Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 Measurements of 7Be and oxygen isotope (18O/16O) ratios from the 1997–1998 SHEBA experiment were used to trace the source of the high heat and freshwater content of the upper ocean observed during the initial occupation of the SHEBA site in October 1997. The evidence suggests that the heating resulted from local input primarily through extended lead coverage in the late spring and summer of 1997 with no requirement of advective input. The freshening was derived from a large ice melt (1.2 m) that was consistent with the thin ice and extensive melt pond coverage (by then frozen) observed at the site. However, a significant contribution to the freshwater budget (0.8 m) included enhanced input from river runoff during the melt season. This obviates the requirement for an unrealistically large ice melt (∼2 m) to account for the freshwater content of the mixed layer, and would have increased the stratification stability of the upper ocean that in turn would have promoted local heating. The question then arises, however, as to the fate of the significant upper ocean heat at SHEBA in the fall 1997 which resulted from an active heating season. Similar evaluation of the fall 1998 SHEBA site indicate that the ice melt was comparable to that of 1997, but the riverine input and stored water column heat were less than in the previous year.
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.
 An objective of the SHEBA (Surface Heat Budget of the Arctic Ocean) field program undertaken between October 1997 and October 1998 was to study the summer surface energy balance of the Arctic Ocean and feedbacks associated with the ice-water albedo system. Downwelling solar radiation through leads is the primary source for the direct heating of the ocean, but the fate of the solar energy absorbed through the leads is not well understood [Rind et al., 1995]. A goal of the program therefore was to understand the partitioning of heat between lateral melting of ice in contact with the leads, conveyance to depth where it melts the ice from below, or transport away from the ice. The nature of the partitioning has direct consequence upon the persistence of the ice cover and the associated summertime albedo [e.g., Ingram et al., 1989].
 Observations during the fall 1997 SHEBA deployment indicated that the upper ocean (<70 m) was less saline and warmer than expected with a temperature above freezing (δT) 2.5 times greater than that which was measured during an earlier program, the Arctic Ice Dynamics Joint Experiment (AIDJEX) in 1975. Additionally, the mixed layer salinity in 1997 was approximately 27.6 ppt, compared to 29.7 ppt in 1975 [Maykut and McPhee, 1995; McPhee et al., 1998]. These observations suggested that there had been considerably greater oceanic heat flux during the 1997 summer than during the earlier program. In addition, the ice conditions observed at SHEBA in 1997 included thinner ice than anticipated (∼1.2 m as opposed to 3–4 m), extensive melt pond (by then frozen) coverage, and melt ponds that had melted through to the underlying ocean (D. Perovich, personal communication, 2000). Because the albedo of open water is considerably less than that of sea ice, it was suggested that the percentage of open water (i.e., leads) in 1997 was triple that of 1975 to allow a greater amount of heating to occur [McPhee et al., 1998]. Consistent with this, measurements and modeling of the naturally occurring, short-lived (T1/2 = 53.3d) tracer 7Be suggested that the heat stored in the SHEBA upper ocean could not have derived from cumulative input over several summers, or advected from distant sources, but rather was a consequence of local heating through leads during the summer of 1997 [Kadko, 2000]. While heat and 7Be input could be readily accommodated by this essentially one-dimensional model, the mixed layer freshening would have required 2 m of in situ ice melting were this the sole source of the freshening [McPhee et al., 1998]. Although this seemed an unrealistically large melt, these authors questioned the possibility that river runoff could have contributed greatly to the observed freshening. Macdonald et al. , however, using oxygen isotope analysis (δ18O), suggested that a large amount of Mackenzie river runoff, plus ice melt from the ice pack edge carried with the Mackenzie plume, was largely responsible for the observed fall 1997 SHEBA freshening. These authors also suggested that heat and stability were delivered by this mechanism as well, although the source and timing of the heat input were not evaluated. In this paper we utilize additional measurements of 7Be and the 18O/16O ratio in the upper ocean to refine these views of the fall 1997 SHEBA observations and compare these results to the observations made at the end of the SHEBA experiment in fall 1998. We will show that in the fall 1997, the enhanced heating was derived locally, predominantly through extended leads with no additional requirement of advective sources. The majority of the freshwater delivered during the melt season was derived from melting ice (1.2 m, or 60% of the total) that would be consistent with the thin ice observed at that time. However, calculations also require a substantial seasonal input of freshwater through river discharge as suggested by Macdonald et al. , and this certainly would have increased the stratification stability of the upper ocean that in turn would have promoted local heating. In the fall 1998 SHEBA period the ice melt was comparable, but the salinity was greater, the integrated water column heat content was less, and the river input was reduced from the previous year.
