Parallel Comparison of the Northern Winter Stratospheric Circulation in Reanalysis and in CMIP5 Models

RAO Jian,REN Rongcai,and YANG Yang

1State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics,Chinese Academy of Sciences,Beijing 100029

2University of Chinese Academy of Sciences,Beijing 100049

Parallel Comparison of the Northern Winter Stratospheric Circulation in Reanalysis and in CMIP5 Models

RAO Jian1,2,REN Rongcai∗1,and YANG Yang1,2

1State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics,Chinese Academy of Sciences,Beijing 100029

2University of Chinese Academy of Sciences,Beijing 100049

A parallel comparison is made of the circulation climatology and the leading oscillation mode of the northern winter stratosphere among six reanalysis products and 24 CMIP5(Coupled Model Intercomparison Project Phase 5)models.The results reveal thattheNCEP/NCAR,NECP/DOE,ERA40,ERA-InterimandJRA25reanalyses arequiteconsistent indescribing the climatology and annual cycle of the stratospheric circulation.The 20CR reanalysis,however,exhibits a remarkable“cold pole”bias accompanied by a much stronger stratospheric polar jet,similar as in some CMIP5 models.Compared to the 1–2 month seasonal drift in most coupled general circulation models(GCMs),the seasonal cycle of the stratospheric zonal wind in most earth system models(ESMs)agrees very well with reanalysis.Similar to the climatology,the amplitude of Polar Vortex Oscillation(PVO)events also varies among CMIP5 models.The PVO amplitude in most GCMs is relatively weaker than in reanalysis,while that in most of the ESMs is more realistic.In relation to the“cold pole”bias and the weaker oscillation in some CMIP5 GCMs,the frequency of PVO events is signif i cantly underestimated by CMIP5 GCMs;while in most ESMs,it is comparable to that in reanalysis.The PVO events in reanalysis(except in 20CR)mainly occur from mid-winter to early spring(January–March);but in some of the CMIP5 models,a 1–2 month delay exists,especially in most of the CMIP5 GCMs.The long-term trend of the PVO time series does not correspond to long-term changes in the frequency of PVO events in most of the CMIP5 models.

CMIP5,northern winter stratospheric circulation,Polar Vortex Oscillation

1.Introduction

Themainvariabilityofthestratosphericcirculationlies in the northern winter season,which can be represented by the leading oscillation of the stratospheric polar vortex between a strong(cold)and weak(warm)state of the vortex.This leading and recurrent oscillation mode is known as the Polar Vortex Oscillation(PVO)(Ren and Cai,2006,2007;Cai and Ren,2006,2007),or the Northern Annular Mode in the stratosphere(NAM)(Thompsonand Wallace,1998).Associated with the occurrence of PVO or NAM events,downward propagation of circulation anomalies exists(Kodera et al., 1990;Baldwin and Dunkerton,1999),as well as simultaneous poleward propagation in the stratosphere synchronized with equatorward propagation of circulation anomalies in the troposphere between the tropics and the polar region (Cai and Ren,2006,2007;Ren and Cai,2007).Due to the intimate coupling of changes between the stratosphere and troposphere,as well as the much longer timescale exhibited by the stratospheric circulation,circulation changes in the stratosphere have been indicated to have signif i cant implications for weather and climate prediction in the troposphere(Thompson and Wallace,1998;Baldwin and Dunkerton,2001;Thompson et al.,2002;Cai,2003;Ren and Cai, 2007).

However,application of the stratospheric effects in weather and climate prediction is still quite limited due to insuff i cient knowledge on the dynamics of the stratosphere–tropospherecoupling.The limitedlengthof the observational data record currently available is an important factor affecting this limitation of understanding.Nevertheless,with improvements in the performance of numerical models in simulating the stratosphere in recent years,numerical models have begun to play an important role in further investigations of stratospheric dynamics and stratosphere–tropospherecoupling processes.An early comprehensive inter-model comparison of the performance of various stratosphere-resolving general circulation models(GCMs)revealed that most of the models generally showed a much stronger and colderstratospheric polar vortex and a less frequent occurrence of“stratospheric sudden warming”(SSW)events(or the negative phase of the PVO/NAM)(Charlton et al.,2007).Based on these results,the frequency of SSW events in observations is about six events per decade,while it is on average only about 1.0–2.6 events per decade in models.This“cold pole”problem has been shown to prevail in many other stratosphere-resolving GCMs(Pawson et al.,2000;Ren et al.,2009).Eyring et al.(2010)found that,for the ensemble means of several GCMs,the polar temperature biases become smaller(＜5 K)and the SSW frequency(f i ve events per decade)becomes much closer to that in observations.Recently,the Coupled Model Intercomparison Project Phase 5 (CMIP5)for the IntergovernmentalPanel on Climate Change (IPCC)Fifth Assessment Report released long-term integration results from more than 50 models of various countries, and in a series of standard scenarios.This provides us with valuable long-term datasets for further studies on the stratosphere.Before adoption of these model datasets in stratospheric studies,systematic and objective assessments on the general performance of all the models in reproducing the climatology and changes of stratospheric circulation are obviously needed.Based on the multi-model results for the historical scenario of CMIP5,the current study systematically evaluates the models’performances in simulating the northern winter stratospheric circulation,including the wintertime climatology,the seasonal evolution,and the polar vortex oscillation process.

The remainder of the paper is organized as follows.Section 2 gives a brief introductionto the CMIP5 models and the historical scenario experiments used in this study.Section 3 presents the reproducibility of the present climatology in the CMIP5 models.In section 4,the CMIP5 models are evaluated based on the annual cycle of the polar stratospheric circulation.Section 5 compares the CMIP5 models in simulating the PVO with reanalysis data.The f i nal section provides further discussion and conclusions.

