The Arctic has experienced dramatic warming, approximately four times faster than the global average, a phenomenon known as Arctic Amplification (AA). This warming is accompanied by a substantial decline in Arctic sea ice extent, particularly pronounced in the western Arctic during summer and the Barents Sea during winter. While the summer western Arctic sea ice decline is linked to the persistent positive Pacific North American pattern, the intensification and warming of the Atlantic inflow, termed "Atlantification," contributes to the winter sea ice decline in the Barents Sea. AA and its associated sea ice loss significantly impact local human and natural systems and influence global-scale circulations. Several mechanisms contribute to AA, including local feedbacks (ice albedo, lapse rate, cloud, and water vapor) and poleward energy transport. Atmospheric moisture transport, particularly through atmospheric rivers (ARs), plays a crucial role in Arctic warming. ARs are long, narrow corridors of intense moisture transport responsible for most poleward atmospheric moisture transport at mid-latitudes. Recent studies indicate ARs transport 70-80% of atmospheric moisture into the Arctic, highlighting their potential contribution to AA. ARs efficiently drive heat extremes and rapid sea ice loss due to their substantial moisture and heat intrusion. ARs are characterized by atmospheric moisture content and wind speed. Given recent Northern Hemisphere warming, more moisture fuels AR formation, and increases in ARs over the Atlantic sector of the Arctic during winter have been observed. These AR increases contribute to sea ice decline over the Barents-Kara Sea. While an upward trend in Arctic-wide annual AR counts has been observed, a comprehensive understanding of the spatial distribution of these trends and the underlying driving mechanisms is lacking. Oceanic internal variability, specifically the Interdecadal Pacific Oscillation (IPO) and the Atlantic Multidecadal Oscillation (AMO), significantly influences large-scale circulations and moisture redistribution. These oscillations exert substantial impacts on regional and global climate, particularly in the Arctic. Their phase shifts, by modulating poleward oceanic and atmospheric energy transport, can accelerate or decelerate Arctic warming and sea ice loss on multi-decadal timescales. AR strength and frequency likely vary with IPO and AMO phase shifts. This study systematically quantifies Arctic AR trends using multiple data sources and investigates the driving mechanisms over the past four decades. The research aims to understand why observed trends differ from model simulations and the role of IPO and AMO in these trends. Improving decadal prediction of IPO and AMO phase shifts may enhance the accuracy of future Arctic AR change projections, ultimately leading to improved predictions of Arctic extreme weather events, warming rates, and the timing of an ice-free Arctic.

Existing research has established the significant role of atmospheric rivers (ARs) in transporting moisture and heat poleward, particularly into the Arctic. Studies have shown that ARs account for a substantial portion (70-80%) of the atmospheric moisture entering the Arctic, linking them directly to Arctic warming and sea ice decline. The impact of ARs on Arctic sea ice loss, especially over the Barents-Kara Sea, has been documented. However, previous studies have largely focused on specific regions or seasons, lacking a systematic assessment of Arctic-wide AR trends and their driving mechanisms. While there's evidence of increasing AR frequency over the Atlantic sector, a consistent understanding of the spatial heterogeneity of these trends and the relative contributions of anthropogenic forcing versus natural variability remains elusive. The influence of large-scale climate oscillations, such as the Interdecadal Pacific Oscillation (IPO) and the Atlantic Multidecadal Oscillation (AMO), on Arctic climate is well-established, but their specific role in modulating AR activity requires further investigation. Prior work has shown that these oscillations can impact Arctic warming and sea ice loss through changes in poleward energy transport, suggesting their potential to influence AR frequency and intensity. This study builds upon this existing literature by comprehensively evaluating the spatial patterns of AR trends, explicitly disentangling the roles of anthropogenic forcing and interdecadal oscillations in shaping these trends, and exploring the implications for future Arctic climate projections.

