1. Introduction
The melting of sea ice in the Arctic in all months of the year during the last few decades has been one of the most alarming signals of global climate change (Overland and Wang 2013; Kay et al. 2011; Screen et al. 2012; Screen and Williamson 2017; Stroeve and Notz 2018). The rate of loss of Arctic sea ice in September has accelerated since the 2000s, raising great concerns that the Arctic will see its first ice-free condition in a few decades (Boé et al. 2009; Wang and Overland 2009; Kay et al. 2011; Mahlstein and Knutti 2012; Massonnet et al. 2012; Liu et al. 2013; Overland and Wang 2013; Hezel et al. 2014; Notz 2015; Screen and Deser 2019; Notz et al. 2020; Årthun et al. 2021; Bonan et al. 2021b; Diebold and Rudebusch 2021; Wang et al. 2021).
A broad consensus has been reached among the scientific community that Arctic sea ice is almost linearly related to the anthropogenic forcing on multidecadal to centennial long time scales (Deser et al. 2010; Screen and Simmonds 2010; Cohen et al. 2014; Simmonds 2015; Notz and Stroeve 2016; Jahn 2018; Screen et al. 2018). Results from the current climate model projections suggest that an ice-free Arctic is very unlikely to occur in summer before 2100 if future warming is limited to 1.5°C or less (Screen and Williamson 2017; Jahn 2018; Niederdrenk and Notz 2018; Screen 2018; Sigmond et al. 2018). But if future greenhouse gas emissions are maintained on the current trajectory or at higher levels, the Arctic may be ice-free by the 2050s (Liu et al. 2013; Overland and Wang 2013; Melia et al. 2015; Jahn et al. 2016; Sigmond et al. 2018; Notz et al. 2020). However, internal variability has played an important part in regulating the variability of sea ice on interannual to decadal time scales and can lead to an occasional ice-free Arctic (Kay et al. 2011; Notz and Marotzke 2012; Swart et al. 2015; Zhang 2015; England et al. 2019; Bonan et al. 2021b; Topál et al. 2020). The inability of the Coupled Model Intercomparison Project (CMIP) multimodel ensemble mean to capture the observed prominent rate of sea ice melting since the 2000s and the interdecadal variability of sea ice indicates a possible low sea ice sensitivity in climate models (Notz and Stroeve 2016; Rosenblum and Eisenman 2017) or other intrinsic process in climate system, part of which is the subject of our research.
Internal drivers of sea ice variability have been suggested to have both atmospheric (Zhang 2007; Deser and Teng 2008; Ogi et al. 2016; Wettstein and Deser 2014; Tokinaga et al. 2017) and oceanic origins (Lee 2012; Screen et al. 2012; Notz 2015; Swart et al. 2015; Grunseich and Wang 2016; Screen and Francis 2016; Ding et al. 2017; Li et al. 2017; Meehl et al. 2018; Wernli and Papritz 2018; Labe et al. 2018; Olonscheck et al. 2019; Screen and Deser 2019). Previous study has successfully identified a linkage between changes in the atmospheric circulation in summer and the following decrease in September sea ice extent (SSIE). A stronger barotropic anticyclonic circulation trend over the Arctic in the troposphere favors a warmer and moister lower troposphere through circulation-driven adiabatic descent, resulting in the thermal melting of sea ice via the emission of more downward longwave radiation (Ding et al. 2017, 2019; Baxter et al. 2019; Topál et al. 2020; Luo et al. 2021). This mechanism is thought to work over temporal scales from interannual to interdecadal time scales. Model experiments and a fingerprint pattern marching method have suggested that this summer circulation trend is primarily internally driven and that this internal variability may have contributed about 30%–50% to the melting of September sea ice during the last few decades (Ding et al. 2017, 2019). Baxter et al. (2019) suggested that this anomalous summer high pressure is not only a local process, but also a response to a remote driver of sea surface temperatures (SSTs) manifested as a Rossby wave train originating from the tropics to the Arctic triggered by a cold SST anomaly in the eastern tropical Pacific. In addition, the Atlantic multidecadal oscillation (AMO), the Atlantic meridional overturning circulation (AMOC), the Pacific decadal oscillation (PDO), and the interdecadal Pacific oscillation (IPO) have been considered as the major internal oceanic processes driving the variability in summer Arctic sea ice through their induced oceanic and atmospheric heat transport on decadal to centennial time scales (Chylek et al. 2009; Day et al. 2012; Yeager et al. 2015; Zhang 2015; Screen and Francis 2016; Castruccio et al. 2019; Screen and Deser 2019).
Internal variability can amplify or diminish the forced signal on decadal time scales and can regulate the occurrence of an ice-free Arctic by around two decades even with the same number of anthropogenic emissions (Kay et al. 2011; Jahn et al. 2016; Jahn 2018; Screen and Deser 2019; Lehner et al. 2020; Bonan et al. 2021b). Understanding the extent to which these internal drivers can influence the Arctic sea ice may help us to evaluate the influence of a changing Arctic environment and potentially reduce the uncertainty in the projections of climate models. Bonan et al. (2021a) estimated that 40%–60% of the total uncertainty in September sea ice area in the next decade can be attributed to internal variability. However, the causes of such internal variations and their relative contributions remains up for debate.
Topál et al. (2020) have demonstrated a robust and qualitative linkage between the internal variations of the atmosphere in June–August (JJA) and September sea ice loss during the next three decades across different models and under different external forcing scenarios. However, they only emphasized a qualitative understanding of the internal drivers of Arctic sea ice during the next three decades; the quantitative contribution of this atmospheric internal variation to projected September sea ice changes remains an unsettled issue.
Separating the internal variability from the original time series requires a precise estimation of the externally forced changes. Many studies have commonly utilized the multimodel ensemble mean of CMIP models as the forced signal. However, different CMIP models include different model assumptions and implementations of external forcing, which serve to conflate the internal variability with model variability (Frankcombe et al. 2015). Thus, in this paper we use the 40 members of the Community Earth System Model Large Ensemble (CESM-LE) (Kay et al. 2015) to avoid this problem. This is a large ensemble run by a single model in which all the simulations are forced by the same external forcing and differ only in their initial atmospheric conditions, therefore enabling us to better distinguish between external forcing and internal variability. In addition, Frankcombe et al. (2018) suggested that ensemble mean result in the same model is a better estimation of the externally forced signal. The spread among the simulations can be roughly treated as the range of different forms of internal variability imposed on the externally forced change in the climate system (Deser et al. 2012a,b; Swart et al. 2015; Jahn et al. 2016; Jahn 2018; Deser et al. 2020). We further use another 100-member large ensemble from the Max Planck Institute Earth System Model (MPI-ESM; Maher et al. 2019) to verify the results obtained from CESM-LE. In this study, we aim to figure out the quantitative contribution of internal atmospheric variability on uncertainty in Arctic sea ice projection by using these two large ensemble models. Through our analysis, we aimed to advance our understanding of the role that summer atmospheric circulation has played in the decadal Arctic sea ice projection in September.
2. Data and methods
a. Reanalysis and sea ice data
We used monthly atmospheric circulation variables during the time period 1979–2017 from the European Center for Medium-Range Weather Forecasts interim reanalysis dataset (ERA-Interim; Dee et al. 2011). The observed sea ice data were provided by the National Snow and Ice Data Center (NSIDC). We used the sea ice concentration (SIC) from CDR algorithm output, which is a combination of the NASA Team algorithm (Cavalieri et al. 1996) and the NASA Bootstrap algorithm (Comiso 2017). The sea ice extent (SIE) was estimated by summing the total area (km2) of Arctic grid cells with at least 15% SIC. The sea ice area (SIA) was defined as the sum of the product of the grid cell area and the corresponding SIC.
b. Model simulations
The 40 members of CESM-LE (Kay et al. 2015) were forced with identical historical greenhouse gases, aerosols, and other radiative forcing for the time period 1920–2005 (Lamarque et al. 2010). For the years 2006–2100, future emissions under the RCP8.5 scenario were applied, with radiative forcing reaching about 8.5 W m−2 at the end of this century (Meinshausen et al. 2011). Because all 40 members are under the same radiative forcing and differ only in their initial conditions, each member can be regarded as an independent realization of the possible climate system, and the difference between the members is caused by the internal variability due to random fluctuations in the initial conditions applied to the different members. The 40-member ensemble mean can be regarded as the response of the model to the cumulative effects of anthropogenic plus other external forcing. The preindustrial (PI) 1850 control run (Kay et al. 2015) in CESM-CAM5 (Hurrell et al. 2013) is also used to examine the atmosphere–sea ice linkage without anthropogenic emissions. Because the PI runs are forced by constant external forcing, the pseudoensemble members can thus be regarded as having been generated purely by internal climate physics of the models. However, it should be noted that the pseudoensemble members were not initialized with perturbed initial conditions and there were overlapping periods between members. Therefore, in its strictest sense, they cannot represent the full scope of the possible realization of internal variability.