2.Description of the CMIP5 models and reanalysis datasets

CMIP5 were carried out by 25 modeling groups representing more than 50 climate models with the aim of furthering understanding of past and future climate change in key areas of uncertainty(Taylor et al.,2012).The changing conditions prescribed in the experiments include atmospheric composition(including CO2)due to anthropogenic and volcanic forcing,solar forcing,concentrations of shortlived species,and natural and anthropogenicaerosols(Taylor et al.,2012).CMIP5 builds on the previous phase(CMIP3) of experiments in two main ways.First,more modeling centers and models are involved.Second,the models generally run at higher spatial resolution or with more comprehensive physical processes.The historical-run scenario denotes that the coupled atmosphere–ocean model simulations are forced by estimates of the changes in atmospheric composition from natural and anthropogenic sources,volcanoes,greenhouse gases(GHGs),and aerosols,as well as the changes in solar output and land cover during the industrial period(1850–2005).Only anthropogenicGHGs and aerosols are prescribed as common forcings in all models,and other forcings,such as changes of land use,may differ from model to model.For earth system models(ESMs),the carbon cycle and natural aerosols are also simulated by models,and therefore feature feedback processes.

The historical-run outputs we used are from 24 fully coupled CMIP5 models,including 12 atmosphere–ocean coupled GCMs and 12 ESMs.Some of the models are further coupled with chemistry modules.Detailed descriptions of all of the adopted 24 models are listed in Table 1,including the countries they are from,the types,the horizontal resolutions, andthenumbersofverticallevelsofthemodels,aswellasthe related references.Most of the models have provided multiple ensemble members,but we only used the f i rst member to capture the temporal variability of the stratospheric circulation effectively.

The reanalysis datasets used include the National Centers for Environmental Prediction–National Center for Atmospheric Research Reanalysis I(NCEP1)(Kalnay et al., 1996),the NCEP–U.S.Department of Energy Reanalysis II(NCEP2)(Kanamitsu et al.,2002),the European Centre for Medium-Range Weather Forecasts 40-Year Reanalysis(ERA40)(Uppala et al.,2005),the European Centre for Medium-Range Weather Forecasts Interim Reanalysis (ERA-I)(Dee et al.,2011),the Japanese 25-year Reanalysis(JRA25)(Onogi et al.,2007),and the Twentieth-Century Reanalysis Project,version 2(20CR)(Compo et al.,2011). Table 2 provides detailed information on these reanalysis datasets.The analysis methods used in this study include least-squares f i tting,linear regression,and empirical orthogonal function(EOF)analysis.

3.Winter climatology

Figure 1 shows the winter mean(December–February, DJF)zonal-mean air temperature(shading)and zonal-mean zonal wind(contours)in each reanalysis.It can be seen that, in the upper troposphere,all the six reanalysis datasets consistently show a subtropical westerly jet near 30°N.While in the stratosphere,the f i rst f i ve reanalysis datasets all show that the strength of the polar jet is～30 m s−1at 10 hPa,located at about 65°N,corresponding to the polar cold center of～200 K in the layer of 30–50 hPa.In contrast,20CR shows a much stronger polar jet(～55 m s−1)and a much colder(～185 K at 20–10 hPa)polar vortex.The poor performance of 20CR in describing the stratospheric circulation maybe relatedto the fact that the dataassimilation is onlyapplied on surface pressure,and the boundary forcing is from monthly-mean sea surface temperature and sea ice distributions(Compo et al.,2011).In this way,the 20CR dataset can provide relatively good estimations of the tropospheric variability,but with larges biases in the stratosphere,also notedin Compo et al.(2011).

Table 1.CMIP5 models evaluated and their attributes.Model types are atmosphere–ocean coupled(GCM),atmosphere–ocean–chemistry coupled(ChmGCM),earth system model(ESM),and earth system model chemistry coupled(ChmESM).

Table 2.Reanalysis datasets used in the evaluations.

Comparing the zonal-mean zonal wind patterns in each CMIP5 model with that in the f i rst f i ve reanalyses,it is seen that,generally,mostof the CMIP5 modelscanreproducereasonably well the strength and the vertical and meridional position of the uppertroposphericsubtropicaljet in the northern winter(Figs.1b–e).However,thereproducibilityofthemodels for the winter stratospheric polar jet and the polar temperature varies substantially from model to model,in terms of their magnitudes and location of action centers relative to the climatology in reanalysis.

To perform an objective evaluation of the performance of the CMIP5 models in reproducing the northern winter stratospheric circulation,several benchmarksare def i ned:the mean temperature in the tropical stratosphere(30°S–30°N, 70–10 hPa),in the midlatitude stratosphere(30°–60°N,100–10 hPa),in the upper polar stratosphere(60°–90°N,30–10 hPa),and in the lower polar stratosphere(60°–90°N,200–50 hPa);and the strength(averaged over 55°–75°N,70–10 hPa) and meridional location of the stratospheric polar jet(maxi-mum westerly).Below,we use box plots to present the distributions of these benchmarks in parallel for reanalysis and for the models.

