The East Greenland Current (EGC) dynamically connects the Arctic Ocean to the North Atlantic on the western side of the Nordic Seas. Observations show that the speed of the EGC considerably varies along the East Greenland Slope (EGS). Here it is shown, using current meter data reported in the literature and climatological hydrographic fields, that velocity and transport variations along the EGS are supported by the cross-isobath component of the density-dependent geostrophic flow relative to the bottom. The relative flow impinging on (leaving) the EGS in a northern (southern) limb of the cyclonic circulation in the Nordic Seas strengthens (weakens) the along-isobath bottom geostrophic flow. Variations of the latter are clearly associated with along-isobath bottom density gradients. Current observations indicate an increase of the along-isobath bottom velocity from 79°N to 75°N equal to about 9 and 10 cm s−1 on the upper (1000 m isobath) and lower (2000 m isobath) EGS, respectively. Corresponding estimates based on bottom density distribution along the 1000 and 2000 m isobaths are grossly consistent with the observations given above though we obtain a higher increase (13 cm s−1) at 1000 m and lower increase (6 cm s−1) at 2000 m. Considering the variability of the system and the poor resolution of the observations we find this to be a very convincing result, demonstrating the power of the geostrophic approximation for such estimates.
 The Nordic (Greenland-Iceland-Norwegian) Seas (Figure 1) have long been recognized to be a key area for the thermohaline circulation of the World Ocean [e.g., Gordon, 1986; Mauritzen, 1996a]. In this area, a dense water production occurs as a response to extreme surface heat losses in the cyclonic subpolar (Greenland Sea) gyre and cooling of Atlantic Water (AW) in the Norwegian Atlantic Current (NwAC) - West Spitsbergen Current (WSC) system. The AW, modified in the NwAC-WSC system and then all around the Arctic Ocean, returns to the Nordic Seas on the western side of Fram Strait. This Arctic AW together with the overlying Polar Surface Water and sea ice as well as the underlying deep waters, formed by the boundary convection on the Arctic Ocean shelves and slopes [e.g., Aagaard et al., 1985], are carried southward by the East Greenland Current (EGC). On their way toward the North Atlantic, the water masses of the EGC interact with the Nordic Seas water masses [Rudels et al., 2002, 2005]. Already in Fram Strait some AW is recirculated by the Return Atlantic Current (RAC) and mixed into the EGC [e.g., Schlichtholz and Houssais, 1999b]. Further south the dense Arctic Intermediate Water from the open sea convection in the Greenland Sea gyre is also mixed into the EGC. The deepest water masses outflowing from the Arctic Ocean mix with the densest product of the convection in the Greenland Sea gyre, the Greenland Sea Deep Water [e.g., Aagaard et al., 1991]. Some Arctic Intermediate Water is also formed in the Iceland Sea. It interacts with these water masses of the EGC which cross the Jan Mayen Fracture Zone at ∼72°N and contribute to the Denmark Strait overflow renewing the deep waters of the World Ocean. Water masses which are too deep to cross the sill at 640 m in Denmark Strait are deflected eastward along the Greenland-Scotland Ridge. Their shallower components may contribute to the overflow feeding the deep waters of the World Ocean in the Faroe Bank Channel (sill at 850 m) while the remainder should participate in the internal circulation of the Nordic Seas.
