Melting and freezing beneath Filchner-Ronne Ice Shelf, Antarctica



[1] We use remote-sensing data sets to evaluate the spatial distribution of melt beneath the Filchner-Ronne Ice Shelf (FRIS). The net melt rate of 83.4 ± 24.8 Gtons/yr is 2.5–5 times lower than previous glaciological estimates, but is similar to existing oceanographic estimates. The spatial distribution, however, differs significantly from standard conceptual and numerical models in which most melt occurs along the grounding lines. Our results suggest most grounding-line melt is refrozen, while the dominant Ice Shelf Water (ISW) source is melting near the ice shelf front, probably associated with tidal action. This suggests that changes in ice shelf extent can impact ISW production rates in the Weddell Sea.

1. Introduction

[2] It is commonly assumed that most (perhaps 50–80%) of total Antarctic Bottom Water (AABW) production (5–15 × 106 m3 s−1) occurs within the Weddell Sea [Broecker et al., 1998]. Melt under the FRIS yields a significant outflow of ISW through the Filchner Trough to the continental shelf break. There, the ISW mixes with Weddell Deep Water to form Weddell Sea Bottom Water, which is one contributor to AABW formation [Foldvik et al., 1985; Schlosser et al., 1990]. We seek to improve upon existing estimates of melt and the associated ISW production rate.

[3] Jacobs et al. [1992] identified three modes of ice shelf melting. The first mode, to which most melt has been attributed and which we refer to as thermohaline convection, is initiated when brine rejection during sea ice formation produces dense High Salinity Shelf Water (HSSW) at the surface freezing point, Tf,0 (∼−1.9°C). The HSSW descends beneath the ice shelf to the deep grounding lines where ice streams first loose contact with the land. Pressure suppresses the melting point by, in some cases, more than 1°C below Tf,0, so that the HSSW is able to efficiently melt ice [Rignot and Jacobs, 2002]. The resulting fresh, buoyant ISW rises following the slope of the ice shelf base and passes along the gradient in the pressure melting point, causing it to supercool at some depth [Jenkins and Doake, 1991]. This leads to the refreezing that produces a layer of marine (refrozen) ice, which exceeds 300-m thickness beneath parts of the Ronne Ice Shelf (RIS) [Thyssen et al., 1993]. In the second mode, melt occurs when warm water at intermediate depths offshore enters the cavity as part of the general circulation. The third mode is associated with the presence of seasonally warmer upper ocean waters just north of the ice front in summer, which can then be advected into the ocean cavity by tidal currents or other phenomena such as topographic edge waves and ocean eddies. In each mode, the melting efficiency depends not only on advective transport of warm water into contact with the ice, but also the availability of turbulent kinetic energy to maintain the vertical flux of ocean heat to the ice base to offset the stabilizing effect of buoyant fresh water production [Makinson, 2002].

[4] Glaciological estimates, based on sparse data, yield large net melt rates of 202 to 440 Gtons/yr (0.45–0.98 m/yr average melt rate) [Doake, 1984; Jacobs et al., 1992]. More recent estimates based on Interferometric Synthetic Aperture Radar (InSAR) data yield local melt rates of 2–14 m/yr at the FRIS grounding lines [Rignot and Jacobs, 2002], but these values apply to only a small fraction of the shelf and exclude areas of refreezing. Oceanographic estimates based on integrating heat and fresh water fluxes along the ice front yield relatively low net melt rates of 45 to 90 Gtons/yr (0.1 to 0.2 m/yr) [Foldvik et al., 2001; Gammelsrød et al., 1994]. Ocean circulation models estimate the spatial distribution of melt [Gerdes et al., 1999; Jenkins and Holland, 2002] and yield net melt rates that are generally consistent with oceanographic observations.

