiliation Horizontal resolution ACCESS1-0 Commonwealth Scientific and Industrial Research Organization/Bureau of Meteorology, Australia 144 × 192 CanAM4 Climate Modelling and Analysis, Canada 96 × 192 CSIRO-Mk3-6-0 Commonwealth Scientific and Industrial Research Organization/Queensland Climate Change Centre 96  × 192 of Excellence, Australia HadGEM2-A Met Office Hadley Centre, UK 144  × 192 IPSL-CM5A-LR Institut Pierre Simon Laplace, France 96  × 96 IPSL-CM5A-MR 96  × 192 IPSL-CM5B-LR 96  × 96 MPI-ESM-LR Max Planck Institute for Meteorology, Germany 96  × 192 MPI-ESM-MR 96  × 192 NorESM1-M Bjerknes Centre for Climate Research, Norwegian Meteorological Institute, Norway 96  × 144 EC-EARTH EC-EARTH consortium published at Irish Centre for High-End Computing, Netherlands/Ireland 160  × 320 GFDL-CM3 Geophysical Fluid Dynamics Laboratory, USA 90  × 144 bcc-csm1-1 Beijing Climate Center, China 64  × 128 bcc-csm1-1-m 160  × 320 CCSM4 National Center for Atmospheric Research, USA 192  × 288 CMCC-CM Centro Euro-Mediterraneo sui Cambiamenti Climatici, Italy 240  × 480 FGOALS-g2 Institute of Atmospheric Physics, Chinese Academy of Sciences, China 60  × 128 FGOALS-s2 108  × 128 GISS-E2-R NASA/GISS (Goddard Institute for Space Studies), USA 90  × 144 MIROC5 Atmosphere and Ocean Research Institute (The University of Tokyo), National Institute for Environ- 128  × 256 mental Studies, and Japan Agency for Marine-Earth Science and Technology, Japan MRI-AGCM3-2H Meteorological Research Institute, Japan 320  × 640 MRI-AGCM3-2S 960  × 1920 MRI-CGCM3 160  × 320 BNU-ESM Beijing Normal University, China 64  × 128 CNRM-CM5 Centre National de Recherches Météorologiques,Centre Européen de Recherche et de Formation 128  × 256 Avancée en Calcul Scientifique, France Inmcm4 Russian Academy of Sciences, Institute of Numerical Mathematics, Russia 120  × 180 N N ๐œ• Ψ ๐œ• Ψ where Ψ is the linear regression of Ψ on the El Niño ฬ„ ฬ„ northerliesโˆถ V ∼ < 0 or southerlies โˆถ V ∼ > 0 ๐œ• p ๐œ• p 3.4 index (ENSO variability) and Ψ is an uncorrelated near the surface. The strength of the HC is defined by the residual (non-ENSO variability referred to as mid-latitude maximum MSF between 30°S and 30°N (Oort and Yienger eddy activity). Here we use the same method to evaluate 1996). The edge of the HC in each hemisphere is defined as the relative importance of ENSO and mid-latitude eddies in the latitudinal position where the value of MSF at 500 hPa is the interannual variability of HC intensity in the Northern zero in the subtropics. Hemisphere and of the HC edge in each hemisphere over 30 years (1979–2008). 2.2.2 Decomposition of the variability in HC strength and extent 2.2.3 Wave activity flux and diabatic heating Caballero (2007) emphasized the dominant role of mid-lat- itude eddies in driving the interannual variability of HC in Wave activity flux is used to track the source of mid-latitude boreal winter, by assessing the relative contributions of ENSO eddies and their propagation from sources to sinks, which is and non-ENSO variability to the total variance of winter hemi- a powerful tool (e.g. see Plumb 1985) to diagnose the impact sphere HC intensity ( Ψ ) over the period 1959–2001. He par- of mid-latitude stationary waves on regional Hadley cells titioned the detrended Ψ (N denotes Northern Hemisphere) (Caballero and Anderson 2009). We calculate wave activity into two components: flux based on 6-hourly output of ERA-Interim reanalysis. This procedure is repeated for individual model simulations N N N Ψ =Ψ +Ψ , (3) e r based on daily data before taking the MME average. 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… cd Fig. 1 Climatological boreal winter (December–January–February, reanalysis (upper panels) and the multi-model ensemble mean (MME, 10 −1 DJF) mass stream function (MSF, units: 10 kg s , contour interval lower panels) of AMIP-type simulations. The green arrows indicate 10 −1 2 × 10 kg s ) and vertical shear of meridional velocity between 200 the air flow direction of the HC in each hemisphere −1 and 850  hPa (i.e., V200 minus V850, units: m s ) in ERA-Interim To understand thermal control of ENSO on HC intensity, Northern Hemisphere from anticlockwise in the Southern diabatic heating estimated as a residual in the thermody- Hemisphere (Fig. 1b). Both panels reveal an obvious asym- namic equation (Nigam et al. 2000) is calculated based on metry of HC between the summer and winter hemispheres 6-hourly frequency of ERA-Interim reanalysis, and daily (Fig. 1a, b). The Hadley cell in the Northern Hemisphere output for each model before MME average. (NHC) is much stronger and wider than that in the Southern Hemisphere (SHC). Similarly, we can use MSF and V200–V850 to evaluate 3 Results the performance of state-of-the-art climate models in simu- lating the HC in boreal winter. Under realistic SST forcing, 3.1 Climatology of HC and inhomogeneous the MME can reproduce reasonably well the hemispheric distribution of RMCs asymmetry of the strength and extent of HC (e.g., a much stronger and “fatter” NHC in the winter hemisphere than We first examine the climatology of the HC in boreal winter SHC in the summer hemisphere) (Fig. 1c, d). and how its structure varies with longitude. Figure  1a, b It is important to keep in mind that the MSF is deduced depict the mean state of the HC in the ERA-Interim rea- from the vertical integral of zonal mean meridional veloc- nalysis based on the MSF and the vertical shear of meridi- ity, and thus limits further insight into the horizontal dis- onal velocity (defined as V200 minus V850, V200–V850). tribution of the HC. Fortunately, V200–V850 performs The positive (negative) contours in Fig. 1a indicate clock- well in representing the climatology of zonal HC and offers wise (anticlockwise) circulation in the Northern (Southern) an opportunity to investigate longitudinal distributions Hemisphere. Likewise, the positive and negative values of RMC and resultant climatology of HC (Fig. 2). Obvi- of V200–V850 can distinguish clockwise motion in the ously, the RMC (Fig.  2a, contours) is far from zonally 1 3 Y. Sun et al. a b Fig. 2 Interannual standard deviation of the HC in the ERA-Interim ent regional meridional cells (RMCs) along the geographic sectors, reanalysis and MME simulation (color shading), illustrated by the including RMCs with appearance similar to HC (red solid rectangles) 10 −1 −1 MSF (units: 10 kg s ) and V200–V850 (units: m s , contour and opposite to that of HC (blue dashed rectangles). The extent of −1 interval: 0.25  m s ) in comparison with the corresponding climato- each rectangle in the following plots is same as that presented here logical distributions (contours). Rectangles in b indicate the differ - uniform as demonstrated by a comparison of the zonal mean distinct roles in the climatology of HC can be captured by V200–V850 with its horizontal distribution (Fig. 2b). The state-of-the-art climate models (contours, Fig. 2c, d). inhomogeneous RMCs across longitudes are of two types: one has almost the same features as the climatological 3.2 Decomposition of RMCs into thermally direct zonal-mean HC in each hemisphere, as seen in the RMC and indirect cells over Eurasia (30°S–30°N, 30°W–155°E; RMC-EA) and the Eastern Pacific (30°S–30°N, 130°W–70°W; RMC-EP); The RMCs in the four geographic sectors show different the other has a spatial pattern in the Northern and South- contributions to the climatology of the HC. Traditionally ern hemispheres almost opposite to the climatology of the HC has been thought of as a thermally direct zonal-mean NHC and SHC, as in the RMCs over the Central Pacific circulation. Here it may raise a question whether the RMCs (30°S–30°N, 155°E–130°W; RMC-CP) and the Western are thermally direct cells with uniform distribution. The Atlantic (30°S–30°N, 70°W–30°W; RMC-WA). Therefore, longitudinal diversity of RMCs consists of thermally direct the RMC-EA and RMC-EP both play a fundamental role monsoon cells and thermally indirect meridional cells. Fig- in maintaining the climatological zonal-mean HC in boreal ure 3a shows the spatial distributions of RMCs and diabatic winter, while RMC-CP and RMC-WA act to oppose the heating in global monsoon (GM) domains. RMCs within climatology. GM domains are associated with monsoonal heating and Examining the spatial patterns of MSF and V200–V850 thus they are thermally direct cells, including large resem- in the MME simulation reveals that the longitudinal diver- blance of RMC over the African–Asian–Australian monsoon sity of RMCs presented in ERA-Interim and their associated (RMC-EA) and the North American monsoon (RMC-EP) as 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Fig. 3 a A simple diagram drawn to reveal the relationship between the spatial distributions of RMCs over four geographic sectors and global monsoon (GM) domains (black dots) based on GPCP precipitation data. The vectors show the cli- matological wintertime atmos- pheric circulation at 850 hPa derived from ERA-Interim rea- nalysis data. The shaded areas indicate summertime diabatic −1 heating (units: K day ) in each hemisphere (boreal summer: May to September; austral sum- mer: November to March); b Taylor diagram to evaluate the performance of climate models participating in CMIP5 and the consistency of the HC in three widely used reanalyses with the HC in ERA-Interim (REF) the climatological HC, and RMC of signs opposite to clima- region (OM) where there is no land–sea thermal contrast tological SHC over the South American monsoon (the south- (southern limb of RMC-CP) (Liu et al. 2009). In contrast, ern limb of RMC-WA) and the pure oceanic monsoon-like the northern limbs of RMC-CP and RMC-WA in the absence 1 3 Y. Sun et al. of monsoon like heating are thermally indirect cells (coun- counterclockwise one in the Northern Hemisphere (Fig. 4c, terclockwise circulation). d, contours). These features are well consistent with the RMCs in ERA-Interim show a longitudinal diversity results derived from V200–V850 (Fig.  2a, b, contours). within the meridional sectors of GM domains. By examin- Besides, the thermally direct and indirect cells of RMCs ing the horizontal distributions of V200–V850 in the MME can be seen more clearly from regional MSF. The thermally simulation (contours, Fig. 2c, d), we can see that MME cap- direct cells associated with monsoonal heating are “Had- tures the decomposition of RMCs into thermally direct and leywise” over EA and EP sectors and “anti-Hadleywise” in indirect cells associated with the presence and absence of the southern parts of CP and WA sectors. The “anti-Hadley- monsoonal heating within the different meridional sectors wise” features in the northern parts of CP and WA sectors of the GM domains. are thermally indirect cells due to the absence of monsoon like heating. 3.3 Interannual variability of HC and longitudinal As for the interannual variation of RMCs, the standard diversity of RMCs variability deviations of MSF in the four sectors are of distinct spatial patterns in terms of magnitude and position of RMC vari- If we examine the interannual standard deviation of the MSF ability (Fig. 4a–d, shading). The RMC-EA has strong vari- in ERA-Interim (Fig. 2a, shading), HC exhibits prominent ability at the poleward limit of the NHC, and weaker vari- variability in the tropics (15°S–10°N) with a larger standard ability in the northern tropics and at the poleward limit of the deviation for NHC than for SHC. The interannual variability SHC (Fig. 4a, shaded areas). Compared with other RMCs of RMCs is not zonally uniform as shown by the interannual (Fig. 4a, b, d, shaded areas), RMC-CP has the strongest vari- standard deviation of V200–V850 in ERA-Interim (Fig. 2b, ability in the tropics of both hemispheres (Fig. 4c, shaded shading). RMC-CP has much stronger variability in the trop- areas). Meanwhile, RMC-EP also exhibits remarkable vari- ics of both the Northern and Southern hemispheres. The ability in the tropics of both hemispheres (Fig. 4b, shaded RMC shows strong interannual variability extending from areas). Unlike the symmetric distributions of RMC-EA and the northern subtropics of WA sector to that of EA. Besides, RMC-CP variability, RMC-WA has one center of variability the relatively weak variability of RMC also can be seen in in the Northern Hemisphere [10°N–20°N], and two in the the southern subtropics of geographic sectors extending Southern Hemisphere [35°S–20°S and 15°S–5°S] (Fig. 4d, from EP to EA. brown and yellow shading). The MME has the capability of reproducing interan- In MME simulation, the thermally direct cells calculated nual variations of HC and longitudinal diversity of RMC as MSF over EA and EP sectors have the typical features of (Fig. 2c, d, shading). For instance, the simulated prominent the climatological HC (Fig. 4e, f, contours). The meridional variability of HC in the tropics (15°S–10°N) (Fig. 2c, shad- stream function obtained from the CP (WA) sector is almost ing), and the remarkable variability of RMC at the poleward opposite to the typical distribution of HC climatology, asso- limit of NHC and SHC are all comparable to ERA-Interim ciated with a thermally direct (indirect) cell in the Southern (Fig. 2d, shading). Nevertheless, there is an evident inter- (Northern) Hemisphere of each sector (Fig. 4g, h, contours). model spread in the magnitude of HC variability, with a The different features of climatological RMCs over the four systematic weaker variability of HC in CMIP5 models than sectors are in good agreement with ERA-Interim. Besides, in ERA-Interim (Fig. 3b). the variability pattern of each RMC in MME simulation has a comparable distribution in ERA-Interim (Fig. 4e–h, 3.4 Inhomogeneity of climatological RMC and its shading), despite a much weaker magnitude of each RMC variability: results from regional MSF variability in MME simulation than in ERA-Interim. This is partly due to the reduction in MME variance caused by the The interannual standard deviation of V200–V850 is helpful averaging operation. to obtain a qualitative understanding of the inhomogeneity of RMC and its variability. It is necessary to further validate 3.5 Links in interannual variability of the HC those results by calculating the MSF over the meridional and RMCs sectors of GM domains. Figure 4 shows the climatology of MSF in each sector and its interannual standard deviation 3.5.1 Dominant modes of HC and RMCs in ERA-Interim and MME simulation. The climatological features of RMC-EA and RMC-EP have a large resemblance In the previous section the interannual variability of to the climatology of HC (Fig. 4a, b, contours). In contrast, HC and longitudinal diversity of RMCs were examined the spatial patterns of RMC-CP and RMC-WA are almost in ERA-Interim and MME. To further demonstrate the inverse, compared to the climatology of HC, associated with variability of HC and its links to RMC, an empirical a clockwise circulation in the Southern Hemisphere and a orthogonal function (EOF) analysis is applied to both the 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… c g d h Fig. 4 Regional characteristics of the HC during boreal winter in of RMCs over four geographic sectors (color shading, units: 10 kg −1 10 −1 ERA-Interim reanalysis data (left panels) and the MME simulation s ). Contours (units: 10 kg s ) indicate the corresponding climato- (right panels), as demonstrated by interannual