A small anticyclonic eddy of extraordinary intensity sits in the center of the Lofoten Basin near 69°40′N, 3°E. This paper gives a first detailed description of its kinematics and lowest-order dynamic balances. Using a 75 kHz vessel-mounted acoustic Doppler current profiler, hydrography, and six RAFOS floats to probe the eddy, we document a solid body core with 7–8 km radius and relative vorticity very close to –f, where f is the local Coriolis parameter. Maximum orbital velocities close to 0.8 ms−1 were observed at 18 km radius. One float, trapped in the core of the eddy for 9 months, indicated undiminished strength throughout that period, possibly even intensifying in winter. Hydrography revealed adiabatic conditions from the bottom of a shallow seasonal thermocline to 1000 m depth in early July 2010. Thermal convection in winter maintains an already deep pycnostad, and may also play a key role in intensifying the eddy, quite possibly through penetrative convection that deepens and sharpens the underlying pycnocline. Heat lost to the atmosphere has to be replenished from warm anticyclonic eddies shed off the eastern branch of the Norwegian Atlantic Current, but the mechanism(s) by which it is added to the eddy remains to be studied. Examination of historical data sets suggests the eddy is a permanent feature of the Lofoten Basin. Two hydrocasts from the 1960s also show a similar adiabatic mixed layer albeit 1°C cooler than in 2010, perhaps reflecting the generally cold winter conditions that prevailed then.
 The Nordic seas, which stretch from the Greenland-Iceland-Scotland ridge in the south to the Fram St. in the north are characterized by a remarkable east-west organization of water: cold water flowing south along Greenland and warm water flowing north along Norway. Thus, the Greenland and Iceland seas are cold. In striking contrast, warm water from the northeast Atlantic flows in along two branches in close proximity just west of Norway. These two topographically guided branches of the Norwegian Atlantic Current delimit or define the extent of warm water in the Nordic seas: the inner branch that flows along the continental slope of Norway to the Barents Sea and Svalbard. The outer strongly baroclinic branch flows north along the eastern margin of the Norwegian Basin, and then along the southwestern and northwestern margin of the Lofoten Basin. Whereas the Norwegian Basin is a quiet, cold sea strongly shaped by cold water emanating from the Greenland and Iceland seas, the Lofoten Basin, is strikingly warm and dynamic in character. The consequence of this “inverted” arrangement is that this northern basin pools and loses more heat per unit area than any other region in the Nordic seas [Isachsen et al., 2007; Segtnan et al., 2011]. Figure 1 shows the eastern and western branches of the warm Norwegian Atlantic Current (NwAC) and the depth of the specific volume anomaly surface δ = 21 × 10−8 m3 kg−1 representative of the main pycnocline. Its great depth in the Lofoten Basin reflects the pooling of warm water [Rossby et al., 2009a]. Because the Lofoten Basin is bounded by the two warm, nutrient-rich flows of North Atlantic origin, it and the Barents Sea to its east, are highly productive bodies of water that have seen intense fishing since the middle ages. Consequently, the Lofoten Basin and Barents Sea are of great interest to society. Hydrographic and fish stock surveys have been conducted regularly for many decades, and these continue today, especially of the near-coastal waters. The reader is referred to Skjoldal  for an excellent overview of the ecology of the Nordic seas.
 Figure 1 also reveals a conspicuous localized deepening of the pycnocline near the center of the Lofoten Basin at 70°N 3°E. A number of papers have examined this area in some detail. Alekseev et al.  use hydrography to show that there exists a strong positive dynamic height anomaly in this area, and Ivanov and Korablev [1995a, 1995b] describe this feature as an intrathermocline lens, a small and anticyclonic eddy with a deep pycnocline, a thick, weakly stratified core, and a less-developed surface pycnocline than the surrounding waters. They envisioned penetrative convection as the mechanism by which the eddy is intensified in winter, and assumed without elaboration that heat is advected into the area (in summer). Köhl  argued on the basis of altimetric data and numerical simulations that this anticyclonic structure is fed by anticyclonic eddies shed from instabilities in the inner branch, where the continental slope becomes exceptionally steep offshore of the Lofoten Islands (Lofoten Escarpment). More recently, Rossby et al. [2009a], using a 50 year hydrographic database, showed that (i) the thermocline is consistently deeper in the central Lofoten Basin than anywhere else in the Nordic seas and (ii) there appears to be an eddy-active path back to the Lofoten Escarpment pointing to the inner branch as the most likely source of its energy (Figure 1). However, Volkov et al.  suggest that the eddies formed at the Lofoten Escarpment reach the center of the Lofoten Basin following a cyclonic path.