2. Melt Rates

[5] We estimate basal melt-freeze using standard glaciological methods [Jenkins and Doake, 1991] applied to greatly expanded data sets. The required data are ice thickness, surface accumulation, and ice flow velocity. While thickness and accumulation data sets have been compiled [Giovinetto and Zwally, 2000; Lythe and Vaughan, 2001; Vaughan et al., 1999], ice velocity has only been sparsely sampled. We used RADARSAT InSAR data [Joughin, 2002] to produce an ice velocity map (Figure 1) for nearly the entire FRIS.

Figure 1.

Ice flow speed (color and 100 m/yr white contours) for the FRIS determined using InSAR and speckle-tracking [Joughin, 2002] with tidal corrections on the floating ice [Padman et al., 2002]. Surface elevation 100-m contours (black) are plotted over the SAR basemap image [Jezek, 1999]. The red lines show inflow flux gates and the yellow lines show gates with negligible flux. Outflow was measured through the orange gates. Blue and red lines surround the region downstream of FIS (Table 2). The velocity data show the September 2000 ice front, while the basemap shows the September 1997 front.

[6] At a point on an ice shelf in steady state, the horizontal divergence of the volume flux equals the combined surface and basal accumulation [Jenkins and Doake, 1991]. If the velocity, thickness, and surface accumulation are known, then basal accumulation (melt-freeze) can be estimated. This yields two approaches for estimating net melt. In the first, the inflow and outflow are evaluated on the ice shelf perimeter. This is the more accurate method since there is some flexibility in defining the perimeter, which allows use of the most accurate thickness data (e.g., radio echo-sounding thicknesses). With the second method, the flux divergence is integrated over the ice shelf. This method relies on less accurate gridded thicknesses, but also yields the melt/freeze distribution (Figure 2). We note that with both methods we are estimating the melt/freeze rate that maintains the ice shelf in equilibrium. Deviations from a steady state would cause our estimates to differ from the actual melt/freeze rates.

Figure 2.

Basal melt rates (color) determined under assumptions of conservation of mass and of a steady-state ice shelf [Jenkins and Doake, 1991]. The color bar saturates at magnitudes >5 m/yr. Negative values imply melt and positive values freezing. The yellow line separates the region with strong melt at the Ronne Shelf front from the interior (see text). Net melt for the region outlined in blue was determined using flux gates (Table 2). Light blue lines show inferred ocean circulation paths [Nicholls et al., 2001].

[7] The first method yielded a net melt of 83.4 ± 24.8 Gtons/yr (0.19 ± 0.06 m/yr) (Table 1). We used the second method to estimate the melt-freeze distribution over much of the shelf (Figure 2), integration of which yields a net melt rate of 54 Gtons/yr. Downstream of Foundation Ice Stream (FIS), we did not have adequate data to estimate melt locally, so we used flux gates surrounding the region (Figure 1) to estimate melt of 24.8 Gtons/yr (Table 2). This yields a combined melt estimate of 78.7 Gtons/yr. Although this estimate neglects a small fraction of the shelf, it agrees well with the flux-gate only estimate.

Table 1. Melt From Flux Gate Estimates
GateFlux (Gton/yr)Uncertainty (Gton/yr)
  • a

    Inflow to the shelf is the sum of the fluxes from individual ice streams (Figure 1, red gates). Gridded accumulation data [Giovinetto and Zwally, 2000; Vaughan et al., 1999] were integrated over the area outlined by the yellow, red, and orange lines in Figure 1, which includes ice rises and areas that contribute ice to the shelf through sheet flow. Total outflow was determined through gates (Figure 1, orange lines) at the shelf front. The errors at individual gates are largely determined by the accuracy of the ice thickness data, which varies from 10-30%. A value of 15% is used for the uncertainty in the accumulation rate estimates.

  • a

    From Rignot and Jacobs [2002].