This study utilizes multiple observational datasets and model simulations to investigate Arctic atmospheric river (AR) trends and their driving mechanisms. Two reanalysis products, ERA5 and MERRA-2, are used to ensure robustness. Both datasets are regridded to a spatial resolution of 1° × 1° and analyzed for the period 1980-2021, using daily data averaged from 6-hourly data. Observed sea surface temperature (SST) data are obtained from the Met Office Hadley Centre's HadSST dataset. To understand the influence of anthropogenic forcing, the researchers utilize the 50-member CESM2 Large Ensemble (LENS2), a fully coupled climate model simulation driven by historical and SSP370 forcing. This ensemble allows for the separation of forced changes from internal variability. To assess the role of observed SST and sea ice variability, a 10-member atmosphere-only CESM2 ensemble (GOGA) is employed, driven by observed SST and sea ice data from 1880 to 2021. The comparison between LENS2 and GOGA helps isolate the impact of observed SST/sea ice variability on AR trends. Three additional large ensembles (CMIP6, ACCESS, and CNRM) are analyzed for robustness testing. Atmospheric rivers are detected using an integrated water vapor transport (IVT)-based algorithm, a modified version of an established algorithm widely used in AR research and recommended by the Atmospheric River Tracking Method Intercomparison Project (ARTMIP). This algorithm identifies ARs based on criteria including IVT magnitude, meridional IVT, IVT direction, and length. The algorithm is refined for Arctic application, adjusting the length requirement to 1500 km to account for the typical lifecycle stage of ARs in the Arctic. The dynamical and thermodynamical contributions to AR trends are separated using a scaling method, which removes the influence of moisture variability on wind fields, allowing for the isolation of dynamic effects on AR trends. Maximum covariance analysis (MCA) is applied to the covariance matrix between internally generated Arctic AR trends and global SST trends to identify the leading modes of co-variability, revealing the influence of large-scale SST patterns on interdecadal AR trends. Regression analysis is used to quantify the relationship between AR trends, the IPO, and the AMO, investigating their individual impacts on AR frequency. To assess how IPO and AMO influence ARs, regression analysis is performed on AR frequency variability against the IPO and AMO indices separately, focusing on both circulation and moisture contributions. Finally, the influence of IPO and AMO on near-future AR projections (under SSP370) is evaluated by removing their linear association from individual ensemble members, assessing the reduction in projection uncertainty.

This study reveals a significant spatial heterogeneity in observed Arctic atmospheric river (AR) trends from 1981 to 2021. Contrary to model simulations showing a uniform positive trend in AR frequency, observations indicate a much faster increase in AR frequency over the Atlantic sector (Greenland Sea and Baffin Bay) compared to the Pacific sector. The increase in the Atlantic sector is approximately twice that of the Pacific sector, reaching as high as 0.9% per decade. This difference is primarily attributed to stronger atmospheric moistening over the Atlantic sector, intensifying both mean and extreme integrated water vapor transport (IVT) trends. Decomposition of observed trends into dynamical and thermodynamical components shows that the faster increase in AR frequency over the Atlantic sector is predominantly due to increased moistening, with circulation changes partially offsetting the positive trend. The observed pattern is robust and consistent across multiple reanalysis datasets (ERA5 and MERRA-2) and insensitive to different AR detection algorithms. Climate model simulations driven by anthropogenic forcing alone (LENS2) produce a uniform positive AR trend across the Arctic, highlighting a discrepancy with observations. However, incorporating observed SST and sea ice data into atmosphere-only model experiments (GOGA) successfully reproduces the observed spatial pattern of AR trends, indicating the significant influence of historical SST/sea ice variability. Maximum covariance analysis (MCA) applied to LENS2 reveals two leading modes of covarying Arctic AR trends and global SST trends. The first mode, explaining 65% of the covariance, shows a strong increase in ARs across most of the Arctic, associated with positive IPO and basin-wide North Atlantic warming. The second mode (12% of covariance) exhibits a dipole pattern—increased AR frequency in the Atlantic sector and decreased frequency in the Pacific sector—linked to a positive AMO and a negative IPO-like pattern. This second mode aligns with the observed negative IPO and positive AMO shifts. Regression analysis further confirms a significant positive correlation between the Arctic mean AR trends and both IPO and AMO trends. The relationship between Pacific sector AR trends and the IPO is less pronounced, possibly due to smaller area extent and the influence of other internal atmospheric variability. This suggests that the combined effects of a negative IPO and a positive AMO phase likely contributed to the observed AR trends. Analyzing the mechanisms by which IPO and AMO influence AR trends, the study finds that the positive IPO increases AR frequency over the Pacific sector but slightly reduces it over the Atlantic sector, while a positive AMO leads to widespread increases in AR frequency across most of the Arctic, particularly over the Greenland Sea. These effects are largely due to changes in moisture, with associated circulation changes tending to decrease AR frequency in specific regions. Specifically, positive IPO and AMO induce SLP anomalies that modify surface wind patterns, partially offsetting the moisture-driven increases in AR frequency. The near-future AR projections under the SSP370 scenario show an accelerated increase in AR frequency, driven by enhanced atmospheric moistening. Removing the influence of IPO and AMO from the projections reduces the projection uncertainty by about 24%, mostly attributed to the AMO's influence. This reduction highlights the importance of improved predictions of future IPO and AMO evolution for enhancing the accuracy of Arctic AR projections.

