Contribution of sea-ice loss to Arctic amplification is regulated by Pacific Ocean decadal variability

The pace of Arctic warming is about double that at lower latitudes—a robust phenomenon known as Arctic amplification. Many diverse climate processes and feedbacks cause Arctic amplification, includingpositive feedbacks associated with diminished sea ice. However, the precise contribution of sea-ice loss to Arctic amplification remains uncertain. Through analyses of both observations andmodel simulations, we show that the contribution of sea-ice loss to wintertime Arctic amplification seems to be dependent on the phase of the PacificDecadalOscillation (PDO).Our results suggest that, for the samepatternandamountof sea-ice loss, consequentArctic warming is larger during the negative PDO phase, relative to the positive phase, leading to larger reductions in the poleward gradient of tropospheric thickness and to more pronounced reductions in the upper-level westerlies. Given the oscillatory nature of the PDO, this relationship has the potential to increase skill in decadal-scale predictability of Arctic and sub-Arctic climate. Our results indicate that Arctic warming in response to the ongoing long-term sea-ice decline is greater (reduced)duringperiodsof negative (positive)PDOphase.We speculate that the observed recent shift to the positive PDO phase, ifmaintained and all other factors being equal, could act to temporarily reduce the pace of wintertimeArctic warming in the near future. Arctic amplification (AA) is a robust feature in observations of the recent past, palaeo-climate reconstructions of the distant past, and model projections of the future. The majority of near-surface AA can be explained by feedbacks associated with a diminished sea-ice cover. Higher in the atmosphere, however, the contribution of sea-ice loss to AA is less well constrained, in part because the atmospheric response to seaice loss is apparently nonlinear and state-dependent. By statedependent we mean that a similar sea-ice anomaly can lead to a different atmospheric response depending on the background ocean–atmospheric state. So far, such state dependencies have generally been attributed to random internal variability. However, known cycles in the ocean–atmosphere coupled system could have a predictable modulating influence on the atmospheric response to sea-ice loss. Here, for the first time, we present evidence suggesting that the Pacific Decadal Oscillation (PDO) modulates the atmospheric response to sea-ice loss. The PDO is a dominant pattern of sea surface temperature (SST) anomalies that typically persists in predominantly one phase for longer than ten years (sometimes with temporary reversals to the opposite state) and has wide-ranging effects on global weather and the Pacific ecosystem. The PDO is not a single phenomenon, but is instead the result of a combination of different physical processes, including stochastic variability of the Aleutian Low, remote tropical forcing and local North Pacific air–sea interactions (see Supplementary Discussion), which can operate on different timescales to drive similar PDO-like SST anomaly patterns (Supplementary Fig. 1). The winter PDO index (Fig. 1a) was predominantly negative from winter 1948/49 to 1975/76, mainly positive until winter 2006/07, then negative again in most winters between 2007/08 and 2012/13. In winter 2013/14, the PDO shifted abruptly back to a positive phase and was followed in winter 2014/15 by the most positive PDO value in the 67-year record. Meanwhile, winter Arctic sea-ice area (Fig. 1b) has declined steadily since the late 1970s, one of the most visible indications of human-induced global warming. The time series of the PDO and sea-ice area indices are only weakly correlated (r=−0.25). Although the PDO does not seem to be a strong driver of winter sea-ice area variability in a pan-Arctic sense, our analysis suggests that the PDOphase affects how the atmosphere responds to sea-ice variability. Figure 1c,d shows composite-meandifferences in air temperature between low ice (LI) and high ice (HI) years during negative PDO (PDO−) and positive PDO (PDO+), respectively. During both PDO phases, negative anomalies in sea-ice area are significantly associated with warmer Arctic air temperatures. The composite anomalies exhibit the classical latitudinal and vertical profile of AA, with greater warming at higher latitudes and at lower altitudes. However, the magnitude of sea-ice-related Arctic warming below 500 hPa is significantly larger during PDO− than during PDO+ (Fig. 1e). At 500 hPa the Arctic-averaged (70–90N) temperature anomaly is 0.7 C and 0.3 C in PDO− and PDO+, respectively. Corresponding values at 700 hPa are 1.0 C and 0.4 C, and at 850 hPa are 1.2 C and 0.5 C. These results suggest that Arctic warming associated with reduced sea ice is 75–150% greater during PDO− than in PDO+. Larger ice-loss-related Arctic warming is also found during the positive phase of the North Pacific Index (NPI) relative to its negative phase (Supplementary Fig. 2), and also to a lesser extent during the negative phase of the El Niño Southern Oscillation (ENSO) relative to its positive phase (Supplementary Fig. 3). Compared to the PDO, the NPI more directly measures changes in the Aleutian Low, whereas the ENSO index more directly measures changes in tropical Pacific SST (see Supplementary Discussion). Returning to the PDO influence, it is important to emphasize that the composite sea-ice anomalies are non-identical in the two PDO phases: the difference between LI and HI years is larger for PDO− (Fig. 2a,c), largely owing to the fact that the cases are not evenly distributed in time (the mean year for each case is 1964, 1974, 1996 and 1995 for HI PDO−, HI PDO+, LI PDO− and LI PDO+, respectively). A priori, we