 A cyclonic motion in the Nordic Seas is clearly seen, for instance, in drifter data [Jakobsen et al., 2003]. The circulation is traditionally thought to be mainly driven by a positive wind stress curl in the area. A northward wind-driven transport should be compensated by a southward transport in the EGC playing the role of a return boundary current. Aagaard  and Jónsson  obtained an annual mean flat-bottom Sverdrup transport of 35 Sv (1 Sv = 106 m3 s−1), in agreement with an estimate of the EGC transport based on observations from a drifting ice island [Aagaard and Coachman, 1968]. These estimates are larger than the annual mean transport of 21 Sv in the EGC reported by Woodgate et al.  from a current meter section occupied from summer 1994 to summer 1995 at 75°N (Figure 1). Calculations made by these authors show that only 14 Sv of the observed annual mean transport can be attributed to the flat-bottom Sverdrup dynamics (estimate for the most relevant choice of the drag coefficient). The difference probably represents a density-driven throughflow from the Arctic Ocean to the North Atlantic. The outflow through Denmark Strait is up to only ∼6 Sv [Hansen and Østerhus, 2000] while an upper ocean (0–700 m) inflow through Fram Strait is 7 Sv according to the yearlong (summer 1984 to summer 1985) current meter measurements in the EGC at 79°N (Figure 1) carried out by Foldvik et al. . A large transport in the EGC is therefore an intrinsic feature of the Nordic Seas. Even the total inflow through Fram Strait of 12–13 Sv reported by Schauer et al.  from recent measurements at 79°N is considerably smaller than the EGC transport at 75°N.
 Density gradients are an important dynamical agent since not only they determine the vertical shear of the geostrophic flow, but also contribute to the bottom geostrophic flow wherever the ocean bottom is sloping. The latter contribution is associated with along-isobath variations of the bottom density as emphasized in several studies on slope currents [e.g., Shaw and Csanady, 1983; Csanady, 1988]. The link between the bottom flow and density has recently been elucidated in a particularly elegant manner by Walin et al.  and Nilsson et al. . In particular, Walin et al.  suggested, using theoretical considerations and numerical simulations, that a reasonably strong cyclonic circulation could exist in the Nordic Seas even in the absence of wind-forcing. The circulation could be maintained by the inflow of AW across the Greenland-Scotland Ridge and a downstream buoyancy loss due to the heat exchange with the atmosphere.
 The influence of density gradients on the depth-independent flow in the Nordic Seas has also been evidenced in a series of papers on the dynamics of the EGC along the East Greenland Slope (EGS) in Fram Strait based on theoretical considerations and hydrographic data. First, the depth-independent flow in this area was linked to along-isobath variations of the potential energy estimated using quasi-synoptic data from the MIZEX 84 experiment [Schlichtholz and Houssais, 1999c]. The MIZEX 84 data were then used to emphasize the role of the bottom density variations along the EGS north of ∼77.5°N [Schlichtholz, 2002]. The latter were found to be highly correlated with the corresponding variations of the bottom flow obtained from an inverse model published earlier [Schlichtholz and Houssais, 1999a]. Finally, it was argued, using climatological data from the PHC (Polar Science Center Hydrographic Climatology [Steele et al., 2001]), that also the mean density field implies a downstream increase of the magnitude of the bottom flow along the EGS in Fram Strait [Schlichtholz, 2005].
 Here it is shown, using the PHC climatology and published current meter data, that the depth-independent flow in the EGC is likely to be controlled by along-isobath density variations from Fram Strait to Denmark Strait. The study is organized as follows. First, in section 2, velocity and transport changes along the EGS from 79°N to 75°N are estimated from current measurements. Then, in section 3, equations for geostrophic flow are given. Variations of geostrophic flow along the EGS from 79°N to 69°N determined from the density distribution are reported in section 4. The results are discussed and summarized in section 5.
