Climate-related change of sea level is one of the most challenging concerns for the humankind. Here we present a comparative analysis of the interannual variability of sea level in the extratropical North Atlantic and North Pacific oceans based on the high-accuracy TOPEX/Poseidon, Jason-1 and ERS-1/2 measurements from November 1992 to June 2003. We found indications of the interdecadal variability of the sea level in the North Pacific possibly related to the Pacific Decadal Oscillation and suggested that the observed decadal rise of sea level in the subpolar and eastern North Atlantic may have been related to the interdecadal change. While the North Atlantic subtropical and subpolar gyres decelerated, the opposite occurred in the North Pacific. The year-to-year variations of sea level showed coherence between the North Atlantic Oscillation, El Niño/La Niña and Pacific Decadal Oscillation events and respective gyre-scale changes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The global mean sea level observed by TOPEX/Poseidon (T/P) altimeters over 1993–2003 was rising at the rate of about 3.1 mm/year [Cazenave and Nerem, 2004]. Satellite altimetry has shown that the change of sea level is characterised by an uneven spatial structure with positive trends in one regions and negative in the other [Cazenave et al., 2004]. Thermal expansion was reported to explain about 60% of the observed trends over 1993–2003 [Willis et al., 2004]. The observed increase of the sea level in the subpolar North Atlantic [Volkov and van Aken, 2004] and associated slowdown of the subpolar cyclonic circulation [Flatau et al., 2003; Häkkinen and Rhines, 2004] cause anxiety about the future state of the North Atlantic Meridional Overturning Circulation, which is a key element of the global thermohaline circulation.
 The long-term variations of sea level, from years to decades, are mainly caused by large-scale changes in advection and buoyancy fluxes. The North Atlantic Oscillation (NAO) in the extratropical North Atlantic [Hurrell et al., 2003], El Niño/Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO) [Hare, 1996] in the tropical and extratropical North Pacific respectively are known as the leading modes of the interannual-interdecadal variability in the NA and in the NP.
 The NAO represents large-scale alterations in the sea level pressure differences between the subtropical and subpolar regions of the North Atlantic [Hurrell et al., 2003]. The positive/negative NAO phase is characterised by strong/weak westerly winds in mid-latitudes and intensified/relaxed wind driven baroclinic transport in the subtropical and subpolar gyres [Curry and McCartney, 2001]. The NAO is most pronounced in winter accounting for more than one-third of the sea level pressure variance in this season [Cayan, 1992].
 The PDO describes the long-term sea surface temperature (SST) oscillation in the northern Pacific [Hare, 1996]. The positive (warm) PDO phase is associated with below normal SSTs in the western and central North Pacific and above normal SSTs along the North American coast. The negative (cool) PDO phase corresponds to an opposite SSTs distribution. The PDO is often considered as a long-lived El Niño/La Niña-like pattern of climate variability in the Pacific because the two oscillations have similar relationships regarding the deviations of SSTs [Zhang et al., 1997]. However, PDO events have persisted for 20–30 years [Mantua et al., 1997; Minobe, 1997], while El Niño/La Niña events last typically from 6 to 18 months. Another characteristic distinguishing the PDO from ENSO is that the strongest PDO signal is located in the northern Pacific Ocean, whereas the strongest ENSO signal is located in the tropical Pacific.
 In this paper, we present a comparative analysis of the interannual change of the sea level in the extratropical parts of the North Atlantic and North Pacific observed with satellite altimetry for over a decade. The observed change is compared with the NAO and PDO climate indices.
2. Data and Method
 We analysed the maps of sea level anomalies (SLA), obtained from the merged Topex/Poseidon-Jason-1+ERS-1/2 altimetry data from October 1992 to July 2003 [Le Traon et al., 1998]. Jason-1 data replaced T/P data in August 2002. The accuracy of the merged data is about 4 cm. The interannual SLA signal was obtained by low pass filtering the SLA data with a running average with a window equal 1 year.