 The high eddy kinetic energy (EKE) levels in the Lofoten Basin have been documented in several Lagrangian studies [Poulain et al., 1996; Jakobsen et al., 2003; Rossby et al., 2009b; Koszalka et al., 2011]. Lagrangian techniques have the advantage that they can map out the spatial structure of EKE in some detail. A comparison of EKE between the southern and northern basins of the Norwegian Sea is instructive. While both basins are subject to comparably intense and variable winds in winter, the average surface EKE of the southern basin is only a fraction of that in the Lofoten Basin. Knowing that the high heat loss must be fed by the warm boundary currents, the high EKE levels implicate mesoscale processes by which this heat may be exported to the interior of the basin. Rossby et al. [2009b] noted that RAFOS floats in the eastern branch of the NwAC were likely to be expelled into the Lofoten Basin in the vicinity of the Lofoten Escarpment. Recently, Koszalka et al. , using a new surface drifter data set and a novel analysis technique (cluster analysis), were able to highlight several distinct features of the Lofoten Basin including a striking anticyclonic feature located at 70°N 3°E, consistent with the earlier hydrographic findings. From the above studies, one can conclude that there exists a local yet permanent anticyclonic eddy of rather small dimensions that it is maintained by warm anticyclonic eddies that break away from the eastern branch at the Lofoten Escarpment.
 Whereas past studies have focused on the hydrographic properties of what we will call the Lofoten Basin Eddy (LBE), here we will examine its kinematic properties using a shipboard acoustic Doppler current profiler (ADCP), and acoustically tracked floats. The former maps out the velocity structure in considerable detail while the latter give us insight into the eddy's movement and temporal development. Several CTD casts were also taken to map out the temperature and salinity field. The shipboard survey took place in early July 2010, a fortuitous time in that it followed an extremely cold winter such that the eddy must have been subject to large heat losses during the preceding winter months. The next section lists the data sets and outlines the analysis methods that will be used in this study. The section thereafter presents the findings. A discussion followed by a brief summary concludes the paper.
2. Data and Methods
 The fieldwork comprises two parts, a shipboard study conducted on the RV Håkon Mosby, and a Lagrangian study using six RAFOS trajectories lasting up to 11 months following their deployment during the shipboard study. The key tool of the shipboard survey was a 75 kHz Teledyne RD Instruments ocean surveyor ADCP that profiled currents approximately every 3.3 s to about 700 m with 16 m vertical resolution. The single ping accuracy of velocity according to the manufacturer to is ±0.12 ms−1 (http://www.rdinstruments.com/surveyor.aspx). This ensured that the 5 min averaged ADCP velocities (18 × 5=90 pings) were accurate to ±0.12/√90 ms−1 = ±0.013 ms−1. A Kongsberg Seapath GPS compass gave vessel heading to better than 0.1°T such that velocity error due to heading error = sin(0.1°) × 5 ms−1 (=10 Kt) vessel speed = 0.009 ms−1. The calibration parameters for the ADCP (gain and heading) had been done earlier (in the bottom track mode), and were part of the system setup for this cruise. Root mean squared (RMS) velocities in the tidal/inertial band are O(0.04) ms−1 measured at a mooring nearby (69°39′N, 6°58′E; C. Richards, personal communication, 2012). The hydrographic data consist of nine conductivity, temperature and depth (CTD) casts to 1500 m depth. The calibration of the instrument, Sea-Bird Electronics 911plus, is such that the pressure, temperature, and salinity data are accurate to ±0.5 dbars, ±0.002°T, and ±0.003 psu, respectively.
