Important role of stratosphere-troposphere coupling in the Arctic mid-to-upper tropospheric warming in response to sea-ice loss

Arctic surface air temperature has risen at 3–4 times the global average over the past 40 years, a phenomenon known as Arctic amplification. This amplification has profound local impacts and potential remote effects on Northern Hemisphere mid-latitude weather and climate, though the mechanisms and strength of such teleconnections remain debated. An important uncertainty concerns the vertical extent of Arctic warming: recent work indicates that deep warming (from the surface to the upper troposphere) may be necessary to trigger robust mid-latitude circulation changes, while shallow warming (confined near the surface) yields weaker remote responses. Evidence links deep Arctic warming to mid-latitude circulation anomalies, including Eurasian winter cooling and the ‘warm Arctic, cold Eurasia’ pattern, with modeling suggesting stronger responses to deep versus shallow warming. The causes of near-surface Arctic warming include local feedbacks (albedo, longwave cloud, lapse-rate) and remote transports (atmospheric and oceanic). However, the drivers of mid-to-upper tropospheric Arctic warming remain less clear, and the contribution from sea-ice loss is disputed. Hypotheses include vertical diffusion of surface heating from sea-ice loss, modulation by ocean–atmosphere coupling, and regulation by Pacific decadal variability via poleward warm/moist advection. Other studies argue internal variability dominates Arctic warming aloft. This study investigates the contribution of Barents–Kara Sea (BKS) sea-ice loss to Arctic mid-to-upper tropospheric warming, emphasizing the role of stratosphere–troposphere dynamical coupling.

Prior studies have explored mechanisms linking Arctic sea-ice loss to atmospheric circulation. Some identify enhanced upward wave activity due to sea-ice loss leading to polar vortex weakening and downward coupling into the troposphere, potentially impacting mid-latitudes. Simplified and comprehensive models suggest deep Arctic warming is more effective than shallow warming in forcing mid-latitude anomalies (e.g., Eurasian cooling). Explanations for warming aloft include vertical mixing of near-surface heating, amplification via ocean–atmosphere coupling, and modulation by Pacific decadal variability. Conversely, other analyses attribute warming aloft primarily to internal variability rather than sea-ice loss. There is consensus that the signal-to-noise ratio for remote responses to sea-ice forcing is low, necessitating large ensembles. This study builds on that literature by explicitly diagnosing stratosphere–troposphere coupling processes and their role in generating Arctic warming aloft in response to BKS sea-ice loss, using reanalysis, PAMIP multi-model ensembles, and targeted WACCM-SC simulations with and without stratospheric variability.

