An Atlantic-driven rapid circulation change in the North Pacific Ocean during the late 1990s

Interbasin interactions have been increasingly emphasized in recent years due to their roles in shaping climate trends and the global warming hiatus in the northern hemisphere. The profound influence from the North Atlantic on the Tropical Pacific has been a primary focus. In this study, we conducted observational analyses and numerical modeling experiments to show that the North Atlantic has also strongly influenced the Extratropical North Pacific. A rapid and synchronous change in the atmospheric and oceanic circulations was observed in the North Pacific during the late 1990s. The change was driven by the transbasin influence from the Atlantic Ocean. During the positive phase of the Atlantic Multidecadal Oscillation (AMO) since the 1990s, the anomalously warm North Atlantic triggers a series of zonally symmetric and asymmetric transbasin teleconnections involving the Inter-tropical Convergence Zone (ITCZ), Walker and Hadley circulations, and Rossby wave propagation that lead to a decrease in wind stress curls over the Pacific subtropics, resulting in an abrupt weakening in the North Pacific subtropical gyre (NPSG) and the Kuroshio Current.

www.nature.com/scientificreports www.nature.com/scientificreports/ To further investigate the changes in the ocean circulation in the western North Pacific, we divided the time interval 1993-2013 into two periods: 1993-1998 and 1999-2013. Five different datasets were used to calculate surface velocity differences between the two periods, Argo drifters, altimetry observations from the Archiving, Validation and Interpretation of Satellite Oceanographic (AVISO) satellite, and three ocean reanalysis products (HYCOM, JCOPE-2, and GODAS) that are independent of each other in their assimilation methods and simulation settings. Figure 2a,b show the surface velocity differences in the Kuroshio region based on Argo drifters and altimeter-derived geostrophic currents, respectively. Both exhibit negative values (shown in blue band) along the main stream of the Kuroshio. A weakened Kuroshio is also evident in all the three reanalysis products (Fig. S1).
Surface ocean circulation variability is usually associated with variability in the surface wind field 4 . Figure 3 shows the changes of surface wind stress (vector) and wind stress curl anomaly (WSCA; color) between the two periods averaged from four atmospheric reanalysis products. A weakening of surface westerlies is evident over the North Pacific during the 1999-2013 period in this multi-analysis mean as well as in each individual reanalysis (Fig. S2). The weakened westerlies result in positive WSCAs in the subtropics, which should weaken the North Pacific subtropical gyre (NPSG), and subsequently the Kuroshio 4 . There are also positive WSCAs over the North Pacific that should act to weaken the subpolar gyre and in turn the western boundary currents. The mechanism responsible for the wind changes warrants further examination.
To further confirm and examine oceanic and atmospheric changes in the North Pacific around 1998-99, we have extended the first period from 1984 to 1998 based on various data sets for statistical tests. Table S1 and Fig. S3 summarize differences for ocean circulation characteristics, while Fig. S4 shows differences for atmospheric parameters. All differences are statistical significance above the 99% confidence level based on t-test, indicating drastic changes in oceanic and atmospheric environments take place around 1998-99.
Recent studies [10][11][12][13] have discovered that the Atlantic Ocean has acted as a pacemaker for global climate in recent decades, contributing to the global warming hiatus and multidecadal fluctuations in the global mean surface temperature, especially in the Northern Hemisphere. The Atlantic Multidecadal Oscillation (AMO), with a period of 65-80 years, is the leading mode of decadal variability mode in the Atlantic Ocean that exerts profound www.nature.com/scientificreports www.nature.com/scientificreports/ impacts not only on the Atlantic and North American climate but also the Pacific climate variability 14,15 , including the Southeast and East Asian summer monsoons 16 and El Niño and South Oscillation (ENSO) 17 . Figure 4a displays the time series of the AMO index during the period 1980-2013, and shows that the AMO shifted to its positive phase in the mid-1990s. This phase change time is a few years before the abrupt changes in Pacific Ocean circulation were observed (as shown in Fig. 1). Previous studies observed baroclinic responses in  the western subtropical gyre and Kuroshio Extension relative to atmospheric forcing using historical hydrography and satellite altimetry data [18][19][20][21] . As mentioned, it takes a few years for surface wind pattern changes to induce upper ocean circulation changes. Therefore, it is possible that the mid-1990s phase change in the AMO can be a cause for the late-1990s changes in the Pacific Ocean circulation. To examine this possible transbasin influence of the Atlantic on Pacific, we performed numerical experiments with the NCAR Community Atmospheric Model, version 3.0, (CAM3.0) 22 coupled to a mixed-layer slab ocean model (SOM). SSTAs were prescribed in the North Atlantic (0-70° N) to represent the positive AMO phase in one