2.2. Estimation of Eddy Energy and Vorticity
 The analysis methods used here follow directly from Rossby et al. . Potential energy in the eddy can be obtained directly from the geopotential of a fluid parcel Φ = gz, where Φ represents the work required against gravity to raise a unit mass a distance z. Relative to a field at rest outside, Φ will be positive for anticyclones, and negative in cyclones. What makes this statement so useful is that Φ(r,z) can be determined directly from a radial integration of the velocity field terms in the gradient wind equation (the motion is assumed steady and axisymmetric):
 Multiplying Φ by ρ and integrating in the vertical, we get the potential energy of a water column per unit area relative to some background profile Φb:
where the integral runs from a reference depth to the surface and ΔΦ, a function of z and r, represents the departure of the observed geopotential surface from its background value at the same depth. To estimate potential energy per unit area for a layer, between the surface (s) and some selected depth (i), say, the reference depth is not needed:
 A further integration across the eddy gives its total potential energy (TPE):
 This quantity can therefore be deduced from the velocity field only; it is positive in anticyclonic and negative in cyclonic eddies.
 The total kinetic energy (TKE) in the eddy is obtained directly from the directly measured velocity field:
where the external integral assumes the velocity field decays to zero outside the eddy. In practice, some judgment must be exercised to decide where that cutoff should be. The relative vorticity follows directly from the velocity field. Again, assuming rotational symmetry, we have
2.3. Lagrangian Trajectories
 The RAFOS float deployments were part of a larger study of the Lofoten Basin circulation, which will be reported elsewhere. For this purpose, seven sound sources have been deployed around the basin to provide an underwater acoustic navigation system [Rossby et al., 1986, 2009b]. The floats discussed here were deployed at the center and 17 km from the center of the eddy and were programmed such that two would surface after 2 months and the remainder after 11 months. Six floats returned excellent data.
 We begin with the ADCP data because, as will be seen, these played the key role in locating the exact center of the eddy. This is followed by the hydrographic survey. The RAFOS trajectory data give us considerable insight into the intensity of the eddy, its temporal development and movement over the following months beginning in July 2010.
3.1. ADCP Data
 The shipboard survey of the LBE took place 3–4 July. Figure 2 shows the vessel track and the ADCP vectors at 44 and 700 m depth, respectively. The track, running clockwise around, through the eddy center, and finally off to the east, spans 53 h. Assuming axial symmetry the route was set up to enable us to determine the center of the eddy from the intersection of lines normal to the measured velocity vectors at several widely separated points. The estimated center at 70.07°N, 3.65°E puts it very close to its historical mean. CTD casts were taken at the sites marked with a dot including one at 70.07°N, 3.58°E, very close to the ADCP estimated center. The vectors reveal a well-defined anticyclonic circulation. The longer red than blue arrows show the circulation is more intense at 700 m depth than near the surface. Figure 3 shows the north-south velocity along the line extending east from the eddy center. The velocity field shows strong radial organization with gradual increase in magnitude from the surface to the ∼700 m maximum range of the ADCP.
 The velocity data allow one to diagnose the eddy in considerable detail. By integrating equation (1) as a function of radius at various depths, we can map out the geopotential anomaly field Φ(r,z) throughout the eddy, Figure 4.
 The relative vorticity field, determined directly from the velocity field using equation (6), is shown in Figure 5. The profile at origin has been deleted due to the indeterminateness of v/r as r → 0. An interesting feature about the figure besides the tight center with its increasingly negative vorticity at depth are the diffuse bands of positive vorticity at ∼50 and 80 km distance from the center. Similar vorticity bands have been observed in another intense anticyclonic eddy in the Sargasso Sea at 30°N, 70°W [Rossby et al., 2011]. One can also determine relative vorticity at the center by fitting a biquadratic stream function Ψ to all vectors within 7 km radius where solid body rotation appears to apply. This can be done independently for each layer. The resulting profile is shown in Figure 6. It shows that the eddy intensifies with depth, especially between 200 and 300 m, and exceeds −0.9f below 500 m depth (f is the local Coriolis parameter). Relative vorticity is weaker at the surface consistent with the reverse shear from the shallow pycnocline there relative to the surrounding waters (next section).