3.1.Temperature in the tropical and midlatitude stratosphere

Figure 2 shows the benchmarks for the mass-weighted area mean temperature in the tropical stratosphere(Ttrp, 30°S–30°N,70–10 hPa,Figs.2a and b)and in the midlatitude stratosphere(Tmid,30°–60°N,100–10 hPa,Figs.2c and d),and for the reanalysis(Figs.2a and c)and for the CMIP5 models(Figs.2b and d).The mean values ofTtrpin NCEP1,NCEP2,ERA40,REA-I and JRA25 are quite consistent at around 211.8 K,but in 20CR it is higher(213.6 K,Fig.2a).As shown in Fig.2b,the CMIP5 models capture the tropical stratospheric temperature with varying degrees of success.The mean values ofTtrpin 20 of the 24 models (BCC-CSM1-1,BCC-CSM1-1-M,CCSM4,CNRM-CM5, FGOALS-s2,GFDL-CM3,HadCM3,INMCM4,IPSLCM5A-LR,IPSL-CM5A-MR,IPSL-CM5B-LR,MIROCESM,MIROC-ESM-CHEM,MIROC5,MRI-CGCM3,MPIESM-LR,MPI-ESM-MR,MPI-ESM-P,NorESM1-M and WACCM)exhibit a warm bias relative to the ensemble mean of the f i rst f i ve reanalysis datasets.The largest warm bias inTtrpis from GFDL-CM3(～216 K).The other four models exhibit a cold bias ofTtrp,with the largest cold bias from CSIRO-Mk3.6.0(～208 K).

Similarly,the mean values ofTmidin NCEP1,NCEP2, ERA40,ERA-I,and JRA25 are also very consistent(～215 K),but 20CR again shows a large positive departure (～217.5 K).The mean values ofTmidin 17 of the 24 models(CSIRO-Mk3.6.0,BCC-CSM1-1,BCC-CSM1-1-M,CCSM4,FGOALS-s2,GFDL-CM3,HadCM3,INMCM4,IPSL-CM5A-LR,IPSL-CM5A-MR,IPSL-CM5B-LR,MIROC-ESM,MIROC-ESM-CHEM,MPI-ESM-MR, MRI-CGCM3,NorESM1-M and WACCM)are also overestimated relative to the f i rst f i ve reanalyses.The largest warm bias in Tmidis again from GFDL-CM3(～218 K),and the othersevenmodelsexhibita coldbiasofTmid,withthelargest cold bias being from FGOALS-g2(～211.5 K).

3.2.Temperature in the upper and lower polar stratosphere

The benchmarksof the mass-weightedarea meantemperature in the upper(Tpl,up,60°–90°N,30–10 hPa,Figs.3a and b)and lower(Tpl,lw,60°–90°N,200–50 hPa,Figs.3c and d) polar stratosphere are shown in Fig.3,including the distributions of Tpl,upand Tpl,lwfor both the reanalysis(Figs.3a and c)and the CMIP5 models(Figs.3b and d).Note that the valuesof Tpl,upandTpl,lwcan def i nethe intensity of thestratospheric polar vortex.The mean values of Tpl,upand Tpl,lware～211.5 K and～214 K,respectively—fairly consistent among the f i rst f i ve reanalyses(NCEP1,NCEP2,ERA40, ERA-I,and JRA25).The mean values of Tpl,up(～198K)and Tpl,lw(～207.5 K)in 20CR are both much smaller.The“cold pole”problem,especially for the lower polar stratosphere, also prevails in most of the CMIP5 models.For example,the mean values of Tpl,upand Tpl,lwin some models(e.g.,BCCCSM1-1,BCC-CSM1-M,CCSM4,CNRM-CM5,FGOALS-g2)areas lowas～205Kand～210K,respectively,botheven exceedingthe lower interquartilerange ofthe ensemblemean of the f i rst f i ve reanalysis datasets.The much larger cold deviation of Tpl,upand Tpl,lwin FGOALS-g2 might be related to the systematic cold biases of the model,because all the mean values of Ttrp,Tmid,Tpl,up,and Tpl,lwexhibit considerable cold biases.In contrast,although the“cold pole”problem alsoexists in most of the ESMs,the cold biases of Tpl,upand Tpl,lwin some of the ESMs and ChmESMs(MIROC-ESM, MIROC-ESM-CHEM,MPI-ESM-LR,MPI-ESM-MR,MPIESM-P and WACCM)are relatively much smaller than those in most of the GCMs.

3.3.Strength and the central latitudinal location of the polar jet

The strength of the polar night jet is def i ned as the mass-weighted area mean zonal wind over(55°–75°N,70–10 hPa).The mean value of Upnis～20 m s−1in NCEP1, NCEP2,ERA40,ERA-I,and JRA25(Fig.4a).Consistent with the positive(negative)deviation of temperature in the midlatitude(polar)stratosphere in 20CR,the mean value of Upnin 20CR(～37 m s−1)is nearly double that in the other reanalysis datasets.Similarly,the overestimated (much stronger)polar jet also prevails in most of the CMIP5 models(e.g.,BCC-CSM1-1,BCC-CSM1-1-M,CCSM4, FGOALS-s2,GFDL-CM3,MRI-CGCM3);while across the ESMs(IPSL-CM5A-LR,IPSL-CM5A-MR,IPSL-CM5BLR,MIROC-ESM,MIROC-ESM-CHEM,MPI-ESM-LR, MPI-ESM-LR,MPI-ESM-P,NorESM1-M and WACCM), the meanUpnis relativelyclose to that in reanalysis(Fig.4b).

The mean central latitude of the polar jet in NCEP1 is around 65°N,consistent with all the other reanalysis datasets,including 20CR,despite 20CR showing a much stronger polar jet(Fig.4c).The polar jet in most of the models(BCC-CSM1-1,BCC-CSM1-1-m,CCSM4,FGOALS-s2,GFDL-CM3,INMCM4,MIROC-ESM,MIROC-ESMCHEM,MIROC5,MPI-ESM-LR,MPI-ESM-MR,MPIESM-P,MRI-CGCM3,NorESM1-M and WACCM)is located at a similar central latitudinal position,with a mean value falling in the interquartile range of the reanalysis ensemble(20CR excluded);while in some other models (CNRM-CM5,CSIRO-Mk-3.6.0,FGOALS-g2,GISS-E2-H and GISS-E2-R,HadCM3 and IPSL-CM5A-MR),the polar jet tends to lie further equatorward(Fig.4d).

