Coherent variability of the Norwegian Atlantic Slope Current derived from TOPEX/ERS altimeter data



[1] This study deals with large-scale variability of the Norwegian Atlantic Current (NwAC) transporting warm and saline Atlantic water toward the Arctic Ocean. We concentrate on the eastern branch of the NwAC, the Norwegian Atlantic Slope Current (NwASC), an about 3500 km long, nearly barotropic shelf edge current. Comparison of variations of cross-slope sea level gradient based on the CLS TOPEX/ERS altimeter data and the NwASC from direct current measurements 1995–2002 in the Svinøy section at about 63°N shows significant coherence for periods of 1 to 12 months. Motivated by this the analysis is extended to the entire NwASC using the local cross-slope sea level gradient from the altimeter data as a proxy. The leading NwASC mode represents along-slope current variations of similar sign from the Irish-Scottish shelf to the Fram Strait that can be interpreted as a quasi-steady, direct response to the large-scale wind field.

1. Introduction

[2] The Norwegian Atlantic Current (NwAC) is the northeastern extension of the Gulf Stream, transporting warm and saline Atlantic water (AW) to the polar regions (Figure 1). This current plays a key role in modifying the local climate and in transformation of water masses that return to the Atlantic to contribute in the meridional overturning circulation [e.g., Rahmstorf, 1999], so understanding the variability of this flow is of great importance.

Figure 1.

Schematic map of the investigation area with the flow of Atlantic Water and the 500, 1000, 2000, and 3000 m depth contours included. The Svinøy section is the solid line running NW from Norwegian coast at 63°N. The variability of the NwASC derived from the altimeter data is estimated along the f/H contour that crosses the Svinøy section (63°N) at 500 m depth, basically following the indicated pathway of the NwASC from west of Ireland to Spitsbergen. Abbreviations are explained in the text.

[3] The origin of the NwAC is in the eastern North Atlantic, where the North Atlantic Current (NAC) splits into two branches entering the Norwegian Sea close to the eastern coast of Iceland, and through the Faroe-Shetland Channel [Fratantoni, 2001; Orvik and Niiler, 2002]. From there the flow continues as the two-branch NwAC through the entire Norwegian Sea toward the Arctic [Poulain et al., 1996; Orvik and Niiler, 2002].

[4] In this study we concentrate on the eastern branch of the NwAC, the Norwegian Atlantic Slope Current (NwASC), a nearly barotropic shelf edge current about 3500 km long extending from the Irish-Scottish shelf to the Arctic Ocean, with branches through both the Fram Strait and Barents Sea. The NwASC is thus the major conduit between the Atlantic and the Arctic Ocean.

[5] Variability of the NwASC has been studied west of Scotland-Shetland [Gould et al., 1985; Gordon and Huthnance, 1987; Burrows et al., 1999], in the Svinøy section [Mysak and Schott, 1977; Orvik et al., 2001; Skagseth and Orvik, 2002; Orvik and Skagseth, 2003a] and west of Spitsbergen [Hanzlick, 1983]. Fluctuations at time scales ranging from days to years have been found. The dominant periods are 2–30 days which appear to be an indirect response to local winds or free waves [Gordon and Huthnance, 1987; Skagseth and Orvik, 2002], as wind modulating the sea level along-stream the NwASC on monthly to annual time scales [Skagseth, 2004], and on inter-annual time scales as baroclinic response to wind stress curl in the northern North Atlantic [Orvik and Skagseth, 2003a].

[6] The main focus of this study is on the spatial structure of the coherent variability in the NwASC and its relation to the atmospheric forcing. Particularly relevant in this respect are forced continental shelf waves in response to the local along-slope wind [Gordon and Huthnance, 1987; Skagseth and Orvik, 2002]. If the scale of the atmospheric system is large compared to the width of the continental shelf, and including a coastal boundary [Gill and Schumann, 1974], a shallow onshore Ekman transport will generate a deeper offshore return current through piling up of water against the coast. The response to the wind field will be a quasi-steady forced flow basically parallel to the isobaths. Assuming that the NwASC is a barotropic shelf edge current in presumed geostrophic balance [Gill and Schumann, 1974; Poulain et al., 1996; Orvik and Niiler, 2002], variations of the NwASC should be mirrored in the across-slope sea level gradient which can be derived from TOPEX/ERS sea level anomaly (SLA) data.

[7] In this study we first substantiate the applicability of using altimeter data as proxy for variations of the NwASC. Then the analysis is extended to the entire NwASC where we seek coherent structures of the along-stream current. Finally, these variations in the current are related to the atmospheric forcing.

2. Data and Methodology

[8] The variations in SLA for the period 1995 to 2002 is obtained from the CLS merged TOPEX/ERS data set given as anomalies from the long-term means [Ducet et al., 2000]. The data are provided as weekly means on a modified 1/3° Mercator projection with the meridional resolution inversely proportional to the cosine of the latitude. The resolution varies from 28 km at 50°N to 6 km at 80°N. Corrections are made for inverse barometric, dry tropospheric and tidal effects. Mapping errors are less than or about 20% of the signal variance [LeTraon and Ogor, 1998].