2. Velocity and Transport Estimates From Current Meter Arrays
 To illustrate velocity changes along the EGS, estimates of the annual mean flow at 79°N [Foldvik et al., 1988, Table 2] and 75°N [Woodgate et al., 1999, Table 1] are used. At 79°N, moorings named FS-1, FS-2, and FS-3 were located in the water depth (H) of 1094, 1678, and 2359 m, respectively. The lowest instrument at FS-1 and FS-3 (FS-2) was placed 25 (300) m above the bottom (Figures 2a and 2b). At 75°N, four moorings numbered 410-2, 411-2, 412-4, and 413-4 were located on the EGS, in the water depth of 413, 985, 2240, and 3022 m, respectively, and one mooring (414-3) in the Greenland Basin. The lowest instrument at the three westernmost moorings was placed 52 m above the bottom and nearly so at the other two moorings (Figures 2c and 2d). Figure 2 shows distributions of the flow components in the along-isobath (y) direction, v, and the cross-isobath (x) direction, u. The components are obtained by projecting the observed mean velocity, u, on the local bottom depth gradients estimated from the 5-min gridded Earth topography, ETOPO5 [U.S. National Geophysical Data Center, 1988]. Two pairs of moorings are selected for further analysis, each in a comparable water depth and with the lowest instrument at a comparable distance to the bottom. The first, upper slope pair (FS-1 and 411-2) is approximately on an isobath (∼1000 m) which farther south crosses the Jan Mayen Fracture Zone and then turns eastward along the Greenland-Scotland Ridge (Figure 1). The second, lower slope pair (FS-3 and 412-4) is approximately on an isobath (∼2000 m) which turns eastward at the Jan Mayen Fracture Zone. Values of the along-isobath component of the depth-averaged velocity, (here and henceforth the overbar denotes vertical averaging from the ocean bottom to the ocean surface), the near-bottom velocity, ub, and their difference, r, for the upper and lower slope pair of moorings are given in Table 1. The table also contains estimates of the along-slope transport (M = ∫Hdx, Mb = ∫vbHdx, and Mr = M − Mb) between the ∼1000 and ∼2000 m isobaths at 79°N and 75°N. In the north, the transports are obtained by integration of the along-slope flow from mooring FS-1 to mooring FS-3 (Figure 2a). In the south, the along-slope flow is integrated from mooring 411-2 to mooring 412-4 (Figure 2c) and multiplied by a factor of 0.7 (cosine of the average angle between meridians and isobaths) to account for the non-meridional run of the EGS at 75°N.
Table 1. Estimates of the Along-Isobath Velocity (v) and Along-Slope Transport (M) on the EGS at 79°N and 75°N From the Current Meter Dataa
∼1000 m isobath
∼2000 m isobath
Transport (∼1000–2000 m isobaths)
v is in cm s−1 while M is in 106 m3 s−1 (sign convention as in Figures 2a and 2c). v on the ∼1000 (∼2000) m isobath represents values measured at moorings FS-1 and 411-2 (FS-3 and 412-4) indicated by arrows in Figure 2a and 2c. M is is obtained by integration of the flow at 79°N (75°N) between moorings FS-1 and FS-3 (411-2 and 412-4). The overbar denotes vertical averaging while indices b and r indicate the bottom flow and the flow relative to the bottom, respectively. Δ stands for the difference (given in parenthesis) between values at 79°N and 75°N.
 The velocity distributions demonstrate that the along-isobath flow on the EGS is generally much larger than the corresponding cross-isobath flow, especially at 75°N where v (Figure 2c) exceeds u (Figure 2d) by an order of magnitude throughout the entire water column. At 79°N, the disproportion between the magnitude of v (Figure 2a) and u (Figure 2b) is not so large, especially in the upper layer at the two easternmost moorings because of the westward flow in the RAC. As a result of a small ratio of the u- and v-component of the near-bottom flow (∼0.05 at 75°N and ∼0.2 at 79°N), the estimates of vb in Table 1 are actually the same as the observed velocities of bottom currents given originally by Woodgate et al.  and Foldvik et al. . Even though the cross-isobath bottom flow is small, it may provide a significant term to the vorticity balance of the EGC [e.g., Schlichtholz, 2005]. Here we focus on the along-isobath component of ub which, on both sections, is extreme at moorings on the ∼1000 m isobath. On that isobath vb increases from 2.6 cm s−1 at 79°N to 11.5 cm s−1 at 75°N, i.e., by Δvb ≈ 9 cm s−1. The corresponding increase for the along-isobath component of is larger (Δ ≈ 16 cm s−1) because of an increase of r from 0.2 cm s−1 to 7.2 cm s−1. On the ∼2000 m isobath, vb increases from 0.4 cm s−1 at 79°N to 10.8 cm s−1 at 75°N while increases from 3.8 cm s−1 to 14 cm s−1, so that Δ ≈ Δvb ≈ 10 cm s−1. Practically the same results for velocity changes along the upper (lower) EGS, i.e., Δvb ≈ 9 cm s−1 (10 cm s−1) and Δ ≈ 16 cm s−1 (11 cm s−1), are obtained from the current observations interpolated along the zonal sections onto the 1000 and 2000 m isobaths at 79°N and 75°N (assuming that vb and on the 1000 m isobath at 79°N are equal to the corresponding values at FS-1).