CESM-LE is chosen in this study to investigate the uncertainty in the near-term projection of September sea ice related to internal variability due to its most realistic simulations of the spatial pattern and magnitude of sea ice internal variability compared with other models (Kay et al. 2011; Jahn et al. 2016; Labe et al. 2018; England et al. 2019). The observed climatological September SIC and the negative correlation between the JJA 200-hPa geopotential height (Z200) and SSIE are well reproduced in CESM-LE (Figs. 1a,b), although the atmosphere–sea ice coupling is weaker in CESM-LE than that in observations (Figs. 1c,d). However, when compared with other large ensemble models, CESM-LE produced the most realistic pattern and magnitude of the linkage between the atmospheric circulation and sea ice (Topál et al. 2020). The reason why we choose CESM-LE rather than CESM2-LE is because we conducted this research before the CESM2-LE became available in summer of 2021. In addition, CESM1 has a better performance in simulating SIC than CESM2 as identified by DeRepentigny et al. (2020) and Shen et al. (2021). To help confirm that the results derived from the CESM-LE are model independent, the output from another large ensemble simulation carried out by MPI-ESM (Maher et al. 2019) was also used. This is a low-resolution configuration updated from the coupled oceanic and atmospheric general circulation model submitted to CMIP5 and is a 100-member ensemble integrated from 1850 to 2005 with the observed natural and anthropogenic forcings and performed with RCP8.5 forcings from 2006 to 2099.
Evaluation of CESM-LE in simulating climatological September sea ice concentration and the correlation of JJA Z200 with September SIE index. (a) NSIDC_CDR and (c) CESM-LE ensemble mean climatological September sea ice concentration over the Arctic. Also shown is regressed detrended JJA Z200 with respect to the detrended September SIE during 1979–2017 for (b) NSIDC_CDR and (d) the CESM-LE ensemble mean. Vectors are plotted when regressions are statistically significant at the 5% level.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Evaluation of CESM-LE in simulating climatological September sea ice concentration and the correlation of JJA Z200 with September SIE index. (a) NSIDC_CDR and (c) CESM-LE ensemble mean climatological September sea ice concentration over the Arctic. Also shown is regressed detrended JJA Z200 with respect to the detrended September SIE during 1979–2017 for (b) NSIDC_CDR and (d) the CESM-LE ensemble mean. Vectors are plotted when regressions are statistically significant at the 5% level.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Evaluation of CESM-LE in simulating climatological September sea ice concentration and the correlation of JJA Z200 with September SIE index. (a) NSIDC_CDR and (c) CESM-LE ensemble mean climatological September sea ice concentration over the Arctic. Also shown is regressed detrended JJA Z200 with respect to the detrended September SIE during 1979–2017 for (b) NSIDC_CDR and (d) the CESM-LE ensemble mean. Vectors are plotted when regressions are statistically significant at the 5% level.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
c. Identifying the internal mode related to the projection uncertainty
Two methods were used to search for the internal mode related to the uncertainty in the projection of sea ice in CESM-LE. First, a simple but very efficient method is implemented to tease apart internal variability from the forced component. Because we aim to identify an internal atmospheric process responsible for the summer sea ice variability as demonstrated in Ding et al. (2019), we focused on the difference between the members of the ensemble that we can select a certain number of extreme members of the ensemble with slow SSIE trends and a certain number of extreme members with fast SSIE trends respectively. Since all the ensemble members are simulated with the same external forcing, the difference in the linear trends of the JJA atmospheric circulation between the corresponding slow and fast groups indicated the leading internal mode in the atmosphere responsible for the sea ice projection uncertainty. The selection criterion of the number of ensemble members belonging to the fast and slow groups was the 10th percentile of the total number of ensemble members (CESM-LE: 4 members; MPI-ESM: 10 members).
Second, given the subjectivity in the selecting the number of fast and slow ensemble members in the above method, following Deser et al. (2017), an empirical orthogonal function (EOF) method is applied to the trends in Arctic SIC in September during the time period 2016–45 across the ensemble members. That is, the time series in a conventional EOF was replaced by the member index. The leading principal mode of the EOF represents the major uncertainties in the projection of sea ice among the ensemble members. The regression of the atmospheric circulation trend derived from different ensemble members on the standardized corresponding leading principal component (PC1) through the member index denotes the internal variability in the atmosphere responsible for the uncertainty in the projected sea ice change in September.
d. Cross-validation method
To examine whether the uncertainty in the near-term SSIE projection can be reduced with a more accurately projected JJA Z200, we, chose one of the 40 ensemble members, member e, as the “pseudo reality” with a JJA Z200 trend of
The selected members in group A met the criterion that their simulated JJA Z200 trends fell within the range of
e. Quantifying the contribution of the internal JJA atmospheric circulation to the SSIE trend projection
In each ensemble member, deviations from the ensemble mean represent the different realization of internal variability imposed on the forced component.
f. Index trends
The chance of an ice-free Arctic (COIA) index was used to denote the probability that the Arctic will reach an ice-free condition (defined as the SSIE less than 1 × 106 km2) during the next 30 years (by 2045). It was obtained by dividing the number of ensemble members that showed SSIE trends (106 km2 per 30 years) of less than a certain value f(i) during the time period 2016–45 via the total ensemble member. The value f(i) was calculated as SSIE2016 − 1, where SSIE2016 is the modeled SSIE value in 2016 in each member i (106 km2) and 1 (106 km2) is the ice-free condition. If the absolute value of the SSIE trend in member i surpasses the f(i), then member i is counted in the ice-free member.
3. Results
a. Projected September sea ice change during the next few decades
The externally forced changes in September sea ice projected by the CESM-LE ensemble mean show a decreasing trend over the whole Arctic region during the time period 2016–45, rather than only in the marginal seas, as seen during the observational period. The most pronounced decreasing trend in the SIC occurs over the Beaufort Sea with a value of 27% per decade (P < 0.01) under the RCP8.5 emission scenario (Fig. 2a). However, the large standard deviation (SD) across the 40 members over the Beaufort Sea, East Siberian Sea, and Laptev Sea illustrates the large uncertainty in the projected SIC trend (Fig. 2b). Although driven by the same external forcing, the 40 ensemble members do exhibit very different magnitudes of SIC changes, as illustrated by the sharp contrast between the composites of the four members with the fastest SIC trends (hereinafter Fast4; Fig. 2c) and the four members with the slowest trends (hereinafter Slow4; Fig. 2d). Although substantial sea ice loss occurs both in Fast4 and Slow4, it tends to be smaller in magnitude and centered mainly around the Canadian Archipelago in Slow4 compared with Fast4. The SSIE trends among the 40 members over the time period 2016–45 all show negative trends, which indicates an unquestionable decrease in SSIE in the next few decades; however, the spread among the ensemble members ranges from −2.0 × 106 to −0.6 × 106 km2 decade−1, with some realizations projecting a much faster sea ice loss and some showing a slower loss, which suggests that internal variability can augment or diminish the externally forced changes in SSIE in any individual model realization (Fig. 3a).
Arctic sea ice concentration changes under the RCP8.5 scenario. September Arctic sea ice concentration trends (106 km2 decade−1) during 2016–45 for (a) the CESM1-CAM5 40-member ensemble mean, (b) the intermember SD, (c) the composite of the four members with the fastest SSIE trends, and (d) the composite of the four members with the slowest trends. The area with the largest uncertainty in sea ice trend is highlighted in blue box (73°–85°N, 120°E–130°W).
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Arctic sea ice concentration changes under the RCP8.5 scenario. September Arctic sea ice concentration trends (106 km2 decade−1) during 2016–45 for (a) the CESM1-CAM5 40-member ensemble mean, (b) the intermember SD, (c) the composite of the four members with the fastest SSIE trends, and (d) the composite of the four members with the slowest trends. The area with the largest uncertainty in sea ice trend is highlighted in blue box (73°–85°N, 120°E–130°W).
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Arctic sea ice concentration changes under the RCP8.5 scenario. September Arctic sea ice concentration trends (106 km2 decade−1) during 2016–45 for (a) the CESM1-CAM5 40-member ensemble mean, (b) the intermember SD, (c) the composite of the four members with the fastest SSIE trends, and (d) the composite of the four members with the slowest trends. The area with the largest uncertainty in sea ice trend is highlighted in blue box (73°–85°N, 120°E–130°W).