 Berylium-7, produced from cosmic rays, has a radioactive mean life (defined as the 1/e decay level, or half-life/0.693) of 77 days and thus is well suited for studying seasonal phenomenon. Deposited upon the Earth's surface by precipitation, 7Be “tags” water in contact with the ocean surface, thus providing evidence of air/sea interaction during the previous seasonal period. Once deposited on the ocean, it is homogenized within the surface mixed layer rapidly with respect to its decay rate [e.g., Silker, 1972; Young and Silker, 1980; Aaboe et al., 1981; Kadko and Olson, 1996; Kadko, 2000]. The nature of 7Be as a tracer is dictated by the environment in which it is deposited. In the low particle environment of oligotrophic oceanic gyres, it has been shown that 7Be is quite soluble. Under these circumstances, it is removed predominantly from the surface ocean by decay and mixing processes allowing it to be used as a water mass tracer [e.g., Silker, 1972; Aaboe et al., 1981; Bloom and Crecelius, 1983; Kadko and Olson, 1996]. This is in contrast to the high particle environments of coastal and estuarine areas where beryllium is taken up substantially by the particulate phase and is thus prone to removal by particle settling. Under those conditions, 7Be cannot be used as a water mass tracer but is applicable to studies of metal uptake on suspended matter and sediment transport [e.g., Aaboe et al., 1981; Olsen et al., 1985, 1986]. This dual nature behavior of 7Be is significant to the results of this study.
 In the low particle open ocean, the mixed layer depth is a critical parameter which largely determines the 7Be surface activity (equivalent to concentration, here expressed as dpm/m3, or, disintegrations per min per m3) because a given 7Be input is concentrated in shallow mixed layers and diluted in deeper mixed layers. In snow, the 7Be activity is approximately 2 orders of magnitude greater than in the ocean mixed layer, because a given 7Be flux is diluted in mixed layers typically tens of meters deep, but deposited in snow layers (over a 77-day period) which are at least an order of magnitude shallower, thereby concentrating the nuclide relative to the surface ocean [e.g., Cooper et al., 1991]. Leads also receive 7Be through direct atmospheric input, but because of their generally limited spatial coverage, receive most of this isotope from the surrounding snow and ice melt that contains 7Be deposited previously [Kadko, 2000; Eicken et al., 2002].
 The application of variations in the 18O/16O ratio (δ18O) and salinity of natural waters is an established method for defining contributions of various water types to water masses. Atmospheric waters are depleted in the heavier isotope with a well-defined latitudinal dependence that allows δ18O to be used as a tracer of meteoric input [e.g., Dansgaard, 1964]. This technique has been used extensively in high-latitude oceanographic studies [e.g., Weiss et al., 1979; Östlund and Hut, 1984; Cooper et al., 1997; Macdonald et al., 1995, 1999].
 Vertical profiles of 7Be were collected during the SHEBA drift experiment 1997–1998 (Figure 1) by the methods described by Kadko . Briefly, in the fall of 1997 and 1998, hydroholes were melted through approximately 2 m of ice through which samples for 7Be were collected by pumping. During the summer of 1998, samples were pumped through lead openings. In this process, 700 L of seawater were passed through iron-impregnated acrylic fibers packed in cylindrical cartridges [e.g., Krishnaswami et al., 1972; Lal et al., 1988; Lee et al., 1991; Kadko and Olson, 1996]. The water was taken at various depths through a 1.5″ hose at the end of which was a CTD system. The pumping was achieved with a centrifugal pump, powered by a gasoline-fueled generator on the ice, at a rate of approximately 14 L/min. In most cases, double or triple samples were collected at any one depth and later combined. On one cast (October 8, 1997), where only single cartridges were used at each depth, three samples from different depths in the mixed layer were combined as representative of the mixed layer activity. The extraction efficiency of the Fe-fiber, tested with stable beryllium in the lab or by counting two cartridges placed in series, was 0.90 ± 0.02.