4.Annual cycle

Figure 5 shows the annual cycle of the stratospheric zonal-mean zonal wind at 10 hPa.The extratropical zonal wind at 10 hPa clearly exhibits an annual cycle from a summer easterly to a winter westerly,which is quite consistent among NCEP1,NCEP2,ERA40,ERA-I,and JRA25,particularly the maximum tropical easterly in late January,the maximum polar westerly in early December,and the transitions between westerlies and easterlies in the extratropics. Unlike in these reanalyses,the polar jet in 20CR peaks until late January,exhibiting a seasonal drift of 1–2 months.In other words,not only the strength of the polar jet is overestimated,but there also exists a temporal delay of the winter westerly center in 20CR.Meanwhile,an elusive zonal-mean westerly exists over the equatorthroughoutthe year in 20CR.

The CMIP5 models capture the seasonal variation of stratospheric zonal-mean zonal wind with varying degrees of success.The simulated annual cycle in most of the CMIP5 models is quite similar to that in the f i rst f i ve reanalyses,including the transition between the wintertime westerly and the summertime easterly in the extratropics in both the southern and northern hemispheres.However,the strength of the northern polar jet,the easterly in the subtropics,and the seasonal timing of the maximum circumpolar westerly vary from model to model.For example,a common problem seems to exist in some of the GCMs(CCSM4,CNRM-CM5, CSIRO-Mk3.6.0,FGOALS-g2,FGOALS-s2,GFDL-CM3, MIROC5,and MRI-CGCM3)and some of the ESMs(MPIESM-LR,MPI-ESM-MR,MPI-ESM-P and NorESM1-M)in that the simulated polar night jet also peaks 1–2 months later relative to that in the reanalysis data.

Figure 6 further shows a quantitative measure of the performance of the CMIP5 models in reproducing the annual cycle of the extratropical zonal wind and the polar temperature relative to the f i rst f i ve reanalysis datasets.Figure 6a is a Taylor diagram for the simulated mass-weighted zonal wind in the circumpolar region(55°–75°N,70–10 hPa),and Fig. 6b is the same but for the polar temperature(75°–90°N,100–20 hPa).The correlation coeff i cient of the mass-weighted zonal-wind/temperaturebetween each CMIP5 model and the reanalysis ensemble(except 20CR)is denoted by the cosine of the azimuth angle,and the ratio of the correspondingstandard deviation in every CMIP5 model to that in the reanalysis ensemble is represented by the radial distance.The radial distance of each model indicates that the seasonal variation of the circumpolar zonal wind and the polar temperature in some models(e.g.,BCC-CSM1-1,BCC-CSM1-1-M,CCSM4,FGOALS-s2,GFDL-CM3,IPSL-CM5B-LR) are obviously stronger than that in the reanalysis ensem-ble(REF in Fig.6);while in some other GCMs(e.g., CSIROC-Mk3.6.0,GISS-E2-H,GISS-E2-R,HadCM3 and INMCM4),they are relatively much weaker.In contrast,the annual cycles of polar stratospheric circulation in most of the ESMs(e.g.,IPSL-CM5A-LR,IPSL-CM5A-MR,MIROCESM,MIROC-ESM-CHEM,MPI-ESM-LR,MPI-ESM-MR, MPI-ESM-P,NorESM1-M and WACCM)are reproduced much more realistically.The normalized standard deviations in theseESMs arecloser tothat ofthe 20CR-excludedreanalysis ensemble(REF in Fig.6).In general,the correlation coeff i cientsofthezonalwindbetweenallmodelsandthe20CR-excluded reanalysis ensemble can reach 0.9(Fig.6a),which is also true for the polar temperature(Fig.6b),indicating a well-reproduced transition of the stratospheric circulation in the polar region between summertime and wintertime.

5.Polar Vortex Oscillation events

The leading oscillation process,the NAM or PVO(Cai andRen,2007;RenandCai,2006),is alwaysrelatedtoanoscillation between a stronger and a weaker stratospheric polar vortex accompanied with radical changes of the circumpolar jet between a stronger and a weaker westerly(or even easterly)state.Here,we perform EOF analysis on the monthly zonal-mean zonal wind anomalies north of 20°N for each of the reanalyses and each of the model historical runs.Following Ren and Yang(2012)and Liu et al.(2012),we name the leading mode as the PVO mode.Rather than def i ning PVO events based on the standard deviations(STDs)of the PVO time series as in their studies,we identify PVO events based on a criterion that represents the average oscillation changes of the stratospheric zonal wind for unit STD of the PVO intensities in the reanalysis datasets,which is obtained by regressing the spatial pattern of the PVO mode against the standardized PVO time series.As a result,the threshold for PVO events in reanalysis is unit STD,which on average corresponds to a central intensity of 8–9 m s−1for the leading zonal-mean zonal wind oscillation(i.e.,PVO time series multiplied by the leading PVO mode).And the PVO events in models are identif i ed only when the central intensity of the leadingzonal wind oscillation in the extratropicsachieves ±10 m s−1,±15 m s−1,or±20 m s−1.In this way,the def initions of PVO events are uniform in terms of their intensity among the reanalyses and the CMIP5 models.