 Downstream changes of can be caused either by variations in the bottom slope with no transport changes or by a net mass exchange with adjacent areas. The topographic effect is not negligible in the EGC since the ratio of the cross-slope area between the 1000 m and 2000 m isobaths at 79°N and 75°N is ∼2. However, this ratio can explain less than a half of the average velocity increase along the EGS. Nearly two-thirds of the increase should be linked to a change in the along-slope transport. Estimates of M based on the velocity integration between the selected moorings are equal to 2.7 Sv at 79°N and 4.6 Sv at 75°N (Table 1). The downstream change of M is therefore equal to ΔM = 1.9 Sv. This change is due to the transport associated with the bottom flow which increases from 1.2 Sv at 79°N to 3.3 Sv at 75°N, i.e., by ΔMb = 2.1 Sv. The small negative difference of 0.2 Sv between ΔM and ΔMb is certainly within the error of transport estimates based on low-resolution and ‘non-synoptic’ data. Estimates of the transport between the 1000 and 2000 m isobaths based on the interpolated velocities yield comparable results, with ΔM (=2.1 Sv) slightly exceeding ΔMb (=1.8 Sv).
3. Basic Formulae for Geostrophic Currents
 Good approximations to the horizontal and vertical components of the momentum equation for ocean currents away from boundary layers and equator are the geostrophic and hydrostatic balances,
where f ≡ Coriolis parameter, g ≡ acceleration due to gravity, z ≡ vertical coordinate (positive upward), ≡ vertical unit vector, ≡ horizontal gradient operator, ρ0 ≡ constant density, ρ ≡ density anomaly with respect to a depth-dependent reference density ρr(z), p ≡ pressure anomaly with respect to the static distribution associated with ρr, and ug ≡ geostrophic velocity, i.e., horizontal velocity driven by the horizontal pressure gradient.
3.1. Geostrophic Velocity
 Various decompositions of the geostrophic velocity can be obtained by combining equations (1) and (2). For instance, integration of the latter upwards from the bottom (z = −H) to a given level z and insertion of the result into the former yields
where Φ ≡ p∣z=−H is the bottom pressure anomaly. Equation (3) is, for instance, a frictionless limit of equation (6) by Schlichtholz and Houssais [1999c]. It separates the dependence of the geostrophic flow on the local density distribution (contribution uρ) from the dependence on external and remote factors affecting the bottom pressure distribution (contribution uΦ). The velocity uρ is depth-dependent while the velocity uΦ is depth-independent.
 An important feature of the density-dependent flow is that it has a depth-independent contribution, hereafter referred to as the slope velocity (uρb). The slope velocity appears wherever the bottom slope (s = ∣H∣) and the bottom density anomaly (ρb≡ρ∣z=−H) are both different from zero. This follows from the rule of interchange between the differential and integral operators applied to the formula for uρ in equation (3) which gives
The formula for ugr (geostrophic velocity relative to zero flow at the bottom) appearing in equation (4) is classically obtained by upward integration of the thermal wind equation. The latter is derived by cross-differentiation of equations (1) and (2), and hence relates the vertical shear of ug to the horizontal density gradient. While ugr can have both cross- and along-isobath components, uρb is aligned with isobaths as emphasized by Walin et al.  and Nilsson et al. . In topography following coordinates we have
where vρb is positive (negative) if the deep water is to the right (left) when looking downstream while α ≡ −gs(fρ0)−1 combines the dependence of vρb on the bottom slope and the Coriolis parameter.