[9] For validations of the SLA satellite altimeter data as a proxy for the NwASC, we use direct current measurements from 100 m depth in the core of the NwASC in the Svinøy section at about 63°N (Figure 1). Single-point measurements located over the 500 m isobath have been shown to capture the variations of the NwASC on time scales longer than 7 days [Orvik and Skagseth, 2003b]. The measurements extend from April 1995 to December 2002.

[10] As a proxy for the wind field we use the NCEP/NCAR reanalysis daily mean sea level pressure (SLP) fields with spatial resolution of 2.5° [Kalnay et al., 1996]. The SLP and current meter data are re-sampled to weekly means, in accordance with the SSH data.

[11] We apply the methodology of using variations of across-slope sea level gradient to determine the along-slope current (hereafter referred to as altimeter current, Valt) in a presumed geostrophic balance, as Valt = equation image, where g is the acceleration due to gravity, f is the Coriolis parameter, and n is the across-slope distance i.e., directed normal to the NwASC. To validate the applicability of using SLA variations as a proxy for the NwASC, we compare the calculated altimeter current with direct current measurements (Vobs) in the Svinøy section (Figure 2).

Figure 2.

Comparison between Vobs from the moored instrument and the derived Valt in the Svinøy section including a) the original time series (note the relative shift of 20 cm/s between the series applied for illustrative purpose), b) spectra, c) coherence analysis, d) 1–12 month band-pass filtered data, and e) one-year low-pass filtered data. The 95% significance level for the coherence amplitude estimates is obtained by Monte-Carlo testing of 1000 synthetic series of similar length and the same lag-1 auto-correlation as the original data.

[12] In light of topographic steering of geostrophic currents, we use the f/H contour (H is the depth) crossing the Svinøy section (63°N) at 500 m depth to capture along-slope variability of the NwASC. This contour coincides with the current record in the core of the flow, corresponding to a depth of 430 m upstream west of Ireland (50°N), and 553 m downstream west of Spitsbergen (80°N) (Figure 1). The f/H line is extracted from the ETOPO-5′ topographic data set and re-sampled to a spatial resolution of about 30 km. Then by considering the local orientation along the f/H line and using the four nearest grid points the altimeter current is calculated.

[13] Empirical Orthogonal Function (EOF) decomposition is applied to extract the dominant mode of variability in the altimeter current (Figure 3) and in the SLP data (Figure 4). The co-variations of the leading SLP - and Valt mode are then investigated (Figure 5).

Figure 3.

EOF analysis of NwASC represented by Valt including a) the leading spatial mode and b) the associated principal component. For comparison we show also the altimeter current in the Svinøy section analyzed in Figure 2.

Figure 4.

EOF analysis of the SLP including a) the spatial pattern and b) the associated principal component (multiplied by −1). The variance explained by this mode is 34.0%. For comparison we also include the principal component associated with the leading NwASC mode analysed in Figure 3. Note that Valt PC1 is shifted by −2.

Figure 5.

Comparison between SLP PC1 and Valt PC1 including a) the spectra and b) the coherence analysis with the 95% significance levels obtained by the same method as for Figure 2.

3. Results

[14] The original series of Valt and Vobs, in the Svinøy section reveal coincident variability (r = 0.55) over a broad range of time scales (Figure 2a). In the frequency domain, the spectra coincide fairly well on 1–12 month time scales, while Vobs becomes relatively more energetic at shorter periods compared to Valt (Figure 2b). The squared coherence estimates are basically above the 95% significance level, and the phase differences are within the ±45° range for all periods (Figure 2c). In the period range of 1–12 months variations in Vobs and Valt show relatively large coincident fluctuations (r = 0.67 and the standard deviation (σ) = 9 cm/s, Figure 2d). On inter-annual time scales the correlation between these series is low and the fluctuations are relatively small (r = 0.17 and σ = 2–3 cm/s, Figure 2e).

[15] Motivated by the significant correlations between Vobs and Valt in the Svinøy section for periods of 1–12 months, the analysis is extended to the entire NwASC by applying an EOF decomposition (Figure 3). Only the leading mode that captures 33% of the variance is significant. The subsequent modes each represent less than 6% of the variance and contribute over only relatively small areas and will therefore not be discussed here. The leading spatial mode (Valt EOF1) represent a coherent structure of similar sign along the NwASC, from about 54°N to 80°N. Superimposed onto this are some smaller scale structures, having maximum amplitudes near Svinøy section (63°N) and Lofoten (68°N), and an intermediate minimum over the Vøring Plateau (66°N). The associated principal component (Valt PC1) coincides with the variability of Valt in the Svinøy section (r = .70) including both a prominent annual cycle and shorter time scale fluctuations.

[16] The leading EOF mode of SLP captures 34% of the variance (Figure 4). The spatial field (SLP EOF1) represents a North Atlantic Oscillation (NAO) [Hurrell, 1995] -like pattern with the main centre of action over Iceland, extending into the Norwegian Sea. The associated principal component (SLP PC1) shows a prominent seasonal cycle, and superimposed higher-frequency fluctuations which coincide with Valt PC1 (r = 0.64). The associated spectra are comparable for longer periods, while the variations associated with SLP PC1 exceed that of Valt PC1 at the shorter periods (Figure 5a). With the exception of the shortest resolved periods close to 2 weeks, significant coherences are found for most periods, with phase differences in the range ±45° (Figure 5b).