 Samples for oxygen isotopes (18O/16O) were diverted from the fiber cartridges to plastic vials and sealed and stored for laboratory analysis. The oxygen isotopic composition was determined using an adaptation of the conventional method described by Epstein and Mayeda , whereby the oxygen is determined on CO2 which has been injected into serum bottles containing 1 cm3 of sample and equilibrated at 35°C for 12 hours without shaking. The process is entirely automated with the CO2 being injected and retrieved using an autosampler followed by gas transfer to a dual-inlet mass spectrometer (Europa GEO) through a cryogenic trap (−70°C) to remove water vapor. The precision of this method, determined by measuring 59 replicates of an internal standard, is 0.08‰. All data are calibrated using V-SMOW and are reported in ‰ according to the conventional notation.
 The typical temperature and salinity profiles from SHEBA in October 1997, along with profiles of 7Be, δT (temperature above freezing) and δ18O are shown in Figure 2. The 7Be and δ18O data are presented in Table 1. Figure 2b shows that the mixed layer extended to a depth of approximately 30 m, below which was a layer approximately 0.5°C warmer than the mixed layer. McPhee et al.  noted that this feature was appreciably warmer than a similar layer observed during AIDJEX in 1975. The salinity of the mixed layer in the 1997 SHEBA site was also considerably fresher than that of AIDJEX. The 7Be below the mixed layer marks the warm water as having been in contact with the sea surface within the previous 77 days. Thus the heated layer beneath the October mixed layer at SHEBA was remnant of an earlier, deeper mixed layer, resulting from very active heat input during the late spring-summer of 1997. The 7Be shows that the high heat layer could not have been advected from afar or emplaced cumulatively over several seasons [Kadko, 2000]. As the summer progressed, fresher water input capped this warm layer, isolating it from further 7Be input. The 7Be there decayed as additional 7Be was emplaced in the shallower mixed layer from direct input through leads and from draining meltwater from the ice surface [Kadko, 2000; Eicken et al., 2002]. Because the sun angle decreases in late summer, the δT of the shallower layer is less than that of the remnant mixed layer below and will continuously decrease as the fall season develops. Kadko  derived the SHEBA 7Be profile using a one-dimensional model of mixed layer evolution. This was reasonable because the relatively short half-life of the isotope precludes transport from distant sources or cumulative accumulation over several seasons. Applying the same model parameters derived from the 7Be distribution to an appropriate seasonal heat input, the observed temperature profile was also derived. Compelling evidence of the association of 7Be with summer heating is presented in Figure 3, which shows the δT and 7Be profile through a lead in mid-July 1998. High heat and 7Be were trapped in the upper fresh layer until July 18 when meltwater from the lead began to flow under the adjacent ice [Paulson and Pegau, 2001]. Reflecting this, the data show that heat and 7Be in the upper lead dropped proportionately between July 15 and July 22. Days later, this stratified, freshwater layer and associated heat was driven down into the mixed layer by storm conditions [e.g., Richter-Menge et al., 2001]. These results support the notion, suggested by McPhee et al. , that the high heat content of the 1997 SHEBA site resulted primarily from direct heat input through leads under conditions of expanded lead coverage. As with 7Be, the heat observed in October could not be cumulative (i.e., the heat is reset each winter), and its correspondence with 7Be offers strong evidence that it was derived locally (i.e., predominantly through leads and not from distant transport). The appreciable freshening of the SHEBA mixed layer, however, defies easy explanation by one-dimensional (or local) models as this would have required 2 m of in situ ice melting were this the sole source of the freshening. Unlike heat and 7Be, however, the freshwater need not be constrained to have a local source. Adapting a mass balance approach that has been used previously to investigate the source of freshwater in the surface Arctic Ocean [e.g., Östlund and Hut, 1984; Macdonald et al., 1995, 1999], we will use δ18O and salinity to explore possible sources of the observed freshening in 1997 and compare these to the freshwater budget for SHEBA in 1998.
