Impact of global sea surface temperature on the recent early winter Arctic tropospheric warming in coordinated large ensemble simulations

Luto Wng , , Yongqi Go , Dong Guo , , To Wng , , , Ying Zhng , , Wi Hu ,

a School of Atmospheric Science, Chengdu University of Information Technology, Chengdu, China

b Nansen-Zhu International Research Center, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

c Nansen Environmental and Remote Sensing Center and Bjerknes Centre for Climate Research, Bergen, Norway

d Climate Change Research Center, Chinese Academy of Sciences, Beijing, China

e Joint Laboratory for Climate and Environmental Change, Chengdu University of Information Technology, Chengdu, China

f Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Nanjing University of Information Science and Technology, Nanjing, China

Keywords:Arctic warming Sea surface temperature Multi-model coordinated experiment Eliassen-Palm flux Planetary wave

ABSTRACT Coordinated numerical ensemble experiments with six different state-of-the-art atmosphere models were used to evaluate and quantify the impact of global SST (from reanalysis data) on the early winter Arctic warming during 1982-2014. Two sets of experiments were designed: in the first set (EXP1), OISSTv2 daily sea-ice concentration and SST variations were used as the lower boundary forcing, while in the second set (EXP2) the SST data were replaced by the daily SST climatology. In the results, the multi-model ensemble mean of EXP1 showed a nearsurface (～850 hPa) warming trend of 0.4 °C/10 yr, which was 80% of the warming trend in the reanalysis.The simulated warming trend was robust across the six models, with a magnitude of 0.36-0.50 °C/10 yr. The global SST could explain most of the simulated warming trend in EXP1 in the mid and low troposphere over the Arctic, and accounted for 58% of the simulated near-surface warming. The results also suggest that the uppertropospheric warming (～200 hPa) over the Arctic in the reanalysis is likely not a forced signal; rather, it is caused by natural climate variability. The source regions that can potentially impact the early winter Arctic warming are explored and the limitations of the study are discussed.

1. Introduction

The surface air temperature in the Arctic is warming at twice the global rate (e.g., Serreze and Barry, 2011 ) -a phenomenon known as Arctic amplification. Arctic amplification happened both in the early 20th century ( Tokinaga et al., 2017 ; Svendsen et al., 2018 ) and in recent decades ( Ding et al., 2014 ). The troposphere over the Arctic has also been warming in recent decades, both in reanalysis data and model simulations. Though weak, this tropospheric warming is non-negligible relative to the surface ( Graversen et al., 2008 ; Screen et al., 2012 ;Ding et al., 2014 ; Dai et al., 2019 ).

Fig. 1. (a-d) Zonal-mean temperature trend (shaded; units: °C/10 yr) for (a) ERA-Interim, (b) the multi-model ensemble mean (MMEM) of EXP1, (c) MMEM EXP2,and (d) the difference between MMEM EXP1 and MMEM EXP2 (MMEM DIFF), in November-December. (e-h) Zonal-mean zonal wind climate distribution (thick black contours; units: m s − 1 ; interval: (e-g) 5 m s − 1 , (h) 0.2 m s − 1 ) and trend (shaded; units: m s − 1 /10 yr) for (e) ERA-Interim, (f) MMEM EXP1, (g) MMEM EXP2,and (h) MMEM DIFF, in November-December. The black dotted regions indicate significant values based on the 95% confidence level from a two-tailed Student’s t -test.

Early studies proposed various mechanisms responsible for the Arctic warming, which included local snow and sea-ice feedback (e.g.,Screen and Simmonds, 2010 ), cloud and water vapor ( Taylor et al.,2013 ; Burt et al., 2016 ; Gong et al., 2017 ), and remote transport of heat by the atmosphere and ocean as well as moisture via the atmosphere from the low latitudes (e.g., Sandøet al., 2014 ; Svendsen et al.,2018 ). Screen et al. (2012) were the first to use AGCMs to explore the winter Arctic warming (both surface and tropospheric warming). They found that the winter Arctic surface warming was strongly coupled with the sea-ice decline, which was consistent and confirmed by many other studies. Further, they found the winter tropospheric warming over the Arctic was caused by the remote SST. Ding et al. (2014) suggested that the SST change in the tropical Pacific could explain a significant part of the increased annual atmospheric temperature between 850 hPa and 700 hPa over northeastern Canada and Greenland during 1979-2012.Li et al. (2015) suggested that the SST anomalies in the North Pacific can impact the Arctic polar vortex and therefore contributed to the winter Arctic surface warming during 1994-2013. In this paper, we extend these early studies by using two sets of coordinated numerical experiments performed with six AGCMs and quantify the impact of global SST on the surface and tropospheric warming over the Arctic. Each set of experiments consisted of 130 simulations and each simulation lasted 33 model years from 1982 to 2014. Since the Arctic warming is strongest during winter, our paper mainly focuses on that period.

