Impact of the Asian—Pacific Oscillation on early autumn precipitation over Southeast China:CMIP6 evaluation and projection

Changji Xia ,Wei Hua ,Yu Zhang ,Guangzhou Fan,c

a School of Atmospheric Sciences,Chengdu University of Information Technology,Chengdu,China b Plateau Atmosphere and Environment Key Laboratory of Sichuan Province,Chengdu,China c Meteorological Disaster Prediction and Warning Engineering Laboratory of Sichuan Province,Chengdu,China d Zhenyuan Meteorological Bureau,Qiandongnan Prefecture,Zhenyuan,China

Keywords: Asian-Pacific Oscillation CMIP6 Model Evaluation Projection

ABSTRACT This study evaluated the capability of 32 models of phase 6 of the Coupled Model Intercomparison Project in modeling the influence of the preceding August Asian—Pacific Oscillation(APO)on subsequent early autumn(September) precipitation over Southeast China and associated atmospheric anomalies,as well as its future projection during 2021—2040 (near-term),2041—2060 (mid-term),and 2081—2100 (long-term) under different Shared Socioeconomic Pathways(SSPs:SSP2-4.5 and SSP5-8.5).Results indicated that two-thirds of the individual models yielded positive correlations between the APO and Southeast China precipitation that conformed to the observations.On the basis of the capability to reproduce the significantly positive relationship between the APO and Southeast China precipitation,three models were chosen as the “best”model ensemble (BMME).The BMME effectively simulated both the APO-associated precipitation and the atmospheric anomalies,and outperformed the ensemble of the remaining 29 models in terms of the positive correlation between the APO and Southeast China precipitation,and the negative correlations between the meridional displacement of the East Asian jet(EAJ)and the APO and Southeast China precipitation.In general,during three future time periods under both SSPs,the BMME projected persistent negative correlations between the APO and EAJ,and the APO—Southeast China precipitation and EAJ—Southeast China precipitation relationships were projected to weaken.However,considerable discrepancies were evident among the changes projected by the individual models;only the projected changes in the APO—EAJ relationship showed good model agreement.

1.Introduction

The Asian—Pacific Oscillation(APO),as a seesaw vibration of uppertropospheric temperature between Asia and the North Pacific,is a dominant teleconnection pattern of the Asian—Pacific sector (Zhao et al.,2007).The APO exists not only in summer but also in other seasons(Zhao et al.,2008).Previous studies have demonstrated that anomalous precipitation over China is affected by the APO(Zhou and Zhao,2010;Lin et al.,2019,2021;Xia et al.,2022),and that changes in the APO are strongly correlated with anomalous atmospheric circulations over the Asian—Pacific sector and beyond (Zhao et al.,2008,2012;Zhou et al.,2008;Zhou and Wang,2015;Hua et al.,2019).Generally,the APO plays a prominent role in establishing atmospheric circulation linkages between Asia and the North Pacific.

Simulations produced by models participating in phase 6 of the Coupled Model Intercomparison Project(CMIP6)have created new avenues for studying climatic changes (Eyring et al.,2016).With the gradual release of CMIP6 outputs,many studies have assessed and projected temperatures,precipitation,and atmospheric circulations using CMIP6 models (Agel and Barlow,2020;Chen et al.,2020;Hu et al.,2022).A recent study by Fan and Zhou (2022) indicated reasonable fidelity of CMIP6 models in reproducing the APO pattern during all four seasons.However,less emphasis has been placed on the ability of CMIP6 simulations to reproduce APO-induced precipitation and atmospheric anomalies.Because most previous studies relevant to the APO have focused on summer,we sought to investigate whether CMIP6 models can reproduce the time-lagged linkages between the summer APO and subsequent early autumn precipitation anomaly documented by Xia et al.(2022).

The remainder of this paper is arranged as follows.Section 2 describes the data and methods used in this investigation.Section 3 presents the results of model evaluations and projections.Finally,Section 4 concludes the paper by summarizing the derived findings.