 We used the Empirical Orthogonal Functions (EOF) analysis to identify the dominant modes of the interannual SLA variability. Although no direct or mathematical relationship necessarily exists between the statistical EOFs and any related dynamical modes, coupling the dominant modes with known physical processes can help to understand the forcing mechanisms responsible for a certain pattern of variability. The EOF method finds both spatial patterns and their temporal evolution. In this paper we refer to the spatial patterns as EOFs and to the time series as Principal Components (PCs). Here we investigate only the first empirical mode since higher modes are not significant for such a short record. The EOF-1 is shown as a correlation map produced by computing correlation coefficients between PC-1 and SLA time series at each grid point. The square of the correlation at a particular location scaled by 100% represents the local portion of variance accounted for by a mode.
3. Results and Discussion
 The first empirical mode explains 42% of the variance in the NA Ocean and exhibits a dipole pattern of the sea level variability (Figure 1). Most of variance (over 80%) is explained in the subpolar region and west of Gibraltar. The EOF-1 and PC-1 (Figures 1 and 4a) suggest that a gradual rise of sea level from 1993 to 2003 occurred in the subpolar and eastern areas of the extratropical NA. At the same time sea level was decreasing in the Gulf Stream extension and in the North Atlantic Current (NAC) area, west of the Mid-Atlantic Ridge. In the NP Ocean, EOF-1 also displays a dipole pattern of the sea level variability and accounts for 66% of the variance (Figure 1). In the cores of the subpolar and subtropical gyres EOF-1 explains over 80% of variance. The EOF-1 and PC-1 (Figures 1 and 4b) indicate that in 1998 there was a reversal between the two centres of the interannual variability of sea level: one located in the Kuroshio extension and North Pacific Current (NPC), and another in the Bering Sea. In 1998, sea level started to decrease in the Bering Sea, in the areas of the Alaska Current, the Alaskan Stream and the California Current, accompanied by an increase in the northern part of the subtropical gyre.
 To approximate the linear change of sea level during the decade of altimeter measurements (Dec 1992–Nov 2002) linear trends in the SLA data were estimated by regression. The geostrophic velocity anomalies, associated with sea level trends, were also computed to demonstrate corresponding trends in the surface geostrophic circulation (Figure 2).
 The observed decadal changes of sea level in the NA are substantially different from those in the NP. In the subpolar and eastern areas of the NA sea level was rising while in the areas of the GS and the NAC it was decreasing. The maximum rise was observed in the Labrador Sea (over 7 cm) and in the Irminger Basin (about 10 cm). The maximum decrease (10–15 cm) occurred in the Gulf Stream extension. Similar spatial distribution of the altimetry-derived and thermosteric sea level trends in the NA over 1993–1998 was presented by Cabanes et al.  and later by Lombard et al. . However, on the longer time period from 1955 to 1996, the thermosteric sea level trends for the upper 3000 m were negative in the subpolar regions and positive in the subtropical regions of the NA [Cabanes et al., 2001]. Thus taking into account the observed sea level rise in the subpolar regions of the NA during the last decade, it is possible to suggest an existence of interdecadal gyre-scale variability in the North Atlantic Ocean. Lombard et al.  also showed that from 1950 to 1998 global thermosteric sea level trends exhibited an oscillatory behaviour on the interdecadal time scale.
 The decadal trends of geostrophic velocity anomalies in the NA (Figure 2) suggest that cyclonic circulation in the subpolar gyre (Labrador Sea and Irminger Basin) and anticyclonic circulation in the subtropical gyre declined. The decline of the subpolar NA circulation, manifested in altimeter data, direct current-meter observations, and hydrographic data, was described earlier [Flatau et al., 2003; Häkkinen and Rhines, 2004]. Geostrophic currents in the Labrador Sea and Irminger Basin decelerated with 2–3 cm/s. A slowdown of geostrophic velocities also occurred in the NAC and in the Gulf Stream. The northern wall of the NAC associated with the Subarctic Front as well as the NAC flow around the Newfoundland Rise decelerated with 2–4 cm/s. On the other hand the eastward flow of the NAC along 45°N intensified. Geostrophic velocities reduced in the northern part of the Gulf Stream extension (around 40°N), but intensified in the southern part (around 35°N). This suggests a possible southward shift of the Gulf Stream extension core during the last decade. Decadal changes also occurred in the Azores Current. As suggested by the decadal trends in geostrophic velocity anomalies the core of the Azores Current located at 33°N decelerated while an intensification took place at the northern boundary of the current. Such a change in the geostrophic flow indicates that the Azores Current shifted northward.
