The role of interdecadal climate oscillations in driving Arctic atmospheric river trends

The study investigates why observed Arctic atmospheric river (AR) trends over recent decades show strong regional differences, despite overall Arctic warming and increased atmospheric moisture availability. The Arctic has warmed nearly four times faster than the global mean, with associated sea ice decline, especially in the western Arctic in summer and the Barents Sea in winter. ARs, narrow corridors of intense moisture transport, account for 70–80% of moisture entering the Arctic and can rapidly enhance downward longwave radiation, leading to heat extremes and rapid sea ice loss. While prior work noted increases in ARs over the Atlantic sector and a poleward shift in AR occurrence, a comprehensive assessment of spatial patterns of Arctic AR frequency trends and their drivers on multi-decadal timescales has been lacking. The study hypothesizes that interdecadal oceanic variability, particularly the Interdecadal Pacific Oscillation (IPO) and Atlantic Multidecadal Oscillation (AMO), modulates regional AR trends in addition to anthropogenic forcing.

Prior research has established Arctic Amplification (AA) and identified mechanisms such as ice–albedo, lapse rate, cloud and water vapor feedbacks, and poleward energy transport. ARs are known to dominate poleward moisture transport in midlatitudes and contribute 70–80% of moisture import to the Arctic, with impacts on Arctic heat extremes and sea ice. Observational studies reported increasing ARs/moisture intrusions in the Atlantic sector during winter and a poleward shift of AR occurrence, contributing to Barents–Kara sea ice decline. Internal climate variability modes, IPO and AMO, strongly influence large-scale circulation and energy transport, affecting Arctic warming and sea ice on multi-decadal scales. However, the specific role of IPO/AMO in shaping regional Arctic AR trends had not been systematically quantified.

Data and period: Two reanalyses (ERA5; MERRA-2) regridded to 1°×1°, daily means from 6-hourly data, focusing on 1980–2021 (trends largely assessed for 1981–2021). Observed SST from HadSST. Model experiments: (1) CESM2 Large Ensemble (LENS2), 50-member fully coupled ensemble (historical to 2014, SSP370 thereafter), using the sub-ensemble with smoothed CMIP6 biomass burning (BMB) emissions to avoid spurious early-21st-century Arctic responses. (2) CESM2 atmosphere-only 10-member ensemble (GOGA) driven by observed SST (ERSSTv5) and sea ice (HadISSTi) for 1880–2021. (3) Additional ensembles: CMIP6 multi-model ensemble (23 models), ACCESS-ESM1-5 40-member, and CNRM-CM6-1 30-member (1979/1981–2014/2021; SSP370 where available). AR detection: IVT-based algorithm (Guan & Waliser) with Arctic-oriented refinements. Criteria: threshold is max(monthly 85th percentile of IVT, 100 kg m−1 s−1); mean poleward IVT > 50 kg m−1 s−1; >50% of object grid points within 45° of object-mean IVT direction; length-to-width > 2; minimum length relaxed to 1500 km for Arctic; iterative thresholds not applied for efficiency; tracking improvements to axis identification. AR statistics shown to be robust to algorithm choice (via ARTMIP comparisons) and to IVT formulation. Trend decomposition: Dynamical vs thermodynamical contributions separated using a moisture-scaling method applied seasonally. The dynamical contribution is obtained by scaling moisture to remove its interannual variability (Qs/Qc) and recomputing IVT and AR statistics; the thermodynamical contribution is estimated as the residual (total minus dynamical). Attribution to internal variability: Maximum covariance analysis (MCA) between internally generated Arctic AR frequency trends and global SST trends (60°S–70°N) across the 50 LENS2 members, after removing the ensemble mean forced trends. Intermember regressions of SST trends onto Arctic mean AR trends were performed to relate AR changes to SST patterns. IPO/AMO indices: IPO defined as the 7-year running average of the principal component of EOF1 of detrended Pacific (120°E–70°W, 50°S–60°N) SST anomalies; AMO defined as the 7-year running average of detrended North Atlantic (60°W–0°, equator–70°N) SST anomalies. In observations, linear detrending uses 1850–2021 HadSST; in simulations, ensemble mean is removed. Mechanisms: Regress AR frequency variability and its separated dynamical component onto standardized IPO and AMO indices (historical+SSP370, 1979–2100) to isolate circulation- versus moisture-driven effects. Diagnose associated sea-level pressure and wind anomaly patterns. Projection constraint: For LENS2 near-future (2024–2064), compute Arctic mean AR frequency trend distributions across members. Remove components linearly associated with IPO and AMO in each member via regression to obtain residual trends. Compare standard deviations before and after removal to quantify uncertainty reduction. Statistics: Linear trends, Student’s t-test for significance; spatial averaging over defined Atlantic and Pacific Arctic sectors; comparison of sectoral trend differences against ensemble spreads.