Table 1. Water Column 7Be and δ18O Data From SHEBA
 The δ18O profile from SHEBA on October 8, 1997 is shown in Figure 2c, and the δ18O-S relationship is plotted in Figure 4a. The δ18O-S plot reveals a two end-member mixing line, with a seawater end-member assigned S = 35 and δ18O = 0.00‰. Extrapolation to S = 0 defines an apparent freshwater component of δ18O = −16.57‰, indicating an appreciable contribution from a precipitation source. The freshened mixed layer can be further resolved by assuming it to be composed of three components [Östlund and Hut, 1984; Macdonald et al., 1999]: (1) river runoff, including local precipitation (δ18O = −20.3‰, S = 0), (2) ice melt, with δ18O = −1.9‰, S = 4.5 [Eicken et al., 2002], and (3) a seawater component, in this case from below the fall and remnant spring mixed layers representing the upper halocline of the southern Beaufort sea. The δ18O and S of this component should fall on the δ18O-S relationship shown in Figure 4, and may vary depending on its source. The most likely source is inflow of Pacific waters through the Bering Strait which maintains the upper halocline and strong stratification between relatively fresh surface water and underlying saline water of Atlantic origin [e.g., Cooper et al., 1997; MacDonald et al., 1999] and mixes at shallow depth with freshwater derived from riverine and ice melt sources. Recent time series observations reveal that volume transport, salinity, and temperature of the Pacific input display seasonal and annual variability likely driven by variability of the regional wind field [Coachman and Aagaard, 1988; Roach et al., 1995; Proshutinsky and Johnson, 1997], and thus the S and δ18O values for this end-member should also be expected to vary. In late 1996 the temperature of this source peaked and was followed rapidly by cooler temperatures in 1998, while the salinity was fresher by about 1 psu in late 1996 compared to 1998 (K. Aagaard, unpublished data, personal communication, 2001). The choice of δ18O and S for this component is based on the salinity observed below the remnant mixed layer at SHEBA (∼60 m depth), assuming little impact by ice melt or river input at that depth. The δ18O is calculated from the δ18O-S regression in Figure 4a. For 1997, this component is given by δ18O = −1.86‰ and S = 31.5. Having thus defined the three contributing components, the mixed layer, with δ18O = −3.5‰ and S = 27.6, is therefore composed of seawater (87%), river (8.9%), and ice melt (4.1%). From McPhee et al. , the inflection in the salinity profile at about 40 m marks the salinity at the start of the melt season, such that its difference with the mixed layer salinity represents the freshwater input during the melt season. In 1997 this amounted to 2 m. From our partitioning results, 4.1% of the 30-m mixed layer (1.2 m) was derived from ice melt which amounts to 60% of the 2-m seasonal freshwater input. In addition, there was a substantial freshwater contribution from a riverine source amounting to 40% of the seasonal input (2 m − 1.2 m = 0.8 m). The ice melt is large, alone being greater than the total freshwater input (0.8 m) calculated for the AIDJEX period [McPhee et al., 1998] and is consistent with the thin ice conditions noted during the SHEBA occupation. As suggested by Macdonald et al. , the river input was a major factor in the upper ocean freshwater budget, and our calculations suggest that this equaled the total freshwater input from the AIDJEX study. For the underlying remnant layer, δ18O = ∼−2.5‰ and S = ∼29.6. Applying the same partitioning approach to this layer yields a composition of seawater (93.5%), river (3.5%), and ice melt (3.0%). This is consistent with the remnant layer having developed in the spring, earlier in the melt season before the peak of freshwater generation from any source. The sensitivity of the calculated river input to the salinity chosen for the deep end-member is shown in Figure 4b.