standard deviations logical mean RMC in each sector zonal-mean MSF and to the regional MSF (Zhang and In both ERA-Interim (Fig.  5a) and MME (Fig.  5b), Wang 2015). The leading modes (EOF-1) of the HC and there is one thermally direct anomalous cell in the tropics of the four RMCs are shown in Figs. 5 and 6, respectively. and one indirect anomalous cell in the subtropics of the Time series of the principal component (i.e., PC-1) of HC Northern Hemisphere, which correspond to the ascend- and RMCs are also shown in Fig. 5c (ERA-Interim) and ing and descending motions of the NHC in the tropics Fig.  5d (MME). We choose the sign convention for the and subtropics. The thermally direct cell in the northern EOF so that the patterns are similar for ERA-Interim and tropics is of similar pattern as shown in the recent work MME (Figs. 5, 6). For the sake of convenience, we only of Guo and Li (2016) and of intermediate pattern between depict the spatial structures of HC and RMCs and their ASM and SM (Ma and Li 2008). For instance, the south- relationship during the positive phase. ern extent of this anomalous cell (approximately at 5°S) 1 3 Y. Sun et al. Fig. 5 (Left panels) The leading mode (EOF-1) of interannual varia- principal component (PC-1) derived from ERA-Interim reanalysis tion of the zonal-mean HC meridional stream function in DJF (shad- and the MME simulation, together with time series of EOF-1 in each ing, arbitrary units), superimposed over the mean HC (contours, RMC sector 10 −1 units: 10 kg s ). (Right panels) Time series of the corresponding in Fig. 5a is narrower than ASM (approximately at 10°S) the HC intensity and boundaries, resulting in a stronger and and wider than SM (near the equator). The differences narrower HC during boreal winter. between the present and previous works largely result The structure of the leading mode of RMC-EA variability from the different influences of mid-latitude eddies and is generally the reverse of its climatology (Fig. 6a). As a ENSO on the behaviors of HC. The spatial pattern asso- consequence, the thermally direct RMC is weakened over ciated with mid-latitude eddies is asymmetric about the the EA sector, which tends to reduce the global HC. Fur- equator and shows typical features of climatological HC thermore, a pronounced counterclockwise anomalous cell (approximately at 10°S, see Caballero 2007). In contrast, in the northern subtropics stretches the NHC edge toward the spatial pattern associated with ENSO is generally sym- the equator. The leading mode of RMC-EP also opposes its metric about the equator (Ma and Li 2008). Therefore, the climatology in both hemispheres. Therefore, the thermally combined constraints of mid-latitude eddies and ENSO on direct RMC is also weakened over the EP sector (Fig. 6b). the interannual variability of HC could be responsible for RMC-CP shows the most prominent mode, with a stronger the southern extent of this anomalous cell of the interme- anomalous cell in each hemisphere (Fig. 6c). This implies diate latitudinal position between mid-latitude eddies- and that RMC-CP contributes directly to the strength of the ENSO-related patterns. global HC. RMC-WA has a similar leading mode to RMC- In addition, there is one thermally direct anomalous cell EP, with slightly weaker cells in its spatial pattern (Fig. 6d). in the southern tropics and a much weaker indirect cell in By examining the temporal evolution of dominant HC the southern subtropics. Following the MSF-based defini- and RMCs modes (Fig. 5c), we found that the interannual tions of HC (Oort and Yienger 1996; Hu and Fu 2007), such variability of HC is generally synchronous with that of anomalous cells in the tropics and subtropics directly affect RMCs. This result indicates that the leading mode of global 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… a e b f c g d h Fig. 6 As the left panels in Fig. 5, but for the leading EOF patterns for the RMCs in the four geographic sectors 1 3 Y. Sun et al. HC is strongly linked to the modes of individual RMCs. In Hemisphere. The simulated strong convergence near the contrast to leading HC mode (Fig. 5a), the variability of the equator in the CP sector is comparable to that seen in ERA- RMC in the CP sector plays a dominant role in modulat- Interim. The general increase of RMC-WA in ERA-Interim ing the thermally direct anomalous cells of the two hem- that results from southerly winds in the Northern Hemi- ispheres, contributing to a stronger global HC, while the sphere and northerly winds in the Southern Hemisphere is variability modes in the other sectors play a dominant role also well reproduced in the MME. in the anomalous subtropical cell that contributes to a nar- rower HC. The MME can reproduce the leading mode of 3.5.3 Different roles of RMCs on HC strength and edges HC (Fig. 5b) and the modes of RMCs in the four sectors (Fig. 6e–h). For instance, the MME can capture the reversed As discussed above, the leading mode of the HC has a strong modes of RMC-EA and RMC-EP, which imply a reduction expression in both HC strength and extent, giving either a of global HC (Fig. 6e, f). The leading mode of RMC-CP in stronger and narrower HC or a weaker and broader HC. MME also exhibits a prominent variability similar to that Moreover, the leading mode of HC is certainly associated of ERA-Interim (Fig. 6g). The leading mode of RMC-WA, with RMC variability over the four sectors. Here we cal- which is close to its climatology in the tropics, is also well culate the correlation coefficients to further clarify the ties reproduced in MME (Fig. 6h). between each RMC and NHC intensity (NHCI) as well as the relationship between each RMC and the extent of the 3.5.2 Regional atmospheric circulation associated NHC and SHC (i.e., NHCE and SHCE). with leading modes of RMCs As shown in Table 2a, the leading mode of HC is cer- tainly a good indicator of NHCI variability, since positive In this section, we depict the regional atmospheric circula- correlation coefficients are obvious for the four reanalysis tion directly relevant to the RMC variability. Figure 7 shows datasets and MME simulation (statistically significant at the the regressions of the surface wind vector at 850 hPa onto 1% level). Meanwhile, NHCI also shows significant posi- the time series of each leading PC. A generally weakening tive correlation with RMC-EP, RMC-CP and RMC-WA and of RMC-EA in ERA-Interim can be identified that is linked generally insignificant correlation with RMC-EA (with the to an anomalous anticyclone over the Northwestern Pacific exception of NCEP2 and MME). Given that the temporal (NWPAC) and an anomalous anticyclone over the tropical evolution of HC mode is generally synchronous with that Southern Indian Ocean (TSIOAC). The meridional compo- of RMCs, the signs of the HC mode over the tropics with nents of wind anomalies on the left flank of the NWPAC that of the four RMC modes are compared to identify the and the TSIOAC reduce the northern and southern limbs of dominant sector responsible for NHCI (Figs. 5, 6). Result RMC-EA. The southerly winds in the Northern Hemisphere shows that only the positive correlation between NHCI and and northerly winds in the Southern Hemisphere result in RMC-CP can ensure the leading role of the HC mode in a weakening of the North American winter monsoon and a NHCI. That is, RMC-CP plays a dominant role in the inter- reduction of the southern limb of the RMC-EP. In contrast, annual variability of NHCI. Likewise, we can identify WA the southerly and northerly winds in the WA sector are dif- and EA as the dominant sectors whose RMCs determine ferent, which leads to an overall increase of RMC-WA, since the interannual variability of NHCE. EP, WA and EA are there is a strengthening of the northern limb of RMC-WA the geographic sectors whose RMCs are most related to the and an increase of the South American summer monsoon interannual variability of SHCE. circulation. The northerly winds in the Northern Hemisphere and southerly winds in the Southern Hemisphere meet near 3.6 Mechanisms controlling the HC strength the equator in the CP sector, generating strong convergence and extent and thus strengthens both the northern and southern limbs of the HC. Two mechanisms are proposed to understand the variability The MME captures the overall decrease of RMC-EA, of HC strength and extent. One is related to ENSO and the since there is a general weakening of the Asian–Afri- other to mid-latitude eddies. ENSO is the most prominent can–Australian monsoon circulation. However, there is a mode of tropical climate variability on interannual time strong model bias of cyclonic anomalies over the mid-lat- scales. It exerts a significant impact on the atmospheric itude ocean in the Southern Hemisphere. This model bias circulation at global and regional scales (Ropelewski and is largely associated with a general overestimate of diaba- Halpert 1987). Atmospheric eddies are characterized by tic heating in the southern mid-latitude of the EA sector large-scale Rossby wave perturbations at mid-latitudes that (Fig. 12d). The MME also reproduces the weakened RMC- may propagate into the tropics and affect the HC extent. EP with a weaker limb in the Southern Hemisphere and a In this section, we aim to assess the relative importance of weakened North American winter monsoon in the Northern 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Fig. 7 Regional circulation characteristics associated with the lead- regressed surface wind vectors at 850 hPa (above the 5% significance ing modes of RMCs (vectors) and their connections to ENSO (colors) level) and regressed SST pattern (above the 1% significance level) in ERA-Interim (left panels) and MME (right panels). Plotted are the 1 3 Y. Sun et al. Table 2 (a) Correlation NCEP1 NCEP2 JRA25 ERA-Interim MME coefficients between time series of NHCI and the leading (a) principal components of HC  HC 0.82 0.95 0.89 0.68 0.93 and RMCs (top-to-bottom: HC,  RMC-EA −0.05 0.47 0.16 −0.16 0.82 RMC-EA, RMC-EP, RMC-CP and RMC-WA). The principle  RMC-EP 0.59 0.81 0.72 0.47 0.85 component is calculated  RMC-CP 0.58 0.88 0.75 0.45 0.94 for four re-analysis datasets  RMC-WA 0.73 0.88 0.78 0.57 0.85 (NCEP1, NCEP2, JRA25 and (b) ERA-Interim) and for MME of climate models and (b) same  HC −0.30 (0.63) −0.43 (0.56) −0.24 (0.52) −0.21 (0.20) −0.80 (0.56) as in (a), but for correlation  RMC-EA −0.37 (0.22) −0.44 (0.11) −0.46 (0.04) −0.32 (0.00) −0.78 (0.56) coefficients between NHCE  RMC-EP −0.21 (0.71) −0.26 (0.66) −0.27 (0.62) −0.02 (0.43) −0.78 (0.65) (SHCE) and the principle  RHC-CP −0.34 (0.71) −0.39 (0.63) −0.39 (0.59) −0.18 (0.37) −0.84 (0.69) components  RMC-WA −0.44 (0.75) −0.42 (0.70) −0.43 (0.69) −0.25 (0.42) −0.84 (0.73) The bold fonts indicate correlation coefficients that are statistically significant at the (a) 1% level and (b) 5% level except for values given in bold italic these two proposed mechanisms on NHCI and extent of HC Pacific. The weakening of RMC over the Northwestern in each sector. Pacific would, to some extent, counteract the strengthen- ing of RMC in the Northern Hemisphere, which could be 3.6.1 On the relative role of ENSO and midโ€‘latitude eddies responsible for the regression pattern related to Ψ over the EA sector: resemblance to the climatology of NHC, but sta- To qualitatively assess the relative contribution of ENSO tistically insignificant. and mid-latitude eddies, we first calculate the regression of Figure 8e to h display the counterpart of ERA-Interim in N N MSF and MSF in each sector onto Ψ and Ψ , and then MME simulation. The pattern related to mid-latitude eddies e r compare the regression patterns. in the MME shows an overall increase of NHC like in ERA- The regression pattern of NHCI for mid-latitude eddies Interim (Fig.  8e), while the simulated regression pattern ( Ψ ) has a spatial structure comparable to its climatology related to ENSO has a symmetric structure about the equator (Fig. 8a), while that related to ENSO ( Ψ ) has a structure as seen in ERA-Interim (Fig. 8f). Moreover, the predomi- close to the leading mode of HC (Fig. 8b). The mid-latitude nance of the ENSO and mid-latitude eddy effect on NHCI eddies can explain a large fraction of the total variance of are also seen in the CP sector in the regression patterns and NHCI (74%). ENSO explains only 26%. Figure 8c, d display the composite of V200-V850 (Figs. 8g, h, 9c, d). Neverthe- the regression patterns of the CP sector’s RMC onto Ψ and less, ENSO explains a larger part of the total variance in Ψ . It is clear that both ENSO and mid-latitude eddies drive NHCI (57%) than do the mid-latitude eddies (43%) in the the interannual variability of NHCI in boreal winter. MME. The proportion is reversed in ERA-Interim. The role In addition, an eddy-relevant regression pattern over the of ENSO in driving NHCI is thus overestimated in MME. EA sector is similar to the climatology of NHC (not shown), The underestimate of eddy’s role in NHCI in MME can also but with low significance. This result indicates a possible be seen in the composite of V200–V850 (Fig. 9d) showing contribution from mid-latitude eddies via the EA sector into especially a remarkable underestimate of the strengthening the NHCI. We still do not fully understand this behavior, role of eddies in Northern Africa (0°–30°N, 0°–60°E). but a composite analysis of V200–V850 (Fig. 9) can give A linear regression is also used to assess the relative role us some hints. The composite of V200–V850 also shows of ENSO and mid-latitude eddies on the extent of the HC dominant influences of both ENSO and mid-latitude eddies in the two hemispheres. The time series of NHCE related through the CP sector on NHCI (Fig. 9a, b), which is con- to ENSO and to mid-latitude eddies are both positively sistent with previous regression patterns shown in Fig. 8c, correlated with NHCE in ERA-Interim, with correlation d. In contrast, the composite of V200–V850 over the EA coefficients of 0.43 and 0.90 (significant at the 1% level), sector has different features between ENSO and eddy cases. respectively. These statistically significant relations can also A significantly weakening of RMC during ENSO events is be seen in the northern subtropics for the regression patterns seen in both hemispheres [30°S–30°N, 60°–120°E], while a of HC against mid-latitude eddies and ENSO (Fig. 10a, b). significant strengthening of RMC during eddy cases is seen Furthermore, the EA sector is identified as the main sector in the Northern Hemisphere [0°–30°N, 0°–60°E], associated where mid-latitude eddies have significant influence on the with a significant decrease of RMC over the Northwestern interannual variability of wintertime NHCE (Fig. 10c), while 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Fig. 8 Regression of MSF anomalies in ERA-Interim onto N N a Ψ , b Ψ and regression of a r e regional MSF anomalies (CP sector) in ERA-Interim onto c N N Ψ and d Ψ . The right panels r e show results from MME. The dotted areas indicate where regression coefficients are significant at the 5% level (red: positive values and blue: nega- tive values) WA is identified as the main sector where ENSO exerts its In the MME, the ENSO-related regression pattern and significant impact on the interannual variability of NHCE that for mid-latitude eddies in the northern subtropics in boreal winter (Fig. 10d). Besides, it is important to note (Fig. 10e, f) are both comparable to those in ERA-Interim that mid-latitude eddies explain 82% of the total variance (Fig. 10a, b). However, in MME there is a general under- of NHCE and ENSO only 18%. Therefore, the meridional estimation of the role of mid-latitude eddies in NHCE, propagation of mid-latitude eddies through EA into the trop- with the fraction of total variance decreasing to 48%, ics plays a fundamental role in the interannual variability of smaller than that explained by ENSO (which increases to NHCE, while ENSO plays a secondary role. 52%). This reverse is largely associated with the dominant 1 3 Y. Sun et al. b d −1 Fig. 9 Composite analysis of V200–V850 (units: m s ) in ERA- events follow the criterion of Welhouse et  al. (2016). The criterion N N Interim between a El Niño and La Niña events, b between strong for strong-eddy cases is years with Ψ > 0.75 ๐œŽ and Ψ < −0.75๐œŽ for r r and weak eddy cases. Dotted areas denote significance at 5% level. weak-eddy cases.  