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
(a) Time series of September sea ice extent from 1920 to 2100. The 40-member ensemble mean (solid lines), one SD across 40 members (shading), and the minimum and maximum value of the 40 members (dashed lines) for historical simulations (gray) and RCP8.5 projection (red). The green line represents the observed SSIE derived from NSIDC CDR dataset. The composite of 4 members with the fastest and slowest SSIE trends are shown in blue and orange solid lines, respectively. The insert figure in (a) is the histogram of the SSIE trends projected by the 40 members for the period of 2016–45 (106 km2 decade−1). The vertical gray, blue, and red dashed lines represent the 40-member ensemble mean, the four fastest SSIE trend members, and the four slowest SSIE trend members, respectively. (b) Scatterplot of the original SIE trends (x axis; 106 km2 decade−1) and the internal component of the SIE trends (y axis; 106 km2 decade−1) derived from CESM-LE members (number indicated beside circle) during the period 1979–2014 (pink) and 2016–45 (green). Dark pink and dark green circles in (b) denote the four members with the strongest positive Z200 trend during 1979–2014 and 2016–45, respectively.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
(a) Time series of September sea ice extent from 1920 to 2100. The 40-member ensemble mean (solid lines), one SD across 40 members (shading), and the minimum and maximum value of the 40 members (dashed lines) for historical simulations (gray) and RCP8.5 projection (red). The green line represents the observed SSIE derived from NSIDC CDR dataset. The composite of 4 members with the fastest and slowest SSIE trends are shown in blue and orange solid lines, respectively. The insert figure in (a) is the histogram of the SSIE trends projected by the 40 members for the period of 2016–45 (106 km2 decade−1). The vertical gray, blue, and red dashed lines represent the 40-member ensemble mean, the four fastest SSIE trend members, and the four slowest SSIE trend members, respectively. (b) Scatterplot of the original SIE trends (x axis; 106 km2 decade−1) and the internal component of the SIE trends (y axis; 106 km2 decade−1) derived from CESM-LE members (number indicated beside circle) during the period 1979–2014 (pink) and 2016–45 (green). Dark pink and dark green circles in (b) denote the four members with the strongest positive Z200 trend during 1979–2014 and 2016–45, respectively.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
(a) Time series of September sea ice extent from 1920 to 2100. The 40-member ensemble mean (solid lines), one SD across 40 members (shading), and the minimum and maximum value of the 40 members (dashed lines) for historical simulations (gray) and RCP8.5 projection (red). The green line represents the observed SSIE derived from NSIDC CDR dataset. The composite of 4 members with the fastest and slowest SSIE trends are shown in blue and orange solid lines, respectively. The insert figure in (a) is the histogram of the SSIE trends projected by the 40 members for the period of 2016–45 (106 km2 decade−1). The vertical gray, blue, and red dashed lines represent the 40-member ensemble mean, the four fastest SSIE trend members, and the four slowest SSIE trend members, respectively. (b) Scatterplot of the original SIE trends (x axis; 106 km2 decade−1) and the internal component of the SIE trends (y axis; 106 km2 decade−1) derived from CESM-LE members (number indicated beside circle) during the period 1979–2014 (pink) and 2016–45 (green). Dark pink and dark green circles in (b) denote the four members with the strongest positive Z200 trend during 1979–2014 and 2016–45, respectively.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
To investigate the relative importance of internal variability to changes in the SSIE during both the past and future periods, we separated the internal part of the change in SSIE from the original SSIE time series for the periods 1979–2014 and 2016–45, respectively, based on Eq. (1). The spread in the SSIE trend caused by internal variability only can range from −0.42 × 106 to 0.3 × 106 km2 decade−1 during the time period 1979–2014. However, with stronger external forcing, the spread during the time period 2016–45 is much larger, ranging from −0.6 × 106 to 0.87 × 106 km2 decade−1, which is around double that during 1979–2014 (Fig. 3b). This larger spread of internal variability among the individual realizations during the time period 2016–45 implies considerable potential for the internal variability to modulate the decrease in sea ice decline during the next three decades, leading to a large uncertainty in the year in which the Arctic will be ice free.
b. Physical linkage between September sea ice and the preceding JJA atmospheric circulation during 2016–45
Observational records have shown that, during the past three decades, the internal variability of September sea ice in the Arctic has been primarily driven by the rising geopotential height in JJA. This has a barotropic structure that favors the sea ice melting through circulation-driven adiabatic descent that warms and moistens the lower troposphere (Ding et al. 2014, 2017, 2019; Baxter et al. 2019). This summertime Arctic high pressure is suggested to be dominated by the internal variability and can explain 30%–50% of the observed multidecadal decrease in Arctic sea ice. Whether this identified atmospheric–sea ice mutual effect will still persist in future was investigated by Topál et al. (2020); however, the quantitative contribution of this internal atmospheric circulation to the September sea ice change has so far not been addressed.
The major atmospheric climate model connected with the uncertainty in the projected Arctic SIC trend during the period 2016–45 is illustrated by a comparison of the difference in the trend in JJA zonal mean geopotential height between Fast4 and Slow4 (Fig. 4a). It is demonstrated as a tendency toward high pressure with an equivalent barotropic high pressure over the Arctic during JJA (Fig. 4a). The absence of this strong barotropic high pressure trend in the ensemble mean of 40 members, which denotes the response to external forcing, indicates an internal source for this circulation trend (Fig. 4b).
Trends in JJA large-scale circulation associated with September SIE trend spread under the RCP8.5 scenario during the period 2016–45: (a) fast-minus-slow difference and (b) CESM-LE ensemble mean of JJA zonal mean height.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Trends in JJA large-scale circulation associated with September SIE trend spread under the RCP8.5 scenario during the period 2016–45: (a) fast-minus-slow difference and (b) CESM-LE ensemble mean of JJA zonal mean height.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Trends in JJA large-scale circulation associated with September SIE trend spread under the RCP8.5 scenario during the period 2016–45: (a) fast-minus-slow difference and (b) CESM-LE ensemble mean of JJA zonal mean height.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
An EOF analysis conducted on the intermember spread further confirms that this internal atmospheric mode is responsible for the spread in the projected change in SSIE (Fig. 5a). The results showed that the first leading principal component (PC1) accounts for 37.1% of the intermember variance in SIC. EOF1, which represents the main pattern of change in sea ice and is mostly influenced by the internal variability, shows the most significant change in SIC over the Beaufort Sea, East Siberian Sea, and Laptev Sea—the same as in Fig. 2b. By regressing the changes in the 200- and 700-hPa geopotential height (denoted as Z200 and Z700) during the time period 2016–45 onto PC1 across members, we found that the first EOF mode is linked to a tendency toward high pressure with an equivalent barotropic structure in the troposphere during JJA over the Arctic, with higher Arctic pressure favoring accelerated sea ice loss (Figs. 5b,c). As a result of the strong coupling between the circulation and vertical velocity, the anticyclonic circulation aloft can force adiabatic descent, contributing to warming of the lower atmosphere, which, in turn, accelerates the loss of sea ice (Fig. 5d).