2.Data and methods

This study adopted the atmospheric variables and precipitation of 32 CMIP6 “r1i1p1f1”monthly historical simulations and two Shared Socioeconomic Pathways (SSPs: SSP2-4.5 and SSP5-8.5) simulations(O’Neill et al.,2014;Gidden et al.,2019).A brief summary of the CMIP6 models considered in this study is presented in Table 1.Additional details of each of the models can be obtained from https://esgfnode.llnl.gov/search/cmip6/.

Table 1 Details of the 32 CMIP6 models employed in this study.

Monthly mean atmospheric reanalysis data at 2.5° × 2.5° resolution,provided by the National Centers for Environmental Prediction/National Center for Atmospheric Research (Kalnay et al.,1996),were exploited for this study.Monthly precipitation data at 0.25°×0.25°resolution were extracted from the CN05.1 dataset (Wu and Gao,2013).

For comparison with observations,bilinear interpolation was employed to interpolate the monthly variables of the CMIP6 models before analysis,i.e.,2.5°×2.5°,to a resolution consistent with that of the atmospheric circulation variables.Specifically,the observed CN05.1 monthly precipitation data and the 32 sets of simulated CMIP6 monthly precipitation data were interpolated to a 1.5°×1.5°grid.

We focused on the relationship between the preceding August APO and subsequent early autumn (September) precipitation and atmospheric circulations on an interannual scale,and therefore linear trends of all the data mentioned above were removed before analysis.Correlation and regression analyses were conducted to analyze the relationship between the APO and associated atmospheric variables.Statistical significance was examined using the Student’st-test.

Three indices were employed in our study:

(1) The August APO index (hereafter APOI),as defined by Xia et al.(2022),which is calculated from the arithmetic difference between the vertically (500—200 hPa) and regionally averaged eddy temperature (T′) over Asia and the Pacific,i.e.,

(2) The Southeast China September precipitation index (SCPI),defined as the area-averaged precipitation over Southeast China(23°—33°N,110°—120°E).

(3) The East Asian jet (EAJ) meridional displacement index(EAJI),defined as the arithmetic difference between southern(30°—40°N,120°—150°E) and northern (40°—50°N,120°—150°E)area-averaged 200hPa zonal winds (Lu,2004),i.e.,EAJI=A positive(negative)normalized EAJI means that the East Asian jet moves toward the south(north).

The CN05.1 dataset covers the period from 1961 to the present and the CMIP6 historical outputs cover the period 1850—2014,and therefore the evaluation period in this study was set as 1961—2014.Additionally,the period 1995—2014 was selected as the baseline period to be consistent with the Sixth Assessment Report of the Intergovernmental Panel on Climate Change,and the periods 2021—2040,2041—2060,and 2081—2100 under different SSPs were respectively taken to represent the nearterm,mid-term,and long-term.

3.Results

3.1.Evaluation

Fig.1 depicts the August APO pattern and APO-associated September circulation anomalies.Fig.1(a) illustrates the 1961—2014 August climatologicalT′distribution in the upper troposphere,which exhibits an obvious dipole mode ofT′over the Asian—Pacific sector,i.e.,the APO pattern(Zhao et al.,2007).The correlations between the APOI and the September precipitation anomalies over China,displayed in Fig.1(b),are consistent with the results of Xia et al.(2022)obtained for the period 1961—2020.The preceding APO shows significant positive correlation with subsequent early autumn Southeast China precipitation anomalies,with a correlation coefficient between the APOI and SCPI of 0.46(above the 95%confidence level).The observed APO-related atmospheric circulation anomalies are also presented in Fig.1(c,d).At 200 hPa(Fig.1(c)),a stronger APOI signifies enhanced meridional shear of the zonal wind and northward migration of the EAJ.The right-hand side of the EAJ entrance region positioned over Southeast China is conducive to intensification of ascending motion.The observed EAJI—APOI relationship shows negative correlation of -0.32 (above the 95% confidence level),indicating that a stronger(weaker)August APO signifies northward(southward)migration of the September EAJ.Additionally,the observed SCPI—EAJI relationship shows a negative correlation of-0.23(above the 90%confidence level).The regressed 500-hPa geopotential height(GPH)and 850-hPa winds against the APOI are displayed in Fig.1(d).The APOassociated 500-hPa GPH anomalies exhibit a seesaw distribution,with positive and negative GPH anomalies dominating in the North Pacific and Asian continent regions,respectively.Additionally,the Asian continent and Pacific regions are controlled by a cyclonic anomaly and an anticyclonic anomaly at the lower level,respectively,accompanied by anomalous southerly winds prevailing over Southeast China.As outlined in the Introduction,the APO signifies warming (cooling) over Asia and cooling (warming) over the North Pacific,and therefore the physical processes of the APO influencing the GPH and wind field in the lower and upper troposphere satisfy the static equilibrium(Zhao et al.,2008;Zhou and Zhao,2010).These APO-induced atmospheric anomalies provide advantageous conditions for strengthened precipitation over Southeast China,which agrees with the findings of previous related studies(Lin et al.,2021;Xia et al.,2022).The possible mechanism is that the anomalous thermal effect of the preceding August APO can persist into the following September(autumn),and induces anomalous atmospheric circulations and precipitation(Lin et al.,2021;Xia et al.,2022).