 It is instructive to compare this result to the SHEBA site in October 1998 (Figure 5). Because of the SHEBA drift, such comparison is not strictly a local time series but reflects rather a broader regional evaluation. Qualitatively, the T, S, δT, δ18O, and 7Be profiles between the 2 years are similar, suggesting comparable seasonal mixed layer evolution in both time periods. However, the maximum δT from 1998 was only about 50% of the 1997 occupation, which corresponded to an observed lead coverage in 1998 of 5% for the summer period, ranging to as high as 18% only toward summer's end [Perovich et al., 2002]. This is comparable to the 9–10% lead coverage estimated for the 1975 AIDJEX melt season [McPhee et al., 1998] but much less than the >20% estimated for periods in late June–July 1997 from SSM/I ice concentrations near a drifting buoy experiment in the Beaufort Sea [Yang et al., 2001, Figure 4b]. On the other hand, the 7Be inventories between the 2 years were almost identical, the 1998 site being 90% of the 1997 site. Unlike heat, the 7Be input is not strongly dependent on the fractional lead coverage as much as it is on melt runoff from the ice which carries most of the seasonal 7Be fallout into the upper ocean [Kadko, 2000; Eicken et al., 2002]. If the 7Be atmospheric flux is relatively invariant from year to year, then the seawater 7Be inventory should be similar between comparable annual seasons. Table 3 shows that the ΣHeat/Σ7Be ratio of 1997 is nearly twice that of 1998, indicating that the lead coverage in 1997 was substantially greater than in 1998.
 The 1998 mixed layer salinity was 29.7 compared to 27.6 from 1997 and similarly, the δ18O from 1998 is less negative (−1.57‰), reflecting a diminished riverine source for that year. The δ18O-S relationship (Figure 4a) again indicates a two-end-member mixing line but with the extrapolated apparent freshwater end-member less enriched in precipitation than the fall 1997 case. Using a deep seawater component with δ18O = −0.73‰ and S = 32.5, the partitioning model yields a mixed layer composition of seawater (90.6%), river (4%), and ice melt (5.4%). This corresponds to melting of 1.1 m of ice, which is comparable to the measured summer ice loss in 1998 resulting from a combined 56-cm surface and 62-cm bottom ablation [Perovich et al., 2003]. This agreement between observed and modeled ice melt lends credence to the 1997 model results presented above for which no melt-season observations are available. The total seasonal freshwater equivalent thickness was 1.5 m, yielding a river input of 0.4 m for the 1998 melt season. For the underlying remnant mixed layer, δ18O = −0.86‰ and S = ∼32.1. Partitioning yields a composition of seawater (98.7%), river (0.6%), and ice melt (0.7%). Modeling results are summarized in Tables 2 and 3.
Table 2. The δ18O and S Values Used in Modeling
Deep component (1997)
Deep component (1998)
Table 3. Year Comparison
Total freshwater equivalent thickness from summer input (ice melt plus river).
 We have not identified the source of the riverine input in the preceding modeling approach. MacDonald et al.  suggested that enhanced MacKenzie River discharge was responsible for the 1997 freshening, but it has also been suggested that water from Russian rivers was carried much farther eastward in the 1990s relative to previous years [Steele and Boyd, 1998]. The tracers presented in this paper cannot distinguish between the possible riverine inputs.
 The remnant mixed layer δT of SHEBA in October 1997 was 2.5 times that of AIDJEX, which suggested that the heat flux through the 1997 summer leads (neglecting contributions through ice and melt ponds) was approximately triple that at AIDJEX [McPhee et al., 1998; Kadko, 2000]. Evidence from 7Be measurements and modeling supported the notion that this heat was indeed locally emplaced. River outflow could not have transported the 7Be (and thus heat) to the SHEBA site because the short half-life limits the transport distance of the isotope, and its particle reactive nature precludes its escape from river watersheds and the particle-rich nearshore environment. For example, Cooper et al.  found that the export inventory of 7Be in stream water (during snowmelt) from a site in north Alaska was only about 8% of the total inventory expected from snow within the watershed, and Olsen et al.  found that river input delivered only 5% of the atmospherically deposited 7Be inventory to the James River Estuary in Virginia. This attests to the ability of drainage basin soils and vegetation to retain 7Be. In addition, the mean residence time for particle-reactive nuclides (e.g., 7Be) in coastal zones decreases sharply from open ocean values as the result of increased particle concentration (from higher biological productivity and sediment load). For a site in the mid-Atlantic coastal zone, Olsen et al.  reported a mean residence