Data and indices: - Reanalysis: ERA-5 atmospheric variables (1979–2021), 1°×1° horizontal resolution, 37 vertical levels (1000–1 hPa). - Sea surface temperature (SST) and sea-ice concentration (SIC): HadISST1, monthly, 1°×1°. - Multi-model ensembles: PAMIP (CMIP6 contribution) experiments pdSST-pdSIC (1979–2008 climatology SST/SIC) and pdSST-futBKSeasSIC (future BKS SIC representative of +2 K warming), 1200 members from eight models. Results scaled by 1.16 (ratio of observed BKS SIC index difference between low- and high-SIC years to PAMIP forcing). BKS definition and compositing/regression: - BKS region: 65°–85°N, 20°E–90°E. - BKS SIC index: early winter (Nov–Dec) area-mean SIC over BKS, linearly detrended (1979–2020). - High/low SIC years: selected via ±1 standard deviation threshold; 8 high-SIC years (1988, 1997, 1998, 2002, 2003, 2010, 2014, 2019) and 4 low-SIC years (1984, 2012, 2016, 2020). - Atmospheric fields regressed onto the detrended BKS SIC time series, then multiplied by the composite difference (low minus high SIC; 2-std sea-ice loss) to estimate anomalies associated with strong BKS SIC reductions. Model experiments (WACCM-SC 4.0): - Four bespoke experiments: two free-running (HICE: high BKS SIC; LICE: low BKS SIC) and two with stratospheric nudging (HICE_ndg; LICE_ndg). A control (CTRL) with climatological SIC (1982–2001) provided the climatological state for nudging. - Forcing for HICE/LICE: composites of HadISST1-selected BKS high vs low SIC years (high: 1988, 1997, 1998, 2002, 2003, 2010, 2014; low: 1984, 2012, 2016). The composited SIC differences prescribed through the full seasonal cycle; co-located SSTs adjusted accordingly; elsewhere, SST fixed to 1982–2001 climatology. - Integration length: 210 years for each experiment; first 3 years discarded as spin-up. Identical initial and repeating seasonal boundary conditions across experiments except for prescribed SIC/SST differences. Stratospheric nudging protocol: - Nudging region: polar stratosphere above 100 hPa and north of 66°N; nudging coefficient 1 there, tapering to 0 between 100–200 hPa (vertically) and 66°–60°N (meridionally) to avoid discontinuities; elsewhere coefficient 0. - Variables nudged: horizontal wind and temperature toward CTRL climatology every 3 model hours. - Assumption: nudging effect independent of sea-ice state; differences between LICE_ndg and HICE_ndg isolate sea-ice-forced response without interactive stratospheric variability. Diagnostics and frameworks: - Transformed Eulerian Mean (TEM) framework employed to separate mean circulation and eddy feedbacks. Computed Eliassen–Palm (E–P) flux components (meridional F_x and vertical F_z), E–P flux divergence, residual meridional and vertical velocities, and residual streamfunction χ using standard formulations (Holton 2004; Andrews et al. 1987). The relation between χ and the vertical gradient of E–P flux divergence (elliptic operator) used to interpret overturning anomalies. - Temperature tendency decomposition (TEM): zonal-mean temperature tendency expressed as sum of adiabatic contribution due to residual vertical motion and diabatic heating (J/C_p). Adiabatic term computed on daily data, then averaged in the Arctic mid-to-upper troposphere (600–225 hPa; 225 hPa is the first level below the nudging region) north of 65°N. Diabatic term directly available from WACCM-SC (not available in ERA-5). Statistical testing: - Bootstrap resampling (1000 iterations) to assess significance of composite differences; 90% confidence intervals from 5th/95th percentiles; regions hatched when significant. Additional analyses: - In free-running WACCM, years classified by polar vortex strength using 50 hPa temperature averaged north of 66°N: inactive (within ±0.5σ of CTRL), active (> +0.5σ or < −0.5σ), to implicitly validate nudging results via strat-state dependence of responses. Constraints/availability: - In PAMIP, only air temperature and zonal wind available with sufficient outputs for this analysis; other diagnostics (e.g., TEM quantities) not analyzed from PAMIP due to data limitations.

- Reanalysis linkage: ERA-5 regressions onto BKS SIC show Arctic warming anomalies throughout the troposphere, strongest near the surface and in the upper troposphere, alongside mid-latitude cooling—consistent with observed trends and prior studies. - Forced response across models: PAMIP multi-model ensemble mean exhibits statistically significant deep Arctic warming in response to future BKS sea-ice loss, with a pattern similar to ERA-5 but weaker amplitude. Six of eight models show deep warming (not always significant), reflecting low signal-to-noise. - Targeted model experiments: Free-running WACCM-SC simulations (LICE minus HICE) produce deep Arctic warming throughout the troposphere, qualitatively matching reanalysis and PAMIP patterns but with reduced magnitude. - Magnitude attribution: Simulated Arctic mid-to-upper tropospheric warming (north of 65°N, 600–225 hPa) induced by sea-ice loss is approximately 20% of that inferred from ERA-5 composites (PAMIP scaled by 1.16 for comparability), implying sea-ice loss explains about one-fifth of observed warming aloft; the remainder likely arises from other forcings and internal variability. - Essential role of stratosphere–troposphere (S–T) coupling: When polar stratospheric variability is suppressed via nudging, warming is confined to the lower troposphere and is absent aloft; the mid-to-upper troposphere instead exhibits cooling. Free-running years with an active polar vortex show significant deep warming, whereas inactive-vortex years show warming confined near the surface—implicitly validating the nudging experiments. - Dynamical mechanism: BKS sea-ice loss enhances upward and poleward propagation of wave activity (E–P flux), yielding convergence anomalies in the sub-polar lower stratosphere and upper troposphere, weakening the stratospheric polar vortex and decelerating sub-polar westerlies, which descend into the troposphere. This eddy feedback produces clockwise residual overturning anomalies in the sub-polar middle/upper troposphere and lower stratosphere, with anomalous descent on the poleward flank over the Arctic—causing adiabatic warming aloft. In nudged runs, absent lower-stratospheric convergence shifts the vertical gradient of eddy divergence, inducing counterclockwise overturning and anomalous ascent aloft over the Arctic—leading to cooling. - Thermodynamic budget: Adiabatic heating due to anomalous descent is the dominant contributor to mid-to-upper tropospheric warming anomalies; in ERA-5, the adiabatic warming tendency is about five times stronger than in WACCM, mirroring the magnitude disparity of warming. Diabatic tendencies in WACCM are weaker aloft, consistent with most sea-ice–related diabatic processes being confined to the lower troposphere.