3.2. Hydrographic Data
 Figure 7 shows the measured temperature profiles at the three innermost sites, 0, 17, and 20 km from the center of the eddy. The center profile is very close to adiabatic between 100 and 1000 m depth. The next two profiles show a well-defined pycnostad, but being outside the core they are not adiabatic. All three profiles are terminated by a sharp pycnocline underneath. They also show that the surface thermocline is thinnest at the center of the eddy.
 Figure 8 shows a contour plot of all nine CTD stations taken in the vicinity of the eddy as a function of depth and distance from CTD 813 (the center CTD station in Figure 2), assuming the eddy has rotational symmetry. (Three of the casts appear close together at 70–75 km. Their similarity even at that range attests to the rotational symmetry of the eddy.) The top plots one and two highlight the very thin thermocline and cooler and denser surface waters at the center of the eddy relative to the surrounding waters (6.5 versus ∼8°C), whereas at 1000 m depth, the core temperature is 5° higher than the surrounding water. One sees clearly the pycnostad in the center and the reverse slopes of the shallow pynocline relative to the main pycnocline. This explains qualitatively the weakening of relative vorticity at the surface.
 The difference in dynamic height between R = 0 and R = 90 km (bottom left) is about 0.47–0.3 = 0.17 dyn, m. The ADCP gives a difference of 0.3 m at the surface (Figure 4), such that at 1500 m there is 0.13 m excess at the center relative to 90 km. This means that the dynamic method integrated to 1500 m depth misses a large part of the total mass circulating in the eddy; the eddy has a deep component rotating in the same direction. To estimate the required reference field at 1500 m so that this plus the dynamic estimate matches the ADCP estimate, we fit Gaussian functions to both dynamic height functions, Figure 9, which shows both the original estimates and the fits. (See figure legend for details.) Both dynamic height curves have been shifted so that they pass through 0 near 60 km radius. This was done to reduce influence from the surrounding waters and results in a good curve fit to the data. The difference between the two fits is shown in black. The Gaussian function is often used to characterize the structure of coherent eddies. Thus, assuming a velocity field of the form , we find that with A = 0.53 and R = 31 km, the corresponding dynamic height field matches the difference field well. This v(r) is the reference velocity field and is shown in Figure 9 as the negative dashed curve (in ms−1). We return to the question of the eddy's vertical extent in the discussion below.
3.3. Lagrangian Observations
 Six floats were deployed in the eddy. They were ballasted for the specific volume anomaly surface δ = 21 × 10−8 m3 kg−1, but we did not realize at the time that this surface could go so much deeper than shown in Figure 1. Since the isopycnal capability of the floats did not go that deep, the floats equilibrated at 700–900 m depths, and deepened over time (for reasons not yet fully resolved, but most likely due to creep). As a result, the floats orbit in the adiabatic core at first, and eventually either sink into the stratified layer below, or escape into the surrounding waters. Once outside, they may shoal several hundred meters due to the high density of the surrounding cold water. The traditional first step with RAFOS float data is to determine the float trajectories. This proved to be extremely difficult if not impossible to do for the floats deployed in the eddy center due to their unexpectedly tight and rapid looping. The reason is that floats in the center of the eddy have such a small orbit that the time of arrival variations are barely larger than the inherent scatter of the arrival times. That coupled with the orbital period just below the Nyquist frequency precluded tracking the small orbit floats. Floats deployed away from the core center have a large enough orbit and a longer period rendering them trackable. Figure 10 shows the 11 month trajectory of float 931. Deployed at a radius of 18 km, essentially at the radius of maximum velocity, it loops near that radius for the first 6 months after which it moves to the outer perimeter of the eddy before final escape about 2 months later.