Projected leading uncertainty modes and related large-scale circulation patterns. (a) The first leading mode derived from intermember empirical orthogonal function (EOF) analysis on projected sea ice concentration trends under RCP8.5 during the period 2016–45. The value in the top-right corner is explained intermodel variance by corresponding mode. Member spread patterns of (b) JJA 7200 and (c) JJA Z200 trend during 2016–45 associated with PC1. (d) Correlation of JJA Arctic surface air temperature trend with JJA Z200 trend across the CESM1-CAM5 40 members. Stippling indicates statistical significance at the 5% level under Student’s t test.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Projected leading uncertainty modes and related large-scale circulation patterns. (a) The first leading mode derived from intermember empirical orthogonal function (EOF) analysis on projected sea ice concentration trends under RCP8.5 during the period 2016–45. The value in the top-right corner is explained intermodel variance by corresponding mode. Member spread patterns of (b) JJA 7200 and (c) JJA Z200 trend during 2016–45 associated with PC1. (d) Correlation of JJA Arctic surface air temperature trend with JJA Z200 trend across the CESM1-CAM5 40 members. Stippling indicates statistical significance at the 5% level under Student’s t test.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Projected leading uncertainty modes and related large-scale circulation patterns. (a) The first leading mode derived from intermember empirical orthogonal function (EOF) analysis on projected sea ice concentration trends under RCP8.5 during the period 2016–45. The value in the top-right corner is explained intermodel variance by corresponding mode. Member spread patterns of (b) JJA 7200 and (c) JJA Z200 trend during 2016–45 associated with PC1. (d) Correlation of JJA Arctic surface air temperature trend with JJA Z200 trend across the CESM1-CAM5 40 members. Stippling indicates statistical significance at the 5% level under Student’s t test.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
The 40 ensemble members do indeed yield different JJA atmospheric circulations during the time period 2016–45, as revealed by the diverse trends in the JJA Z200 index (Fig. 6), which was calculated as the area-averaged Z200 over the Arctic. The projected SSIE trends are negatively correlated with the Z200 index trends across 40 realizations (r = −0.51, P < 0.01; Fig. 6). In addition, evidence for this atmosphere-related change in sea ice can also be found in the preindustrial control simulations, which represents the climate in the absence of anthropogenic forcing. We created a total of 290 pseudoensemble members using all consecutive 30-yr periods of the SSIE and JJA Z200 index from the 319-yr integration (Coats et al. 2013; Rosenblum and Eisenman 2017; Ding et al. 2019; Luo et al. 2021). Figure 6 shows that the relationship between the change in Z200 and the rate of decrease in sea ice (PI, r = −0.6) is similar to that derived from the all-forcing simulations. This negative correlation and the associated physical process are consistent with the findings of previous studies that focused on the atmospheric circulation and sea ice linkage during the last and next three decades (Ding et al. 2017, 2019; Topál et al. 2020; Luo et al. 2021). We note that choice of either SIE or SIA to represent the sea ice cover does not affect our conclusions (see Fig. S1 in the online supplemental material).
Trends in JJA Arctic Z200 index against trends of SSIE during 2016–45 in 40-Forced. Hollow circles denote trends for each consecutive 30-yr period from the PI run in CESM-CAM5. Solid colored circles denote trends (2016–45) from 40 members in CESM-LE.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Trends in JJA Arctic Z200 index against trends of SSIE during 2016–45 in 40-Forced. Hollow circles denote trends for each consecutive 30-yr period from the PI run in CESM-CAM5. Solid colored circles denote trends (2016–45) from 40 members in CESM-LE.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Trends in JJA Arctic Z200 index against trends of SSIE during 2016–45 in 40-Forced. Hollow circles denote trends for each consecutive 30-yr period from the PI run in CESM-CAM5. Solid colored circles denote trends (2016–45) from 40 members in CESM-LE.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
c. Quantifying the contribution of JJA Z200 to the SSIE projection uncertainty
So far, based on the previous studies, we have reached a qualitative understanding of the internal atmospheric driver of sea ice loss on multidecadal time scales, but its quantitative contribution has not yet been addressed.
We identify that the near-term SSIE projection uncertainty can indeed be reduced with the improved representation of JJA Z200 in models based on a cross-validation method (see section 2). In which, we chose one of the 40 ensemble members as a “pseudo reality” in turn and in each case, the remaining 39 members are grouped based on their simulated JJA Z200 trends during the time period 2016–45. The selected members from groups A to C have increasing errors in their simulated JJA Z200 trends. Compared with the SD of the SSIE trends of the full ensemble, the average SDs of SSIE trends from groups A to C were reduced by around 20%, 14%, and 2%, respectively (Fig. 7).
The contribution of JJA Z200 to the September SIE future projection. The SD distribution in 40 cases of different groups. The blue dots represent the September SIE SD in group A for each case that has Z200 trends during the period 2016–45 within the range of
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
The contribution of JJA Z200 to the September SIE future projection. The SD distribution in 40 cases of different groups. The blue dots represent the September SIE SD in group A for each case that has Z200 trends during the period 2016–45 within the range of
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
The contribution of JJA Z200 to the September SIE future projection. The SD distribution in 40 cases of different groups. The blue dots represent the September SIE SD in group A for each case that has Z200 trends during the period 2016–45 within the range of
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
To further quantify the extent to which the SSIE projection uncertainty can be explained by the change in JJA Z200internal from 2016 to 2045, we removed the SSIE variations that are linearly dependent on the Z200internal index in each ensemble member to exclude all of the influence of Z200internal [see section 2 and Eq. (2)]. This method is based on two assumptions: first, the JJA Z200 trends can be used as the fingerprints of the internal and anthropogenic drivers of SSIE changes and the linear additivity of internal and external components; second, the atmosphere–sea ice coupling is similar on both interannual and multidecadal time scales.
The similar regression slopes between the SSIE and JJA Z200 trends in 40-Forced and PI (Fig. 6) confirms the validity of the first assumption, although there might be intricate mutual effects between internal and anthropogenic component that are not fully captured by the models. The second assumption is justified by a comparison of Figs. 5 and 8, which shows that the spatial patterns of the atmosphere–sea ice linkage on multidecadal time scales (Fig. 5) strongly resemble those shown in the year-to-year time scales [Fig. 8; Topál et al. 2020; Luo et al. 2021)].
Linear detrended correlation of JJA (a) Z200, (b) zonal mean geopotential height, and (c) temperature with SSIE for 2016–45. Correlations are calculated for each member individually and then averaged out.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Linear detrended correlation of JJA (a) Z200, (b) zonal mean geopotential height, and (c) temperature with SSIE for 2016–45. Correlations are calculated for each member individually and then averaged out.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Linear detrended correlation of JJA (a) Z200, (b) zonal mean geopotential height, and (c) temperature with SSIE for 2016–45. Correlations are calculated for each member individually and then averaged out.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
We then compared the spread of the SSIE(i, t) and SSIEnon-Z200(i, t) trends among the 40 realizations to quantify the contribution of the change in JJA Z200internal to the SSIE trend during the time period 2016–45. The histograms and the box-and-whisker plots suggest that the uncertainty in the SSIE trends narrow after removing the influence of JJA Z200internal, with the SD and 5th and 95th percentile range of the projected SSIE trends among the 40 CESM-LE members reduced by about 12.3% and 16.5%, respectively (Fig. 9a).
Histograms for the SSIE trend during 2016–45 with and without Z200 influence. (a) Histogram (bars) and Gaussian distribution (lines) of SSIE derived from 40 CESM1-CAM5 ensemble members. The gray bars denote number of ensemble members with different SSIE trends from original 40 members. The pink bars denote the number of ensemble members with different SSIE trends without the influence from the internal variability of Z200. The gray and pink dots and horizontal lines represent ensemble mean and one SD range of the 40 members, respectively, which are represented by the corresponding color. (b) The gray bars and gray curves are the same as in (a). The blue and orange bars and curves represent the number of ensemble members with different SSIE trends influenced with the same amplitude of +2 SD and −2 SD of Z200internal, respectively. The gray, blue, and orange dots and horizontal lines have the same meaning as in (a). The COES (COEF) index is defined as the chances of extreme slow (fast) future SSIE loss and is obtained by dividing the number of ensemble members with adjusted SSIE trends greater (less) than or equal to the 10th (90th) percentile of the original SSIE trends by the total number of ensemble members.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Histograms for the SSIE trend during 2016–45 with and without Z200 influence. (a) Histogram (bars) and Gaussian distribution (lines) of SSIE derived from 40 CESM1-CAM5 ensemble members. The gray bars denote number of ensemble members with different SSIE trends from original 40 members. The pink bars denote the number of ensemble members with different SSIE trends without the influence from the internal variability of Z200. The gray and pink dots and horizontal lines represent ensemble mean and one SD range of the 40 members, respectively, which are represented by the corresponding color. (b) The gray bars and gray curves are the same as in (a). The blue and orange bars and curves represent the number of ensemble members with different SSIE trends influenced with the same amplitude of +2 SD and −2 SD of Z200internal, respectively. The gray, blue, and orange dots and horizontal lines have the same meaning as in (a). The COES (COEF) index is defined as the chances of extreme slow (fast) future SSIE loss and is obtained by dividing the number of ensemble members with adjusted SSIE trends greater (less) than or equal to the 10th (90th) percentile of the original SSIE trends by the total number of ensemble members.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Histograms for the SSIE trend during 2016–45 with and without Z200 influence. (a) Histogram (bars) and Gaussian distribution (lines) of SSIE derived from 40 CESM1-CAM5 ensemble members. The gray bars denote number of ensemble members with different SSIE trends from original 40 members. The pink bars denote the number of ensemble members with different SSIE trends without the influence from the internal variability of Z200. The gray and pink dots and horizontal lines represent ensemble mean and one SD range of the 40 members, respectively, which are represented by the corresponding color. (b) The gray bars and gray curves are the same as in (a). The blue and orange bars and curves represent the number of ensemble members with different SSIE trends influenced with the same amplitude of +2 SD and −2 SD of Z200internal, respectively. The gray, blue, and orange dots and horizontal lines have the same meaning as in (a). The COES (COEF) index is defined as the chances of extreme slow (fast) future SSIE loss and is obtained by dividing the number of ensemble members with adjusted SSIE trends greater (less) than or equal to the 10th (90th) percentile of the original SSIE trends by the total number of ensemble members.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
To demonstrate how the magnitude of the change in JJA Z200internal will influence the SSIE trends, the near-future JJA Z200internal was assumed to be known and then a so-called Z200internal constraint was used to adjust the projected SSIE trend [see section 2 and Eq. (3)]. After removing the influence of Z200internal on SSIE in each realization through linear regression, a fixed Z200internal trend was added back. The trend of standardized Z200internal index during a phase transition from +1 (−1) SD to −1 (+1) SD from 2016 to 2045 was used for a significant negative-to-positive (positive-to-negative) Z200internal shift.