time of 10–25 days for 7Be with respect to particle uptake, which is small compared to the 77-day radioactive mean-life. Aaboe et al.  found that the water of Long Island Sound was nearly depleted of its atmospherically deposited 7Be inventory, suggesting a particle removal time of only about 4 days. From the Cooper et al.  study, typical snow 7Be concentrations were 12,000 dpm/m3 while stream water concentrations were only about 600 dpm/m3. This value would be further reduced by radioactive decay during the residence time of water within the watershed, loss by particle uptake within the high particle environment of the coastal areas, and radioactive decay and dilution during transport of this freshwater to the SHEBA site. Ignoring radioactive decay in the watershed, and using a conservative estimate of the particle removal residence time of 25 days in the coastal regime, then an additional 75% of the 7Be would be lost. Transport across the approximately 600 km from the Alaska coast to SHEBA in 1997 at an assumed rate of 10 cm/s would have required about 70 days, leading to a decay loss of an additional 60%, or a lowering of concentration to 60 dpm/m3. Finally, calculations presented here (Table 3) suggest that the remnant mixed layer was composed of 3.5% river water, leading to a maximum river 7Be component of 2 dpm/m3. This is below the detection limit of our method. The consistency of Σ7Be between fall 1997 and 1998 also suggests that the variable river input (halved from 1997 to 1998) has no effect on the 7Be input. As discussed in the background section, it is important to note that the removal of 7Be is significant only in the high particle, coastal environment. Offshore, away from the influence of high particle flux, 7Be added along with heat through lead openings to the upper ocean tags surface water and serves to trace the fate of heat on a seasonal timescale. This conservative behavior of 7Be (with respect to particle uptake) within the Beaufort Gyre is more fully discussed in Appendix A.
 Several studies of the SHEBA heat budget also suggest that local heating, alone, was sufficient to account for the October 1997 observations. However, it is noted [e.g., Perovich et al., 2003] that there remains considerable uncertainty in these heat budget calculations because parameters such as albedo, heat transmittance through melt ponds and ice, turbulent heat flux, and cloud effects have spatial and temporal variability or are presently not well characterized. It is for this reason that tracers, such as 7Be, with relatively less complex input functions, are valuable for tagging water masses and associated heat. In one SHEBA study, Paulson and Pegau  constructed a heat budget for the lead described earlier (Figure 3). They found that for the period July 4–18, 1998, the net heat flux (taking into account measured net incoming radiation, albedo, and molecular conduction out of the base of the fresh layer) into the 12,000 m2 lead supplied the necessary heat to account for both the measured heat storage in the lead and the lateral ice-melting rate. No additional source of heat was required. After July 18, meltwater flowed beneath the surrounding ice, and on July 28, a passing storm mixed the heat and freshwater below the ice floes, precluding continuation of this budget [Richter-Menge et al., 2001]. Perovich et al.  considered the heat budget for the extraordinarily large amount of bottom ice melting (62 cm) for the period June 3 to October 4, 1998. They found that the necessary heat was derived locally, through a combination of solar input through leads and transmittance through melt ponds and ice such that an advective, nonlocal component was not required. These authors pointed out that while in the past, heat input through melt ponds and ice have been neglected, these sources might account for approximately 30% of the required energy. From another perspective, the estimated amount of energy expended in surface, bottom, and lateral melting on July 20, 1998, at SHEBA was 7.8 MJ m−2 [Perovich et al., 2003]. For the summer period of 1998 encompassing June 6 to August 1, the average incoming shortwave radiation was 239 W/m2 [Paulson and Pegau, 2001; C. A. Paulson, personal communication, 2001) and the spatially and temporally averaged albedo was approximately 0.5 [Perovich et al., 2002]. Thus a crude estimate yields 10.3 MJ m−2 d−1 were available from solar heating to account for the observed ice melt. These calculations, although each having uncertainty, indicate that the presence of 4% river water in the mixed layer was not important to the 1998 heat budget, and it is unlikely therefore that freshwater input in 1997, although approximately double that of 1998, constituted a significant source of heat to SHEBA. The high observed heat content of 1997 is consistent with a higher lead coverage during the summer of that year.