 Despite the challenge of tracking the floats, the tracking or time-of-arrival (TOA) data can still be used to determine the orbital period, and this has been done for all floats while trapped in the LBE. As an example, Figure 11 shows the TOA data received by float 926 from one of the sources. Using these time series, we have estimated looping period and radius by means of a least-square sine/cosine fit to typically 7–8 days of data at a time. This cannot be done at all times due to missing data or due to poor fits, perhaps because the float is undergoing a radial shift. Nonetheless, the TOA data from all floats suffice to plot the radial structure of velocity, Figure 12. Only those period/amplitude estimates for which the RMS error in the fit was less than 1.5 s (∼2.2 km error in distance) were used in the figure. The two top plots show looping period and radius of the various floats as a function of time for the first 250 days. This can be seen by following a float, say 921 (red) or 924 (green). Note how it remains at roughly the same radius (top right) and loops with about the same period (top left) for over 90 days. They both show that the floats tend not to move about radially in the eddy; whatever period and radius they had at deployment tends to prevail thereafter. This is reminiscent of the slowness with which the 2 year SOFAR float in meddy “Sharon” [Richardson et al., 1989] moved about radially in the core. The bottom two plots show period and computed velocity as a function of radius. They both show that floats inside a radius <∼7 km had a looping period very close to the 25.5 h twice inertial period (indicated by the dotted line). The maximum looping velocity of ∼0.8 ms−1 near 1000 m depth is quite large compared to typical surface eddy velocities in the Lofoten Basin [Orvik and Niiler, 2002; Koszalka et al., 2011]. The floats in the core have depths between ∼750 and 900 m at the start and gradually sink over time toward 1000 m depth. Regardless of depth, the three floats at small radii exhibit a looping motion extremely close to –f, which is the theoretical limit for anticyclonic relative vorticity [Holton, 1992]. Recall that the relative vorticity consists of −v/r (what you see from a float's orbit) and −∂v/∂r (which is not apparent). In solid body rotation the latter equals the former and thus the relative vorticity becomes −2v/r or 2ω (where ω is the angular velocity). Only float 921 remains close to the center, but quality TOA data are few when elevated ambient noise levels due to heavy winter weather prevent signal detection. Nonetheless, the TOA data show the LBE to be at undiminished strength more than 200 days later in midwinter. In fact, the sustained short looping period from float 921 (red stars and circles) between days 170 and 230 suggest that the eddy may be reintensifying in winter (they lie closer to the –f/2 line in the bottom right than other floats at similar radius), which would be consistent with Ivanov and Korablev's [1995a] thesis that intense cooling and resultant penetrative convection deepens the pycnocline and increases the baroclinicity of the eddy.
 The pressure-temperature record from 921 is instructive, Figure 13. It shows a rock-solid temperature at 4.96°C until day 150 (early December) at which time small cold fluctuations begin to appear. The cause of the gradual sinking of the float during this time remains to be determined, but it is instrumental and not due to oceanic effects. A cooling event occurs on days 180–200. Given the January date, these cooling events might reflect cooling from above, but their (5) day duration at 1000 m depth argues against convection driven by heat loss at the surface, since this would be characterized by fast time scales and significant vertical displacements [D'Asaro et al., 1996]. It seems more likely that at this radius the float is making contact with a band or layer of cooler water stirred in from outside the eddy. Finally, around day 270, about 9 months after deployment (∼1 April 2011), the temperature drops precipitously and the float shoals, both indicating the float's departure from the core waters.
 The TOA data from float 921 can also be used to track the movement of the eddy as a whole, what we might call eddy wobble. Since this motion is much slower than that of the looping, it can be determined using an 8 day low-pass filtered version of the tracking data from float 921, which remained in the LBE the longest. The result, shown in Figure 14, shows that the eddy center is constrained to a roughly 1° latitude box and thus coincident with the deepest part of the Lofoten Basin. That the LBE sits at the deepest spot of the Lofoten Basin most likely reflects the tendency for anticyclonic eddies to drift toward areas of low ambient potential vorticity (f/H) to compensate for their loss of negative relative vorticity due to Rossby wave radiation (J. Nycander, personal communication, 2012).