 In the NP the decadal increase of sea level (over 10 cm) was associated with the Kuroshio extension, the NPC and with the Western Subarctic Gyre. At the same time sea level lowered in the Bering Sea, Alaska gyre and near the Californian coast (Figure 2). Such a decadal trend is well captured by the first empirical mode of the interannual SLA (Figures 1 and 4b). The pattern of the decadal sea level change in the NP from December 1992 to November 2002 (Figures 1 and 2) is almost opposite to that obtained by Cabanes et al. . Using satellite altimetry observations and monthly mean temperature fields for the upper 500 m they showed that during 1993–1998 sea level was rising in the subpolar NP and along the North American coast and decreasing in the areas occupied by the Kuroshio Current and the NPC. Thus the results, obtained in this work, suggest that after 1998 there was a drastic change in the NP, which forced sea level trends to change sign. This change is well portrayed by the PC-1 of the interannual SLA in the NP (Figure 4b). After 1998, sea level started to rise in the subtropical gyre and to decrease in the subpolar gyre. The PC-1 of the interannual SLA in the NP appeared to be coherent with the yearly moving average of the monthly PDO index for the same time interval (Figure 4b). This is not surprising as sea level is directly dependent upon SST. A flip from positive to negative PDO phases may have taken place in 1998, coinciding with the demise of the 1997/1998 El Niño and the beginning of a La Niña episode [Mantua, 1999]. The La Niña events are characterised by the westward propagation of the positive SST anomalies in the tropical Pacific, which are ultimately entrained into the Kuroshio Current and into the NPC. In order to verify whether this flip of the PDO phase really occurred and it was not just a short-term fluctuation caused by a strong 1997/1998 El Niño event we need at least several more years of observations. The positive trends in the thermosteric sea level in the eastern and subpolar NP and the negative trends in the NP subtropical gyre from 1950 to 1996 [Cabanes et al., 2001; Lombard et al., 2005] are possibly the result of the PDO-related interdecadal variability. The period between 1942–43 and 1976–77 corresponds to the negative (cool) phase and the period between 1976–77 and 1998 corresponds to the positive (warm) phase (N. Mantua, http://jisao.washington.edu/pdo). Such a sequence of negative and positive PDO phases would result in positive sea level trends in the subpolar and eastern NP, and negative sea level trends in the NP subtropical gyre.
 The change, which occurred in 1998, induced a strong intensification of the geostrophic flow of the Kuroshio extension at around 35°N with 10 cm/s and a 2–8 cm/s increase of geostrophic velocities in the NPC (Figure 2). The cyclonic circulation in the Alaska gyre intensified with an increase in geostrophic velocities in the Alaskan Current of about 2 cm/s. Intensification also took place in the California Current where the southward geostrophic flow increased by about 1–2 cm/s. An anomalous anticyclonic geostrophic flow developed in the Western Subarctic Gyre suggesting a deceleration of the Alaskan Stream along the Aleutian Islands and its possible deviation south-westwards.
 The spatial patterns of the observed sea level trends in both oceans (Figure 2) are identical to the spatial patterns of the EOFs-1 of the interannual SLA data (Figure 1). Thus the first empirical modes seem to describe long-term, possibly interdecadal, variations of sea level in both oceans. In order to find the dominant modes of SLA changes at shorter interannual time scales, a linear trend approximated by regression was removed from the interannual SLA prior to the EOF analysis. In this case the first empirical mode of the remaining interannual SLA explained 47% of variance in the NA, and 51% of variance in the NP. The spatial EOFs-1 of the detrended interannual SLA displayed dipole gyre-scale oscillatory patterns in both oceans (Figure 3).