5.1.Spatial pattern and intensity of the leadingoscillation mode

TheregressedspatialpatternoftheDJF zonal-meanzonal wind anomalies against the PVO time series is shown in Fig. 7 for the reanalyses and the CMIP5 models.It can be seen that the dipole pattern,or the out-of-phase relationship of the zonal wind anomalies between the subtropics and the circumpolar region,are largely consistent among the reanalysis datasets.This is also true for the oscillation amplitudes, as indicated by the comparable action centers in the panels of Fig.7(a).The oscillation center lies north of 67.5°N, close to the climatological location of the DJF polar jet,and the oscillation amplitudes are all～8 m s−1in the reanalysis datasets,including 20CR.Specif i cally,the maximum zonalmean zonal wind anomalies for unit PVO STD are 7.8,8.0, 8.7,8.2,8.5,and 9.0 m s−1in NCEP1,NCEP2,ERA40, ERA-I,JRA25,and 20CR,respectively.It can be seen from Figs.7b–e that the oscillation amplitudes are reproduced with varying degrees of success in the CMIP5 models.The PVO intensity in some of the GCMs is much weaker,with the central value of the zonal-wind oscillation for unit STD of the leading time series being only 1.6 m s−1in CSIROMk3.6.0,2.5 m s−1in MIROC5,6.9 m s−1in CCSM4,6.4 m s−1in CNRM-CM5,6.8 m s−1in FGOALS-g2,7.5 m s−1in GFDL-CM3,5.5 m s−1in GISS-E2-H,5.6 m s−1in GISS-E2-R,4.7 m s−1in HadCM3,and 5.7 m s−1in INMCM4.Relatively,the oscillation amplitudes are more realistic in most of the ESMs(e.g.,BCC-CSM1-1,BCC-CSM1-1-M,ISPL-CM5A-LR,ISPL-CM5A-MR,IPSL-CM5B-LR, MIROC-ESM,MIROC-ESM-CHEM,MPI-ESM-LR,MPIESM-MR,MPI-ESM-P,NorESM1-M and WACCM).The central value of the zonal-wind oscillation for unit STD of the leading time series is around 8–9 m s−1in these ESMs.

5.2.Frequencies of PVO events

Based on the results of Cai and Ren(2006,2007),PVO events occurs 1–2 times in each winter season in the NCEP2 reanalysis.Table 3 shows the average frequency of positive/negative PVOs in each reanalysis dataset and in each CMIP5 model based on the different thresholds(10,15 and 20 m s−1)for PVO events.When 10 m s−1is chosen as the threshold for PVO events,there are on average about 7–8 PVO events in one decade in NCEP1,NCEP2,ERA-I,and JRA25.In 20CR,however,there are only 4–5 PVO events in one decade.

The reproducibility of the PVO frequency varies among CMIP5 models(Table 3).Specif i cally,whether a higher (20 m s−1)or a lower(10 m s−1)threshold is used,not even one PVO event can be identif i ed in CSIRO-Mk3.6.0 and MIROC5,again indicating that the zonal-wind oscillation intensity in these models is relatively very weak(Fig. 7).Similar problems also exist in some other GCMs.For the thresholds of 10,15 or 20 m s−1,the frequency of positive/negative PVO events in the reanalysis ensemble(20CR excluded)is about 6.8/7.6,3.4/4.2,and 1.5/2.7 per decade, respectively,and it is only about 4.9/5.6,2.2/3.2,and 0.9/1.8 per decade,respectively,in the model ensemble.This seems to indicate that the polar vortex oscillation in most of the CMIP5 models,particularly the CMIP5 GCMs,is generally much weaker and can barely reach the observed PVO intensity.In other words,aside from the common problem of the cooler or stronger stratospheric polar vortex,the underestimated frequency of positive/negative PVO events is another challenge for the CMIP5 models,especially the CMIP5GCMs.Incontrast,thefrequencyofpositive/negative PVO events that reach the thresholds in most of the ESMs (e.g.,BCC-CSM1-1-m,IPSL-CM5A-LR,IPSL-CM5A-MR, MIROC-ESM,MIROC-ESM-CHEM,MPI-ESM-LR,MPIESM-MR,MPI-ESM-P and WACCM)is comparable with the observation(Table 3).

Next,we compare the seasonal distributions of PVO events among each reanalysis dataset and each CMIP5 model.To do this,the month-by-month distributions of the average numbers of PVO events in each reanalysis dataset (Fig.8a)and each CMIP5 model(Figs.8b–e)are shownin Fig.8.Positive/negative PVO events are denoted with dark/light gray bars.From the distribution of PVO events in NCEP1,NCEP2,ERA40,ERA-I,and JRA25,it can be seen that,typically,most PVO events occur during mid-winter to early spring(January–March),with only a few PVOs occurring in November and April.The seasonal distributions of PVO events in NCEP1,NCEP2,ERA40,ERA-I,and JRA25areratherconsistent,especiallyamongNCEP2,ERAI,and JRA25,and between NCEP1 and ERA40.Whereas, in 20CR,the PVO events occur mainly in March–May(Fig. 8a).This seasonal drift problem for the occurrence of PVO events is also common in the CMIP5 GCMs(e.g.,CCSM4, CNRM-CM5,FGOALS-g2,FGOALS-s2,GFDL-CM3 and MRI-CGCM3),and even in several of the ESMs(e.g.,BCCCSM1-1,BCC-CSM1-1-m and IPSL-CM5A-LR,Figs.8b– 8e).This seasonaldriftmayberelatedto the1–2monthdelay of the winter extratropical westerly center in these models relative to the reanalysis datasets(Fig.5).In contrast,the PVO frequency and its seasonal distribution in most of the ESMs(e.g.,IPSL-CM5A-MR,IPSL-CM5B-LR,MIROCESM,MIROC-ESM-CHEM,MPI-ESM-LR,MPI-ESM-MR, MPI-ESM-P,NorESM1-M and WACCM)are well reproduced.

Table 3.Frequencies of positive and negative PVO events in each reanalysis dataset and each CMIP5 model,based on different intensity thresholds of the leading oscillation of zonal-mean zonal wind anomalies.Reanalysis ensemble(20CR excluded)and model ensemble results are also listed in the last two rows.