is the standard deviation of Ψ c, d Same as a, b, but for the MME simulation. The selected ENSO influence of ENSO on NHCE in the MME simulation evi- maximum shifting northward to north of 35°S. Neverthe- dent through the EA sector (Fig. 10g). WA has been identi- less, the combined effects of regression patterns over EP fied as one sector where ENSO plays the significant role on and WA may contribute to the significant regression pat- NHCE (Fig. 10h), and this is consistent with ERA-Interim. tern in the southern subtropics (Fig. 11a). Similarly, variability in SHCE also results from varia- It is important to note that SHCE, determined by the zero- tions in both mid-latitude eddies and ENSO (Fig. 11a, b: MSF position at 500 hPa, is insignificantly correlated with ERA-Interim). Mid-latitude eddies and ENSO explain 61 RMC-EA (ERA-Interim: Fig.  11e) and its leading mode and 39% of the total variance, respectively. Both EP and (Table 2). However, 500 hPa is not necessarily the correct WA are identified as the main sectors where mid-latitude pressure level on which to define the HC edge by the zero- eddies in the Southern Hemisphere play the dominant role MSF contour (Hu et al. 2011). Moreover, SHCE is signifi- in the interannual variability of SHCE (Fig. 11c, d: ERA- cantly related to RMC-EA in most parts of the southern Interim). If we compare the eddy-related regression pat- subtropics. In this sense, EA remains a sector where ENSO tern shown in Fig. 11a with that of the EP and WA sectors, can exert an impact on SHCE (ERA-Interim: Fig. 11e). In the significant areas in the southern subtopics of the EP general, SHCE varies in phase with RMC-EP and RMC-WA sector shift southward to south of 35°S. The regression and EA. pattern of the WA sector is generally insignificant with 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Fig. 10 As Fig. 8, but for the regression patterns derived from the variation of NHCE, used to highlight the distinctive roles of ENSO and mid-latitude eddies in the variation of NHCE In the MME simulation, mid-latitude eddies and ENSO of the WA sector related to mid-latitude eddies (Fig. 11i) contribute equally to the SHCE variance (Fig. 11f, g). The does not match well its counterpart in ERA-Interim mid-latitude eddy-dominant sector EP (Fig. 11h) and the (Fig. 11d). ENSO-dominant sector EA (Fig.  11j) show very similar regression patterns in the southern subtropics as those in ERA-Interim (Fig. 11c, e). However, the regression pattern 1 3 Y. Sun et al. Fig. 11 As Fig. 10, but for regression patterns with SHCE, used to highlight the distinctive f roles of ENSO and mid-latitude eddies in the variation of SHCE 3.6.2 Mechanisms of ENSO and midโ€‘latitude eddies on HC eddies in each sector to the interannual variability of NHCI and HC edges. It remains necessary now to unravel their edges and its intensity physical mechanisms. We first discuss the physical processes of ENSO influ- In previous section, by means of linear regression. we assessed the relative contribution of ENSO and mid-latitude ence on NHCI and HC edges. A significant diabatic 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… −1 Fig. 12 As Fig.  9, but for composite analysis of the tropical diabatic matology (contours, intervals 10 m s , middle), HadISST anomalies heating (TDH) anomalies at 400  hPa (units: K/day, top), 250  hPa- (units: °C, bottom) during ENSO events − 1 zonal wind anomalies (shading, units: m s ) associated with its cli- 1 3 Y. Sun et al. heating associated with ENSO is confined in the CP sec- while there are no significant mid-latitude eddies penetrating tor away from the equator (Fig. 12a). HC is sensitive to into tropical North Africa in MME (Fig. 13b). Similarly, the latitudinal position of diabatic heating. The heating wave penetrations into tropical EA are significant in ERA- off the equator would strengthen the intensity of HC in Interim (Fig.  13c), but insignificant in MME (Fig.  13d). the Northern Hemisphere as shown by Lindzen and Hou These results from the wave activity flux calculation coin- (1988). Therefore, the thermal control of ENSO on the cide well with those shown in Fig. 10c, g. Significant equa- interannual variability of NHCI is largely associated with torward propagations of mid-latitude eddies in the Southern CP diabatic heating off the equator. Hemisphere hardly reach the EP sector (Fig.  13e), which ENSO can further affect the interannual variation of the may partly explain the southward shift of significant areas edges of HC in the Northern and Southern Hemispheres in Fig. 11c. Unlike EP, tropical WA (Fig. 13e) receives sig- (Fig. 12b, c). Based on the zonal-mean concept, the impacts nificant wave activity flux from the Southern mid-latitudes, of ENSO on HC edges can be exerted through shifting the which may contribute to the northward shift of the subtropi- latitudinal position of subtropical Jets (Ceppi and Hartmann cal maximum in Fig. 11d. 2013) and altering the SST gradient between the tropics A direct but weak wave activity flux from mid-latitude and midlatitudes (Adam et al. 2014). Here we highlight that eddies into EP is simulated in MME (Fig.  13f), which is the impact of ENSO on NHCE is largely associated with a consistent with what is shown from the regression pattern southeastward shift of the North America jet in the WA sec- in Fig. 11h. tor (Fig. 12b). This result is also coherent with the fact that WA was identified as an ENSO-dominant factor for NHCE. The ENSO-induced meridional gradient of SST in the 4 Summary and conclusions Southern Hemisphere can affect the interannual variability of SHCE. For instance, the increase of SST gradient between HC is generally considered as a thermally direct circulation the tropics (0–20°S) and higher latitudes (20–45°S), conse- within the framework of zonal average in the tropics, sub- quence of warmer SST anomalies in the tropical Southern tropics and mid-latitudes of the globe. However, if we divide Indian Ocean and cooler SST anomalies in the mid-latitudes the tropical belt (30°S to 30°N) into 4 sub-domains, fol- of the Southern Atlantic (Fig. 12c), results in a narrower lowing roughly the global monsoon system, we obtain four SHC. The significant increase of ENSO-induced SST gra- RMCs including both thermally direct and indirect cells. A dient in the EA sector is consistent with the fact that EA significant portion of this study was devoted to investigat- was identified as a responsible sector for ENSO’s control on ing roles of different geographic sectors in the interannual SHCE in both ERA-Interim and MME. variability of NHCI and edges of HC in both the Northern The thermal control of ENSO via the CP sector off- and Southern Hemispheres, We paid a particular attention equatorial diabatic heating on NHCI is reproduced in MME to the underlying physical mechanisms through statistical (Fig.  12d), but with a stronger magnitude, compared to analyses and dynamic diagnostics. We used ERA-Interim ERA-Interim (Fig. 12a, d). In addition, ENSO affects NHCE and SST-driving climate models throughout the work. Our through meridional shift of the North America jet, which key findings are summarized as follows: is obvious in ERA-Interim, and well captured in MME (Fig. 12e). 1. Climatology of HC and inhomogeneity of climatological Wave activity flux is presented to track the source areas RMCs The thermally direct HC consists of a clockwise of mid-latitude eddies and their propagation into the tropics. NHC and counterclockwise SHC (Fig. 1). The RMCs Caballero and Anderson (2009) showed that it is an efficient within the meridional sectors show a rich longitudinal diagnostic tool for relating mid-latitude eddies and HC. It diversity, including thermally direct RMCs over the EA is complementary to the above linear regression methodol- and EP sectors with typical features as in the climatol- ogy to identify privileged sectors where mid-latitude eddies ogy of HC (i.e., “Hadleywise”) and the thermally direct exert impacts on NHCI and edges of HC. Figure 13 displays southern limbs of RMC-CP and RMC-WA that oppose the wave propagation paths, we can identify the geographic the SHC climatology (i.e., “anti-Hadleywise”), and sectors where mid-latitude eddies propagate into the trop- the thermally indirect northern limbs of RMC-CP and ics. We can compare them with those sectors responsible for RMC-WA that oppose the NHC climatology (i.e., “anti- NHCI and edges of HC in both hemispheres. As we can see Hadleywise”) (Figs. 2, 3a, 4). in Fig. 13a, there are significant propagations of mid-latitude 2. Interannual variability of HC linked to the variability eddies through CP and North Africa into the northern trop- of RMCs The leading mode of HC variability is associ- ics, which is consistent with what shown in Figs.  8c and ated with a stronger and narrower HC (Fig.  5) or the 9b. Compared with ERA-Interim, the penetration of mid- inverse. The mode of variability of HC is in phase with latitude waves into tropical CP is overestimated in MME, the principal components of the main variability modes 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… 2 −2 Fig. 13 As Fig.  12, but for wave activity flux (WAF, units: m s ) in each plot indicate the horizontal streamfunction anomalies (units: 2 −1 anomalies associated with the eddy-related variability of NHCI and m s ). Dotted areas in a–d indicate WAF of significance at the 5% edges of HC in the Northern and Southern Hemispheres. Contours level, while WAF of significance at the 10% level is shown in e, f of the four RMCs (Fig. 6). The meridional components near the CP sector equator intensifies NHC and SHC of wind anomalies on the left flank of the NWPAC and (Fig. 7). the TSIOAC result in an overall decrease of RMC-EA. 3. Distinctive effects of the four RMCs on HC strength and A reduction of the northern limb of RMC-EP largely extent By analyzing the spatial features of four RMC is associated with a weakening of the North American modes and diagnosing their correlation with NHCI and winter monsoon. In contrast, a general increase of RMC- HC edges (Table 2), we found that CP and EA are identi- WA is associated with a strengthening of the northern fied as the dominant sectors where variabilities of RMCs limb of RMC-WA and an increase of the South Ameri- determine the interannual variability of NHCI (Figs. 8, can summer monsoon circulation. Strong convergence 9). RMC-EA and RMC-WA are the dominant contribu- tors to the interannual variability of NHCE (Fig. 10). 1 3 Y. Sun et al. The combined effects of RMC-EP and RMC-WA and References RMC-EA largely determine the interannual variability Adam O, Schneider T, Harnik N (2014) Role of changes in mean of SHCE (Fig. 11). temperatures versus temperature gradients in the recent widen- 4. The mechanism of ENSO and mid-latitudes eddies con- ing of the Hadley circulation. J Clim 27(19):7450–7461 trolling NHCI and HC edges Both mid-latitude eddies Bordoni S, Schneider T (2008) Monsoons as eddy-mediated regime and ENSO can influence the interannual variability of transitions of the tropical overturning circulation. Nat Geosci 1:515–519. https ://doi.org/10.1038/ngeo2 48 NHCI and extent in each hemisphere. The mid-latitude Caballero R (2007) Role of eddies in the interannual variabil- eddies explain a larger fraction of the total variance of ity of Hadley cell strength. Geophys Res Lett. https ://doi. HC strength and extent than does ENSO. org/10.1029/2007G L0309 71 ENSO can impose its thermal control on the interan- Caballero R, Anderson BT (2009) Impact of midlatitude stationary waves on regional Hadley cells and ENSO. Geophys Res Lett. nual variability of NHCI through CP diabatic heating https ://doi.org/10.1029/2009G L0396 68 off the equator (Fig.  12a). ENSO can affect the interan- Ceppi P, Hartmann DL (2013) On the speed of the Eddy-Driven jet nual variability of NHCE by shifting the North America and the width of the Hadley cell in the southern hemisphere. J jet southeastward in the WA sector (Fig. 12b). ENSO- Clim 26:3450–3465. https://doi.or g/10.1175/jcli-d-12-00414.1 Corvec S, Fletcher CG (2017) Changes to the tropical circulation induced SST gradient increases in the southern parts in the mid-Pliocene and their implications for future climate. of EA leads to a narrower SHC (Fig. 12c). This is the Clim Past 13:135–147. https://doi.or g/10.5194/cp-13-135-2017 consequence of warmer SST anomalies over the tropical Dee DP et al (2011) The ERA-Interim reanalysis: configuration and Indian Ocean and Atlantic SST cooler in the subtropics performance of the data assimilation system. Q J R Meteorol Soc 137:553–597. https ://doi.org/10.1002/qj.828 and mid-latitude. Dima IM,Wallace JM (2003) On the seasonality of the Hadley cell. The penetration of mid-latitude eddies through CP J Atmos Sci 60:1522–1527. https ://doi.or g/10.1175/1520- and North Africa into the northern tropics can impact 0469(2003)060<1522:OTSOT H>2.0.CO;2 the interannual variability of NHCI (Fig.  13a). Mid- Feng J, Li JP (2013) Contrasting Impacts of two types of ENSO on the Boreal Spring Hadley circulation. J Clim 26:4773–4789. latitude eddies in the Northern Hemisphere propagate https ://doi.org/10.1175/jcli-d-12-00298 .1 equatorward and penetrate into the northern subtropics Feng R, Li JP, Wang JC (2011) Regime change of the boreal summer of EA, and thus impose a signature on the interannual Hadley circulation and its connection with the tropical SST. J variability of NHCE (Fig. 13c). EP and WA are eddy- Clim 24:3867–3877 Feng J, Li JP, Xie F (2013) Long-term variation of the principal dominant sectors on the interannual variability of SHCE, mode of boreal spring Hadley Circulation linked to SST over with a stronger wave activity flux from the southern the Indo-Pacific Warm Pool. J Clim 26:532–544 mid-latitude eddies into WA than into EP (Fig. 13e). Feng J et al (2015) Simulation of the equatorially asymmetric mode 5. Generally speaking, MME from the state-of-the-art cli- of the Hadley circulation in CMIP5 models. Adv Atmos Sci 32(8):1129–1142. https ://doi.org/10.1007/s0037 6-015-4157-0 mate models performs well in simulating HC behaviors, Feng J et al (2016) Contrasting responses of the hadley circulation including the variability of HC, its climatology and the to equatorially asymmetric and symmetric Meridional Sea sur- changes in HC strength and edges, despite