- Observed spatial pattern (1981–2021): Significant AR frequency increases concentrated over the Atlantic sector (Greenland Sea, Baffin Bay) up to ~0.9% decade−1; weaker and more localized increases over the Pacific sector (notably Chukchi Sea). - Sectoral rates: Area-averaged AR frequency increased by ~0.42% decade−1 (ERA5) and 0.49% decade−1 (MERRA-2) over the Atlantic sector versus ~0.19% decade−1 (ERA5) and 0.29% decade−1 (MERRA-2) over the Pacific sector—about twice as fast in the Atlantic sector. - Decomposition: The faster Atlantic-sector increase is primarily thermodynamical (greater atmospheric moistening; positive IWV trends); dynamical changes partly offset Atlantic increases, especially in winter and summer. - Forced response vs observations: Coupled ensembles (LENS2, CMIP6, ACCESS, CNRM) simulate a largely uniform, moisture-driven Arctic-wide increase in AR frequency with negligible dynamical contribution—contrasting with observed sectoral differences. - Role of observed SST/sea ice: Atmosphere-only CESM2 (GOGA) reproduces stronger Atlantic and weaker Pacific AR trends and negative dynamical contribution over the Atlantic, implicating observed SST/sea ice variability as key drivers. Observed Atlantic–Pacific sector trend differences fall within GOGA intermember spread but outside LENS2 spread (ensemble mean near zero difference). - IPO/AMO linkage: MCA of internal variability shows (Mode 1, 65% covariance) positive IPO-like Pacific warming and North Atlantic warming associated with broad AR increases; (Mode 2, 12% covariance) positive AMO over the North Atlantic with negative IPO-like Pacific pattern associated with increased ARs in the Atlantic sector and decreased ARs in the Pacific sector—mirroring observations. - Correlations: Across LENS2 members, Arctic mean AR trends correlate significantly with IPO trends (r ≈ 0.35, p = 0.01) and AMO trends (r ≈ 0.37, p = 0.009). Sectoral relationships are consistent but weaker/stronger as expected; Pacific correlation improves when an outlier is removed. - Mechanisms: Positive IPO moistens the Arctic and increases ARs, especially over the Pacific sector, while associated circulation anomalies (SLP patterns yielding northerly anomalies in key regions) locally reduce ARs. Positive AMO increases ARs broadly via moistening, with circulation resembling negative NAO that enhances ARs over Baffin Bay and suppresses them over the Barents Sea and parts of the Beaufort/Laptev sector. - Near-future projections and constraint: Under SSP370 (2024–2064), AR frequency increases further with strong moistening. Removing IPO and AMO influences from member-wise AR variations reduces the intermember standard deviation of Arctic mean AR trends from 0.11 to 0.084 % decade−1—a ~24% reduction in projection uncertainty; most of the reduction arises from removing AMO influence (~23.4%), with minor contribution from IPO (~1.5%).

The study addresses the question of what drives spatially heterogeneous multi-decadal Arctic AR trends by demonstrating that internal oceanic variability, specifically the phase evolution of the IPO and AMO, modulates regional AR frequency beyond the anthropogenic forced moistening. Observations show stronger increases over the Atlantic sector and weaker changes over the Pacific sector, a pattern not captured by the uniform forced response in coupled models. Atmosphere-only simulations driven by observed SST/sea ice reproduce the observed sectoral contrast, and MCA/intermember regressions link AR trends to IPO/AMO-like SST trend patterns. Mechanistically, both IPO and AMO increase ARs via Arctic moistening, while associated circulation anomalies locally suppress or enhance ARs, yielding the observed dipole. These findings reconcile the discrepancy between observed and simulated ensemble-mean trends and highlight the crucial role of interdecadal variability in shaping AR-driven moisture and energy transport into the Arctic. The results are relevant for projections of Arctic warming, extreme events, and sea ice loss, as ARs strongly influence surface energy budgets and sea ice variability. By constraining future AR projections using predictable components of IPO/AMO, uncertainties in near-term AR trends can be reduced, improving assessments of Arctic climate impacts and the timing of an ice-free Arctic.

The paper shows that Arctic atmospheric river frequency increased about twice as fast over the Atlantic sector as over the Pacific sector from 1981 to 2021, primarily due to stronger atmospheric moistening in the Atlantic sector. While coupled climate models simulate a uniform anthropogenically forced increase in ARs, the observed regional contrast is explained by interdecadal internal variability: a negative IPO and positive AMO phase combination that enhances ARs over the Atlantic and suppresses them over parts of the Pacific. Atmosphere-only simulations driven by observed SST/sea ice and MCA/regression analyses corroborate the strong IPO/AMO control via thermodynamical and circulation pathways. For the near future (SSP370), ARs are projected to increase further; removing IPO/AMO-linked variability reduces projection uncertainty by ~24%, largely due to AMO. These findings underscore the value of improved decadal prediction of IPO/AMO evolution to constrain Arctic AR projections and, by extension, projections of Arctic warming, extreme weather, and sea ice loss. Future work should refine understanding of the relative roles of IPO versus other atmospheric internal variability over the Pacific sector and assess model dependence of circulation responses.

- The attribution of Pacific-sector AR variability to IPO versus other atmospheric internal variability is not fully resolved; the authors note that their relative roles at different timescales warrant further study. - While multiple ensembles are analyzed, some model dependence exists: not all ensembles reproduce identical circulation anomalies (e.g., AMO-induced circulation anomalies differ in the CNRM ensemble), indicating structural uncertainty. - The AR detection uses a specific IVT-based algorithm (with Arctic-specific adjustments); although cross-checks suggest robustness across methods (ARTMIP), residual methodological sensitivity cannot be entirely excluded. - Biomass burning forcing artifacts in CESM2 necessitated using a smoothed BMB sub-ensemble; while justified, it highlights sensitivity of Arctic simulations to forcing choices. - Data availability and time coverage vary across ensembles (e.g., lack of SSP370 extension in CNRM beyond 2014), potentially limiting some comparative analyses.