Misc. Gates7.92.4
Total Inflow190.18.4
Shelf Accumulation109.017.2
Total Input to Shelf299.019.2
Ronne Front133.913.4
Filchner Front81.88.2
Total Outflow215.715.7
Net Acc. (-Melt)83.424.8
Table 2. Melt Determined From Results Shown in Figure 2
Region/GateFlux Gtons/yr
  1. a

    Melt downstream of FIS (Figure 2: blue outline) was calculated using flux gates. Accumulation is for the area outlined in blue and red in Figure 1 and includes the catchment area not accounted for through the ice stream gates. The uncertainty in the thickness and thickness gradients is spatially variable and not well quantified, and thus, we haven't estimated the uncertainty in the local melt rates or the resulting net melt.

a: Total Inflow44.3
b: Accumulation15.5
c: Total Outflow35.0
d: Melt for Missing Region c(a+b)24.8
e: Melt for Region with Data (Figure 2)54.0
Net Accumulation (-Melt) d + e78.7

[8] Our net melt rate (83.4 ± 24.8 Gtons/yr) is lower than earlier glaciological shelf-wide estimates [Doake, 1984; Jacobs et al., 1992] by a factor of roughly 2.5 to 5. Downstream of FIS (Figure 2), the estimate of 24.8 Gtons/yr is half that of an estimate of 48 Gtons/yr for nearly the same area [Lambrecht et al., 1999]. The large difference between estimates can likely be attributed to the improved sampling and accuracy of our velocity data.

[9] We find better agreement when our results are compared with oceanographic estimates (45 [Gammelsrød et al., 1994] to 90 [Foldvik et al., 2001] Gtons/yr). Schlosser et al. [1990] estimated melt of 144 Gtons/yr from δ18O in the Filchner ISW outflow. Analyses of δ18O, however, give the total melt (excluding refreezing) rather than the net melt (including refreezing) [Gammelsrød et al., 1994]. We estimate a total melt of 152 Gtons/yr (neglecting any refreezing in the area with no velocity data downstream of FIS), which agrees well with the δ18O estimate.

[10] The melt-freeze map (Figure 2) uses thicknesses derived from radar altimeter data assuming hydrostatic equilibrium [Lythe and Vaughan, 2001]. Radar altimeters perform poorly in high-slope regions (e.g., near some grounding lines), so there may be minor errors in the map. We believe, however, the overall melt-freeze pattern is robust. There is good qualitative agreement between the melt-freeze map and results from ocean circulation models [Gerdes et al., 1999; Jenkins and Holland, 2002], with the exception of the strong melt that we find near the ice front. The areas of strong freezing (Figure 2) lie just upstream of areas where marine ice has been observed [Grosfeld et al., 1998; Thyssen et al., 1993].

3. Discussion

[11] We hypothesize that the spatial pattern of melt-freeze results from two distinct modes of melting. With this hypothesis in mind, we partitioned the portion of RIS where we have data into two subregions along approximately the zero melt-freeze contour (Figure 2, yellow line). On the landward side the total melt is 50.4 Gtons/yr, much of which occurs at the grounding lines [Rignot and Jacobs, 2002]. This melt is offset, however, by freezing of 55.6 Gtons/yr for a net freezing of 5.2 Gtons/yr. An additional 24.8 Gtons/yr of ice is melted downstream of FIS. Oceanographic models [Gerdes et al., 1999; Jenkins and Holland, 2002] and observations [Nicholls et al., 2001] indicate that much of this melt is the source for the area of refreezing to the east of Henry Ice Rise. The combined melt of 19.6 Gtons/yr beneath the southern RIS near the grounding lines is likely a response to thermohaline convection (“mode-1”).

[12] On Filchner Ice Shelf, melting of 20.6 Gtons/yr is offset by 16.1 Gtons/yr of refreezing along the east coast of Berkner Island. This is consistent with observations and model-based melt estimates [Grosfeld et al., 1998] in response to thermohaline convection.

[13] The combined Filchner and Ronne estimates for the grounding-line regions and adjacent ice shelf (excluding the RIS front) is only 24.3 Gtons/yr. This melt estimate is relatively small because of the strong refreezing beneath the central portion of the Ronne Ice Shelf. Thus, the primary impact of circulation due to pure thermohaline convection beneath the southern 2/3 of the RIS appears to be largely the redistribution of ice, rather than being the dominant source of ISW as previously suggested [Jacobs et al., 1992].