2. Data and methods

2.1. Models and reanalysis data

Among the AGCMs employed, five of them (CAM4, IAP4, IFS,LMDZOR, WACCM) are forced by prescribed sea-ice concentration and SST from OISSTv2 ( Reynolds et al., 2007 ), whereas a different dataset ( Hurrell et al., 2008 ) is used to force the sixth model (AFES).The external forcings of all experiments followed the CMIP5 protocol ( https://www.wcrp-climate.org/wgcm-cmip/wgcm-cmip5 ). For more details on the experimental setup, readers are referred to Ogawa et al. (2018) . The multi-model ensemble mean of all six models was calculated. The ERA-Interim reanalysis dataset ( Dee et al., 2011 )was used for model-data comparison. To facilitate the comparison, all model outputs and ERA-Interim were interpolated and regridded onto a 2.5°×1.5° latitude-longitude grid.

The Arctic in this paper refers to the region poleward of 67°N, and we use November-December to represent early winter ( Screen et al., 2012 ).The results in January-February are similar to those in November-December unless otherwise specified.

2.2. Diagnostic analysis methods

The Eliassen-Palm (EP) flux was used to characterize the planetary flux (wavenumber 1-3) activity ( Andrews and Mcintyre, 1976 ; Li et al.,2015 ). The EP flux function is defined as

3. Results

3.1. Warming trends

Fig. 2. Temperature trend (shaded; units: °C/10 yr) for (a) ERA-Interim, (b) the multi-model ensemble mean (MMEM) of EXP1, (c) MMEM EXP2, and (d) the difference between MMEM EXP1 and MMEM EXP2, at 850 hPa in November-December. The black dotted regions indicate significant values based on the 95%confidence level from a two-tailed Student’s t -test.

As mentioned above, the two sets of experiments only differed in their SST forcing. The resemblance between EXP1 and EXP2 indicates the impacts of the Arctic sea-ice concentration, the external forcings and the internal variability, and the difference between them evidences the impact of the global SST ( Koenigk et al., 2018 ). We show the zonal warming trends in ERA-Interim, EXP1, EXP2, and their difference, respectively, in Fig. 1 (a-d). ERA-Interim ( Fig. 1 (a)) shows two warming centers over the Arctic. One is at the surface and near-surface, and the other at around 200 hPa. The area-averaged near-surface warming trend over the Arctic is 0.6 °C/10 yr. EXP1 ( Fig. 1 (b)) captures well the surface and near-surface warming trend of 0.4 °C/10 yr over the Arctic.The simulated surface and near-surface warming trends over the Arctic are weaker than in ERA-Interim, especially over areas north of 80°N,which means there is a southward shift in the warming center compared to ERA-Interim. This is a common problem for AGCMs in their simulation of Arctic warming ( Cohen et al., 2019 ). The simulated near-surface warming trend is robust across the six models (figure not shown) and varies between 0.36 and 0.50 °C/10 yr across the ensemble mean of each model. It should be noted that the models can successfully capture the significant upper-level warming trend over the tropics apparent in ERAInterim. The warming trend in EXP2 ( Fig. 1 (c)) is significant at the surface and near-surface over the Arctic, with a magnitude of 0.2 °C/10 yr.The difference between EXP1 and EXP2 indicates the global SST can explain most of the simulated warming trend in the troposphere -similar to early studies (e.g., Screen et al., 2012 ). The sea-ice change is strongly coupled with the surface warming trend over the Arctic -also consistent with early studies (e.g., Dai et al., 2019 ). It can be seen that the global SST also contributes to the near-surface warming.

Fig. 1 (e-h) shows the reanalysis and simulated ensemble mean of the zonal-mean zonal wind climatology and the trend. In terms of the climatology, the reanalysis and the simulated locations and intensities of the westerly jet are similar at 200 hPa around 35°N. In ERA-Interim( Fig. 1 (e)), the westerly jet shows a southward shift during 1982-2014,which also holds for EXP1 and EXP2 ( Fig. 1 (f,g)). Further, ERA-Interim( Fig. 1 (e)) also shows that the zonal wind below 200 hPa in the tropical region (0°-20°N) and the midlatitudes (40°-60°N) has a significant weakening trend during the same period. However, at high latitudes,the whole tropospheric zonal wind has an increasing trend, especially below 500 hPa. It should be noted that the simulated trend of zonal wind in EXP1 and EXP2 is not statistically significant and much weaker than in ERA-Interim. This indicates that the global SST has no robust and significant impact on the trend of the westerly jet.