a systematic face temperature structures. J Clim 29:8949–8963. https ://doi. underestimate of the magnitude of variability. However, org/10.1175/JCLI-D-16-0171.1 Feng J et al (2017) The responses of the Hadley circulation to dif- there is an overestimate for ENSO contribution to the ferent meridional SST structures in the seasonal cycle. J Geo- interannual variability of HC extent. phys Res Atmos 122:7785–7799. https://doi.or g/10.1002/2017J D0269 53 Gastineau G et al (2011) Some atmospheric processes governing the Acknowledgements We acknowledge the World Climate Research Pro- large-scale tropical circulation in idealized aquaplanet simula- gram’s Working Group on Coupled Modelling, which is responsible for tions. J Atmos Sci 68:553–575. https ://doi.or g/10.1175/2010j CMIP, and we thank the climate modeling groups (listed in Table 1 of as343 9.1 this paper) for producing and making available their model output. This Guo Y-P, Li J-P (2016) Impact of ENSO events on the interannual vari- work was jointly supported by National Natural Science Foundation ability of Hadley circulation extents in boreal winter. Adv Clim of China (41661144009, 41505076), China Scholarship Council, and Change Res 7:46–53. https://doi.or g/10.1016/j.accre.2016.05.001 Chinese-French project. Hadley G (1735) Concerning the cause of the general trade-winds: by Geo Hadley, Esq F. R. S. Philos Trans R Soc Lond 39:58–62. https ://doi.org/10.1098/rstl.1735.0014 Open Access This article is distributed under the terms of the Crea- Held IM (2000) The general circulation of the atmosphere, paper pre- tive Commons Attribution 4.0 International License (http://creat iveco sented at 2000 Woods Hole Oceanographic Institute Geophysical mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- Fluid Dynamics Program, Woods Hole Oceanogr. Inst., Woods tion, and reproduction in any medium, provided you give appropriate Hole, Mass. http://gfd.whoi.edu/proce eding s/2000/PDFvo l2000 credit to the original author(s) and the source, provide a link to the .html Creative Commons license, and indicate if changes were made. Held IM, Hou AY (1980) Nonlinear axially symmetric circulations in a nearly inviscid atmosphere. J Atmos Sci 37:515–533 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Hu Y, Fu Q (2007) Observed poleward expansion of the Hadley cir- Riehl H et al (1950) The intensity of the net meridional circulation. culation since 1979. Atmos Chem Phys 7:5229–5236. https://doi. Q J R Meteor Soc 76:182–188. https ://doi.org/10.1002/qj.49707 org/10.5194/acp-7-5229-200763280 8 Hu Y, Zhou C, Liu J (2011) Observational evidence for poleward Ropelewski CF, Halpert MS (1987) Global and regional scale precipi- expansion of the Hadley circulation. Adv Atmos Sci 28(1):33–44. tation patterns associated with the El Niño/Southern oscillation. https ://doi.org/10.1007/s0037 6-010-0032-1 Mon Weather Rev 115: 1606–1626. https://doi.or g/10.1175/1520- Jayakumar A, Kumar V, Krishnamurti TN (2013) Lead time for 0493(1987)115<1606:GARSP P>2.0.CO;2 medium range prediction of the dry spell of monsoon using multi- Schneider T (2006) The general circulation of the atmosphere. Annu models. J Earth Syst Sci 122(4):991–1004 Rev Earth Planet Sci 34:655–688. https ://doi.org/10.1146/annur Kalnay E et al (1996) The NCEP/NCAR 40-Year reanalysis project. ev.earth .34.03140 5.12514 4 Bull Am Meteorol Soc 77:437–471. https:/ /doi.org/10.1175/1520- Stachnik JP, Schumacher C (2011) A comparison of the Hadley cir- 0477(1996)077<0437:TNYRP >2.0.CO;2 culation in modern reanalyses. J Geophys Res Atmos. https://doi. Kanamitsu M et al (2002) NCEP–DOE AMIP-II Reanalysis (R-2). org/10.1029/2011J D0166 77 Bull Am Meteor Soc 83:1631–1643. https ://doi.or g/10.1175/ Sun Y, Zhou T (2014) How does El Niño affect the interannual vari- bams-83-11-1631 ability of the Boreal summer Hadley circulation? J Clim 27:2622– Karnauskas KB, Ummenhofer CC (2014) On the dynamics of the Had- 2642. https ://doi.org/10.1175/jcli-d-13-00277 .1 ley circulation and subtropical drying. Clim Dyn 42:2259–2269. Sun Y et al (2013) Observational analysis and numerical simulation of https ://doi.org/10.1007/s0038 2-014-2129-1 the interannual variability of the boreal winter Hadley circulation Knutti R et  al (2009) Challenges in combining projections from over the recent 30 years. Sci China Earth Sci 56:647–661. https:// multiple climate models. J Clim 23:2739–2758. https ://doi.doi.org/10.1007/s1143 0-012-4497-x org/10.1175/2009J CLI33 61.1 Sun Y et al (2017) Simulation of long-term trends in Hadley circulation Lindzen RS, Hou AY (1988) Hadley circulations for zonally averaged during boreal winter using an ocean data assimilation scheme with heating centered off the equator. J Atmos Sci 45(17):2416–2427 the coupled general circulation model FGOALS-s2 (in Chinese). Liu J et al (2009) Centennial variations of the global monsoon precipi- Chin Sci Bull 62:1–9. https ://doi.org/10.1360/N9720 17-00409 tation in the last millennium: results from ECHO-G model. J Clim (in press) 22:2356–2371. https ://doi.org/10.1175/2008J CLI23 53.1 Tanaka HL et al (2004) Trend and interannual variability of Walker, Ma J, Li J (2008) The principal modes of variability of the boreal monsoon and Hadley circulations defined by velocity potential in winter Hadley cell. Geophys Res Lett 35:L01808. ht tp s :/ /d oi . the upper troposphere. https:/ /doi.org/10.3402/tellus a.v56i3.14 410 org/10.1029/2007G L0318 83 Toma VE, Webster PJ (2010) Oscillations of the intertropical conver- Nguyen H et al (2013) The Hadley circulation in reanalyses: climatol- gence zone and the genesis of easterly waves. Part I: diagnostics ogy, variability, and change. J Clim 26:3357–3376. https ://doi. and theory. Clim Dyn 34(4):587–604 org/10.1175/jcli-d-12-00224 .1 Walker CC, Schneider T (2006) Eddy influences on Hadley circula- Nguyen H et al (2017) Variability of the extent of the Hadley circula- tions: simulations with an idealized GCM. J Atmos Sci 63:3333– tion in the southern hemisphere: a regional perspective. Clim Dyn. 3350. https ://doi.org/10.1175/JAS38 21.1 https ://doi.org/10.1007/s0038 2-017-3592-2 Walker JM, Bordoni S, Schneider T (2015) Interannual variability in Nigam S, Chung C, DeWeaver E (2000) ENSO diabatic heating in the large-scale dynamics of the South Asian summer monsoon. J ECMWF and NCEP–NCAR reanalyses, and NCAR CCM3 simu- Clim 28:3731–3750 lation. J Clim 13:3152–3171 Welhouse LJ et al (2016) Composite analysis of the effects of ENSO Onogi K et al (2007) The JRA-25 reanalysis. J Meteorol Soc Jpn Ser II events on Antarctica. J Clim 29:1797–1808 85:369–432. https ://doi.org/10.2151/jmsj.85.369 Williamson DL et  al (2013) The Aqua-Planet experiment (APE) Oort AH,Yienger JJ (1996) Observed interannual variability in the response to changed meridional SST profile. J Meteor Soc Japan Hadley circulation and its connection to ENSO. J Clim 9:2751– 91(A):57–89 2767. https://doi.or g/10.1175/1520-0442(1996)009<2751:OIVIT Zhang G, Wang Z (2015) Interannual variability of tropical cyclone H>2.0.CO;2 activity and regional Hadley circulation over the Northeast- Plumb RA (1985) On the three-dimensional propagation of stationary ern Pacific. Geophys Res Lett 42:2473–2481. https ://doi. waves. J Atmos Sci 42:217–229org/10.1002/2015G L0633 18 Quan XW, Diaz HF, Hoerling MP (2004) Change in the tropical Hadley Zhou BT, Shi Y, Xu Y (2016) CMIP5 simulated change in the intensity cell since 1950. In: Diaz HF, Bradley RS (eds) Hadley circulation: of the Hadley and Walker circulations from the perspective of present, past and future, vol 21. Springer, Dordrecht, pp 85–120 velocity potential. Adv Atmos Sci 33:808–818 1 3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Climate Dynamics Springer Journals

Regional meridional cells governing the interannual variability of the Hadley circulation in boreal winter

Free
23 pages

Loading next page...
 
/lp/springer_journal/regional-meridional-cells-governing-the-interannual-variability-of-the-VqH0yokWdT
Publisher
Springer Journals
Copyright
Copyright ยฉ 2018 by The Author(s)
Subject
Earth Sciences; Geophysics/Geodesy; Climatology; Oceanography
ISSN
0930-7575
eISSN
1432-0894
D.O.I.
10.1007/s00382-018-4263-7
Publisher site
See Article on Publisher Site

Abstract

The Hadley circulation (HC) has conventionally been considered as thermally direct with uniform zonal distribution. How- ever, the meridional circulation in the tropics is far from uniform, including the thermally direct cells associated with the global monsoon heating and indirect cells in the absence of diabatic heating. This study aims at assessing the thermally direct and indirect cells in different regions, identifying the geographic sectors responsible for interannual variability of HC strength and boundaries, and unraveling the underlying mechanism for the interannual variability of HC edges and intensity. Results derived from ERA-Interim reanalysis and climate models show that the climatological HC in wintertime (December–January–February) obscures longitudinal diversity of regional meridional cells (RMC), including thermally direct RMCs over Eurasia and the Eastern Pacific, thermally direct southern limbs of RMCs over the Central Pacific and Western Atlantic with opposite circulation. For each of the regions, El Niño-South Oscillation and mid-latitude eddies are assessed in terms of their relative contributions to the interannual variability of HC intensity and extent. Their underneath physical mechanisms are thoroughly investigated. Keywords Hadley circulation · Global monsoon · Strength · Extent · Interannual variability · ENSO · Mid-latitude eddies · Jet · Reanalysis · CMIP5 model 1 Introduction and momentum. The intensity and extent of the HC have varied significantly over recent decades (Quan et al. 2004; The Hadley circulation (hereafter HC) is conventionally Hu and Fu 2007; Gastineau et al. 2011; Sun et al. 2013), depicted as the dominant large-scale atmospheric circula- with very pronounced interannual variability. tion in the Earth’s tropics and subtropics, as described by a An early attempt to understand the interannual vari- conceptual model with zonal average over the globe (Hadley ability of HC was reported in Oort and Yienger 1996 who 1735). It plays an important role in the transport and balance used radiosonde data and pinpointed a strong connection to of fundamental climate variables, such as moisture, energy, ENSO. Most recent studies have used reanalysis datasets to examine the seasonal cycle and interannual variability of the HC. For instance, a good agreement on the interan- * Yong Sun nual variability of the annual mean HC was reported among sunyong@mail.iap.ac.cn multi-reanalysis datasets (Nguyen et al. 2013). But HC has pronounced seasonality (Dima and Wallace 2003), annual State Key Laboratory of Numerical Modeling mean does not reveal its complete picture. A few other stud- for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy ies were reported to examine the interannual variability of of Sciences, Beijing 100029, China a specific season (Tanaka et al. 2004; Stachnik and Schu- Laboratoire des Sciences du Climat et de macher 2011). l’Environnement/IPSL, CEA-CNRS-UVSQ, Université Sun et al. (2013) showed that there is a mode of HC inter- Paris-Saclay, Gif-sur-Yvette, France annual variability that is driven by ENSO, and this mode Laboratoire de Météorologie Dynamique, CNRS, Sorbonne is thought to be linked with the interannual variability of Université, Paris, France HC strength associated with HC anomalies that are simi- Institute of Meteorology and Climate Research, Karlsruhe lar (opposite) to its mean state in the tropics (subtropics). Institute of Technology, Karlsruhe, Germany Vol.:(0123456789) 1 3 Y. Sun et al. Furthermore, Sun and Zhou (2014) investigated the behavior ENSO-induced meridional temperature gradient plays a key of HC in boreal summer. There are two prominent variability role in setting the edges of the HC. There is also an indirect modes for the developing and decaying phases of El Niño, effect of ENSO through ENSO-induced changes in mid-lat- respectively. The interannual variabilities of its northern itude eddies, as demonstrated by Guo and Li (2016) on the branch in both modes are largely under the control of the interannual variability of the northern HC edge. East Asian summer monsoon (EASM) variability, while that After an overview of numerous studies on global-scale of its southern branch results from the atmospheric response HC, it is necessary to emphasize that HC is decomposed to meridional shift of diabatic heating from the developing to of regional meridional cells (RMC) over a few geographic decaying phases of El Niño (Sun and Zhou 2014). A series sectors. They are collectively responsible for HC strength of recent studies have decomposed the HC variability into and extents, and there is a rich longitudinal diversity in the an equatorially asymmetric mode (ASM) and a symmetric tropics. We want to study the contributions from RMC of mode (SM) in terms of variability of annual mean (Feng different sectors to HC. We should point out that the lack et al. 2015), boreal winter (Ma and Li 2008) and boreal sum- of adequate metrics to gauge the regional characteristics of mer HC (Feng et al. 2011). The ASM mainly responds to HC increases the complexity and challenges to study the the SST warming over the Indo-Pacific warm pool and thus relationship between HC and RMC. explains the long-term variability of HC (Feng et al. 2013), The thermally direct summer monsoon cells in some while the SM linked to ENSO is responsible for the interan- geographic sectors (e.g. East Asia) are of inverse sense to nual variability of HC (Ma and Li 2008). the climatological HC in the summer hemisphere (North- By comparing the impacts of different meridional struc- ern Hemisphere) (Riehl et al. 1950). The interannual vari- ture of tropical SST on annual mean HC, ASM and SM ability of RMC over EASM sector plays a dominant role have been identified as the contrasting responses to equatori- in the interannual variability of the northern branch of the ally asymmetric and symmetric meridional SST structures HC in boreal summer (Sun and Zhou 2014). However, there (Feng et al. 2016). It is also true when the responses of the are no studies to assess the direct linkage between HC and HC to different meridional SST structures in the seasonal the global monsoon circulation. Additionally, it remains cycle are taken into account (Feng et al. 2017). A recent unknown whether the thermal direct circulation in associa- case study was reported in Feng and Li (2013) who investi- tion with the southern hemispheric monsoon regimes also gated different impacts of two ENSO types (canonical ENSO makes a reverse contribution to the southern Hadley cell in and ENSO Modoki) on the interannual variability of HC in austral summer. boreal spring. They revealed the important role of the merid- Most of these previous studies are based on the mass ional structure of sea surface temperature (SST) on the vari- stream function (MSF) that strictly satisfies the mass con- ation of HC. It has been reported that SST changes can also servation constraint. This is a powerful tool for diagnosing explain part of variability in HC strength and extent (Corvec the zonally-averaged HC. However, it neglects the regional and Fletcher 2017). The