4. Discussion

[17] By using across-slope SLA data from satellite altimeter to derive Valt used as a proxy for long-term variability of the entire NwASC, the applicability and accuracy of the method are crucial. According to Figure 2, Vobs and Valt in the Svinøy section show good resemblance on the 1 to 12 month time scale, while the discrepancies are substantial on both shorter and longer (inter-annual) time scales. This justifies the applicability of the altimeter methodology for periods of 1 to 12 months, while variations on shorter and longer periods are not properly resolved.

[18] The discrepancy for periods of less than one month is to a large degree caused by the one-week time resolution used in this study (a Nyquist period of two weeks). A dominant part of the variability of the NwASC is due to the local along-slope wind forcing of 3–6 day periods [Skagseth and Orvik, 2002] and thus not resolved by the one-week averages used here. An additional uncertainty influencing the shorter periods is the different sampling of Valt, representing an average over ∼25 km relative to the single point Vobs. In addition, at all periods the results are influenced by the uncertainty of Valt estimates of about ±5 cm s−1. The variations in the 1–12 month range have a σ = 9 cm s−1 and can therefore be resolved. In contrast, the inter-annual variations are too small (σ of 2–3 cm/s) to be resolved by the altimeter (see Figures 2d–2e).

[19] The spatial structure of Valt EOF1 (Figure 3a) shows that the along-stream structure of the NwASC coincides fairly well with the steepness of the shelf edge slope. Near the Svinøy section (63°N) and west of the Lofoten (68°N) the slope becomes steep, and due to topographic steering of geostrophic currents along isobaths, the converging isobaths will accelerate the flow. The flow will weaken over diverging isobaths such as near the Vøring Plateau (66°N). This is in accordance with Poulain et al. [1996] who, from near-surface drifter observations, found the strongest currents to be over the steep slope between 68°N and 70°N west off Lofoten. The reduced importance of the mode north of Lofoten agrees with the NwASC here branching of into the Barents Seas. The southern limit Valt EOF1 coincides with the latitude where the eastward flowing branch of the NAC encounters the continental slope off Ireland (Figure 1).

[20] According to Figure 3a, a coherent along-stream structure appears along the entire NwASC. The similarity between time series Valt PC1 and Valt (r = 0.70) in the Svinøy section (Figure 3b) confirms the along-stream simultaneity. Then comparing Valt PC1 and SLP PC1 (Figure 4) they also show coincident variations (r = 0.64) and a prominent annual cycle. The similarity is confirmed through a significant coherence with negligible phase difference (Figure 5). Considering the combined effect of the leading Valt mode (Figure 3) and the leading SLP mode (Figures 4), a strengthened NwASC occur in phase with a positive phase NAO-like SLP, and vice versa. This coincidence on 1–12 month time scales supports a model where the fluctuations are quasi-steady, direct responses to the large-scale wind field [Gill and Schumann, 1974]. Because they are generated through piling up of water against the coast resulting in sea level variations of the order tens of cm, they are detectable by the satellite altimeter.

[21] The prominent variations of Valt which reflect direct quasi-steady response to the wind field for 1 to 12 month time scales, suggests that the lagged inter-annual response to the wind field in the North Atlantic [Orvik and Skagseth, 2003a] is not properly resolved by the altimeter method. Presumably this flow is established through conversion from a baroclinic to a barotropic shelf edge current along the Irish-Scottish shelf. The resulting change in the sea level will be of the order of a few cm along the shelf break, which is not detectable by the altimeter. In any case, Figure 2 shows that the inter-annual variations are weaker than the seasonal signal, and thus less detectible by satellite altimeter.

5. Concluding Remarks

[22] In this study we have investigated the possibilities of using across-slope SLA gradients from satellite altimeter as a proxy for the variability of the NwASC. We conclude that the methodology is robust in resolving the fluctuations in the 1–12 month time range, and that the fluctuations are strongly linked to the large-scale wind field. For inter-annual time scales, the signal is apparently too weak to be captured by the altimeter data.

[23] The leading mode of the NwASC shows a prominent seasonality connected to the wind forcing in accordance with the findings locally in the Svinøy section [Skagseth and Orvik, 2002]. We demonstrate that the large-scale variability of the NwASC, the main conduit of heat and salt between the North Atlantic Ocean and the Artic Ocean, is a coherent mode extending from the shelf edge west of Ireland to the Fram Strait. In general, causal mechanisms are needed to evaluate the fate of systems in light of a climate change and as such the identified wind-forced mode of the NwASC is important.


[24] Funding is received from the Norwegian Research Council project NOCLIM. Thanks to P. Jaccard and S. Myking for invaluable contribution in running the Svinøy section mooring program, to F. Cleveland for providing Figure 1, and to A. Jenkins and two anonymous reviewers for useful comments on the manuscript. This is publication Nr. A56 of the Bjerknes Centre for Climate Research.