[14] We find that roughly two thirds (54 Gtons/yr) of the net FRIS melt is generated at shallow depths (mean 375 m) near the RIS front. The general pattern of melt in this region resembles maps of mean tidal current speed [Makinson and Nicholls, 1999; Robertson et al., 1998]. Some of this increase in melt rate near the ice front may result from advection of seasonally warm near-surface waters under the ice shelf by wind-forced and tidal currents [Makinson and Nicholls, 1999]. We attribute most of this melt, however, to the ability of vertical mixing by tidal stress at the ice base to overcome the stabilizing effects of melt water that is produced by any of the three modes discussed above. Jenkins and Doake's [1991] estimate of high melt rates at a location along the ice shelf front further supports our conclusions. In further support of this argument, we note that current oceanographic circulation models [Gerdes et al., 1999; Jenkins and Holland, 2002] that do not include added mixing from tides do not reproduce the strong melt we observe at the front.

[15] Nicholls et al. [2001] used data from two boreholes along the coast of Berkner Island to infer a through-flowing current near the base of the ice shelf (Figure 2), which they believe to be the source of the Filchner ISW outflow. Their temperature-salinity data suggest the parent water mass is HSSW from the Berkner Bank that is converted to ISW in the region of intense mixing near the ice front. They also found that much of the water originating from the western RIS front is not buoyant enough to escape the Filchner Trough, and instead, recirculates to melt ice at the grounding lines. Their results are consistent with our finding of strong melt at the ice front. Thus, this type of ISW is a likely source for much of the Filchner-Trough ISW outflow, and there is no need to postulate a significant contribution to the outflow from grounding line melt.

[16] If tides near the ice front are a critical factor in the net melt under the FRIS, large changes in shelf extent through iceberg calving may significantly alter ISW production. Tidal current speed near the ice front correlates well with bathymetry. The strongest currents are in regions where the water column thickness is small, principally where the eastern RIS front extends over the shallow water of the Berkner Bank. By causing a decrease in the area subject to high melt, iceberg calving may cause an abrupt reduction in melt that is followed by a steady recovery over decades as the ice front re-advances over the shallow continental shelf. Our melt estimate was computed just after several large calving events (1998–2000), so it may reflect the minimum melt in a multi-decadal cycle. The region that calved was about 3% (16,000 km2) of the total FRIS area. Assuming a mean melt rate for this region of 2 m/yr (Figure 2), the total melt just prior to calving would have been roughly 29 Gtons/yr (35%) larger. This assumes that there is sufficient oceanic heat available to melt the additional ice.

[17] The rapid thinning of ice near the RIS front as the shelf ice advances over the shallow Berkner Bank suggests the potential for tidal control of the mean ice front position. The shelf front ice is weakened by melt-induced thinning and the area subject to strong tidal currents increases as the ice front advances, likely accelerating melt and limiting the northward advance of ice through increased calving.

[18] Our estimate of FRIS melt contradicts the larger earlier glaciological estimates and lies within the range of previous oceanographic estimates. With the improved data, the uncertainty on the total melt has been reduced to ±30% (one standard deviation). The significant differences between our distribution of melt-freeze (Figure 2) and that generally assumed in oceanographic interpretations [Jacobs et al., 1992] implies that additional processes such as tidal mixing must be incorporated into models of circulation beneath the shelf. We propose that a relatively minor change to ice shelf extent could have a significant influence on net ISW production, given the strong melt near the RIS front. In turn, this observation points to the need to take processes such as tides into account when assessing the likely response of the ice shelf, and ISW and AABW production, in future climate scenarios.


[19] I. J. performed this work at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. L. P. was funded by NASA (NAG5-7790) and the NSF Office of Polar Programs (OPP-9896006). We thank D. Vaughan, M. Giovinetto, K. Jezek and the Alaska SAR facility for data that went into deriving these results.