 The EOF-1 pattern of the detrended interannual SLA in the NA (Figure 3) suggests that an increase/decrease of sea level in the subtropical gyre, associated with a corresponding intensification/relaxation of the subtropical anticyclonic circulation, was accompanied by a decrease/increase of sea level in the subpolar gyre, associated with a strengthening/weakening of the subpolar cyclonic circulation. The PC-1 (Figure 4a) indicates that sea level in 1995–1997 and in 2001–2001 raised in the subpolar gyre and lowered in the subtropical gyre. In 1997–1998 sea level was at its decadal maximum in the subpolar gyre, and in 1999–2000 it was at its decadal maximum in the subtropical gyre.
 The EOF-1 and PC-1 (Figures 3 and 4b) of the detrended interannual SLA in the NP depicted a simultaneous drop of sea level in the subtropical gyre and a rise in the Bering Sea peaking in 1997. The latter coincided with an increase of sea level in the subpolar gyre of the NA illustrated by the first empirical mode of the detrended interannual SLA (Figures 3 and 4a). After 1997 an increase of sea level in the subtropical gyre and a decrease in the subpolar gyre of the NP took place. Thus the interannual variability of the detrended SLA during 1993–2003 in both oceans was dominated by gyre scale changes.
 The EOF-1 and PC-1 of the detrended interannual SLA in the NA appeared to be coherent with the variations of the winter (December through January) NAO index (J. Hurrell, http://www.cgd.ucar.edu/∼jhurrell) (Figure 4a). When the NAO index changed from positive to negative values in 1995–1996 and in 2000–2001, sea level responded with a rise in the subpolar and a decrease in the subtropical gyres. In the NP, the removal of the decadal trend did not greatly change the spatial pattern depicted by the EOF-1 of the interannual SLA (Figure 1), but it filtered out a part of the interdecadal signal possibly associated with PDO. The PC-1 of the detrended SLA in the NP (Figure 4b) highlights the change of sea level due to an unusually strong 1997/1998 El Niño event. The 1997 depression in the PC-1 is associated with the lowest SLA observed in the northern part of the subtropical gyre from 1993 to 2003. The similarity of the EOFs-1 of the detrended and non-detrended interannual SLA in the NP is due to the similar spatial patterns of the ENSO- and PDO-induced deviations in SSTs [Zhang et al., 1997].
 We demonstrated that the major part of the variance of the interannual SLA in the NA and NP is possibly associated with the interdecadal variations of sea level. We showed that in 1993–2003 the subpolar and subtropical gyres of the NA decelerated with corresponding sea level rise in the former and decrease in the latter, while an opposite occurred in the subpolar and subtropical gyres of the NP. The interannual change of the sea level in the NA during the decade of 1993–2003 was characterised by a dipole inter-gyre oscillation pattern, associated with the winter NAO index. The interannual change of the sea level in the NP was found coherent with PDO, which possibly switched from positive to negative phase in 1998. The results of this work suggest that the recently reported local trends of sea level are not necessarily related to the global sea level rise, but may be a part of interdecadal fluctuations. Thus the interdecadal variability is a likely reason for the recent decline of the surface circulation in the extratropical NA and for the positive sea level trends in the subpolar and eastern NA and in the subtropical NP.
 The authors thank Anny Cazenave and an anonymous reviewer for comments that helped to improve the manuscript. The TOPEX/Poseidon and ERS-1/2 sea level anomaly maps have been produced by the CLS Space Oceanography Division as part of the Environment and Climate EU ENACT project with support from CNES and distributed by AVISO operations centre. This research was supported by the Space Research Organization Netherlands (SRON) under project EO-032.