intensity of HC is proportional to diversity of the tropical atmospheric circulation. The verti- SST gradient between the tropics and subtropics (William- cal shear of the meridional wind between 200 and 850 hPa son et al. 2013), while the edges of HC are inversely pro- (V200–V850) is also a good indicator for HC and to some portional to meridional SST gradient (Adam et al. 2014). extent can remedy the shortcomings of MSF. It has been A rich literature can also be found in developing theories widely used as an alternative measure of boreal winter HC and mechanisms to understand the dynamics of HC. Under in earlier studies to depict the climatology of zonal HC (Oort the assumption of angular momentum conservation, Held and Yienger 1996; Quan et al. 2004), the interannual vari- and Hou (1980) provided a thermal model to explain suc- ation of HC related to ENSO (Oort and Yienger 1996; Sun cessfully the HC behavior. But this early theory neglected et al. 2013), and the strengthening trend of HC related to completely the role of mid-latitude eddies in the dynamics of the interdecadal intensification of ENSO amplitude (Quan HC. The effect of eddies has been highlighted in the subse- et al. 2004). A recent study has used V200-V850 to separate quent work (Held 2000), together with local Rossby number regions where the HC is reinforced (i.e., the features are R used to measure the relative importance of thermal forc- “Hadleywise”) from those where the HC is weakened (i.e., ing and eddy forcing on HC (Schneider 2006). they are “anti-Hadleywise”) and highlighted the dynamics ENSO and mid-latitude eddies both affect the HC strength of the HC and its inextricable linkage to subtropical drying (Oort and Yienger 1996; Walker and Schneider 2006), but a in a future scenario (Karnauskas and Ummenhofer 2014). In larger role is attributed to mid-latitude eddies for the interan- addition, a very recent study has attempted to investigate the nual variability of HC strength in boreal winter (Caballero regional variability of the extent of the HC in the Southern 2007; Caballero and Anderson 2009). ENSO can however Hemisphere by calculating the stream function via the verti- strongly affect the interannual variability of the extent and cal integral of the divergent meridional wind (Nguyen et al. boundaries of the HC in boreal winter, suggesting that the 2017). The strengthening trend of the HC in boreal winter 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… has also been validated from the perspective of the velocity one of the latest generation of atmospheric reanalyses pro- potential (Zhou et al. 2016). However, the stream function duced by the European Centre for Medium-Range Weather and velocity potential are obtained from the incompressible Forecasts (ECMWF), with a higher horizontal resolution of and irrotational components of air flow, respectively. Obvi- 1.25° × 1.25° (Dee et al. 2011), and (4) the Japanese 25-year ously, neither includes full details of the horizontal wind Reanalysis (JRA25) obtained from the Japan Meteorological and may not give a complete description of the regional Agency (JMA) (Onogi et al. 2007). diversity of the HC (Sun et al. 2017). Moreover, there are In general, the four reanalyses are consistent. Throughout increasing attempts to depict features of monsoon circulation the paper, we use the ERA-Interim reanalysis to demonstrate using regional MSF via a direct integral of zonally averaged the diversity of HC climatology and the implications for HC meridional wind over the specific monsoon sector (Bordoni intensity and edges, except where multi-reanalyses are used and Schneider 2008; Toma and Webster 2010; Jayakumar to confirm the differences in the HC strength and extent in et al. 2013; Walker et al. 2015). Out of consideration above, each of the four geographic sectors (Table 2). in the present paper we attempt to study the longitudinal We also examine model performance on simulating HC diversity of RMC within global monsoon domains by calcu- behavior using the Atmospheric Model Intercomparison lating the MSF over specific sectors from the original wind Project (AMIP) simulations involved in the Coupled Model fields and scaled by multiplying the coefficient (Δθ is Intercomparison Project phase 5 (CMIP5) (Table 1). The longitudinal range of each sector) (Zhang and Wang 2015). multi-model ensemble (MME) is used to evaluate model Given the general context and state-of-the-art for HC, performance, despite the fact that MME tends to reduce the firstly we would like in the present study to focalize on the variance and overestimate correlation, compared with each longitudinal diversity of RMCs within different monsoon single model (Knutti et al. 2009). domains. We are not only interested in long-term mean cli- matology, but also the interannual variability of HC (inten-2.2 Methodology sity and its southern and northern edges) and the associated RMC in each sector. Secondly, we would like to qualitatively 2.2.1 Mass stream function assess the relative contribution of ENSO and mid-latitude eddies in each sector on HC edges and intensity at interan- The meridional–vertical MSF is an important metric of nual time scale. Finally, our aim is also to assess the perfor- the HC (Oort and Yienger 1996). The MSF can be easily mance of models forced by realistic SST in reproducing the obtained via a mass-weighted vertical integral of the zonal interannual variability of HC and its connection to that of mean meridional velocity. In spherical coordinates, the MSF, RMC within each global monsoon sector. Ψ, at each pressure level, p, and latitude,  , can be expressed Due to the expected strong influence of monsoon on HC, as: transition seasons (spring and autumn) are not our ideal choice. But the boreal winter provides a good case. Moreo- 2๐œ‹ a cos ๐œƒ Ψ(๐œƒ , p)= V(๐œƒ , p) ⋅ dp, (1) ver, ENSO usually reaches its strongest amplitude during g boreal winter (December–January–February, DJF). Finally selecting boreal winter also offers us the possibility of com- where a is the Earth’s radius, g the gravitational accelera- parison with previous work (Caballero 2007; Ma and Li tion, and V the zonal mean meridional velocity. For mass 2008; Caballero and Anderson 2009). conservation, the zonal mean meridional velocity and verti- This paper is organized as follows. Descriptions of the cal motion ( ) must satisfy: data and methodology are presented in Sect. 2. Section 3 g g ๐œ• Ψ ๐œ• Ψ gives the main results and a summary and conclusions are V = and =− . (2) 2๐œ‹ a cos ๐œƒ ๐œ• p 2๐œ‹ a cos ๐œƒ ๐œ•๐œƒ provided in Sect. 4. We should keep in mind that V ( ) has the same (opposite) Ψ Ψ sign as . By examining the climatology of the MSF 2 Data and methodology (Fig. 1), it is clear that the MSF can depict the thermodynamic 2.1 Reanalysis datasets and model simulation circulation, characterized by ascending motion near the equa- ๐œ• Ψ tor ∼− < 0 and descending motion in the subtropics ๐œ•๐œƒ We first evaluate the consistency of four reanalysis datasets ๐œ• Ψ ∼− > 0 as well as poleward flow in depicting the climatology of HC and its interannual vari- ๐œ•๐œƒ ability. They are (1) the NCEP/NCAR Reanalysis (NCEP1) ๐œ• Ψ ๐œ• Ψ ฬ„ ฬ„ southerliesโˆถ V ∼ > 0 or northerliesโˆถ V ∼ < 0 at ๐œ• p ๐œ• p (Kalnay et al. 1996), (2) the NCEP–DOE AMIP-II Reanal- upper levels and equatorward flow ysis (NCEP2) (Kanamitsu et al. 2002), (3) ERA-Interim, 1 3 ฬ„๐œ” ฬ„๐œ” ฬ„๐œ” ฬ„๐œ” ฬ„๐œ” Y. Sun et al. Table 1 List of CMIP5 models used in this study Model Affiliation Horizontal resolution ACCESS1-0 Commonwealth Scientific and Industrial Research Organization/Bureau of Meteorology, Australia 144 × 192 CanAM4 Climate Modelling and Analysis, Canada 96 × 192 CSIRO-Mk3-6-0 Commonwealth Scientific and Industrial Research Organization/Queensland Climate Change Centre 96  × 192 of Excellence, Australia HadGEM2-A Met Office Hadley Centre, UK 144  × 192 IPSL-CM5A-LR Institut Pierre Simon Laplace, France 96  × 96 IPSL-CM5A-MR 96  × 192 IPSL-CM5B-LR 96  × 96 MPI-ESM-LR Max Planck Institute for Meteorology, Germany 96  × 192 MPI-ESM-MR 96  × 192 NorESM1-M Bjerknes Centre for Climate Research, Norwegian Meteorological Institute, Norway 96  × 144 EC-EARTH EC-EARTH consortium published at Irish Centre for High-End Computing, Netherlands/Ireland 160  × 320 GFDL-CM3 Geophysical Fluid Dynamics Laboratory, USA 90  × 144 bcc-csm1-1 Beijing Climate Center, China 64  × 128 bcc-csm1-1-m 160  × 320 CCSM4 National Center for Atmospheric Research, USA 192  × 288 CMCC-CM Centro Euro-Mediterraneo sui Cambiamenti Climatici, Italy 240  × 480 FGOALS-g2 Institute of Atmospheric Physics, Chinese Academy of Sciences, China 60  × 128 FGOALS-s2 108  × 128 GISS-E2-R NASA/GISS (Goddard Institute for Space Studies), USA 90  × 144 MIROC5 Atmosphere and Ocean Research Institute (The University of Tokyo), National Institute for Environ- 128  × 256 mental Studies, and Japan Agency for Marine-Earth Science and Technology, Japan MRI-AGCM3-2H Meteorological Research Institute, Japan 320  × 640 MRI-AGCM3-2S 960  × 1920 MRI-CGCM3 160  × 320 BNU-ESM Beijing Normal University, China 64  × 128 CNRM-CM5 Centre National de Recherches Météorologiques,Centre Européen de Recherche et de Formation 128  × 256 Avancée en Calcul Scientifique, France Inmcm4 Russian Academy of Sciences, Institute of Numerical Mathematics, Russia 120  × 180 N N ๐œ• Ψ ๐œ• Ψ where Ψ is the linear regression of Ψ on the El Niño ฬ„ ฬ„ northerliesโˆถ V ∼ < 0 or southerlies โˆถ V ∼ > 0 ๐œ• p ๐œ• p 3.4 index (ENSO variability) and Ψ is an uncorrelated near the surface. The strength of the HC is defined by the residual (non-ENSO variability referred to as mid-latitude maximum MSF between 30°S and 30°N (Oort and Yienger eddy activity). Here we use the same method to evaluate 1996). The edge of the HC in each hemisphere is defined as the relative importance of ENSO and mid-latitude eddies in the latitudinal position where the value of MSF at 500 hPa is the interannual variability of HC intensity in the Northern zero in the subtropics. Hemisphere and of the HC edge in each hemisphere over 30 years (1979–2008). 2.2.2 Decomposition of the variability in HC strength and extent 2.2.3 Wave activity flux and diabatic heating Caballero (2007) emphasized the dominant role of mid-lat- itude eddies in driving the interannual variability of HC in Wave activity flux is used to track the source of mid-latitude boreal winter, by assessing the relative contributions of ENSO eddies and their propagation from sources to sinks, which is and non-ENSO variability to the total variance of winter hemi- a powerful tool (e.g. see Plumb 1985) to diagnose the impact sphere HC intensity ( Ψ ) over the period 1959–2001. He par- of mid-latitude stationary waves on regional Hadley cells titioned the detrended Ψ (N denotes Northern Hemisphere) (Caballero and Anderson 2009). We calculate wave activity into two components: flux based on 6-hourly output of ERA-Interim reanalysis. This procedure is repeated for individual model simulations N N N Ψ =Ψ +Ψ , (3) e r based on daily data before taking the MME average. 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… cd Fig. 1 Climatological boreal winter (December–January–February, reanalysis (upper panels) and the multi-model ensemble mean (MME, 10 −1 DJF) mass stream function (MSF, units: 10 kg s , contour interval lower panels) of AMIP-type simulations. The green arrows indicate 10 −1 2 × 10 kg s ) and vertical shear of meridional velocity between 200 the air flow direction of the HC in each hemisphere −1 and 850  hPa (i.e., V200 minus V850, units: m s ) in ERA-Interim To understand thermal control of ENSO on HC intensity, Northern Hemisphere from anticlockwise in the Southern diabatic heating estimated as a residual in the thermody- Hemisphere (Fig. 1b). Both panels reveal an obvious asym- namic equation (Nigam et al. 2000) is calculated based on metry of HC between the summer and winter hemispheres 6-hourly frequency of ERA-Interim reanalysis, and daily (Fig. 1a, b). The Hadley cell in the Northern Hemisphere output for each model before MME average. (NHC) is much stronger and wider than that in the Southern Hemisphere (SHC). Similarly, we can use MSF and V200–V850 to evaluate 3 Results the performance of state-of-the-art climate models in simu- lating the HC in boreal winter. Under realistic SST forcing, 3.1 Climatology of HC and inhomogeneous the MME can reproduce reasonably well the hemispheric distribution of RMCs asymmetry of the strength and extent of HC (e.g., a much stronger and “fatter” NHC in the winter hemisphere than We first examine the climatology of the HC in boreal winter SHC in the summer hemisphere) (Fig. 1c, d). and how its structure varies with longitude. Figure  1a, b It is important to keep in mind that the MSF is deduced depict the mean state of the HC in the ERA-Interim rea- from the vertical integral of zonal mean meridional veloc- nalysis based on the MSF and the vertical shear of meridi- ity, and thus limits further insight into the horizontal dis- onal velocity (defined as V200 minus V850, V200–V850). tribution of the HC. Fortunately, V200–V850 performs The positive (negative) contours in Fig. 1a indicate clock- well in representing the climatology of zonal HC and offers wise (anticlockwise) circulation in the Northern (Southern) an opportunity to investigate longitudinal distributions Hemisphere. Likewise, the positive and negative values of RMC and resultant climatology of HC (Fig. 2). Obvi- of V200–V850 can distinguish clockwise motion in the ously, the RMC (Fig.  2a, contours) is far from zonally 1 3 Y. Sun et al. a b Fig. 2 Interannual standard deviation of the HC in the ERA-Interim ent regional meridional cells (RMCs) along the geographic sectors, reanalysis and MME simulation (color shading), illustrated by the including RMCs with appearance similar to HC (red solid rectangles) 10 −1 −1 MSF (units: 10 kg s ) and V200–V850 (units: m s , contour and opposite to that of HC (blue dashed rectangles). The extent of −1 interval: 0.25  m s ) in comparison with the corresponding climato- each rectangle in the following plots is same as that presented here logical distributions (contours). Rectangles in b indicate the differ - uniform as demonstrated by a comparison of the zonal mean distinct roles in the climatology of HC can be captured by V200–V850 with its horizontal distribution (Fig. 2b). The state-of-the-art climate models (contours, Fig. 2c, d). inhomogeneous RMCs across longitudes are of two types: one has almost the same features as the climatological 3.2 Decomposition of RMCs into thermally direct zonal-mean HC in each hemisphere, as seen in the RMC and indirect cells over Eurasia (30°S–30°N, 30°W–155°E; RMC-EA) and the Eastern Pacific (30°S–30°N, 130°W–70°W; RMC-EP); The RMCs in the four geographic sectors show different the other has a spatial pattern in the Northern and South- contributions to the climatology of the HC. Traditionally ern hemispheres almost opposite to the climatology of the HC has been thought of as a thermally direct zonal-mean NHC and SHC, as in the RMCs over the Central Pacific circulation. Here it may raise a question whether the RMCs (30°S–30°N, 155°E–130°W; RMC-CP) and the Western are thermally direct cells with uniform distribution. The Atlantic (30°S–30°N, 70°W–30°W; RMC-WA). Therefore, longitudinal diversity of RMCs consists of