 The seasonal ice melt equivalent heights during 1997 and 1998 were about 50% greater than the 0.8 m estimated for AIDJEX in 1975 [McPhee et al., 1998]. Although the ice melt for 1997 and 1998 were nearly equivalent, the integrated heat content of the remnant mixed layer in 1997 was twice that measured in 1998. While, as stated above, this implies greater lead coverage in 1997, it is not known why more ice melt did not occur for the greater heat input in 1997. It then brings into question the fate of the remnant heat found at the SHEBA site in 1997. Figure 2 shows that by October 1997, this heat, at 30 m depth and below a sharp pycnocline, was not readily available for further ice melting. In fact, after October 10, the ice column thickness only increased [Perovich et al., 1999]. In a sense, this heat was subducted and not extracted back to the surface during the following winter. Such an observation was also made during the 1975 AIDJEX study [Maykut and McPhee, 1995], but in 1997 the trapped heat was substantially greater. We cannot say when and by what mechanisms this “stored” heat is ultimately extracted for further ice melting. One possibility is that winter storms might be powerful enough to drive deep mixing events capable of penetrating the halocline [Yang et al., 2001]. Alternatively, as suggested by Maykut and McPhee , downwelling associated with Ekman convergence in the anticyclonic Beaufort Gyre would tend to maintain the heat at depth, so that at least a portion of the summer insolation is eventually added to the Pacific Water Layer, instead of the slow upward flux from below often assumed.
 The ice conditions first observed at SHEBA in October 1997 included thinner ice than anticipated and evidence of a very active melt season (D. Perovich, personal communication, 2000). This condition did not reverse itself during the SHEBA experiment during which a net ice loss of 75 cm was measured over the year-long study period ending in October 1998 [Perovich et al., 2003]. It has been suggested that such changes are representative of longer-term trends associated with increased cyclonicity, enhanced ice divergence, and a decrease in the size of the Beaufort Gyre as a consequence of atmospheric pressure changes [e.g., Serreze et al., 1989; Maslanik et al., 1996; Walsh et al., 1996; Proshutinsky and Johnson, 1997; McPhee et al., 1998; Thompson and Wallace, 1998; MacDonald et al., 1999]. Reduced ice area as a consequence of divergence would lead to decreased albedo, and enhanced summer melting and water column heating through positive feed-back mechanisms. While the SHEBA site manifested reduced ice cover and enhanced local heating, recording of “anomalous” observations were also reported elsewhere. For example, the recent time series study from the Bering Strait discussed above suggested anomalous salinity and temperature of the Pacific input in late 1996 (K. Aagaard, unpublished data, personal communication, 2001). Also, the freshening of the SHEBA site by river input in 1997 corresponded in that year to the largest outflow of Mackenzie water in the past decade (Figure 6). Are these observations related by broad regional processes such as those suggested by Proshutinsky and Johnson  and are they indicative of ongoing changes in the Arctic? Large-scale modeling must be used to consider the former, and only continuous monitoring of the Arctic Ocean will answer the latter.
 We have used measurements of 7Be and oxygen isotope (18O/16O) ratios from the 1997–1998 SHEBA experiment to trace the source of the high heat and freshwater content observed in the upper ocean at the SHEBA site in October 1997. The evidence suggests that the heating resulted from local input primarily through extended lead coverage in the late spring and summer of 1997. The freshening was derived from a large ice melt (1.2 m) that was consistent with the thin ice conditions observed at the site. However, a significant contribution to the freshwater budget (0.8 m) included enhanced input from river runoff. This obviates the requirement for an unrealistic ice melt (∼2 m) to account for the freshwater content of the mixed layer and likely increased the stratification stability that in turn promoted heating. However, the question arises as to the fate of the large heat input at SHEBA in 1997 that resulted from an active heating season. The water column heat content at SHEBA in fall 1998 was considerably less than that of the previous year, consistent with a small percentage of lead openings observed throughout most of the melt season. Although the melt equivalents were similar in both years, the freshwater content of the fall 1998 SHEBA site was lower due to a reduced riverine input.