We further present the spatial distributions of the warming trend at 850 hPa in the reanalysis and simulation in Fig. 2 . ERA-Interim shows there are three significant warming centers -over the Barents-Kara Sea,northeastern Eurasia, and northeastern Canada ( Fig. 2 (a)). The simulated warming trend ( Fig. 2 (b)) in EXP1 is qualitatively consistent with that in ERA-Interim, but with weaker intensity. Without SST forcing(EXP2), the warming trend ( Fig. 2 (c)) is weaker than that in EXP1. Both EXP1 and EXP2 do not capture the North Pole-centralized warming over the Arctic at 200 hPa in ERA-Interim (figure not shown). It should be noted, however, that the warming at 200 hPa in ERA-Interim is not statistically significant. The SST forcing ( Fig. 2 (d)) can contribute to the simulated near-surface warming.

Fig. 3. Trend of EP flux cross sections (vectors; scale: 10 8 ; units: m 2 s − 2 /10 yr) and divergence (shaded; units: m s − 1 d − 1 /10 yr) for (a) ERA-Interim, (b) the multimodel ensemble mean (MMEM) of EXP1, (c) MMEM EXP2, and (d) the difference between MMEM EXP1 and MMEM EXP2, during November-December. The black dotted regions indicate significant values based on the 95% confidence level from a two-tailed Student’s t -test.

3.2. EP flux trends and eddy heat transport

The reanalysis and simulated estimates of the EP flux convergence/divergence and the trend are shown in Fig. 3 . ERA-Interim( Fig. 3 (a)) shows that the EP flux at 30°-40°N propagates upward from the near-surface to the upper troposphere and downward at 50°-60°N from the upper troposphere to the near-surface, implying enhanced tropospheric poleward eddy heat transport at 30°-40°N and weakened poleward eddy heat transport at 50°-60°N (also refer to Fig. 4 (a)). The EP flux shows a trend of convergence, which corresponds to the weakening of westerly wind ( Fig. 1 (e)). On the other hand, the EP flux around 70°N also tends to propagate upward, indicating enhanced poleward eddy heat transport ( Fig. 4 (a)). The EP flux in EXP1 shows a downward propagation trend in the troposphere (surface to 200 hPa) between 40°-60°N, indicating weakened poleward eddy heat transport ( Fig. 4 (b)).The EP flux propagates upward in the middle and lower troposphere at 850 hPa to 500 hPa between 70°N and 80°N, indicating increased poleward eddy heat transport. As a result, the Arctic is warming in the middle and lower troposphere ( Fig. 1 (b)). The upward propagation of EP flux between 50°N and 70°N in the troposphere in EXP2 shows an increased trend ( Fig. 3 (c)) compared to that in EXP1 ( Fig. 3 (b,d), implying more poleward eddy heat transport (or a less reduced poleward eddy heat transport trend between 50°N and 70°N) in EXP2 ( Fig. 4 (b)). Forced by the global SST ( Fig. 3 (d)), the 40°-60°N troposphere has a downward propagation trend, implying reduced poleward eddy heat transport. The EP flux in the troposphere between 70°N and 80°N propagates upward,implying enhanced eddy heat transport.

Fig. 4. Zonal mean trend of eddy heat flux (shaded; units: m s − 1 K − 1 /10 yr) for (a) ERA-Interim, (b) the multi-model ensemble mean (MMEM) of EXP1, and (c)MMEM EXP2, during November-December. The black dotted regions indicate significant values based on the 95% confidence level from a two-tailed Student’s t -test.

Fig. 5. (a, c, e) Trend of the non-zonal component of geopotential height (shaded) and the wave activity flux (vectors; units: m 2 s − 2 /10 yr) associated with the non-zonal component of the geopotential height trend pattern for (a) ERA-Interim, (c) the multi-model ensemble mean (MMEM) of EXP1, and (e) MMEM EXP2,at 200 hPa during November-December. (b, d, f) Trend of the non-zonal component of the temperature (shaded) pattern for (b) ERA-Interim, (d) the multi-model ensemble mean (MMEM) of EXP1, and (f) MMEM EXP2, averaged at 700 hPa during November-December. The white dotted regions indicate significant values based on the 95% confidence level from a two-tailed Student’s t -test.