thermally direct the RMC-EA and RMC-EP both play a fundamental role monsoon cells and thermally indirect meridional cells. Fig- in maintaining the climatological zonal-mean HC in boreal ure 3a shows the spatial distributions of RMCs and diabatic winter, while RMC-CP and RMC-WA act to oppose the heating in global monsoon (GM) domains. RMCs within climatology. GM domains are associated with monsoonal heating and Examining the spatial patterns of MSF and V200–V850 thus they are thermally direct cells, including large resem- in the MME simulation reveals that the longitudinal diver- blance of RMC over the African–Asian–Australian monsoon sity of RMCs presented in ERA-Interim and their associated (RMC-EA) and the North American monsoon (RMC-EP) as 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Fig. 3 a A simple diagram drawn to reveal the relationship between the spatial distributions of RMCs over four geographic sectors and global monsoon (GM) domains (black dots) based on GPCP precipitation data. The vectors show the cli- matological wintertime atmos- pheric circulation at 850 hPa derived from ERA-Interim rea- nalysis data. The shaded areas indicate summertime diabatic −1 heating (units: K day ) in each hemisphere (boreal summer: May to September; austral sum- mer: November to March); b Taylor diagram to evaluate the performance of climate models participating in CMIP5 and the consistency of the HC in three widely used reanalyses with the HC in ERA-Interim (REF) the climatological HC, and RMC of signs opposite to clima- region (OM) where there is no land–sea thermal contrast tological SHC over the South American monsoon (the south- (southern limb of RMC-CP) (Liu et al. 2009). In contrast, ern limb of RMC-WA) and the pure oceanic monsoon-like the northern limbs of RMC-CP and RMC-WA in the absence 1 3 Y. Sun et al. of monsoon like heating are thermally indirect cells (coun- counterclockwise one in the Northern Hemisphere (Fig. 4c, terclockwise circulation). d, contours). These features are well consistent with the RMCs in ERA-Interim show a longitudinal diversity results derived from V200–V850 (Fig.  2a, b, contours). within the meridional sectors of GM domains. By examin- Besides, the thermally direct and indirect cells of RMCs ing the horizontal distributions of V200–V850 in the MME can be seen more clearly from regional MSF. The thermally simulation (contours, Fig. 2c, d), we can see that MME cap- direct cells associated with monsoonal heating are “Had- tures the decomposition of RMCs into thermally direct and leywise” over EA and EP sectors and “anti-Hadleywise” in indirect cells associated with the presence and absence of the southern parts of CP and WA sectors. The “anti-Hadley- monsoonal heating within the different meridional sectors wise” features in the northern parts of CP and WA sectors of the GM domains. are thermally indirect cells due to the absence of monsoon like heating. 3.3 Interannual variability of HC and longitudinal As for the interannual variation of RMCs, the standard diversity of RMCs variability deviations of MSF in the four sectors are of distinct spatial patterns in terms of magnitude and position of RMC vari- If we examine the interannual standard deviation of the MSF ability (Fig. 4a–d, shading). The RMC-EA has strong vari- in ERA-Interim (Fig. 2a, shading), HC exhibits prominent ability at the poleward limit of the NHC, and weaker vari- variability in the tropics (15°S–10°N) with a larger standard ability in the northern tropics and at the poleward limit of the deviation for NHC than for SHC. The interannual variability SHC (Fig. 4a, shaded areas). Compared with other RMCs of RMCs is not zonally uniform as shown by the interannual (Fig. 4a, b, d, shaded areas), RMC-CP has the strongest vari- standard deviation of V200–V850 in ERA-Interim (Fig. 2b, ability in the tropics of both hemispheres (Fig. 4c, shaded shading). RMC-CP has much stronger variability in the trop- areas). Meanwhile, RMC-EP also exhibits remarkable vari- ics of both the Northern and Southern hemispheres. The ability in the tropics of both hemispheres (Fig. 4b, shaded RMC shows strong interannual variability extending from areas). Unlike the symmetric distributions of RMC-EA and the northern subtropics of WA sector to that of EA. Besides, RMC-CP variability, RMC-WA has one center of variability the relatively weak variability of RMC also can be seen in in the Northern Hemisphere [10°N–20°N], and two in the the southern subtropics of geographic sectors extending Southern Hemisphere [35°S–20°S and 15°S–5°S] (Fig. 4d, from EP to EA. brown and yellow shading). The MME has the capability of reproducing interan- In MME simulation, the thermally direct cells calculated nual variations of HC and longitudinal diversity of RMC as MSF over EA and EP sectors have the typical features of (Fig. 2c, d, shading). For instance, the simulated prominent the climatological HC (Fig. 4e, f, contours). The meridional variability of HC in the tropics (15°S–10°N) (Fig. 2c, shad- stream function obtained from the CP (WA) sector is almost ing), and the remarkable variability of RMC at the poleward opposite to the typical distribution of HC climatology, asso- limit of NHC and SHC are all comparable to ERA-Interim ciated with a thermally direct (indirect) cell in the Southern (Fig. 2d, shading). Nevertheless, there is an evident inter- (Northern) Hemisphere of each sector (Fig. 4g, h, contours). model spread in the magnitude of HC variability, with a The different features of climatological RMCs over the four systematic weaker variability of HC in CMIP5 models than sectors are in good agreement with ERA-Interim. Besides, in ERA-Interim (Fig. 3b). the variability pattern of each RMC in MME simulation has a comparable distribution in ERA-Interim (Fig. 4e–h, 3.4 Inhomogeneity of climatological RMC and its shading), despite a much weaker magnitude of each RMC variability: results from regional MSF variability in MME simulation than in ERA-Interim. This is partly due to the reduction in MME variance caused by the The interannual standard deviation of V200–V850 is helpful averaging operation. to obtain a qualitative understanding of the inhomogeneity of RMC and its variability. It is necessary to further validate 3.5 Links in interannual variability of the HC those results by calculating the MSF over the meridional and RMCs sectors of GM domains. Figure 4 shows the climatology of MSF in each sector and its interannual standard deviation 3.5.1 Dominant modes of HC and RMCs in ERA-Interim and MME simulation. The climatological features of RMC-EA and RMC-EP have a large resemblance In the previous section the interannual variability of to the climatology of HC (Fig. 4a, b, contours). In contrast, HC and longitudinal diversity of RMCs were examined the spatial patterns of RMC-CP and RMC-WA are almost in ERA-Interim and MME. To further demonstrate the inverse, compared to the climatology of HC, associated with variability of HC and its links to RMC, an empirical a clockwise circulation in the Southern Hemisphere and a orthogonal function (EOF) analysis is applied to both the 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… c g d h Fig. 4 Regional characteristics of the HC during boreal winter in of RMCs over four geographic sectors (color shading, units: 10 kg −1 10 −1 ERA-Interim reanalysis data (left panels) and the MME simulation s ). Contours (units: 10 kg s ) indicate the corresponding climato- (right panels), as demonstrated by interannual standard deviations logical mean RMC in each sector zonal-mean MSF and to the regional MSF (Zhang and In both ERA-Interim (Fig.  5a) and MME (Fig.  5b), Wang 2015). The leading modes (EOF-1) of the HC and there is one thermally direct anomalous cell in the tropics of the four RMCs are shown in Figs. 5 and 6, respectively. and one indirect anomalous cell in the subtropics of the Time series of the principal component (i.e., PC-1) of HC Northern Hemisphere, which correspond to the ascend- and RMCs are also shown in Fig. 5c (ERA-Interim) and ing and descending motions of the NHC in the tropics Fig.  5d (MME). We choose the sign convention for the and subtropics. The thermally direct cell in the northern EOF so that the patterns are similar for ERA-Interim and tropics is of similar pattern as shown in the recent work MME (Figs. 5, 6). For the sake of convenience, we only of Guo and Li (2016) and of intermediate pattern between depict the spatial structures of HC and RMCs and their ASM and SM (Ma and Li 2008). For instance, the south- relationship during the positive phase. ern extent of this anomalous cell (approximately at 5°S) 1 3 Y. Sun et al. Fig. 5 (Left panels) The leading mode (EOF-1) of interannual varia- principal component (PC-1) derived from ERA-Interim reanalysis tion of the zonal-mean HC meridional stream function in DJF (shad- and the MME simulation, together with time series of EOF-1 in each ing, arbitrary units), superimposed over the mean HC (contours, RMC sector 10 −1 units: 10 kg s ). (Right panels) Time series of the corresponding in Fig. 5a is narrower than ASM (approximately at 10°S) the HC intensity and boundaries, resulting in a stronger and and wider than SM (near the equator). The differences narrower HC during boreal winter. between the present and previous works largely result The structure of the leading mode of RMC-EA variability from the different influences of mid-latitude eddies and is generally the reverse of its climatology (Fig. 6a). As a ENSO on the behaviors of HC. The spatial pattern asso- consequence, the thermally direct RMC is weakened over ciated with mid-latitude eddies is asymmetric about the the EA sector, which tends to reduce the global HC. Fur- equator and shows typical features of climatological HC thermore, a pronounced counterclockwise anomalous cell (approximately at 10°S, see Caballero 2007). In contrast, in the northern subtropics stretches the NHC edge toward the spatial pattern associated with ENSO is generally sym- the equator. The leading mode of RMC-EP also opposes its metric about the equator (Ma and Li 2008). Therefore, the climatology in both hemispheres. Therefore, the thermally combined constraints of mid-latitude eddies and ENSO on direct RMC is also weakened over the EP sector (Fig. 6b). the interannual variability of HC could be responsible for RMC-CP shows the most prominent mode, with a stronger the southern extent of this anomalous cell of the interme- anomalous cell in each hemisphere (Fig. 6c). This implies diate latitudinal position between mid-latitude eddies- and that RMC-CP contributes directly to the strength of the ENSO-related patterns. global HC. RMC-WA has a similar leading mode to RMC- In addition, there is one thermally direct anomalous cell EP, with slightly weaker cells in its spatial pattern (Fig. 6d). in the southern tropics and a much weaker indirect cell in By examining the temporal evolution of dominant HC the southern subtropics. Following the MSF-based defini- and RMCs modes (Fig. 5c), we found that the interannual tions of HC (Oort and Yienger 1996; Hu and Fu 2007), such variability of HC is generally synchronous with that of anomalous cells in the tropics and subtropics directly affect RMCs. This result indicates that the leading mode of global 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… a e b f c g d h Fig. 6 As the left panels in Fig. 5, but for the leading EOF patterns for the RMCs in the four geographic sectors 1 3 Y. Sun et al. HC is strongly linked to the modes of individual RMCs. In Hemisphere. The simulated strong convergence near the contrast to leading HC mode (Fig. 5a), the variability of the equator in the CP sector is comparable to that seen in ERA- RMC in the CP sector plays a dominant role in modulat- Interim. The general increase of RMC-WA in ERA-Interim ing the thermally direct anomalous cells of the two hem- that results from southerly winds in the Northern Hemi- ispheres, contributing to a stronger global HC, while the sphere and northerly winds in the Southern Hemisphere is variability modes in the other sectors play a dominant role also well reproduced in the MME. in the anomalous subtropical cell that contributes to a nar- rower HC. The MME can reproduce the leading mode of 3.5.3 Different roles of RMCs on HC strength and edges HC (Fig. 5b) and the modes of RMCs in the four sectors (Fig. 6e–h). For instance, the MME can capture the reversed As discussed above, the leading mode of the HC has a strong modes of RMC-EA and RMC-EP, which imply a reduction expression in both HC strength and extent, giving either a of global HC (Fig. 6e, f). The leading mode of RMC-CP in stronger and narrower HC or a weaker and broader HC. MME also exhibits a prominent variability similar to that Moreover, the leading mode of HC is certainly associated of ERA-Interim (Fig. 6g). The leading mode of RMC-WA, with RMC variability over the four sectors. Here we cal- which is close to its climatology in the tropics, is also well culate the correlation coefficients to further clarify the ties reproduced in MME (Fig. 6h). between each RMC and NHC intensity (NHCI) as well as the relationship between each RMC and the extent of the 3.5.2 Regional atmospheric circulation associated NHC and SHC (i.e., NHCE and SHCE). with leading modes of RMCs As shown in Table 2a, the leading mode of HC is cer- tainly a good indicator of NHCI variability, since positive In this section, we depict the regional atmospheric circula- correlation coefficients are obvious for the four reanalysis tion directly relevant to the RMC variability. Figure 7 shows datasets and MME simulation (statistically significant at the the regressions of the surface wind vector at 850 hPa onto 1% level). Meanwhile, NHCI also shows significant posi- the time series of each leading PC. A generally weakening tive correlation with RMC-EP, RMC-CP and RMC-WA and of RMC-EA in ERA-Interim can be identified that is linked generally insignificant correlation with RMC-EA (with the to an anomalous anticyclone over the Northwestern Pacific exception of NCEP2 and MME). Given that the temporal (NWPAC) and an anomalous anticyclone over the tropical evolution of HC mode is generally synchronous with that Southern Indian Ocean (TSIOAC). The meridional compo- of RMCs, the signs of the HC mode over the tropics with nents of wind anomalies on the left flank of the NWPAC that of the four RMC modes are compared to identify the and the TSIOAC reduce the northern and southern limbs of dominant sector responsible for NHCI (Figs. 5, 6). Result RMC-EA. The southerly winds in the Northern Hemisphere shows that only the positive correlation between NHCI and and northerly winds in the Southern Hemisphere result in RMC-CP can ensure the leading role of the HC mode in a weakening of the North American winter monsoon and a NHCI. That is, RMC-CP plays a dominant role in the inter- reduction of the southern limb of the RMC-EP. In contrast, annual variability of NHCI. Likewise, we can identify WA the southerly and northerly winds in the WA sector are dif- and EA as the dominant sectors whose RMCs determine ferent, which leads to an overall increase of RMC-WA, since the interannual variability of NHCE. EP, WA and EA are there is a strengthening of the northern limb of RMC-WA the geographic sectors whose RMCs are most related to the and an increase of the South