The findings of this study address the research question concerning the driving mechanisms behind observed spatial variations in Arctic atmospheric river (AR) trends. The study demonstrates that while anthropogenic forcing contributes to a general increase in AR frequency, as predicted by climate models, the observed spatial heterogeneity is significantly modulated by the interplay of interdecadal climate oscillations, specifically the IPO and AMO. The observed negative IPO and positive AMO phases during the study period are shown to favor increased AR activity over the Atlantic sector and decreased activity over the Pacific sector. This nuanced understanding highlights the crucial role of internal variability in shaping observed AR trends, which is often overlooked in solely anthropogenic-forcing driven climate projections. The significant reduction in projection uncertainty when the influence of IPO and AMO is removed underscores the importance of accurately predicting these oscillations for improving the reliability of future AR projections and their impact on Arctic warming and sea ice loss. These results emphasize the need for climate models to accurately represent both anthropogenic forcing and internal climate variability to reliably project future Arctic climate change.

This study reveals a significant spatial heterogeneity in observed Arctic atmospheric river (AR) trends, highlighting the importance of considering both anthropogenic forcing and interdecadal climate oscillations (IPO and AMO) in projecting future AR changes. The observed faster increase in AR frequency over the Atlantic sector compared to the Pacific sector is primarily attributed to stronger atmospheric moistening over the Atlantic, modulated by the negative IPO and positive AMO phases. Removing IPO and AMO influences reduces projection uncertainty significantly, indicating that improving predictions of these oscillations is crucial for enhancing future AR projections. This study contributes to a more comprehensive understanding of Arctic climate change and its complex interactions between anthropogenic forcing and natural variability. Future research should focus on improving the representation of these oscillations in climate models and investigating their complex interactions with other Arctic climate processes at finer temporal and spatial resolutions.

The study primarily focuses on the period 1981-2021, limiting the analysis of long-term trends. While multiple datasets and models are used, the inherent uncertainties associated with reanalysis products and climate models need to be considered. The analysis uses a specific AR detection algorithm; results might vary slightly with different algorithms. The study primarily investigates the linear relationship between IPO/AMO and AR trends; non-linear interactions might exist and require further investigation. The focus on IPO and AMO might overlook the influence of other climate modes or processes on Arctic AR variability. The use of a simplified scaling method to disentangle dynamical and thermodynamical contributions could be subject to limitations; other methods should be considered in future studies.