5.3.Long-term changes of PVO

It is known that the global troposphere and the surface have become substantially warmer due to the increase in greenhouse gases(GHGs)in recent decades.Meanwhile, the stratosphere has become steadily cooler,and the northern winter stratospheric polar vortex has become stronger(Randel et al.,1999;Ramaswamy et al.,2001;Langematz et al., 2003;Manzini et al.,2003;Ramaswamy et al.,2006).The strengthening of the polar vortex and the cooling of the polar stratosphere are intimately related to the long-term changes in the low-frequencyvariability in the troposphere,including the AO or the NAM.Next,we turn our attention to the performanceofthe CMIP5 modelsinreproducingthis long-term trend,particularly the long-term trend and the changes in frequency of the stratospheric oscillation events in the CMIP5 models.

Figure 9 shows the existence of a long-term trend of the PVO time series(asterisks)during 1900–2005 in each CMIP5 model,and the corresponding changes in frequency of the PVO events based on the 10 m s−1intensity threshold explained above.It can be seen that some of the models can reproduce a signif i cant positive trend of the PVO time series(e.g.,BCC-CSM1-1-m,CNRM-CM5,CSIROMk3.6.0,FGOALS-s2,HadCM3,MIROC-ESM,MIROCESM-CHEM,MIROC5,MPI-ESM-P,MRI-CGCM3 and NorESM1-M),but exhibit an insignif i cant increase in the frequency of positive PVO events from 1900–50 to 1951–2005. MRI-CGCM3 evenshows a decrease in frequencyof positive PVO events when there is a positive trend of the PVO time series(Fig.9a).Meanwhile,there are some models which show a decrease in frequency of negative PVO events(e.g., CCSM4,FGOALS-s2,MRI-CGCM3,CNRM-CM5,BCCCSM1-1,BCC-CSM1-1-m,NorESM1-M,MIROC-ESMCHEM,MPI-ESM-P,and WACCM),though also insigni ficant.Other models even show an insigni fi cant increase in the frequency of negative PVO events(Fig.9b).As a result,a signi fi cantpositivetrendofthePVOtimeseries canbeidentifi ed in the model ensemble,but accompaniedby insigni fi cant changes in the average frequency of either the positive(Fig. 9a)or the negative(Fig.9b)PVO events.Further diagnosis of the poleward eddy heat fl ux(60°N,50 hPa)by planetarywave activity in winter shows that nearly all the models,except GISS-E2-R[−3.6 K m s−1(10 yr)−1,95%con fi dence level]reproduce insigni fi cant changes in the northward eddy heat fl ux during 1900–2005(not shown).These results basically con fi rm the results from one single model(FGOALS-s2)in Ren and Yang(2012),i.e.that the long-term trend of the polar cooling in recent decades(or the positive trend of the PVO time series)is not related to the decrease in dynamic forcingby planetary-waveactivity(or the occurrenceof PVO events),but rather to the long-term thermodynamic forcing due to the increase in GHGs.

6.Summary

Based on six reanalysis datasets and the historical scenario simulations from 24 CMIP5 models,the northern wintertime stratospheric circulation is diagnosed and systematically assessed.The results indicate that NCEP1,NECP2, ERA40,ERA-I,and JRA25 are quite consistent in describing the general features of the circulation climatology from the stratosphere to the troposphere in the Northern Hemisphere winter,particularlythezonal-meanpatternsof thezonal wind and temperature.The annual cycle of the stratospheric zonalmean zonal wind is also highly consistent among those f i ve reanalysis datasets.As one of the most common problems in the 20CR reanalysis andin some ofthe CMIP5 models(especially the GCMs,e.g.,CCSM4,FGOALS-g2,FGOALS-s2, MRI-CGCM3,and CNRM-CM5),a much stronger polar jet is reproduced,accompanied by a much cooler polar stratosphere.Most of the GCMs show a serious seasonal drift of the zonal-mean zonal wind with a 1–2 month delay of the maximum westerly in the circumpolar region.The simulated seasonal cycle of the stratospheric zonal-mean zonal wind in most of the ESMs agrees very well with the f i rst f i ve reanalyses.The observed PVO,def i ned as the leading mode of the extratropical(20°–90°N)zonal-mean zonal wind in reanalysis,is characterized by a dipole pattern of the zonal-mean zonal wind between the subtropics and the circumpolar region.The amplitude of the circumpolar westerly wind oscillation is reproduced with varying degrees of success in the CMIP5 models.It is generally weaker in some of the GCMs than that in the reanalysis ensemble(excluding 20CR),while it is reproducedmuchmore realistically in most of the ESMs. The frequencyof the PVO events in most of the CMIP5 models is also underestimated,especially in most of the GCMs. Specif i cally,there are on average 6.8/7.6,3.4/4.2,and 1.5/2.7 positive/negative PVO events in one decade in the reanalysis ensemble(excluding 20CR)when an oscillation intensity threshold of 10,15,and 20 m s−1is applied,respectively. The correspondingaverage number of positive/negativePVO events is only 4.9/5.6,2.2/3.2,and 0.9/1.8 in one decade,respectively,in the model ensemble.

In addition,PVO events in NCEP1,NCEP2,ERA40, ERA-I,and JRA25 consistently take place mainly during mid-winter to early spring(January–March),while the peak frequency of PVO events in 20CR and in most GCMs appears 1–2 months later in February–April.The seasonal drift in PVO frequency is consistent with the similar 1–2 months delay,relative to the reanalysis,of the appearance of the strongest winter extratropical westerly center.By contrast, the seasonal drift of the peak frequency of PVO in most of the ESMs is relatively insignif i cant.

The model ensemble shows a positive trend of the PVO time series,accompanied by less signif i cant changes in PVO frequency from 1900–50 to 1951–2005.This verif i es the results from one single model(Ren and Yang,2012),which showed that the long-term trend of the polar cooling in recent decades(or the positive trend of the PVO time series)is not related to the decrease in dynamic forcing by planetarywave activity(or the occurrenceof PVO events),but rather to the long-term thermodynamic forcing due to the increase in GHGs.