experiment (i.e., AMO-positive experiment) and the negative AMO phase in the other experiment (i.e., AMO-negative experiment). The slab ocean is free to interact with the CAM3 model in other ocean basins. Figure 4b shows the SST differences between the two experiments. In the North Atlantic, the SSTA differences resembles the observed AMO pattern 23 , which is characterized by a tropical and a subpolar warming bands separated by a weaker-warming band in between (Fig. 5). Figure 4c shows wind stress and wind stress curl differences between the two experiments. The weakened westerlies and positive WSCA in the subtropical region resemble those observed during 1999-2013 (cf. Fig. 3). The pattern correlations over the subtropical Pacific between the www.nature.com/scientificreports www.nature.com/scientificreports/ model differences and the observed changes in Fig. 3 are 0.49 (P < 0.01) for zonal surface wind anomalies and 0.27 (P < 0.01) for WSCAs. This modeling result adds further support to the suggestion that the AMO phase change is likely a cause for the decelerating westerlies and positive WSCA in the subtropical Pacific region that results in a weakened NPSG and Kuroshio during 1999-2013.
The transbasin influence of the AMO on the Pacific can be explained by zonally-symmetric as well as zonally-asymmetric circulation mechanisms in the atmosphere 24 . The zonal-symmetric circulation mechanism is based on the fact that the Inter-tropical Convergence Zone (ITCZ) 25,26 typically displaces toward the warmer hemisphere 27,28 where the Hadley cell is weaker due to the smaller meridional gradient in surface temperatures 29 . The weaker Hadley cell results in a weaker subtropical high and weaker surface westerlies to the north of the high due to geostrophic balance or reduced eddy transport in the mid-latitudes 27,30 . During the positive AMO phase, an anomalously warm North Atlantic and cold South Atlantic should displace the ITCZ northward and weakened the Northern Hemisphere Hadley not only in the Atlantic but also in the Pacific. A Pacific ITCZ index based on the precipitation anomalies in the northwestern tropical Pacific (averaged over 0-10°N and 130-160°E) 31 confirms that the Pacific ITCZ is displaced southward during the negative AMO phase of the pre-1990 era but northward during the positive AMO phase of the post-1990 era (Fig. S5) 24 . The zonal-mean Hadley circulation also became weaker in the post-1990 period and in the positive AMO phase experiment not only over the North Atlantic but also over the North Pacific (Fig. S6).
The tropical North Atlantic warming associated with the positive AMO phase can also trigger a zonally-asymmetric circulation mechanism that acts to weaken the Pacific subtropical high 32 . In this mechanism, the Atlantic warming induces an anomalous Walker circulation with a descending branch over the tropical central Pacific thus suppressing convection [33][34][35][36][37] . The resultant can then excite a Rossby wave response to higher latitudes inducing an anomalous cyclone over the subtropical Pacific, an anomalous anti-cyclone over the North Pacific, and an anomalous cyclone over the North America, which are not only in upper-level also in low-level. This wavetrain pattern of anomalies in the low-level and upper-level circulation is clearly evident in the positive AMO experiment (Fig. 4c,d) and can also explain the weakened subtropical Pacific high, Aleutian Low, and mid-latitude surface westerlies observed during the post-1990 period (Fig. 3).
As schematized in Fig. 5, our observational analyses and climate model experiments suggest that the change of the AMO to a positive phase in the middle 1990s can trigger a series of zonally symmetric and asymmetric www.nature.com/scientificreports www.nature.com/scientificreports/ transbasin processes to give rise to a late-1990s abrupt change of the North Pacific circulation that is manifested as a weakening of the NPSG and Kuroshio. This Atlantic or AMO control of the North Pacific circulation was not noted in earlier periods of instrumental observations. This can be an indication of an emerging change of climate dynamics due to global warming that deserves attention.

Materials and Methods
Observations and reanalysis data. The monthly-mean SSTs of the ERSST (Extended Reconstructed Sea Surface Temperature, version 5) were provided by the NCEI/NOAA (National Centers for Environmental Information/National Oceanic and Atmospheric Administration, https://data.nodc.noaa.gov) with 2° × 2° horizontal resolution since 1854 38 Fig. 1c) is calculated as ∫ ⋅ v D dm , where ν is the velocity across the Kuroshio axis, dm is the distance between two neighboring stations (122°E, 27°N; 124°E, 24.5°N) across the Kuroshio (shown in Fig. 1a), and D is the mean depth (400 m) of the upper-ocean (modified from Hwang and Kao 42 ).
Four atmospheric reanalysis products were used, the NCEPr1 (National Centers for Environmental Prediction/National Center for Atmospheric Research reanalysis 1) 43   gov/psd/data/timeseries/AMO/) which is calculated as the detrended SSTAs averaged over the North Atlantic from the equator to the 70°N. Statistical analyses. The correlation of significance test is performed based on the two-tailed Student's t-test. More analyses and discussion on the statistical significance have been included in Supplementary Information.