Most of the 32 CMIP6 models captured the climatological state and spatial distribution of the APO pattern effectively,albeit with slight differences in amplitude evident between certain models (figures not shown).To assess the performance in modeling the APO and associated circulations,we calculated the correlations of the APOI,EAJI,and SCPI for quantitative analysis.As shown in Fig.2(a),modeling of the relationship between the APO and early autumn Southeast China precipitation produced diversity among the individual CMIP6 models.The simulated correlation coefficients between the SCPI and the APOI range from-0.41(CIESM)to 0.31(BCC-CSM2-MR)during 1961—2014.Overall,19(59%) of the 32 models show positive correlations and the remaining 13 (41%) models show negative correlations.Only three models (i.e.,BCC-CSM2-MR,EC-Earth3-CC,and GFDL-ESM4)exhibit positive correlation between the APOI and the SCPI significantly above the 0.1 level.This result indicates that reasonable reproduction of the time-lagged impact of the preceding APO on subsequent Southeast China precipitation remains a challenge for climate models.

Fig.2.(a)Correlation coefficients between the APOI and the SCPI of the 32 CMIP6 models during 1961—2014.The long(short)dotted lines indicate the 90%(95%)confidence level.(b) Scatter diagram displaying the simulated APOI—EAJI and EAJI—SCPI correlation coefficients of individual CMIP6 historical outputs during 1961—2014.The observational result is represented by the red rhombus and individual models are represented by blue dots marked with numbers;see Table 1 for corresponding model names.

Fig.2(b) shows a scatter diagram of the APOI—EAJI and the EAJI—SCPI correlations during 1961—2014 derived from the 32 CMIP6 historical outputs.For the APOI—EAJI relationship,28 (87.5%) of the 32 models yield negative correlations that conform to the observations.The APOI—EAJI correlations range from-0.41(EC-Earth3)to 0.13(CIESM),and 10 (31%) of the 32 models exhibit significant negative correlation(above the 90%confidence level).However,individual models show differences in reproducing an EAJI—SCPI relationship that conforms with the observations,i.e.,only 17 (53%) of the models yield negative correlations.The EAJI—SCPI correlations range from -0.37 (AWI-CM-1-1-MR) to 0.58 (CAS-ESM2-0),and only 5 (16%) of the 32 models exhibit significant negative correlation(above the 90%confidence level).

Considering the capability of each of the individual CMIP6 models in reproducing the APOI—SCPI relationship,three models(i.e.,BCC-CSM2-MR,EC-Earth3-CC,and GFDL-ESM4) emerged as the “best”model ensemble (BMME).The multimodel ensemble (MME) mean of the BMME and the ensemble of the remaining 29 models(AMME)were compared to evaluate the performance of the two MME samples.Additionally,the two MME samples were constructed from the unweighted mean of their ensemble members.

Table 2 summarizes the three indices obtained from the observations,AMME,and BMME.The BMME outperforms both the individual models and the AMME in simulating the relationships between the APO,Southeast China precipitation,and the EAJ,i.e.,all indices calculated by the BMME comply effectively with the observational results.However,the AMME shows relatively poor performance in reproducing the connections between the September precipitation and large-scale atmospheric circulations.