The above analysis is reconfirmed by the trend distribution of the eddy heat flux transfer ( Fig. 4 ). It shows the trend of meridional eddy heat flux (the vertical component of EP flux). ERA-Interim shows increased eddy heat flux over the Arctic between 200 hPa and 400 hPa,and from the surface to about 600 hPa ( Fig. 4 (a)), consistent with the two warming centers over the Arctic ( Fig. 1 (a)). The increased eddy heat flux in EXP1 happens from the surface to 400 hPa ( Fig. 4 (b)), consistent with the simulated warming trend ( Fig. 1 (b)). No increase in eddy heat flux is apparent in EXP2 between 250 hPa and 700 hPa ( Fig. 4 (c)). However, it should be noted that the trends of meridional eddy heat flux to the Arctic troposphere do not pass the significance test in ERA-Interim,EXP1, and EXP2. Further, the EP flux trends in the simulations are much weaker than in ERA-Interim, and can partly reflect the warming trend over the Arctic.

3.3. Wave activity flux

Ding et al. (2014) suggested that the increase in annual temperature at 850-300 hPa during 1979-2012 over northeastern Canada and Greenland was largely driven by the SST in the tropical Pacific via forcing a Rossby wave train. We analyzed the wave activity flux in the reanalysis data and in the multi-model simulations ( Fig. 5 ). In ERA-Interim( Fig. 5 (a)), the wave train originates from the tropical Pacific (～10°-30°N) and propagates following two pathways, with one branch to the north and the other to the northeast. The two branches meet over North America and cross the North Atlantic and the Eurasian continent. The wave train pattern closely resembles the anomalous tropospheric warming trend ( Fig. 5 (b)). In EXP1 ( Fig. 5 (c)), the wave train pattern resembles ERA-Interim: originating from the tropical Pacific and propagating following two pathways. The wave train pattern is intensified over the North Pacific and consistent with the simulated anomalous tropospheric warming trend over northeastern Eurasia and Greenland ( Fig. 5 (d)). As expected, the wave activity is weaker in EXP2 ( Fig. 5 (e)) and with no corresponding warming trend over northeastern Eurasia and Greenland.

Our results confirm that the SST in the tropical Pacific can lead to anomalous Rossby wave propagation, affecting atmospheric circulation and contributing to the warming in northeastern Eurasia and Greenland. However, we are unable to quantify the impact of the tropical SST because of the limitations of the designed experiments.

4. Summary and discussion

We investigated the impact of global sea surface temperature on the early winter Arctic warming during 1982 to 2014 -particularly the tropospheric warming over the Arctic -by using ERA-Interim data and two sets of coordinated experiments with six AGCMs. Results showed that the multi-model ensemble mean of EXP1 had a warming trend of 0.4 °C/10 yr in the near-surface layer and constituted 80% of the reanalysis warming trend. The simulated warming trend was robust across the six models, with a magnitude of 0.36-0.50 °C/10 yr. There were individual simulations out of the 130 members that showed almost no warming trend. However, the ensemble mean of each model reproduced the warming trend (figure not shown). Koenigk et al. (2018) found that the spread across ensemble members was large and many ensemble members were required to reproduce the 2-m air temperature variations over northern Europe seen in the reanalysis data. The global SST could explain most of the simulated warming trend in EXP1 in the mid and low troposphere over the Arctic, and accounted for 58% of the simulated near-surface warming. The results also suggest that the uppertropospheric warming (～200 hPa) over the Arctic in the reanalysis is probably not a forced signal; rather, it is caused by natural climate variability. The global sea surface temperature can impact the tropospheric Arctic warming by the increased meridional eddy heat transport into the Arctic. Further analysis suggested that the tropical and extratropical Pacific SST in early winter might be a source of influence on the atmospheric circulation and impact the Arctic warming. However, the impact of the tropical and extratropical Pacific SST on the early winter Arctic warming cannot be quantified because of the limitations of the experiments employed in our study. Ding et al. (2014) highlighted the importance of tropical Pacific SST on the increased annual temperature.Here, the ensemble mean of the six models did not reproduce the Arctic warming in the upper troposphere (200 hPa), implying the reanalysis warming at about 200 hPa is likely caused by natural climate variability rather than a forced signal.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported by the National Key R&D Program of China[grant number 2017YFE0111800 ] and the National Natural Science Foundation of China [grant numbers 41790472 and 41661144005 ]. The second author was also partly supported by the EU H2020 Blue-Action project [grant number 727852 ].

Acknowledgments

The authors are grateful to the NordForsk GREENICE project, which supported the coordinated experiments.