Equation (7) is an equivalent of equation (9) in Walin et al.  or equation (10) in Nilsson et al. . It shows that the constant of integration of the thermal wind equation, the bottom geostrophic velocity (ugb ≡ ug∣z=−H), is equal to the sum of uρb and uΦ, i.e.,
 It should be noted that the decomposition in equation (8) is not unique as ρb depends on the reference density (ρr∣z=−H). For instance, adding a constant Δρ to ρr(z) results in a local change of vρb by the value of the product of α and Δρ. The change requires a compensation of the same magnitude but opposite sign in vΦ since vgb does not depend on the choice of ρr. However, the difference between vρb at two points (say B and A) on the same isobath can be determined from hydrographic data with less ambiguity than vρb itself. The formula for Δvρb(B, A) ≡ vρb(B) − vρb(A) obtained from equation (5) is
where Δρb ≡ ρb(B) − ρb(A) is the bottom density difference between the points of interest, Δα ≡ α(B) − α(A) is the corresponding difference for the environmental parameter α, while ρc (αc) is the average value of ρb (α) at B and A. Only the second contribution to Δvρb in equation (9), ρcΔα, depends on the reference density. This contribution will always be zero if α is the same at B and A. It will be zero also if the average density at these points is chosen as the reference density. The first contribution to Δvρb in equation (9), αcΔρb, represents that part of the variation of vgb which is uniquely determined by the local density field and can be used for diagnostic estimates.
3.2. Geostrophic Transport
 Vertical averaging of the relative velocity defined in equation (4) yields
where ρ0χ ≡ g∫−H0ρzdz is the anomaly of the potential energy per unit area. Similar formulae can be found, e.g., in the work of Rattray  or Schlichtholz and Houssais [1999c]. The second term on the right-hand side (RHS) of equation (10) has the same magnitude but opposite sign to the slope velocity (uρb) from equation (4). Therefore the total density-dependent transport per unit width (Hρ = Hgr + Hρb) is a function of the potential energy, but not the bottom density, i.e.,
so that the density-dependent transport is nondivergent on an f-plane, i.e., for f = fc = const.
 Since uρb is aligned with isobaths, the along-isobath gradient of χ fully determines the cross-isobath relative transport (Hgr = Hρ). This transport cumulated between any point y and a reference point y = y0 on a given isobath is
 The last term on the RHS of equation (13) disappears on the f-plane. In that case the relative transport between points A and B on the same isobath depends only on the potential energy difference between these points, i.e., Kgr(A, B) = fc−1[χ(A) − χ(B)].
 Consider the total density-dependent transport in a box limited by two isobaths, a shallower one (H = Hs) and a deeper one (H = Hd), and two cross-slope sections, say SA and SB. Let section SB be located forward with respect to section SA when looking in the direction with the deep water to the right (Figure 3). Section SA intersects the shallower (deeper) isobath at point As (Ad) while section SB intersects the shallower (deeper) isobath at point Bs (Bd). The nondivergence of Hρ in this box can be written in the integral form on the f-plane as
where ΔKgr ≡ Kgr(Ad, Bd) − Kgr(As,Bs) is the difference between relative transports across the deeper and shallower isobaths while ΔMρ ≡ Mρ(SB) − Mρ(SA) is the difference between density-dependent transports across sections SB and SA. Equation (14) can be used to estimate the change of the density-dependent along-slope transport (including the contribution ΔMρb from the slope velocity) from values of χ at As, Ad, Bs, and Bd, or to estimate ΔMρb from the imbalance of the relative transport (ΔKgr + ΔMgr). These estimates are independent of the reference density and the bottom slope.
4. Estimates of Geostrophic Flow From Hydrographic Data
 As already mentioned in the introduction, the density-dependent variables are calculated using the annual mean temperature and salinity from the PHC [Steele et al., 2001]. The latter is a product of merging two data sets of objectively analyzed, heavily smoothed hydrographic fields. The PHC data are available on a 1° lon ×1° lat grid and have a vertical resolution decreasing with depth from 10 m in the upper 30 m layer to 500 m below the 2000 m level.