Table 2 Correlation coefficients of the APOI,SCPI,and EAJI obtained from the observations and two MME samples (i.e.,AMME and BMME).A ?(??) symbol signifies the 90%(95%)confidence level.

We further examined the MME-simulated precipitation and atmospheric circulations associated with the APOI (Fig.3).For the AMME,the APOI shows no significant correlation with Southeast China precipitation (Fig.3(a)).The spatial pattern of the regressed 200-hPa zonal wind anomalies reproduces significant wind shear near 40°N(Fig.3(b)).However,the regressed wind anomalies are underestimated in amplitude.Failing to reproduce the out-of-phase characteristics of the 500-hPa GPH over the Asian—Pacific region and the low-level cyclonic anomalies over Asia,the AMME regressed 500-hPa GPH and 850-hPa wind are also unsatisfactory (Fig.3(c)).For the BMME,the performance in reproducing the APO-related anomalies of precipitation and atmospheric circulations is better (Fig.3(d—f)).In the BMME result,Southeast China precipitation is significantly positively correlated with the APOI,although the areas of positive correlation are larger than those in the observational result (Fig.3(d)).The regression of the 200-hPa zonal wind exhibits not only the spatial characteristics but also the amplitude of the anomalies (Fig.3(e)).Additionally,the regression of the 500-hPa GPH and 850-hPa wind of the BMME is barely satisfactory.The significant anomalous GPH center and cyclonic anomalies are located southward of those in the observational result(Fig.3(f)).

Fig.3.The AMME-simulated(a)correlations between the normalized APOI and early autumn precipitation anomalies over China,(b)regressed 200-hPa zonal wind(units:m s-1;contours),and(c)regressed 500-hPa GPH(units:m;contours)and 850-hPa winds(units:m s-1;vectors)onto the normalized APOI.(d—f)are the same as(a—c)but for the BMME.Dotted areas in(a)and(c)denote the 95%confidence level.Shaded areas in(b),(c),(e),and(f)passed the 95%confidence level.

In brief,the BMME can produce a more reasonable simulation of the time-lagged impact of the APO on early autumn precipitation in Southeast China than that of any of the individual models and the AMME,providing justification for future projection of the relationship between the APO and early autumn Southeast China precipitation.

3.2.Projection

Fig.4 presents the projected correlation coefficients of the three indices for the three future periods (i.e.,near-,mid-,and long-term)derived from the individual models under the SSP2-4.5 and SSP5-8.5 scenarios.For the APOI—SCPI correlation,the correlation coefficients derived from BMME are projected to reduce under both SSP2-4.5 and SSP5-8.5 during the three periods,indicating that the APO—Southeast China precipitation relationship is projected to weaken in the future with respect to 1995—2014.The reduction of the BMME-projected correlation coefficient is the largest under SSP5-8.5 during the mid-term(passing the 95% confidence level) and the smallest under SSP2-4.5 during the long-term.However,the BMME members exhibit diversity in projecting the APOI—SCPI correlation under the different SSPs and future periods.The three BMME members agree on the projected weakening change of the APO—Southeast China precipitation relationship only during the mid-term under SSP2-4.5,and near-term and mid-term under SSP5-8.5.Under the SSP2-4.5 (SSP5-8.5) scenario,50% (47%) of models during the near-term,56% (44%) of models during the mid-term,and 63% (47%) of models during the long-term,project weakened correlations relative to the period 1995—2014 among the 32 CMIP6 models,indicating that the weakening of the projected APOI—SCPI correlation does not reach a good model agreement.

Fig.4.Scatter diagram of correlations between (a) the APOI and the SCPI,(b) the APOI and the EAJI,and (c) the EAJI and the SCPI,during the near-term(2021—2040).Panels (d—f) and (g—i) are the same as (a—c) but for the mid-term (2041—2060) and long-term (2081—2100),respectively.Individual models are represented by blue dots marked with numbers (the red numbers indicate the BMME members;the red triangle represents the BMME result of different future periods;and the red rhombus represents the BMME result during the baseline period of 1995—2014)—see Table 1 for corresponding model names.