After this process, the adjusted SSIE trends were influenced by the same change in JJA Z200internal and external forcing, together with other internal variability responsible for the decadal variation of SSIE. Figure 9b shows that if a JJA Z200internal with an amplitude of +2 SD (−2 SD) during the time period 2016–45 is predicted to be superimposed onto the anthropogenic forcing under the RCP8.5 scenario, then the average rate of SSIE loss among the 40 CESM-LE members will be −5.08 × 106 km2 (30 yr)−1 [−3.65 × 106 km2 (30 yr)−1], which is 0.73 × 106 km2 (30 yr)−1 [0.71 × 106 km2 (30 yr)−1] faster (slower) than trend purely driven by the external forcing. This might result in the COIA (see section 2) under RCP8.5 at the end of 2045 increasing (decreasing) from the original 7.5% to 30% (0%). The possible occurrence of an extreme fast sea ice loss will be impossible if the JJA Z200internal shows a decreasing trend. These calculations demonstrate that the near-future JJA Z200internal trend will probably have a notable impact on the projected melting of SSIE.
The robustness of the above results is further demonstrated by the 100-member grand ensemble of MPI-ESM (Fig. 10). The most obvious difference is between the externally forced SSIE trends derived in these two models. However, the estimated contribution of the Z200internal to the projection uncertainty in SSIE during the next three decades is model-independent (Figs. 9 and 10). After excluding the Z200internal-related linear variation in the SSIE, the uncertainty in the projected SSIE trends (SD) was reduced by around 12.3% (the same as CESM-LE) from 0.69 to 0.57. The 5th and 95th percentile range of the projected SSIE trends is also narrowed by about 16.7%.
Uncertainty in the projection of SSIE trends under RCP8.5 emission scenario in MPI-ESM. (a) As in Fig. 3a, but derived from the 100 MPI-ESM members. (b),(c) As in Fig. 9, but derived from the 100 MPI-ESM members.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Uncertainty in the projection of SSIE trends under RCP8.5 emission scenario in MPI-ESM. (a) As in Fig. 3a, but derived from the 100 MPI-ESM members. (b),(c) As in Fig. 9, but derived from the 100 MPI-ESM members.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
Uncertainty in the projection of SSIE trends under RCP8.5 emission scenario in MPI-ESM. (a) As in Fig. 3a, but derived from the 100 MPI-ESM members. (b),(c) As in Fig. 9, but derived from the 100 MPI-ESM members.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
The chances of an ice-free Arctic by 2045 will increase by around 9% in MPI-ESM if the Z200internal trend is projected with an amplitude of +2 SD (30 yr)−1 during the next three decades, however, in CESM-LE, the chances will increase by about 22.5%, which may be caused by the stronger atmosphere–sea ice coupling or large sea ice sensitivity in the CESM-LE [Topál et al. (2020)]. If we use the SIA to substitute for SIE in CESM-LE (Fig. S2), the uncertainty in the projected SIA trends can be reduced by 17.4%, slightly larger than that derived from SIE. Despite these slight quantitative differences, the future change in sea ice can, to some extent, be attributed to the internal atmospheric circulation. In all cases, a different Z200internal phase transition can enhance or weaken the magnitude of the sea ice melting and affect the occurrence of an ice-free Arctic in the next three decades (Figs. 9 and 10; see also Fig. S2).
d. Teleconnection between the tropical JJA SST and SSIE
Although the JJA Z200internal-related uncertainty can account for about 10%–20% of the projected uncertainty in SSIE during the next three decades, the remaining large projection uncertainties suggest that other internal drivers responsible for the decadal variation of SSIE cannot be ignored. Comparing Figs. 1c and 1d, the modeled SSIE and geopotential height connection primarily locates in the Arctic Ocean, which is in contrast with the observed linkage occupying a broad region extending from the interior of the polar region to Greenland. This suggests that the observed atmosphere–sea ice coupling is possibly not just a local process, but also has teleconnections with remote regions. Different lines of evidence suggest that Arctic sea ice variations in the past several decades can have tropical origins (Lee et al. 2011; Screen et al. 2012; Ding et al. 2014; Trenberth et al. 2014; Screen and Francis 2016; Meehl et al. 2018; Blanchard-Wrigglesworth and Ding 2019; Baxter et al. 2019; Blanchard-Wrigglesworth and Ding 2019; Screen and Deser 2019; Bonan et al. 2020; Topál et al. 2020). Previous studies have pointed out that El Niño–like SSTs in the tropics, which can trigger a large-scale Rossby wave train with a barotropic structure propagating from the tropical Pacific Ocean to high latitudes, are connected with high pressure over the Arctic. Model experiments have shown that about 50% of the change in the circulation and the accompanying warming trend over the Arctic can be attributed to the internal variability originating from the tropical Pacific Ocean (Hoerling et al. 2001; Lee 2012; Ding et al. 2014; Trenberth et al. 2014; Wettstein and Deser 2014). However, there are discrepancies between models on the specific details of how changes in the tropical SST might influence the future loss of sea ice, which can be attributed either to the model physics or internal variability (Bonan et al. 2020; Topál et al. 2020).
In CESM-LE, we found that the correlations between the JJA SST in the tropics and SSIE are opposite on year-to-year and multidecadal time scales and that no clear signal can be found in the tropical Pacific Ocean on year-to-year time scales in either CESM-LE and MPI-ESM (Figs. 11a,c), which suggest that the sea ice–related JJA geopotential height rise on year-to-year time scales over the Arctic might only be a local process with no teleconnection signals from the tropics. In addition, Blanchard-Wrigglesworth and Ding (2019) analyzes the tropical SST–sea ice linkages during 1967–2005 in CESM-LE and notes that this observed teleconnection might primarily be caused by internal variability. By contrast, on multidecadal time scales, the MPI-ESM shows a significant negative correlation between the tropical SST and sea ice, with a warm Pacific resulting in an accelerated rate of sea ice loss (Fig. 11d), although the related atmospheric pattern seems to act as a stationary wave and the observed Rossby wave train propagating into the high latitudes is missing in the model (Ding et al. 2014; Baxter et al. 2019; Topál et al. 2020).
(a),(b) Correlation of linearly detrended JJA SST and Z200 with detrended SSIE index during 2016–45 and (c),(d) correlation of JJA SST trend and Z200 trend with SSIE trend across members for 2016–45 in (top) CESM-LE and (bottom) MPI-ESM. Stippling indicates statistical significance at the 10% level under Student’s t test. Contours are plotted when correlation coefficients are statistically significant at the 10% level.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
(a),(b) Correlation of linearly detrended JJA SST and Z200 with detrended SSIE index during 2016–45 and (c),(d) correlation of JJA SST trend and Z200 trend with SSIE trend across members for 2016–45 in (top) CESM-LE and (bottom) MPI-ESM. Stippling indicates statistical significance at the 10% level under Student’s t test. Contours are plotted when correlation coefficients are statistically significant at the 10% level.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
(a),(b) Correlation of linearly detrended JJA SST and Z200 with detrended SSIE index during 2016–45 and (c),(d) correlation of JJA SST trend and Z200 trend with SSIE trend across members for 2016–45 in (top) CESM-LE and (bottom) MPI-ESM. Stippling indicates statistical significance at the 10% level under Student’s t test. Contours are plotted when correlation coefficients are statistically significant at the 10% level.