Ingeneral,the parallelcomparisonoftheclimatologyand variability of the stratospheric circulation between CIMP5 models and the currently available reanalysis datasets can help model users to understand the model uncertainties in stratospheric processes in CMIP5,and also provide useful information for future model improvements on model ability in describing stratospheric dynamics.The inclusion of the carbon cycle and natural aerosols as well as stratosphereresolved processes may help to signif i cantly improve model performance in simulating the polar stratosphere.Nevertheless,more detailed or specif i c analyses are still needed to further understand the uncertainties in any specif i c stratosphericprocess in each of the CMIP5 models.

Acknowledgements.This work was jointly supported by the National Basic Research Program of China(Grant Nos. 2010CB950400 and 2010CB428603).The authors thank the reviewers and editors for their helpful comments and suggestions. We acknowledge the World Climate Research Program’s Working Group responsible for the Coupled Model Intercomparison Project.

REFERENCES

Baldwin,M.P.,and T.J.Dunkerton,1999:Propagation of the Arctic Oscillation from the stratosphere to the troposphere. J.Geophys.Res.,104,30937–30946,doi:10.1029/1999jd 900445.

Baldwin,M.P.,and T.J.Dunkerton,2001:Stratospheric harbingers of anomalous weather regimes.Science,294,581–584.

Bao,Q.,and Coauthors,2013:The Flexible Global Ocean-Atmosphere-Landsystemmodel,SpectralVersion2: FGOALS-s2.Adv.Atmos.Sci.,30,561–576,doi:10.1007/ s00376-012-2113-9.

Bentsen,M.,and Coauthors,2013:The Norwegian Earth System Model,NorESM1-M—Part 1:Description and basic evaluation of the physical climate.Geoscientif i c Model Development,6,687–720.

Cai,M.,2003:Potential vorticity intrusion index and climate variability of surface temperature.Geophys.Res.Lett.,30,1119, doi:10.1029/2002GL015926.

Cai,M.,and R.C.Ren,2006:40–70 day meridional propagation of global circulation anomalies.Geophys.Res.Lett.,33, L06818,doi:10.1029/2005GL025024.

Cai,M.,and R.C.Ren,2007:Meridional and downward propagation of atmospheric circulation anomalies.Part I:Northern Hemisphere cold season variability.J.Atmos.Sci.,64,1880–1901.

Charlton,A.J.,and Coauthors,2007:A new look at stratospheric sudden warmings.Part II:Evaluation of numerical model simulations.J.Climate,20,470–488.

Collins,M.,S.F.B.Tett,and C.Cooper,2001:The internal climate variability of HadCM3,a version of the Hadley Centre coupled model without f l ux adjustments.Climate Dyn.,17, 61–81.

Compo,G.P.,and Coauthors,2011:The twentieth century reanalysis project.Quart.J.Roy.Meteor.Soc.,137,1–28.

Dee,D.P.,and Coauthors,2011:The ERA-Interim reanalysis: conf i guration and performance of the data assimilation system.Quart.J.Roy.Meteor.Soc.,137,553–597.

Donner,L.J.,and Coauthors,2011:The dynamical core,physical parameterizations,and basic simulation characteristics of the atmospheric component AM3 of the GFDL Global Coupled Model CM3.J.Climate,24,3484–3519.

Dufresne,J.L.,and Coauthors,2013:Climate change projections using the IPSL-CM5 Earth System Model:from CMIP3 to CMIP5.Climate Dyn.,40,2123–2165.

Eyring,V.,T.G.Shepherd,andD.W.Waugh,2010:SPARCreport on the evaluation of chemistry-climate models.SPARCreport No.5,WCRP-132,WMO/TD-No.1526.[Available online at http://www.atmosp.physics.utoronto.ca/SPARC.]

Gent,P.R.,and Coauthors,2011:TheCommunity ClimateSystem Model Version 4.J.Climate,24,4973–4991.

Giorgetta,M.A.,and Coauthors,2013:Climate and carbon cycle changes from 1850 to 2100 in MPI-ESM simulations for the Coupled Model Intercomparison Project phase 5.Journal of Advances in Modeling Earth Systems,5,572–597.

Kalnay,E.,and Coauthors,1996:The NCEP/NCAR 40-year reanalysis project.Bull.Amer.Meteor.Soc.,77,437–471.

Kanamitsu,M.,W.Ebisuzaki,J.Woollen,S.-K.Yang,J.J.Hnilo, M.Fiorino,and G.L.Potter,2002:NCEP–DOE AMIP-II Reanalysis(R-2).Bull.Amer.Meteor.Soc.,83,1631–1643.

Kim,D.,and Coauthors,2012:The tropical subseasonal variability simulated in the NASA GISS general circulation model.J. Climate,25,4641–4659.

Kodera,K.,K.Yamazaki,M.Chiba,and K.Shibata,1990:Downwardpropagation ofupper stratospheric meanzonal windperturbation to the troposphere.Geophys.Res.Lett.,17,1263–1266,doi:10.1029/Gl017i009p01263.

Langematz,U.,M.Kunze,K.Kr¨uger,K.Labitzke,and G.L.Roff, 2003:Thermal and dynamical changes of the stratosphere since 1979 and their link to ozone and CO2 changes.J.Geophys.Res.,108(D1),4027,doi:10.1029/2002JD002069.

Li,L.J.,and Coauthors,2013:The Flexible Global Ocean-Atmosphere-Land system model,Grid-point Version 2: FGOALS-g2.Adv.Atmos.Sci.,30,543–560,doi:10.1007/ s00376-012-2140-6.