American summer monsoon interannual variability of SHCE. circulation. The northerly winds in the Northern Hemisphere and southerly winds in the Southern Hemisphere meet near 3.6 Mechanisms controlling the HC strength the equator in the CP sector, generating strong convergence and extent and thus strengthens both the northern and southern limbs of the HC. Two mechanisms are proposed to understand the variability The MME captures the overall decrease of RMC-EA, of HC strength and extent. One is related to ENSO and the since there is a general weakening of the Asian–Afri- other to mid-latitude eddies. ENSO is the most prominent can–Australian monsoon circulation. However, there is a mode of tropical climate variability on interannual time strong model bias of cyclonic anomalies over the mid-lat- scales. It exerts a significant impact on the atmospheric itude ocean in the Southern Hemisphere. This model bias circulation at global and regional scales (Ropelewski and is largely associated with a general overestimate of diaba- Halpert 1987). Atmospheric eddies are characterized by tic heating in the southern mid-latitude of the EA sector large-scale Rossby wave perturbations at mid-latitudes that (Fig. 12d). The MME also reproduces the weakened RMC- may propagate into the tropics and affect the HC extent. EP with a weaker limb in the Southern Hemisphere and a In this section, we aim to assess the relative importance of weakened North American winter monsoon in the Northern 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Fig. 7 Regional circulation characteristics associated with the lead- regressed surface wind vectors at 850 hPa (above the 5% significance ing modes of RMCs (vectors) and their connections to ENSO (colors) level) and regressed SST pattern (above the 1% significance level) in ERA-Interim (left panels) and MME (right panels). Plotted are the 1 3 Y. Sun et al. Table 2 (a) Correlation NCEP1 NCEP2 JRA25 ERA-Interim MME coefficients between time series of NHCI and the leading (a) principal components of HC  HC 0.82 0.95 0.89 0.68 0.93 and RMCs (top-to-bottom: HC,  RMC-EA −0.05 0.47 0.16 −0.16 0.82 RMC-EA, RMC-EP, RMC-CP and RMC-WA). The principle  RMC-EP 0.59 0.81 0.72 0.47 0.85 component is calculated  RMC-CP 0.58 0.88 0.75 0.45 0.94 for four re-analysis datasets  RMC-WA 0.73 0.88 0.78 0.57 0.85 (NCEP1, NCEP2, JRA25 and (b) ERA-Interim) and for MME of climate models and (b) same  HC −0.30 (0.63) −0.43 (0.56) −0.24 (0.52) −0.21 (0.20) −0.80 (0.56) as in (a), but for correlation  RMC-EA −0.37 (0.22) −0.44 (0.11) −0.46 (0.04) −0.32 (0.00) −0.78 (0.56) coefficients between NHCE  RMC-EP −0.21 (0.71) −0.26 (0.66) −0.27 (0.62) −0.02 (0.43) −0.78 (0.65) (SHCE) and the principle  RHC-CP −0.34 (0.71) −0.39 (0.63) −0.39 (0.59) −0.18 (0.37) −0.84 (0.69) components  RMC-WA −0.44 (0.75) −0.42 (0.70) −0.43 (0.69) −0.25 (0.42) −0.84 (0.73) The bold fonts indicate correlation coefficients that are statistically significant at the (a) 1% level and (b) 5% level except for values given in bold italic these two proposed mechanisms on NHCI and extent of HC Pacific. The weakening of RMC over the Northwestern in each sector. Pacific would, to some extent, counteract the strengthen- ing of RMC in the Northern Hemisphere, which could be 3.6.1 On the relative role of ENSO and midโ€‘latitude eddies responsible for the regression pattern related to Ψ over the EA sector: resemblance to the climatology of NHC, but sta- To qualitatively assess the relative contribution of ENSO tistically insignificant. and mid-latitude eddies, we first calculate the regression of Figure 8e to h display the counterpart of ERA-Interim in N N MSF and MSF in each sector onto Ψ and Ψ , and then MME simulation. The pattern related to mid-latitude eddies e r compare the regression patterns. in the MME shows an overall increase of NHC like in ERA- The regression pattern of NHCI for mid-latitude eddies Interim (Fig.  8e), while the simulated regression pattern ( Ψ ) has a spatial structure comparable to its climatology related to ENSO has a symmetric structure about the equator (Fig. 8a), while that related to ENSO ( Ψ ) has a structure as seen in ERA-Interim (Fig. 8f). Moreover, the predomi- close to the leading mode of HC (Fig. 8b). The mid-latitude nance of the ENSO and mid-latitude eddy effect on NHCI eddies can explain a large fraction of the total variance of are also seen in the CP sector in the regression patterns and NHCI (74%). ENSO explains only 26%. Figure 8c, d display the composite of V200-V850 (Figs. 8g, h, 9c, d). Neverthe- the regression patterns of the CP sector’s RMC onto Ψ and less, ENSO explains a larger part of the total variance in Ψ . It is clear that both ENSO and mid-latitude eddies drive NHCI (57%) than do the mid-latitude eddies (43%) in the the interannual variability of NHCI in boreal winter. MME. The proportion is reversed in ERA-Interim. The role In addition, an eddy-relevant regression pattern over the of ENSO in driving NHCI is thus overestimated in MME. EA sector is similar to the climatology of NHC (not shown), The underestimate of eddy’s role in NHCI in MME can also but with low significance. This result indicates a possible be seen in the composite of V200–V850 (Fig. 9d) showing contribution from mid-latitude eddies via the EA sector into especially a remarkable underestimate of the strengthening the NHCI. We still do not fully understand this behavior, role of eddies in Northern Africa (0°–30°N, 0°–60°E). but a composite analysis of V200–V850 (Fig. 9) can give A linear regression is also used to assess the relative role us some hints. The composite of V200–V850 also shows of ENSO and mid-latitude eddies on the extent of the HC dominant influences of both ENSO and mid-latitude eddies in the two hemispheres. The time series of NHCE related through the CP sector on NHCI (Fig. 9a, b), which is con- to ENSO and to mid-latitude eddies are both positively sistent with previous regression patterns shown in Fig. 8c, correlated with NHCE in ERA-Interim, with correlation d. In contrast, the composite of V200–V850 over the EA coefficients of 0.43 and 0.90 (significant at the 1% level), sector has different features between ENSO and eddy cases. respectively. These statistically significant relations can also A significantly weakening of RMC during ENSO events is be seen in the northern subtropics for the regression patterns seen in both hemispheres [30°S–30°N, 60°–120°E], while a of HC against mid-latitude eddies and ENSO (Fig. 10a, b). significant strengthening of RMC during eddy cases is seen Furthermore, the EA sector is identified as the main sector in the Northern Hemisphere [0°–30°N, 0°–60°E], associated where mid-latitude eddies have significant influence on the with a significant decrease of RMC over the Northwestern interannual variability of wintertime NHCE (Fig. 10c), while 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Fig. 8 Regression of MSF anomalies in ERA-Interim onto N N a Ψ , b Ψ and regression of a r e regional MSF anomalies (CP sector) in ERA-Interim onto c N N Ψ and d Ψ . The right panels r e show results from MME. The dotted areas indicate where regression coefficients are significant at the 5% level (red: positive values and blue: nega- tive values) WA is identified as the main sector where ENSO exerts its In the MME, the ENSO-related regression pattern and significant impact on the interannual variability of NHCE that for mid-latitude eddies in the northern subtropics in boreal winter (Fig. 10d). Besides, it is important to note (Fig. 10e, f) are both comparable to those in ERA-Interim that mid-latitude eddies explain 82% of the total variance (Fig. 10a, b). However, in MME there is a general under- of NHCE and ENSO only 18%. Therefore, the meridional estimation of the role of mid-latitude eddies in NHCE, propagation of mid-latitude eddies through EA into the trop- with the fraction of total variance decreasing to 48%, ics plays a fundamental role in the interannual variability of smaller than that explained by ENSO (which increases to NHCE, while ENSO plays a secondary role. 52%). This reverse is largely associated with the dominant 1 3 Y. Sun et al. b d −1 Fig. 9 Composite analysis of V200–V850 (units: m s ) in ERA- events follow the criterion of Welhouse et  al. (2016). The criterion N N Interim between a El Niño and La Niña events, b between strong for strong-eddy cases is years with Ψ > 0.75 ๐œŽ and Ψ < −0.75๐œŽ for r r and weak eddy cases. Dotted areas denote significance at 5% level. weak-eddy cases.  is the standard deviation of Ψ c, d Same as a, b, but for the MME simulation. The selected ENSO influence of ENSO on NHCE in the MME simulation evi- maximum shifting northward to north of 35°S. Neverthe- dent through the EA sector (Fig. 10g). WA has been identi- less, the combined effects of regression patterns over EP fied as one sector where ENSO plays the significant role on and WA may contribute to the significant regression pat- NHCE (Fig. 10h), and this is consistent with ERA-Interim. tern in the southern subtropics (Fig. 11a). Similarly, variability in SHCE also results from varia- It is important to note that SHCE, determined by the zero- tions in both mid-latitude eddies and ENSO (Fig. 11a, b: MSF position at 500 hPa, is insignificantly correlated with ERA-Interim). Mid-latitude eddies and ENSO explain 61 RMC-EA (ERA-Interim: Fig.  11e) and its leading mode and 39% of the total variance, respectively. Both EP and (Table 2). However, 500 hPa is not necessarily the correct WA are identified as the main sectors where mid-latitude pressure level on which to define the HC edge by the zero- eddies in the Southern Hemisphere play the dominant role MSF contour (Hu et al. 2011). Moreover, SHCE is signifi- in the interannual variability of SHCE (Fig. 11c, d: ERA- cantly related to RMC-EA in most parts of the southern Interim). If we compare the eddy-related regression pat- subtropics. In this sense, EA remains a sector where ENSO tern shown in Fig. 11a with that of the EP and WA sectors, can exert an impact on SHCE (ERA-Interim: Fig. 11e). In the significant areas in the southern subtopics of the EP general, SHCE varies in phase with RMC-EP and RMC-WA sector shift southward to south of 35°S. The regression and EA. pattern of the WA sector is generally insignificant with 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Fig. 10 As Fig. 8, but for the regression patterns derived from the variation of NHCE, used to highlight the distinctive roles of ENSO and mid-latitude eddies in the variation of NHCE In the MME simulation, mid-latitude eddies and ENSO of the WA sector related to mid-latitude eddies (Fig. 11i) contribute equally to the SHCE variance (Fig. 11f, g). The does not match well its counterpart in ERA-Interim mid-latitude eddy-dominant sector EP (Fig. 11h) and the (Fig. 11d). ENSO-dominant sector EA (Fig.  11j) show very similar regression patterns in the southern subtropics as those in ERA-Interim (Fig. 11c, e). However, the regression pattern 1 3 Y. Sun et al. Fig. 11 As Fig. 10, but for regression patterns with SHCE, used to highlight the distinctive f roles of ENSO and mid-latitude eddies in the variation of SHCE 3.6.2 Mechanisms of ENSO and midโ€‘latitude eddies on HC eddies in each sector to the interannual variability of NHCI and HC edges. It remains necessary now to unravel their edges and its intensity physical mechanisms. We first discuss the physical processes of ENSO influ- In previous section, by means of linear regression. we assessed the relative contribution of ENSO and mid-latitude ence on NHCI and HC edges. A significant diabatic 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… −1 Fig. 12 As Fig.  9, but for composite analysis of the tropical diabatic matology (contours, intervals 10 m s , middle), HadISST anomalies heating (TDH) anomalies at 400  hPa (units: K/day, top), 250  hPa- (units: °C, bottom) during ENSO events − 1 zonal wind anomalies (shading, units: m s ) associated with its cli- 1 3 Y. Sun et al. heating associated with ENSO is confined in the CP sec- while there are no significant mid-latitude eddies penetrating tor away from the equator (Fig. 12a). HC is sensitive to into tropical North Africa in MME (Fig. 13b). Similarly, the latitudinal position of diabatic heating. The heating wave penetrations into tropical EA are significant in ERA- off the equator would strengthen the intensity of HC in Interim (Fig.  13c), but insignificant in MME (Fig.  13d). the Northern Hemisphere as shown by Lindzen and Hou These results from the wave activity flux calculation coin- (1988). Therefore, the thermal control of ENSO on the cide well with those shown in Fig. 10c, g. Significant equa- interannual variability of NHCI is largely associated with torward propagations of mid-latitude eddies in the Southern CP diabatic heating off the equator. Hemisphere hardly reach the EP sector (Fig.  13e), which ENSO can further affect the interannual variation of the may partly explain the southward shift of significant areas edges of HC in the Northern and Southern Hemispheres in Fig. 11c. Unlike EP, tropical WA (Fig. 13e) receives sig- (Fig. 12b, c). Based on the zonal-mean concept, the impacts nificant wave activity flux from the Southern mid-latitudes, of ENSO on HC edges can be exerted through shifting the which may contribute to the northward shift of the subtropi- latitudinal position of subtropical Jets (Ceppi and Hartmann cal maximum in Fig. 11d. 2013) and altering the SST gradient between the tropics A direct but weak wave activity flux from mid-latitude and midlatitudes (Adam et al. 2014). Here we highlight that eddies into EP is simulated in MME (Fig.  13f), which is the impact of ENSO on NHCE is largely associated with a consistent with what is shown from the regression pattern southeastward shift of the North America jet in the WA sec- in Fig. 11h. tor (Fig. 12b). This result is also coherent with the fact that WA was identified as an ENSO-dominant factor for NHCE. The ENSO-induced meridional gradient of SST in the 4 Summary and conclusions Southern Hemisphere can affect the interannual variability of SHCE. For instance, the increase of SST gradient between HC is generally considered as a thermally direct circulation the tropics (0–20°S) and higher latitudes (20–45°S), conse- within the framework of zonal average in the tropics, sub- quence of warmer SST anomalies in the tropical Southern tropics and mid-latitudes of the globe. However, if we divide Indian Ocean and cooler SST anomalies in the mid-latitudes the tropical belt (30°S to 30°N) into 4 sub-domains, fol- of the Southern Atlantic (Fig. 12c), results in a narrower lowing roughly the global monsoon system, we obtain four SHC. The significant increase of ENSO-induced SST gra- RMCs including both thermally direct and indirect cells. A dient in the EA sector is consistent with the fact that EA significant portion of this study was devoted to investigat- was identified as a responsible sector for ENSO’s control on ing roles of different geographic sectors in the interannual SHCE in both ERA-Interim and MME. variability of NHCI and edges of HC in both the Northern The thermal control of ENSO via the CP sector off- and Southern Hemispheres, We paid a particular attention equatorial diabatic heating on NHCI is reproduced in MME to the underlying physical mechanisms through statistical (Fig.  12d), but with a stronger magnitude, compared to analyses and dynamic diagnostics. We used ERA-Interim ERA-Interim (Fig. 12a, d). In addition, ENSO affects NHCE and SST-driving climate models throughout the work. Our through meridional shift of the North America jet, which key findings are summarized as follows: is obvious in ERA-Interim, and well captured in MME (Fig. 12e). 