Citation: Journal of Climate 35, 11; 10.1175/JCLI-D-21-0579.1
In CESM-LE, an IPO-like phase change might be linked with the interdecadal variability of the SSIE (Fig. 11b), with the IPO shifting from its cold to warm phase accelerating the sea ice loss, and vice versa. Screen and Deser (2019) proposed that the IPO phase change could be used as a precursor to the seasonal sea ice conditions. However, the relationship is weak in CESM-LE.
We note that atmosphere–sea ice coupling is similar on different time scales (Figs. 5 and 8), but that the tropical SST and sea ice teleconnection is different between year-to-year and multidecadal time scales. The influence of the tropics is dependent on the Pacific basin and model, even on the same time scale (Fig. 11). Although the models cannot reproduce the observed Rossby wave train originating from the tropics to the Arctic, the local Arctic geopotential height anomaly still exist on both year-to-year and multidecadal time scales. However, it should be noted that the atmosphere–sea ice coupling is weaker on multidecadal time scales than that on year-to-year time scales in both CESM-LE and MPI-ESM. Combining with the SST teleconnection pattern, we speculate that this indicates that the change in the atmospheric circulation on year-to-year time scales is more dependent on local feedbacks than the change on multidecadal time scales—that is, the influence of the tropics on the Arctic is more prominent on decadal scales. The weaker atmosphere–sea ice coupling on multidecadal time scales might therefore be related to the deficiencies in current climate models in capturing the key processes connecting the tropics and the Arctic. Thus, we propose further research dedicated to elucidating the relative role of local and remote forcing mechanisms have played in the atmosphere–sea ice coupling over the Arctic. We also need to untangle which time scales and which side of the tropical Pacific Ocean have the most significant contribution to the Arctic sea ice variability.
4. Conclusions and discussion
As previous research has mainly focused on the qualitative understanding of the internal drivers of Arctic sea ice during the next three decades (Topál et al. 2020), the quantitative estimation of the influence of the internal atmospheric circulation on the uncertainty in future variability of SSIE during the next three decades remains an unsettled issue. Ding et al. (2017) have proposed that the tendency toward an equivalent barotropic higher pressure in the troposphere over the Arctic in JJA warms and moistens the surface air above the sea ice through adiabatic descending motion. This regulates longwave radiation and accelerates the loss of sea ice. By analyzing large ensemble projections based on CESM-LE and MPI-ESM and applying the EOF method to the ensemble members of the same model, we present further evidence and demonstrate that the similar internal atmospheric process driving the melting of sea ice during the time period 2016–45 in two large ensemble models, which has also been demonstrated by Topál et al. (2020). However, Topál et al. (2020) mainly used the fast-minus-slow composite method to demonstrate internal driver for sea ice melting, which might have limitations as a result of the subjective choice of fast and slow ensemble members. Our study applies a more objective EOF method to further verify this mechanism, acting as a supplement to their results.
We found that, after removing the SSIE variations that were linearly related to the internal variability of the Z200 index in each realization, the uncertainty in the near-term projection of SSIE decreased by about 12.3%–16.5% in CESM-LE and 12.3%–16.7% in MPI-ESM. Further, using the Z200internal of two reverse phases to constrain the future SSIE trends, we demonstrated that this internal atmospheric change can enhance or weaken the magnitude of the sea ice melting and affect the occurrence of an ice-free Arctic in the next three decades. However, the quantitative estimation of the contribution of this identified internal atmospheric change to decadal sea ice variations might be underestimated. This is because the models are unable to realistically reproduce the observed magnitude of coupling between changes in atmospheric circulation and sea ice (Baxter et al. 2019; Topál et al. 2020; Luo et al. 2021), and because our method only reflects the linear contribution of the internal atmospheric circulation to sea ice without considering the nonlinear, higher-order process. The results presented here can therefore be regarded as the lower bound of the influence of the JJA Z200 related internal variability on sea ice in September. Moreover, the contribution of atmospheric circulation changes to sea ice variability under different trajectories of anthropogenic emissions remains up to debate and need further investigation.
Although the circulation-related uncertainty plays a nonnegligible role in the near-future projections of SSIE, the remaining large projection uncertainties suggest that other internal variabilities can also play an important part. More recent studies have focused on the impact of the internal SST variability in the tropical Pacific Ocean on the Arctic (Lee et al. 2011; Screen et al. 2012; Ding et al. 2014; Baxter et al. 2019; Luo et al. 2021; Screen and Deser 2019; Topál et al. 2020). However, unlike the similar atmosphere–sea ice coupling on year-to-year and multidecadal time scales, the tropical SST and sea ice interaction varies significantly on different time scales. Our analysis showed that the local atmosphere–sea ice interactions still exist without prominent tropical signals in models, but the coupling intensity is weaker on decadal compared with that on year-to-year time scales. This indicates that the year-to-year variability of atmospheric circulation responsible for the interannual variations of Arctic sea ice might mainly be driven by the local internal process (e.g., AO or AD; Liu et al. 2004; Ogi et al. 2016; Caian et al. 2018), while the decadal sea ice change could be more sensitive to remote forcing. The inability of the models to reproduce the observed propagating of the Rossby wave originating from tropical Pacific into the Arctic may lead to weaker atmosphere–sea ice coupling on decadal time scales. Further efforts toward untangling the relative linear contribution of local and remote drivers for September sea ice variation on different time scales and the nonlinear feedback between different internal drivers of sea ice loss is of vital importance and could help to increase near future predictability of Arctic sea ice.
Acknowledgments.
We thank three anonymous reviewers and Editor James Screen for their constructive suggestions, which greatly helped to improve the quality of this manuscript. This work was supported by the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant XDA19070404) and the National Natural Science Foundation of China (Grant 41725018).
Data availability statement.
The sea ice concentration data from the United States National Snow and Ice Data Center (NSIDC) are available from http://nsidc.org/data/seaice/. The CESM-LE simulations can be found at https://www.cesm.ucar.edu/projects/community-projects/LENS/. The MPI-ESM grand ensemble data can be assessed at https://esgf-data.dkrz.de/search/mpi-ge/.
REFERENCES
Årthun, M., I. H. Onarheim, J. Dörr, and T. Eldevik, 2021: The seasonal and regional transition to an ice-free Arctic. Geophys. Res. Lett., 48, e2020GL090825, https://doi.org/10.1175/JCLI-D-11-00466.1.
Baxter, I., and Coauthors, 2019: How tropical Pacific surface cooling contributed to accelerated sea ice melt from 2007 to 2012 as ice is thinned by anthropogenic forcing. J. Climate, 32, 8583–8602, https://doi.org/10.1175/JCLI-D-18-0783.1.
Blanchard-Wrigglesworth, E., and Q. Ding, 2019: Tropical and midlatitude impact on seasonal polar predictability in the Community Earth System Model. J. Climate, 32, 5997–6014, https://doi.org/10.1175/JCLI-D-19-0088.1.
Boé, J., A. Hall, and X. Qu, 2009: September sea-ice cover in the Arctic Ocean projected to vanish by 2100. Nat. Geosci., 2, 341–343, https://doi.org/10.1038/ngeo467.
Bonan, D. B., and E. Blanchard-Wrigglesworth, 2020: Nonstationary teleconnection between the Pacific Ocean and Arctic sea ice. Geophys. Res. Lett., 47, e2019GL085666, https://doi.org/10.1029/2019GL085666.
Bonan, D. B., F. Lehner, and M. M. Holland, 2021a: Partitioning uncertainty in projections of Arctic sea ice. Environ. Res. Lett., 16, 044002, https://doi.org/10.1088/1748-9326/abe0ec.
Bonan, D. B., T. Schneider, I. Eisenman, and R. C. J. Wills, 2021b: Constraining the date of a seasonally ice‐free Arctic using a simple model. Geophys. Res. Lett., 48, e2021GL094309, https://doi.org/10.1029/2021GL094309.
Caian, M., T. Koenigk, R. Döscher, and A. Devasthale, 2018: An interannual link between Arctic sea-ice cover and the North Atlantic Oscillation. Climate Dyn., 50, 423–441, https://doi.org/10.1007/s00382-017-3618-9.
Castruccio, F. S., Y. Ruprich-Robert, S. G. Yeager, G. Danabasoglu, R. Msadek, and T. L. Delworth, 2019: Modulation of Arctic sea ice loss by atmospheric teleconnections from Atlantic multidecadal variability. J. Climate, 32, 1419–1441, https://doi.org/10.1175/JCLI-D-18-0307.1.