4.1. Relative Flow
 To estimate the relative geostrophic flow from equation (10), first χ and ρb are calculated at the PHC grid points. Then the zonal and meridional velocity components are estimated at half a distance between the adjacent grid points and interpolated onto a common 1° lon ×1° lat grid. The distribution of gr obtained in the area west of the Greenwich meridian shows that a large vertically sheared flow in the EGC occurs on the shelf (Figure 4a). However, the largest relative transport (Hgr) is concentrated on the EGS (Figure 4b).
 Distributions of the along-isobath component of gr, estimated by projecting the vectors from Figure 4a on the direction of local isobaths and then interpolated onto the 1000 and 2000 m isobaths, are shown in Figure 5. On the upper slope, the magnitude of gr increases from ∼1 cm s−1 at 79°N to ∼10 cm s−1 at 75°N (Figure 5, circles). The increase Δgr ≈ 9 cm s−1 is close to the corresponding estimate Δr = 7 cm s−1 based on the current meter data (Table 1). On the lower slope, Δgr between 79°N and 75°N is negligible (Figure 5, crosses). This is also in agreement with the observations, although the values of gr at 79°N and 75°N (1–2 cm s−1) are smaller than the corresponding estimates of r from the current meter data (∼3 cm s−1).
4.2. Slope Velocity
 As pointed out in the introduction and further discussed in section 3.1, a suitable variable for studying variations of the density-dependent contribution to the bottom geostrophic velocity is the bottom density. Distributions of ρb on the EGS along the 1000 and 2000 m isobaths (Figure 6a) are obtained by an interpolation of the PHC data followed by a slight along-isobath smoothing intended to suppress point-to-point fluctuations. On both isobaths, ρb shows a southward densification from Fram Strait to a maximum at ∼75°N and then a southward rarification. According to equation (5), such a distribution corresponds to a slope velocity maximum at ∼75°N. The bottom density difference between 79°N and 75°N is Δρb = −0.04 kg m−3 on the upper slope (Figure 6a, circles) and Δρb = −0.02 kg m−3 on the lower slope (Figure 6a, crosses). The bottom density decrease along the 1000 m isobath south of 75°N toward Denmark Strait is as large as the corresponding increase from Fram Strait to 75°N.
 The slope velocity depends not only on ρb, but also on the bottom slope and the Coriolis parameter. Variation of the latter are small in the area considered. For instance, the relative departure of f from a constant value (fc = 1.4 × 10−4 s−1) between 75°N and 79°N is only 2%. Of course, the bottom slope varies more. Distributions of s on the EGS along the 1000 and 2000 m isobaths (Figure 6b) are obtained from the ETOPO5 data. The average value of the slope at 75°N and 79°N is sc = 0.05 on the 1000 m isobath and sc = 0.04 on the 2000 m isobath. These values can be combined with the estimates of Δρb made above and other parameters (ρ0 = 1027 kg m−3 and g = 9.8 m s−2) to yield estimates of the slope velocity change between 79°N and 75°N based on the contribution αcΔρb to Δvρb in equation (9). One obtains αcΔρb = 13 cm s−1 for the upper slope and αcΔρb = 6 cm s−1 for the lower slope. These values are close to the estimates of the bottom velocity change (Δvb) from the current meter data (Table 1). On the lower slope αcΔρb is 40% smaller than Δvb while on the upper slope it is 45% larger than Δvb.
4.3. Density-Dependent Transport
 As discussed in section 3.2, a suitable variable for studying variations of the density-dependent geostrophic transport is the potential energy. Distributions of ρ0χ interpolated from the PHC grid onto the 1000 and 2000 m isobaths of the EGS are shown in Figure 7a. On both isobaths, ρ0χ decreases from Fram Strait to ∼75°N and then increases farther south. As a consequence, the relative geostrophic flow across both isobaths is onshore from 79°N to ∼75°N and offshore farther south. The onshore transport across the deeper isobath, cumulated between 75°N and 79°N using equation (13), is equal to 3.1 Sv (Figure 7b, crosses). The corresponding estimate for the transport across the shallower isobath is only 1.4 Sv (Figure 7b, circles). The difference (ΔKgr = −1.7 Sv) should be compensated by an increase of the density-dependent along-slope transport from 79°N to 75°N since variations of the Coriolis parameter are negligible. Equation (14) then yields ΔMρ = 1.7 Sv. This value is very close (to within 0.2 Sv) to the increase ΔM of the total along-slope transport between the ∼1000 and ∼2000 m isobaths obtained from the current meter data (Table 1).