Appendix A:: The Conservative Nature of 7Be in the Beaufort Gyre
 To test the conservative nature of 7Be, we use two approaches. In the first, we calculate a 7Be mass balance to demonstrate that the atmospheric input matches the water+ice inventory. In the second, we analyze the suspended sediment load released by melting ice and show that this is insufficient to significantly scavenge the water column 7Be inventory.
A1. Mass Balance for 7Be
where Fi refers to the fractional area of these features. The atmospheric flux in the summer of 1998 was 0.0122 ± 0.0070 dpm/cm2d [Kadko, 2000].
 The average ocean inventory from two mid-summer (July) ocean profiles was 0.509 ± 0.13 dpm/cm2 (D. Kadko, unpublished data, 1998). The ocean inventory from October 4, 1998, was 0.315 dpm/cm2 (Table 3), and as the ocean was ice-covered (that is, isolated from 7Be input) by August 26 [Perovich et al., 2003], the October value must be decay-corrected to August 26 and equals 0.523 dpm/cm2. Thus there is reasonable agreement among the three profiles and their average is 0.513 ± 0.094 dpm/cm2. The required flux to support this inventory is 0.0067 dpm/cm2d.
 The average melt pond concentration in mid-summer was 0.55 dpm/cm2 [Eicken et al., 2002]. This requires a flux of 0.00715 dpm/cm2d. The fractional area of melt ponds (Fmp) was approximately 0.2 [Perovich et al., 2002].
 The amount of 7Be measured within the upper 1 m of an ice core in mid-summer [Eicken et al., 2002] was 0.3 dpm/cm2, requiring a flux of 0.0039 dpm/cm2d. The ice covered about 75% of the area (Fice = 0.75).
 The under-ice meltwater [Eicken et al., 2002] is difficult to quantify but might hold a high activity of 7Be, analogous to the upper 1 m of lead water (Figure 3). It is ignored here but likely increases the inventory by several percent. Then, ocean inventory plus (melt pond times Fmp) plus (ice times Fice) equals 0.0067 + 0.00715 × (.2) + 0.0039 × (.75) = 0.0111. This is very close to the input flux of 0.0122 ± 0.0070 dpm/cm2d, indicating the conservative nature (with respect to suspended sediment scavenging) of 7Be in this Arctic gyre. While there is some uncertainty in budgets of this sort, note that the ocean 7Be inventory equals 60% of the total, whereas the ice inventory is 40% of the total. Using a completely different approach, Eicken et al.  found that of the total surface meltwater production, 59% was discharged into the mixed layer, while 39% was retained on the ice. This further shows the solubility of 7Be in this environment.
A2. Suspended Sediment Evaluation
Reimnitz et al.  reported concentrations of particulate matter in the Beaufort gyre. It was found that the average particle concentration in seawater was only 0.77 mg L−1 and that this was far less (50 times less) than the average for the sea ice. This should not be surprising; although significant transport of shelf sediments in Arctic sea ice, including over deep basin waters, occurs, these and other authors point out that the release of sediment from the ice occurs over a timescale of years [Reimnitz et al., 1993; Eicken, 2004]. Furthermore, the yearly release is spread over a melting season and diluted over a water depth of at least tens of meters.
 At that suspended particle load (<1 mg L−1), it has been shown that ≤10% of 7Be is partitioned upon particles [Honeyman and Sanstchi, 1989, 1992; Bloom and Crecelius, 1983]. As this is a small fraction of the total inventory, and particles require days-to-weeks to sink, particulate loss is not a significant component to the 7Be budget in the Beaufort Gyre nor a significant consideration in our conclusions.
 This work was supported by National Science Foundation Polar Programs grants OPP-9701076 and OPP-9815132 and Chemical Oceanography Program grant OCE-9809168. The authors would like to thank K. Aagaard, M. McPhee, S. Pegau, D. Perovich, H. Eicken, S. Emerson, P. Quay, and J. Morison for many valuable discussions and comments. Comments by anonymous reviewers also were quite helpful. M. Jones of the Water Survey of Canada kindly provided the Mackenzie River discharge values. Thanks also to M. Stephens for field and laboratory assistance, V. Gonzalez for help with the oxygen isotope analyses, and to the crew of the CCGC Des Groseilliers and the SHEBA logistics group for their support during the field operation of SHEBA.