Cavalieri, D. J., C. L. Parkinson, P. Gloersen, and H. J. Zwally, 1996: Sea ice concentrations from Nimbus-7 SMMR and DMSP SSM/I-SSMIS passive microwave data, version 1 (updated yearly). NASA National Snow and Ice Data Center Distributed Active Archive Center, accessed 31 March 2020, https://doi.org/10.5067/8GQ8LZQVL0VL.
Chylek, P., C. K. Folland, G. Lesins, M. K. Dubey, and M. Wang, 2009: Arctic air temperature change amplification and the Atlantic Multidecadal Oscillation. Geophys. Res. Lett., 36, L14801, https://doi.org/10.1029/2009gl038777.
Coats, S., J. E. Smerdon, B. I. Cook, and R. Seager, 2013: Stationarity of the tropical Pacific teleconnection to North America in CMIP5/PMIP3 model simulations. Geophys. Res. Lett., 40, 4927–4932, https://doi.org/10.1002/grl.50938.
Cohen, J., and Coauthors, 2014: Recent Arctic amplification and extreme mid-latitude weather. Nat. Geosci., 7, 627–637, https://doi.org/10.1038/ngeo2234.
Comiso, J. C., 2017: Bootstrap sea ice concentrations from Nimbus-7 SMMR and DMSP SSM/I-SSMIS, version 3. NASA National Snow and Ice Data Center Distributed Active Archive Center, accessed 7 July 2020, https://doi.org/10.5067/7Q8HCCWS4I0R.
Day, J. J., J. C. Hargreaves, J. D. Annan, and A. Abe-Ouchi, 2012: Sources of multi-decadal variability in Arctic sea ice extent. Environ. Res. Lett., 7, 034011, https://doi.org/10.1088/1748-9326/7/3/034011.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
DeRepentigny, P., A. Jahn, M. M. Holland, and A. Smith, 2020: Arctic sea ice in two configurations of the CESM2 during the 20th and 21st centuries. J. Geophys. Res. Oceans, 125, e2020JC016133, https://doi.org/10.1029/2020JC016133.
Deser, C., and H. Teng, 2008: Evolution of Arctic sea ice concentration trends and the role of atmospheric circulation forcing, 1979–2007. Geophys. Res. Lett., 35, L02504, https://doi.org/10.1029/2007GL032023.
Deser, C., R. Tomas, M. Alexander, and D. Lawrence, 2010: The seasonal atmospheric response to projected Arctic sea ice loss in the late twenty-first century. J. Climate, 23, 333–351, https://doi.org/10.1175/2009JCLI3053.1.
Deser, C., R. Knutti, S. Solomon, and A. S. Phillips, 2012a: Communication of the role of natural variability in future North American climate. Nat. Climate Change, 2, 775–779, https://doi.org/10.1038/nclimate1562.
Deser, C., A. Phillips, V. Bourdette, and H. Teng, 2012b: Uncertainty in climate change projections: The role of internal variability. Climate Dyn., 38, 527–546, https://doi.org/10.1007/s00382-010-0977-x.
Deser, C., J. W. Hurrell, and A. S. Phillips, 2017: The role of the North Atlantic Oscillation in European climate projections. Climate Dyn., 49, 3141–3157, https://doi.org/10.1007/s00382-016-3502-z.
Deser, C., and Coauthors, 2020: Insights from Earth system model initial-condition large ensembles and future prospects. Nat. Climate Change, 10, 277–286, https://doi.org/10.1038/s41558-020-0731-2.
Diebold, F. X., and G. D. Rudebusch, 2021: Probability assessments of an ice-free Arctic: Comparing statistical and climate model projections. J. Econ., https://doi.org/10.1016/j.jeconom.2020.12.007.
Ding, Q., J. M. Wallace, D. S. Battisti, E. J. Steig, A. J. E. Gallant, H.-J. Kim, and L. Geng, 2014: Tropical forcing of the recent rapid Arctic warming in northeastern Canada and Greenland. Nature, 509, 209–212, https://doi.org/10.1038/nature13260.
Ding, Q., and Coauthors, 2017: Influence of high-latitude atmospheric circulation changes on summertime Arctic sea ice. Nat. Climate Change, 7, 289–295, https://doi.org/10.1038/nclimate3241.
Ding, Q., and Coauthors, 2019: Fingerprints of internal drivers of Arctic sea ice loss in observations and model simulations. Nat. Geosci., 12, 28–33, https://doi.org/10.1038/s41561-018-0256-8.
England, M., A. Jahn, and L. Polvani, 2019: Nonuniform contribution of internal variability to recent Arctic sea ice loss. J. Climate, 32, 4039–4053, https://doi.org/10.1175/JCLI-D-18-0864.1.
Frankcombe, L. M., M. H. England, M. E. Mann, and B. A. Steinman, 2015: Separating internal variability from the externally forced climate response. J. Climate, 28, 8184–8202, https://doi.org/10.1175/JCLI-D-15-0069.1.
Frankcombe, L. M., M. H. England, J. B. Kajtar, M. E. Mann, and B. A. Steinman, 2018: On the choice of ensemble mean for estimating the forced signal in the presence of internal variability. J. Climate, 31, 5681–5693, https://doi.org/10.1175/JCLI-D-17-0662.1.
Grunseich, G., and B. Wang, 2016: Arctic sea ice patterns driven by the Asian summer monsoon. J. Climate, 29, 9097–9112, https://doi.org/10.1175/jcli-d-16-0207.1.
Hezel, P. J., T. Fichefet, and F. Massonnet, 2014: Modeled Arctic sea ice evolution through 2300 in CMIP5 extended RCPs. Cryosphere, 8, 1195–1204, https://doi.org/10.5194/tc-8-1195-2014.
Hoerling, M. P., J. W. Hurrell, and T. Xu, 2001: Tropical origins for recent North Atlantic climate change. Science, 292, 90–92, https://doi.org/10.1126/science.1058582.
Hurrell, J. W., and Coauthors, 2013: The Community Earth System Model: A framework for collaborative research. Bull. Amer. Meteor. Soc., 94, 1339–1360, https://doi.org/10.1175/BAMS-D-12-00121.1.
Jahn, A., 2018: Reduced probability of ice-free summers for 1.5°C compared to 2°C warming. Nat. Climate Change, 8, 409–413, https://doi.org/10.1038/s41558-018-0127-8.
Jahn, A., J. E. Kay, M. M. Holland, and D. M. Hall, 2016: How predictable is the timing of a summer ice‐free Arctic? Geophys. Res. Lett., 43, 9113–9120, https://doi.org/10.1002/2016GL070067.
Karpechko, A. Y., D. Maraun, and V. Eyring, 2013: Improving Antarctic total ozone projections by a process-oriented multiple diagnostic ensemble regression. J. Atmos. Sci., 70, 3959–3976, https://doi.org/10.1175/JAS-D-13-071.1.
Kay, J. E., M. M. Holland, and A. Jahn, 2011: Inter‐annual to multi‐decadal Arctic sea ice extent trends in a warming world. Geophys. Res. Lett., 38, L15708, https://doi.org/10.1029/2011GL048008.
Kay, J. E., and Coauthors, 2015: The Community Earth System Model (CESM) large ensemble project: A community resource for studying climate change in the presence of internal climate variability. Bull. Amer. Meteor. Soc., 96, 1333–1349, https://doi.org/10.1175/BAMS-D-13-00255.1.
Labe, Z., G. Magnusdottir, and H. Stern, 2018: Variability of Arctic sea ice thickness using PIOMAS and the CESM large ensemble. J. Climate, 31, 3233–3247, https://doi.org/10.1175/JCLI-D-17-0436.1.
Lamarque, J.-F., and Coauthors, 2010: Historical (1850–2000) gridded anthropogenic and biomass burning emissions of reactive gases and aerosols: Methodology and application. Atmos. Chem. Phys., 10, 7017–7039, https://doi.org/10.5194/acp-10-7017-2010.
Lee, S., 2012: Testing of the tropically excited Arctic warming mechanism (TEAM) with traditional El Niño and La Niña. J. Climate, 25, 4015–4022, https://doi.org/10.1175/JCLI-D-12-00055.1.
Lee, S., T. Gong, N. Johnson, S. B. Feldstein, and D. Pollard, 2011: On the possible link between tropical convection and the Northern Hemisphere Arctic surface air temperature change between 1958 and 2001. J. Climate, 24, 4350–4367, https://doi.org/10.1175/2011JCLI4003.1.
Lehner, F., C. Deser, N. Maher, J. Marotzke, E. M. Fischer, L. Brunner, R. Knutti, and E. Hawkins, 2020: Partitioning climate projection uncertainty with multiple large ensembles and CMIP5/6. Earth Syst. Dyn., 11, 491–508, https://doi.org/10.5194/esd-11-491-2020.