 This study is the first one to document in some detail the velocity structure of the LBE, but there is evidence that it is a long-term or permanent feature of the Lofoten Basin. The 1957–1958 International Geophysical Year hydrographic atlas of the North Atlantic [Dietrich, 1969] shows a distinct localized deepening of the main thermocline in the vicinity of the LBE (stations 180–182 on section 12). A similar deepening also shows up in the Koltermann and Lüthje atlas . Even the Helland-Hansen and Nansen  study includes a hydrographic station (M. Sars 43 in August 1900 at 69°52′N, 5°15′E that shows a distinct pynostad between 300 and 600 m. Thus, all three studies give evidence of a deep thermocline in the area of the LBE, but none of them have the station density to permit identification, much less a description of a locally intensified anticyclonic eddy.
 The International Council for the Exploration of the Sea (ICES) hydrographic database includes three stations from the 1960s that show a thick pycnostad (two of them essentially adiabatic) to 1 km depth, Figure 15.
 During the 1960s the Lofoten Basin experienced a number of cold winters characteristic of low North Atlantic Oscillation (NAO) conditions. Perhaps this accounts for the lower core temperatures then than now and the decrease in core temperature in Figure 15 between 1964 and summer 1969, which followed two very cold winters in succession [Rossby et al., 2009a]. In fact, all three stations come from zonal sections along 69.5°N (RV Mikhail Somov in March 1964; RV Fritjof Nansen, June 1967; RV Professor Vize, August 1969). The 1964 and 1967 sections have casts to 1000 m only while the 1969 casts reach almost to the bottom, Figure 16. Unfortunately the salinity data in the 1969 section exhibit such erratic scatter (toward higher values) that we replace them using a temperature-salinity relationship defined by the June 1967 section. From a dynamical point of view, which considers only density differences, the exact T/S relationship matters little, but removal of the scatter makes for a more reliable estimate of baroclinic transport in the eddy. Using a 1500 m reference level the transport is 14 Sv compared to 12 Sv in Figure 8, indicating a very similar intensity. Using 2500 m reference level the transport increases to 25 Sv. This section is helpful in gaining a view on the deeper part of the LBE. Assuming that these two eddies have similar strength, it is interesting to note that the dynamic height anomaly at 2500 m depth for the three stations defining the eddy in Figure 16 is 0.237, 0.477, and 0.234, respectively. The dynamic height anomaly in the center of the eddy is 0.24 dyn. m, greater than the two stations 50 km to either side. This excess matches almost perfectly the dynamic height anomaly extremum at the surface estimated from the ADCP, which means (again assuming the eddies are of similar intensity) that there is little velocity signal at 2500 m depth. These two sections took place following cold winters (during the low NAO 1960s) and following the very cold 2010 winter. We do not know what the core of the eddy might look like following a string of warm winters.
 The ICES database for the late 1980s and early 1990s—a period of high NAO winter—comprises some 180 stations in the 69–71°N, 0–5°E box. There are numerous stations with pycnostads to ∼600–800 m depth, but only one adiabatic profile at 4.3°C (cast to 500 m only) in April 1994 (69.62°N, 1.35°E). The lack of stations with adiabatic or weak stratification, such as in Figures 7 or 15, might suggest a weakening of the LBE in warmer winters, but the low station density near 70°N 3°E leaves the question unresolved. In conclusion, the available data suggest that the LBE is always there, but we have little information on how its size and intensity vary over time. Repeat visits to the LBE could shed considerable light on its seasonal to interannual variability.
 Estimating the energy sources and sinks for the LBE is straightforward. We consider first its TPE and EKE, both of which can be estimated directly for the July 2010 survey using equations (4) and (5) given the velocity field to 700 m depth, and Φ(r,z) from equation (1). Thus, TPE to 700 m =1.08 × 1016 J and EKE = 6.0 × 1014 J for a total of 1.14 × 1016 J. Assuming a uniform velocity field to at least 1 km depth we obtain 1.14 × 1016 × 1000/700 = 1.6 × 1016 J ∼ 342 × 106 Watt years. Mechanisms by which this energy can be dissipated include (i) internal friction and (ii) frictional losses at the bottom (ignore since no bottom contact). We don't know of any internal friction measurements in the LBE, but using 4 × 10−9 Watts/kg (estimated from the anticyclonic meddy “Sharon,” Hebert et al. ) as a guide, it would take many decades for internal friction to spin down the eddy.