1. Climatology of HC and inhomogeneity of climatological Wave activity flux is presented to track the source areas RMCs The thermally direct HC consists of a clockwise of mid-latitude eddies and their propagation into the tropics. NHC and counterclockwise SHC (Fig. 1). The RMCs Caballero and Anderson (2009) showed that it is an efficient within the meridional sectors show a rich longitudinal diagnostic tool for relating mid-latitude eddies and HC. It diversity, including thermally direct RMCs over the EA is complementary to the above linear regression methodol- and EP sectors with typical features as in the climatol- ogy to identify privileged sectors where mid-latitude eddies ogy of HC (i.e., “Hadleywise”) and the thermally direct exert impacts on NHCI and edges of HC. Figure 13 displays southern limbs of RMC-CP and RMC-WA that oppose the wave propagation paths, we can identify the geographic the SHC climatology (i.e., “anti-Hadleywise”), and sectors where mid-latitude eddies propagate into the trop- the thermally indirect northern limbs of RMC-CP and ics. We can compare them with those sectors responsible for RMC-WA that oppose the NHC climatology (i.e., “anti- NHCI and edges of HC in both hemispheres. As we can see Hadleywise”) (Figs. 2, 3a, 4). in Fig. 13a, there are significant propagations of mid-latitude 2. Interannual variability of HC linked to the variability eddies through CP and North Africa into the northern trop- of RMCs The leading mode of HC variability is associ- ics, which is consistent with what shown in Figs.  8c and ated with a stronger and narrower HC (Fig.  5) or the 9b. Compared with ERA-Interim, the penetration of mid- inverse. The mode of variability of HC is in phase with latitude waves into tropical CP is overestimated in MME, the principal components of the main variability modes 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… 2 −2 Fig. 13 As Fig.  12, but for wave activity flux (WAF, units: m s ) in each plot indicate the horizontal streamfunction anomalies (units: 2 −1 anomalies associated with the eddy-related variability of NHCI and m s ). Dotted areas in a–d indicate WAF of significance at the 5% edges of HC in the Northern and Southern Hemispheres. Contours level, while WAF of significance at the 10% level is shown in e, f of the four RMCs (Fig. 6). The meridional components near the CP sector equator intensifies NHC and SHC of wind anomalies on the left flank of the NWPAC and (Fig. 7). the TSIOAC result in an overall decrease of RMC-EA. 3. Distinctive effects of the four RMCs on HC strength and A reduction of the northern limb of RMC-EP largely extent By analyzing the spatial features of four RMC is associated with a weakening of the North American modes and diagnosing their correlation with NHCI and winter monsoon. In contrast, a general increase of RMC- HC edges (Table 2), we found that CP and EA are identi- WA is associated with a strengthening of the northern fied as the dominant sectors where variabilities of RMCs limb of RMC-WA and an increase of the South Ameri- determine the interannual variability of NHCI (Figs. 8, can summer monsoon circulation. Strong convergence 9). RMC-EA and RMC-WA are the dominant contribu- tors to the interannual variability of NHCE (Fig. 10). 1 3 Y. Sun et al. The combined effects of RMC-EP and RMC-WA and References RMC-EA largely determine the interannual variability Adam O, Schneider T, Harnik N (2014) Role of changes in mean of SHCE (Fig. 11). temperatures versus temperature gradients in the recent widen- 4. The mechanism of ENSO and mid-latitudes eddies con- ing of the Hadley circulation. J Clim 27(19):7450–7461 trolling NHCI and HC edges Both mid-latitude eddies Bordoni S, Schneider T (2008) Monsoons as eddy-mediated regime and ENSO can influence the interannual variability of transitions of the tropical overturning circulation. Nat Geosci 1:515–519. https ://doi.org/10.1038/ngeo2 48 NHCI and extent in each hemisphere. The mid-latitude Caballero R (2007) Role of eddies in the interannual variabil- eddies explain a larger fraction of the total variance of ity of Hadley cell strength. Geophys Res Lett. https ://doi. HC strength and extent than does ENSO. org/10.1029/2007G L0309 71 ENSO can impose its thermal control on the interan- Caballero R, Anderson BT (2009) Impact of midlatitude stationary waves on regional Hadley cells and ENSO. Geophys Res Lett. nual variability of NHCI through CP diabatic heating https ://doi.org/10.1029/2009G L0396 68 off the equator (Fig.  12a). ENSO can affect the interan- Ceppi P, Hartmann DL (2013) On the speed of the Eddy-Driven jet nual variability of NHCE by shifting the North America and the width of the Hadley cell in the southern hemisphere. J jet southeastward in the WA sector (Fig. 12b). ENSO- Clim 26:3450–3465. https://doi.or g/10.1175/jcli-d-12-00414.1 Corvec S, Fletcher CG (2017) Changes to the tropical circulation induced SST gradient increases in the southern parts in the mid-Pliocene and their implications for future climate. of EA leads to a narrower SHC (Fig. 12c). This is the Clim Past 13:135–147. https://doi.or g/10.5194/cp-13-135-2017 consequence of warmer SST anomalies over the tropical Dee DP et al (2011) The ERA-Interim reanalysis: configuration and Indian Ocean and Atlantic SST cooler in the subtropics performance of the data assimilation system. Q J R Meteorol Soc 137:553–597. https ://doi.org/10.1002/qj.828 and mid-latitude. Dima IM,Wallace JM (2003) On the seasonality of the Hadley cell. The penetration of mid-latitude eddies through CP J Atmos Sci 60:1522–1527. https ://doi.or g/10.1175/1520- and North Africa into the northern tropics can impact 0469(2003)060<1522:OTSOT H>2.0.CO;2 the interannual variability of NHCI (Fig.  13a). Mid- Feng J, Li JP (2013) Contrasting Impacts of two types of ENSO on the Boreal Spring Hadley circulation. J Clim 26:4773–4789. latitude eddies in the Northern Hemisphere propagate https ://doi.org/10.1175/jcli-d-12-00298 .1 equatorward and penetrate into the northern subtropics Feng R, Li JP, Wang JC (2011) Regime change of the boreal summer of EA, and thus impose a signature on the interannual Hadley circulation and its connection with the tropical SST. J variability of NHCE (Fig. 13c). EP and WA are eddy- Clim 24:3867–3877 Feng J, Li JP, Xie F (2013) Long-term variation of the principal dominant sectors on the interannual variability of SHCE, mode of boreal spring Hadley Circulation linked to SST over with a stronger wave activity flux from the southern the Indo-Pacific Warm Pool. J Clim 26:532–544 mid-latitude eddies into WA than into EP (Fig. 13e). Feng J et al (2015) Simulation of the equatorially asymmetric mode 5. Generally speaking, MME from the state-of-the-art cli- of the Hadley circulation in CMIP5 models. Adv Atmos Sci 32(8):1129–1142. https ://doi.org/10.1007/s0037 6-015-4157-0 mate models performs well in simulating HC behaviors, Feng J et al (2016) Contrasting responses of the hadley circulation including the variability of HC, its climatology and the to equatorially asymmetric and symmetric Meridional Sea sur- changes in HC strength and edges, despite a systematic face temperature structures. J Clim 29:8949–8963. https ://doi. underestimate of the magnitude of variability. However, org/10.1175/JCLI-D-16-0171.1 Feng J et al (2017) The responses of the Hadley circulation to dif- there is an overestimate for ENSO contribution to the ferent meridional SST structures in the seasonal cycle. J Geo- interannual variability of HC extent. phys Res Atmos 122:7785–7799. https://doi.or g/10.1002/2017J D0269 53 Gastineau G et al (2011) Some atmospheric processes governing the Acknowledgements We acknowledge the World Climate Research Pro- large-scale tropical circulation in idealized aquaplanet simula- gram’s Working Group on Coupled Modelling, which is responsible for tions. J Atmos Sci 68:553–575. https ://doi.or g/10.1175/2010j CMIP, and we thank the climate modeling groups (listed in Table 1 of as343 9.1 this paper) for producing and making available their model output. This Guo Y-P, Li J-P (2016) Impact of ENSO events on the interannual vari- work was jointly supported by National Natural Science Foundation ability of Hadley circulation extents in boreal winter. Adv Clim of China (41661144009, 41505076), China Scholarship Council, and Change Res 7:46–53. https://doi.or g/10.1016/j.accre.2016.05.001 Chinese-French project. Hadley G (1735) Concerning the cause of the general trade-winds: by Geo Hadley, Esq F. R. S. Philos Trans R Soc Lond 39:58–62. https ://doi.org/10.1098/rstl.1735.0014 Open Access This article is distributed under the terms of the Crea- Held IM (2000) The general circulation of the atmosphere, paper pre- tive Commons Attribution 4.0 International License (http://creat iveco sented at 2000 Woods Hole Oceanographic Institute Geophysical mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- Fluid Dynamics Program, Woods Hole Oceanogr. Inst., Woods tion, and reproduction in any medium, provided you give appropriate Hole, Mass. http://gfd.whoi.edu/proce eding s/2000/PDFvo l2000 credit to the original author(s) and the source, provide a link to the .html Creative Commons license, and indicate if changes were made. Held IM, Hou AY (1980) Nonlinear axially symmetric circulations in a nearly inviscid atmosphere. J Atmos Sci 37:515–533 1 3 Regional meridional cells governing the interannual variability of the Hadley circulation… Hu Y, Fu Q (2007) Observed poleward expansion of the Hadley cir- Riehl H et al (1950) The intensity of the net meridional circulation. culation since 1979. Atmos Chem Phys 7:5229–5236. https://doi. Q J R Meteor Soc 76:182–188. https ://doi.org/10.1002/qj.49707 org/10.5194/acp-7-5229-200763280 8 Hu Y, Zhou C, Liu J (2011) Observational evidence for poleward Ropelewski CF, Halpert MS (1987) Global and regional scale precipi- expansion of the Hadley circulation. Adv Atmos Sci 28(1):33–44. tation patterns associated with the El Niño/Southern oscillation. https ://doi.org/10.1007/s0037 6-010-0032-1 Mon Weather Rev 115: 1606–1626. https://doi.or g/10.1175/1520- Jayakumar A, Kumar V, Krishnamurti TN (2013) Lead time for 0493(1987)115<1606:GARSP P>2.0.CO;2 medium range prediction of the dry spell of monsoon using multi- Schneider T (2006) The general circulation of the atmosphere. Annu models. J Earth Syst Sci 122(4):991–1004 Rev Earth Planet Sci 34:655–688. https ://doi.org/10.1146/annur Kalnay E et al (1996) The NCEP/NCAR 40-Year reanalysis project. ev.earth .34.03140 5.12514 4 Bull Am Meteorol Soc 77:437–471. https:/ /doi.org/10.1175/1520- Stachnik JP, Schumacher C (2011) A comparison of the Hadley cir- 0477(1996)077<0437:TNYRP >2.0.CO;2 culation in modern reanalyses. J Geophys Res Atmos. https://doi. Kanamitsu M et al (2002) NCEP–DOE AMIP-II Reanalysis (R-2). org/10.1029/2011J D0166 77 Bull Am Meteor Soc 83:1631–1643. https ://doi.or g/10.1175/ Sun Y, Zhou T (2014) How does El Niño affect the interannual vari- bams-83-11-1631 ability of the Boreal summer Hadley circulation? J Clim 27:2622– Karnauskas KB, Ummenhofer CC (2014) On the dynamics of the Had- 2642. https ://doi.org/10.1175/jcli-d-13-00277 .1 ley circulation and subtropical drying. Clim Dyn 42:2259–2269. Sun Y et al (2013) Observational analysis and numerical simulation of https ://doi.org/10.1007/s0038 2-014-2129-1 the interannual variability of the boreal winter Hadley circulation Knutti R et  al (2009) Challenges in combining projections from over the recent 30 years. Sci China Earth Sci 56:647–661. https:// multiple climate models. J Clim 23:2739–2758. https ://doi.doi.org/10.1007/s1143 0-012-4497-x org/10.1175/2009J CLI33 61.1 Sun Y et al (2017) Simulation of long-term trends in Hadley circulation Lindzen RS, Hou AY (1988) Hadley circulations for zonally averaged during boreal winter using an ocean data assimilation scheme with heating centered off the equator. J Atmos Sci 45(17):2416–2427 the coupled general circulation model FGOALS-s2 (in Chinese). Liu J et al (2009) Centennial variations of the global monsoon precipi- Chin Sci Bull 62:1–9. https ://doi.org/10.1360/N9720 17-00409 tation in the last millennium: results from ECHO-G model. J Clim (in press) 22:2356–2371. https ://doi.org/10.1175/2008J CLI23 53.1 Tanaka HL et al (2004) Trend and interannual variability of Walker, Ma J, Li J (2008) The principal modes of variability of the boreal monsoon and Hadley circulations defined by velocity potential in winter Hadley cell. Geophys Res Lett 35:L01808. ht tp s :/ /d oi . the upper troposphere. https:/ /doi.org/10.3402/tellus a.v56i3.14 410 org/10.1029/2007G L0318 83 Toma VE, Webster PJ (2010) Oscillations of the intertropical conver- Nguyen H et al (2013) The Hadley circulation in reanalyses: climatol- gence zone and the genesis of easterly waves. Part I: diagnostics ogy, variability, and change. J Clim 26:3357–3376. https ://doi. and theory. Clim Dyn 34(4):587–604 org/10.1175/jcli-d-12-00224 .1 Walker CC, Schneider T (2006) Eddy influences on Hadley circula- Nguyen H et al (2017) Variability of the extent of the Hadley circula- tions: simulations with an idealized GCM. J Atmos Sci 63:3333– tion in the southern hemisphere: a regional perspective. Clim Dyn. 3350. https ://doi.org/10.1175/JAS38 21.1 https ://doi.org/10.1007/s0038 2-017-3592-2 Walker JM, Bordoni S, Schneider T (2015) Interannual variability in Nigam S, Chung C, DeWeaver E (2000) ENSO diabatic heating in the large-scale dynamics of the South Asian summer monsoon. J ECMWF and NCEP–NCAR reanalyses, and NCAR CCM3 simu- Clim 28:3731–3750 lation. J Clim 13:3152–3171 Welhouse LJ et al (2016) Composite analysis of the effects of ENSO Onogi K et al (2007) The JRA-25 reanalysis. J Meteorol Soc Jpn Ser II events on Antarctica. J Clim 29:1797–1808 85:369–432. https ://doi.org/10.2151/jmsj.85.369 Williamson DL et  al (2013) The Aqua-Planet experiment (APE) Oort AH,Yienger JJ (1996) Observed interannual variability in the response to changed meridional SST profile. J Meteor Soc Japan Hadley circulation and its connection to ENSO. J Clim 9:2751– 91(A):57–89 2767. https://doi.or g/10.1175/1520-0442(1996)009<2751:OIVIT Zhang G, Wang Z (2015) Interannual variability of tropical cyclone H>2.0.CO;2 activity and regional Hadley circulation over the Northeast- Plumb RA (1985) On the three-dimensional propagation of stationary ern Pacific. Geophys Res Lett 42:2473–2481. https ://doi. waves. J Atmos Sci 42:217–229org/10.1002/2015G L0633 18 Quan XW, Diaz HF, Hoerling MP (2004) Change in the tropical Hadley Zhou BT, Shi Y, Xu Y (2016) CMIP5 simulated change in the intensity cell since 1950. In: Diaz HF, Bradley RS (eds) Hadley circulation: of the Hadley and Walker circulations from the perspective of present, past and future, vol 21. Springer, Dordrecht, pp 85–120 velocity potential. Adv Atmos Sci 33:808–818 1 3

Journal

Climate DynamicsSpringer Journals

Published: May 28, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

Itโ€™s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off