Liu,Y.Z.,R.C.Ren,and B.He,2012:Comparison of SAMIL and BCC AGCM simulations of the polar vortex oscillation in the northern hemisphere winter.Chinese J.Atmos.Sci.,36, 1191–1206.(in Chinese)

Manzini,E.,B.Steil,C.Bruhl,M.A.Giorgetta,and K.Kruger, 2003:A new interactive chemistry-climate model:2.Sensitivity of the middle atmosphere to ozone depletion and increase in greenhouse gases and implications for recent stratospheric cooling.J.Geophys.Res.,108,4429,doi:10.1029/ 2002JD002977.

Marsh,D.R.,M.J.Mills,D.E.Kinnison,J.F.Lamarque,N. Calvo,and L.M.Polvani,2013:Climatechange from1850 to 2005 simulated in CESM1(WACCM).J.Climate,26,7372–7391.

Onogi,K.,and Coauthors,2007:The JRA-25 reanalysis.J.Meteor.Soc.Japan,85,369–432.

Pawson,S.,and Coauthors,2000:The GCM-Reality Intercomparison Project for SPARC(GRIPS):Scientif i c issues and initial results.Bull.Amer.Meteor.Soc.,81,781–796.

Ramaswamy,V.,M.D.Schwarzkopf,W.J.Randel,B.D.Santer,B.J.Soden,and G.L.Stenchikov,2006:Anthropogenic and natural inf l uences in the evolution of lower stratospheric cooling.Science,311,1138–1141.

Ramaswamy,V.,and Coauthors,2001:Stratospheric temperature trends:Observations and model simulations.Rev.Geophys., 39,71–122,doi:10.1029/1999rg000065.

Randel,W.J.,F.Wu,J.M.Russell,and J.Waters,1999:Spacetime patterns of trends in stratospheric constituents derived from UARS measurements.J.Geophys.Res.,104,3711–3727,doi:10.1029/1998jd100044.

Ren,R.C.,and M.Cai,2006:Polar vortex oscillation viewed in an isentropic potential vorticity coordinate.Adv.Atmos.Sci., 23,884–900,doi:10.1007/s00376-006-0884-6.

Ren,R.C.,and M.Cai,2007:Meridional and vertical out-ofphase relationships of temperature anomalies associated with the Northern Annular Mode variability.Geophys.Res.Lett., 34,L07704,doi:10.1029/2006GL028729.

Ren,R.C.,and Y.Yang,2012:Changes in winter stratospheric cir-culation in CMIP5 scenarios simulated by the climate system model FGOALS-s2.Adv.Atmos.Sci.,29,1374–1389,doi: 10.1007/s00376-012-1184-y.

Ren,R.C.,G.X.Wu,C.Ming,and J.J.Yu,2009:Winter season stratospheric circulation in the SAMIL/LASG general circulation model.Adv.Atmos.Sci.,26,451–464,doi: 10.1007/s00376-009-0451-z.

Rotstayn,L.D.,M.A.Collier,M.R.Dix,F.Yan,H.B.Gordon,S. P.O’farrell,I.N.Smith,and J.Syktus,2009:Improved simulation of Australian climate and ENSO-related rainfall variability in a global climate model with an interactive aerosol treatment.Int.J.Climatol.,30,1067–1088.

Taylor,K.E.,R.J.Stouffer,and G.A.Meehl,2012:An overview of CMIP5 and the experiment design.Bull.Amer.Meteor. Soc.,93,485–498.

Thompson,D.W.J.,and J.M.Wallace,1998:The Arctic Oscillation signature in the wintertime geopotential height and temperature f i elds.Geophys.Res.Lett.,25,1297–1300,doi: 10.1029/98gl00950.

Thompson,D.W.J.,M.P.Baldwin,and J.M.Wallace,2002: Stratospheric connection to Northern Hemisphere wintertime weather:Implications for prediction.J.Climate,15,1421–1428.

Uppala,S.M.,and Coauthors,2005:The ERA-40 re-analysis. Quart.J.Roy.Meteor.Soc.,131,2961–3012.

Voldoire,A.,and Coauthors,2013:The CNRM-CM5.1 global climate model:Description and basic evaluation.Climate Dyn., 40,2091–2121.

Volodin,E.M.,N.A.Dianskii,and A.V.Gusev,2010:Simulating present-day climate with the INMCM4.0 coupled model of the atmospheric and oceanic general circulations.Izvestiya, Atmospheric and Oceanic Physics,46,414–431.

Watanabe,M.,and Coauthors,2010:Improved climate simulation by MIROC5:Mean states,variability,and climate sensitivity. J.Climate,23,6312–335.

Wu,T.W.,and Coauthors,2013:Global carbon budgets simulated by the Beijing Climate Center Climate System Model for the last century.J.Geophys.Res.,118,4326–4347,doi: 10.1002/jgrd.50320.

Xin,X.,T.Wu,and J.Zhang,2013:Introduction of CMIP5 experiments carried out with the climate system models of Beijing Climate Center.Advances in Climate Change Research, 4,41–9.

Yukimoto,S.,and Coauthors,2012:A new global climate model of the meteorological research institute:MRI-CGCM3, model description and basic performance.J.Meteor.Soc. Japan,90A,23–64.

:Rao,J.,R.C.Ren,and Y.Yang,2015:Parallel comparison of the northern winter stratospheric circulation in reanalysis and in CMIP5 models.Adv.Atmos.Sci.,32(7),952–966,

10.1007/s00376-014-4192-2.

(Received 26 August 2014;revised 7 November 2014;accepted 10 December 2014)

∗Corresponding author:REN Rongcai Email:rrc@lasg.iap.ac.cn

©Institute of Atmospheric Physics/Chinese Academy of Sciences,and Science Press and Springer-Verlag Berlin Heidelberg 2015