The study addresses whether Arctic sea-ice loss can drive deep Arctic warming necessary for influencing mid-latitude circulation. Results demonstrate that BKS sea-ice loss can contribute to Arctic warming aloft primarily through stratosphere–troposphere dynamical coupling: enhanced upward-propagating waves lead to sub-polar wave convergence, polar vortex weakening, and eddy-driven residual overturning anomalies that induce anomalous descent and adiabatic warming in the Arctic mid-to-upper troposphere. Suppressing stratospheric variability eliminates this pathway, confining warming to the lower troposphere and producing cooling aloft. Thus, distinct mechanisms govern near-surface (local feedbacks and lower-tropospheric processes) versus aloft warming (eddy-driven adiabatic descent tied to S–T coupling). Given that deep warming is implicated in mid-latitude responses, the demonstrated coupling mechanism offers a physical route by which sea-ice loss may influence the broader circulation. However, sea-ice loss accounts for only ~20% of the observed warming aloft, indicating substantial roles for internal variability and other forcings (e.g., tropical SSTs, ENSO, ocean–atmosphere coupling) in setting the full magnitude. These findings refine the Arctic–mid-latitude connection framework by identifying S–T coupling as a key amplifier and vertical extender of sea-ice–forced warming.

This work identifies and quantifies a key dynamical pathway by which Barents–Kara Sea ice loss drives Arctic mid-to-upper tropospheric warming: enhanced upward wave activity and subsequent stratosphere–troposphere coupling induce eddy-driven overturning, anomalous descent, and adiabatic warming aloft. Free-running and PAMIP simulations, together with reanalysis, consistently show deep warming patterns, while nudged experiments confirm that interactive stratospheric variability is essential for warming aloft; without it, warming remains shallow and the upper troposphere cools. Sea-ice loss explains roughly one-fifth of the observed Arctic warming aloft, implying significant contributions from internal variability and other remote forcings. Future research should (i) quantify the relative contributions of sea-ice loss, internal variability, and other boundary forcings to Arctic warming aloft; (ii) assess model fidelity in representing eddy feedbacks and S–T coupling; (iii) examine how non–sea-ice-driven deep warming influences mid-latitude circulation; and (iv) expand diagnostics (e.g., diabatic budgets in reanalyses) to further constrain mechanisms.

- Signal-to-noise: Individual model responses exhibit low signal-to-noise; smaller ensembles risk over-interpretation. Large multi-model means are more robust but still show weaker amplitudes than reanalysis. - Scope of forcing: The analysis isolates atmospheric responses to BKS sea-ice loss; other forcings (e.g., tropical SSTs, ENSO, oceanic variability) and internal variability are not explicitly co-varied, limiting attribution. Sea-ice loss accounts for ~20% of observed warming aloft, leaving substantial unexplained variance. - Data limitations: Diabatic heating rates are unavailable from ERA-5, precluding a full observational budget decomposition. PAMIP lacked complete variables/daily outputs for TEM diagnostics, restricting cross-dataset comparisons to temperature and zonal wind. - Nudging assumptions: The nudging approach assumes independence of nudging effects from sea-ice state and perfectly removes interactive stratospheric variability. Residual interactions and sensitivity to nudging configuration (region/strength) may influence results. - Forcing differences and scaling: PAMIP sea-ice forcing differs from the observationally based and WACCM setups; scaling (1.16) was applied to facilitate comparison, introducing additional uncertainty. - Regional focus and selection: Emphasis on BKS region and specific composited high/low SIC years may limit generalizability across other Arctic sectors or periods.