Li, D., R. Zhang, and T. R. Knutson, 2017: On the discrepancy between observed and CMIP5 multi-model simulated Barents Sea winter sea ice decline. Nat. Commun., 8, 14991, https://doi.org/10.1038/ncomms14991.
Liu, J., J. A. Curry, and Y. Hu, 2004: Recent Arctic sea ice variability: Connections to the Arctic Oscillation and the ENSO. Geophys. Res. Lett., 31, L09211, https://doi.org/10.1029/2004GL019858.
Liu, J., M. Song, R. M. Horton, and Y. Hu, 2013: Reducing spread in climate model projections of a September ice-free Arctic. Proc. Natl. Acad. Sci. USA, 110, 12 571–12 576, https://doi.org/10.1073/pnas.1219716110.
Luo, R., Q. Ding, Z. Wu, I. Baxter, M. Bushuk, Y. Huang, and X. Dong, 2021: Summertime atmosphere–sea ice coupling in the Arctic simulated by CMIP5/6 models: Importance of large-scale circulation. Climate Dyn., 56, 1467–1485, https://doi.org/10.1007/s00382-020-05543-5.
Maher, N., and Coauthors, 2019: The Max Planck Institute Grand Ensemble: Enabling the exploration of climate system variability. J. Adv. Model. Earth Syst., 11, 2050–2069, https://doi.org/10.1029/2019MS001639.
Mahlstein, I., and R. Knutti, 2012: September Arctic sea ice predicted to disappear near 2°C global warming above present. J. Geophys. Res., 117, D06104, https://doi.org/10.1029/2011JD016709.
Massonnet, F., T. Fichefet, H. Goosse, C. M. Bitz, G. Philippon-Berthier, M. M. Holland, and P.-Y. Barriat, 2012: Constraining projections of summer Arctic sea ice. Cryosphere, 6, 1383–1394, https://doi.org/10.5194/tc-6-1383-2012.
Meehl, G. A., C. T. Y. Chung, J. M. Arblaster, M. M. Holland, and C. M. Bitz, 2018: Tropical decadal variability and the rate of Arctic sea ice decrease. Geophys. Res. Lett., 45, 11 326–11 333, https://doi.org/10.1029/2018GL079989.
Meinshausen, M., and Coauthors, 2011: The RCP greenhouse gas concentrations and their extensions from 1765 to 2300. Climatic Change, 109, 213–241, https://doi.org/10.1007/s10584-011-0156-z.
Melia, N., K. Haines, and E. Hawkins, 2015: Improved Arctic sea ice thickness projections using bias-corrected CMIP5 simulations. Cryosphere, 9, 2237–2251, https://doi.org/10.5194/tc-9-2237-2015.
Niederdrenk, A. L., and D. Notz, 2018: Arctic sea ice in a 1.5°C warmer world. Geophys. Res. Lett., 45, 1963–1971, https://doi.org/10.1002/2017GL076159.
Notz, D., 2015: How well must climate models agree with observations? Philos. Trans. Roy. Soc., 373A, 20140164, https://doi.org/10.1098/rsta.2014.0164.
Notz, D., and J. Marotzke, 2012: Observations reveal external driver for Arctic sea-ice retreat. Geophys. Res. Lett., 39, L08502, https://doi.org/10.1029/2012GL051094.
Notz , D., and J. Stroeve, 2016: Observed Arctic sea-ice loss directly follows anthropogenic CO2 emission. Science, 354, 747–750, https://doi.org/10.1126/science.aag2345.
Notz, D., and Coauthors, 2020: Arctic sea ice in CMIP6. Geophys. Res. Lett., 47, e2019GL086749, https://doi.org/10.1029/2019GL086749.
Ogi, M., S. Rysgaard, and D. G. Barber, 2016: Importance of combined winter and summer Arctic Oscillation (AO) on September sea ice extent. Environ. Res. Lett., 11, 034019, https://doi.org/10.1088/1748-9326/11/3/034019.
Olonscheck, D., T. Mauritsen, and D. Notz, 2019: Arctic sea-ice variability is primarily driven by atmospheric temperature fluctuations. Nat. Geosci., 12, 430–434, https://doi.org/10.1038/s41561-019-0363-1. .
Overland, J. E., and M. Wang, 2013: When will the summer Arctic be nearly sea ice free? Geophys. Res. Lett., 40, 2097–2101, https://doi.org/10.1002/grl.50316.
Rosenblum, E., and I. Eisenman, 2017: Sea ice trends in climate models only accurate in runs with biased global warming. J. Climate, 30, 6265–6278, https://doi.org/10.1175/JCLI-D-16-0455.1.
Screen, J. A., 2018: Arctic sea ice at 1.5 and 2°C. Nat. Climate Change, 8, 362–363, https://doi.org/10.1038/s41558-018-0137-6.
Screen, J. A., and I. Simmonds, 2010: The central role of diminishing sea ice in recent Arctic temperature amplification. Nature, 464, 1334–1337, https://doi.org/10.1038/nature09051.
Screen, J. A., and F. Francis, 2016: Contribution of sea-ice loss to Arctic amplification is regulated by Pacific Ocean decadal variability. Nat. Climate Change, 6, 856–860, https://doi.org/10.1038/nclimate3011.
Screen, J. A., and D. Williamson, 2017: Ice-free Arctic at 1.5°C? Nat. Climate Change, 7, 230–231, https://doi.org/10.1038/nclimate3248.
Screen, J. A., and C. Deser, 2019: Pacific Ocean variability influences the time of emergence of a seasonally ice‐free Arctic Ocean. Geophys. Res. Lett., 46, 2222–2231, https://doi.org/10.1029/2018GL081393.
Screen, J. A., C. Deser, and I. Simmonds, 2012: Local and remote controls on observed Arctic warming. Geophys. Res. Lett., 39, L10709, https://doi.org/10.1029/2012GL051598.
Screen, J. A., and Coauthors, 2018: Consistency and discrepancy in the atmospheric response to Arctic sea-ice loss across climate models. Nat. Geosci., 11, 155–163, https://doi.org/10.1038/s41561-018-0059-y.
Shen, Z., A. Duan, D. Li, and J. Li, 2021: Assessment and ranking of climate models in Arctic sea ice cover simulation: From CMIP5 to CMIP6. J. Climate, 34, 3609–3627, https://doi.org/10.1175/JCLI-D-20-0294.1.
Sigmond, M., J. C. Fyfe, and N. C. Swart, 2018: Ice-free Arctic projections under the Paris Agreement. Nat. Climate Change, 8, 404–408, https://doi.org/10.1038/s41558-018-0124-y.
Simmonds, I., 2015: Comparing and contrasting the behaviour of Arctic and Antarctic sea ice over the 35 year period 1979–2013. Ann. Glaciol., 56, 18–28, https://doi.org/10.3189/2015aog69a909.
Stroeve, J., and D. Notz, 2018: Changing state of Arctic sea ice across all seasons. Environ. Res. Lett., 13, 103001, https://doi.org/10.1088/1748-9326/aade56.
Swart, N. C., J. C. Fyfe, E. Hawkins, J. E. Kay, and A. Jahn, 2015: Influence of internal variability on Arctic sea-ice trends. Nat. Climate Change, 5, 86–89, https://doi.org/10.1038/nclimate2483.
Tokinaga, H., S.-P. Xie, and H. Mukougawa, 2017: Early 20th-century Arctic warming intensified by Pacific and Atlantic multidecadal variability . Proc. Natl. Acad. Sci. USA, 114, 6227–6232, https://doi.org/10.1073/pnas.1615880114.
Topál, D., Q. Ding, J. Mitchell, I. Baxter, M. Herein, T. Haszpra, R. Luo, and Q. Li, 2020: An internal atmospheric process determining summertime Arctic sea ice melting in the next three decades: Lessons learned from five large ensembles and multiple CMIP5 climate simulations. J. Climate, 33, 7431–7454, https://doi.org/10.1175/JCLI-D-19-0803.1.
Trenberth, K. E., J. T. Fasullo, G. Branstator, and A. S. Phillips, 2014: Seasonal aspects of the recent pause in surface warming. Nat. Climate Change, 4, 911–916, https://doi.org/10.1038/nclimate2341.
Wang, B., X. Zhou, Q. Ding, and J. Liu, 2021: Increasing confidence in projecting the Arctic ice-free year with emergent constraints. Environ. Res. Lett., 16, 094016, https://doi.org/10.1088/1748-9326/ac0b17.