 The distribution of Kgr along a given isobath shows divergences and convergences of the density-dependent along-slope transport between this isobath and the coast, where χ = 0. In particular, the distribution in Figure 7b (circles) indicates that the onshore relative geostrophic transport across the 1000 m isobath between 79°N and 75°N (equivalent downstream increase of Mρ) is fully compensated by the corresponding offshore transport between 75°N and 69°N (equivalent downstream decrease of Mρ).
5. Discussion With Concluding Remarks
 Since the ice cover shields a major portion of the EGC from the atmosphere [e.g., Aagaard and Coachman, 1968; Mauritzen, 1996b], the density distribution on the EGS should mainly be influenced by lateral exchanges with deep basins. As already mentioned in the introduction, dense intermediate and deep water masses are formed in the Greenland Sea gyre and spread onto the EGS. One can therefore expect to find a local density maximum on the lower as well as upper EGS when looking in the along-isobath direction. A density maximum on the EGS at ∼75°N has been evidenced in the present study using climatological distributions of the bottom density (Figure 6a) and potential energy (Figure 7a) based on the PHC data [Steele et al., 2001]. These are important variables since their along-isobath variations determine the corresponding variations of the density-dependent part of the bottom geostrophic velocity and the along-slope variations of the total density-dependent transport as expressed by equations (9) and (14), respectively. The presence of dense waters on the EGS at ∼75°N implies that the downstream feeding of the depth-independent along-isobath flow in the EGC by the depth-dependent cross-isobath (onshore) flow ceases at this latitude and that farther south it is the depth-independent flow which feeds the depth-dependent (offshore) flow. Moreover, a same magnitude of the bottom density change on the upper EGS from 75°N toward Fram Strait and Denmark Strait indicates that a significant difference in the depth-independent flow between 79°N and 69°N, if exists, cannot be related to density variations. A corresponding compensation in the potential energy change further demonstrates that the transport variations in the EGC between 79°N and 69°N are part of a density-dependent internal cyclonic circulation in the Nordic Seas.
 A check against long-term current observations on the EGS at 79°N from Foldvik et al.  and 75°N from Woodgate et al.  has revealed that the along-isobath density variations are indeed relevant for the dynamics of the EGC. The current meter data yield an estimate for the downstream bottom velocity increase equal to ∼9 and ∼10 cm s−1 along the 1000 and 2000 m isobath, respectively (Table 1). The corresponding geostrophic estimates based on the bottom density change are higher on the upper slope and lower on the lower slope (Figure 3), but their average value perfectly fits the observations. Similarly, a very good agreement is found between estimates of the downstream transport increase between the 1000 and 2000 m isobaths. The total increase of ∼2 Sv obtained from the current measurements is mainly associated with the bottom flow. The corresponding transport increase calculated from the potential energy distribution is lower only by ∼15%. Such a good agreement may seem somehow fortuitous given the nonsynopticity of the data and sampling errors. However, all estimates are consistent. Even the differences between the southward bottom velocity increase on the individual isobaths obtained from the current and bottom density data can be explained by the geostrophic dynamics. If interpreted as changes of the flow associated with the bottom pressure distribution, they represent a southward increase of this flow on the 2000 m isobath and a northward increase on the 1000 m isobath, both equal to ∼4 cm s−1. The associated transport increase should be small and southward as obtained by the difference between the estimates from the current meter and potential energy data.
 The author acknowledges the Polar Science Center (Seattle, Washington, United States) for providing the hydrographic data and the Academic Computer Center in Gdansk TASK for computational support. Reviewers comments helped to improve the presentation greatly.