 The LBE contains a huge reservoir of available thermal energy thanks to the deep pool of high temperature water in its core as well as the weak stratification that permits its heat to be lost to the atmosphere in winter. The excess heat comes from a general deepening of the main thermocline represented by the 4°C isotherm (from 700 to 1100 m). Assuming a temperature anomaly of +3°C relative to the surrounding waters, the LBE has per unit area a thermal excess of 3–4°C × 4 × 106 J°C−1m−3 × 400 m deepening = 4.8 × 109 Jm−2. The fact that the eddy has a thinner seasonal thermocline than the surrounding waters means it is preconditioned to expose the central core in winter to deep convection. A recent estimate for the mean heat loss rate in the Lofoten Basin is 100 Wm−2 [Segtnan et al., 2011]. In reality the heat loss would be much larger in wintertime and have opposite sign in summer. Nonetheless, at this annual rate only 4.8 × 107 s ∼ 1.5–2 year would be required to lose this heat excess. This very rapid heat loss rate suggests that the eddy has to be maintained quite steadily for it not to be wiped out. If the LBE is permanent, as we have suggested, these numbers imply a very active process of regeneration and loss on an annual cycle. In light of this, it seems rather remarkable that the relative vorticity of the LBE stays so close to its theoretical extremum (−f) over the 9 month period of observation (with float 921).
 The supply of heat must be the anticyclonic eddies (ACE) that drift toward the LBE from the Lofoten escarpment as noted by Köhl  and Rossby et al. [2009a]. A budget calculation of the supply rate must necessarily be approximate at this stage since we have no information on what might be an average ACE nor the rate at which they merge (if they do) with the with LBE. Figure 18 in Rossby et al. [2009a] shows a hydrographic section through an ACE, which is nearly as wide as the LBE in Figure 8. The thermostad is warmer between 6 and 7°C while the deeper isotherms 5 and 4°C dip at most 300 m. The size of this eddy, if typical, suggests that only a very few would be needed to maintain the LBE. How this heat is actually added to and redistributed in the LBE will be very interesting to study.
5. Concluding Remarks
 The LBE is a small feature, about 1/10 the scale of the basin itself. Its core of homogeneous water has a 7–8 km radius (comparable to the local radius of deformation, Koszalka et al. ) and reaches to >1000 m depth. The measurements reveal the core to be in solid body with a relative vorticity very close to –f, and maximum orbital velocities near 0.8 ms−1 at 18 km radius. The direct measurement of velocity has been used to determine its potential and kinetic energy and the hydrographic casts to determine its heat reservoir. Heat loss to the atmosphere in winter due to thermal convection clearly plays the key role in weakening the eddy over time. However, the float data do not show the eddy spinning down in winter; on the contrary, there is an indication the LBE may reintensify in the winter months, perhaps through penetrative convection that deepens and sharpens the underlying pycnocline. The supply of heat is advected to the eddy via anticyclonic eddies shed off the eastern branch of the NwAC, but the mechanism(s) by which this heat is added to the LBE remains to be studied. Based on the historical database the LBE appears to be a permanent feature, but it is also clear that the core temperature varies in time, perhaps as a low-pass filtered record of past atmospheric conditions. There is much to be learned here.
 We are grateful to the Institute of Marine Research for providing shiptime for the field program. We also thank the Bjerknes Centre for Climate Research for its support. Jim Fontaine was responsible for the expert ballasting of the floats. This research was funded by the National Science Foundation under grant NSF-0850609. We thank the three anonymous reviewers for their very helpful comments and suggestions contributing to a much improved paper. This is publication A426 from the Bjerknes Centre for Climate Research.