For the APOI—EAJI correlation,the BMME projects that the August APO will still be negatively correlated with the September EAJ,except in the near-term under SSP5-8.5,suggesting that the APO—EAJ relationship will be stable,or increases slightly or decrease slightly under both SSPs.Additionally,regardless of the scenario and future period,at least two BMME members project a negative APO—EAJ relationship.Under the SSP2-4.5 (SSP5-8.5) scenario,63% (84%) of models during the nearterm,69%(75%)of models during the mid-term,and 72%(81%)models during the long-term,project negative APO—EAJ correlations among the 32 CMIP6 models,implying that the sustained reverse variation between the APO and EAJ in the future reaches a good model agreement and that SSP5-8.5 has more models that agree with this relationship than SSP2-4.5 has.

For the EAJI—SCPI correlation,the BMME-projected results project reduced correlations between the EAJ and Southeast China precipitation,except in the near-term under SSP5-8.5 and long-term under SSP2-4.5.Notably,regardless of the scenario and future period,all BMME members consistently agree on the sign of the weakening in the EAJ—Southeast China precipitation correlation.However,the individual models show considerable discrepancy in projecting the EAJI—SCPI correlation.Under the SSP2-4.5 (SSP5-8.5) scenario,47% (47%) of models during the near-term,34% (56%) of models during the mid-term,and 31%(66%)models during the long-term,project the reduction of EAJ—Southeast China precipitation correlations among the 32 CMIP6 models—that is,the projected decreasing of the EAJI—SCPI correlation does not reach a good model agreement either.

4.Conclusions

The capability of 32 CMIP6 models in modeling the impact of the preceding August APO on subsequent early autumn Southeast China precipitation was evaluated by comparing historical outputs with reanalysis data for the period 1961—2014.The future changes between the APO and the associated atmospheric anomalies during 2021—2040,2041—2060,and 2081—2100,under different SSPs,were also projected,relative to 1995—2014.

All 32 models effectively captured the APO structure,and approximately two-thirds of the individual models(19 out of 32)exhibited positive APOI—SCPI correlation.Additionally,we assessed the correlation coefficients between different indices and the APO-associated circulation anomalies based on two MME samples(i.e.,the BMME and AMME).The BMME performed better than the individual models and the AMME in simulating the impact of the APO on early autumn Southeast China precipitation,yielding significant in-phase APOI—SCPI(out-of-phase APOI—EAJI and EAJI—SCPI) relationships that conformed to the observations.Furthermore,the AMME-simulated regressions of APO-associated circulations underestimated the amplitude of the atmospheric anomalies,while the regression results obtained from the BMME were more consistent with the regression pattern of the observations,both in spatial distribution and in amplitude of the regressed anomalies.

For the future projection,the BMME projected that the linkage of the APO to the EAJ will persist under the SSP2-4.5 and SSP5-8.5 scenarios during all three future periods.In general,the BMME-projected correlations of the APO—Southeast China precipitation connection and the EAJ—Southeast China precipitation connection decreased in the future.Additionally,the projected changes in APO-associated atmospheric circulations showed considerable discrepancy among the individual models.The sustained APO—EAJ relationship reaches a good model agreement,while the weakening of the APO—Southeast China precipitation and EAJ—Southeast China precipitation relationships do not.Therefore,the simulated results should be interpreted with caution.In view of the large uncertainty among climate models,some bias correction approaches(Li et al.,2010;Yue et al.,2021)have been documented to be effective in reducing model uncertainties,which should be introduced to provide more reliable projection results in future works that explore the APO-associated atmospheric circulations and precipitation.Additionally,previous studies indicate that the summer APO and its linkage to East Asian summer rainfall will weaken in the future (Zhou,2016;Zhou et al.,2018),reflecting the responses of the summer APO and associated circulation anomalies to future warming.The detailed mechanisms responsible for the projected changes elucidated in this study remain to be analyzed in depth in future work.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was jointly supported by the National Natural Science Foundation of China [grant number 42275022],the Second Tibetan Plateau Scientific Expedition and Research Program [grant number 2019QZKK010203],the Sichuan Natural Science Foundation [grant number 2022NSFSC1092],and the Scientific and Technological Innovation Capacity Improvement Project of Chengdu University of Information Technology[KYQN202202].