COMPILED REPORTS OF THE
U.S. ICE CORE RESEARCH WORKSHOP

2.1 GISP II: SITE SELECTION PANEL REPORT
Prepared by:S. Hodge (Chairman), I Bolzan, E. Waddington and 1. Whillans

Introduction:

This report outlines the conclusions reached by the Site Selection Panel. These conclusions are based solely on scientific considerations and purposely do not consider logistical, financial or political factors.

Although much numerical modeling work has already been done, considerably more could still be applied. In fact, more computer calculations will be done in the next few weeks to clear up some points raised during our discussions, and the results will be incorporated into the next version. This current report thus represents only our initial recommendations.

We have tried to consider all factors which would influence the choice of drill sites. These factors, however, are all based on the fundamental assumption that detailed knowledge of the ice flow is necessary to date the deep ice. Direct dating techniques would probably cause us to change some conclusions and rearrange priorities, but this would not change the basic fact that an ice sheet is a dynamic, deforming medium, and so even with ideal independent dating techniques, it would still be necessary to understand how, and from where, the ice has flowed throughout the history covered by the ice cores.

We assumed that two holes are to be drilled. Although not entirely independent, we have attempted to distinguish between the criteria used to select the general location of the pair of holes ("site" criteria) and the criteria used to determine the separation of the two holes ("separation" criteria).

Issues considered:
     - surface topography 
     - bed topography 
     - ice thickness 
     - bed roughness 
     - accumulation rate 
     - horizontal gradients in accumulation rate 
     - length of flowline special flow effects near the ice divide 
     - ice divide migration 
     - basal melting departures from two-dimensional "flowline" flow 
     - age of 20 mm and 5 mm thick layers (thinnest layers for continuous and intermittent counting, respectively, at Camp Century)

Area considered:

Our analysis concerns a 180 x 180 Ian square centered on the 1987 Summit Camp. Ice radar soundings were made over this entire area, along lines spaced 12.5 Ian apart in both eastwest and north-south directions, and accumulation data were obtained throughout the same region. The actual summit of the ice sheet turned out to be about 30 km north, and 10 krn east, of this camp. 

Divide versus flank flow:

Throughout this report we refer to divide flow and flank flow. Using finite element modeling, Raymond (1983) showed that the ice flow within several ice thicknesses of an ice divide is distinctly different from that at greater distances. Flow within this band, which is very narrow relative to the entire width of the ice sheet, is predominantly vertical, whereas,-flow outside this region is simple laminar flow, which is predominantly horizontal. Figure 1, produced as part of our calculations, illustrates these two types of flow.

In the past, it was difficult to model the flow of ice in the vicinity of an ice divide and this was one reason that there was reluctance to drill at such a site. Numerical modeling of ice flow, however, has now advanced to the point where this is no longer a valid objection. In principle, either region can be modeled equally effectively and so "more complicated" flow should not be considered as a disadvantage of an ice divide drill site, relative to a flank site. Numerical models:

Two distinctly different numerical models were used in our calculations. The models complement each other well and are in good agreement where results could be compared. Figure 2 shows the predicted age-depth relationship for both divide flow and flank flow. Note that the two models give identical results for flank flow (one model cannot be used for divide flow), and that outside the divide flow region the relationship is essentially independent of distance from the divide. 

Site criteria:

- oldest ice
- best resolution (thick annual layers)
- most reliable dating 

Site choice:

Based on the modeling calculations and examination of the ice radar records, we conclude that:

- For the oldest ice, the entire region is suitable, and that basal melting was probably unimportant at any time throughout the last glacial cycle. No particular area stands out as clearly older than the rest.

-For the best resolution, the southwest quadrant gives the thickest layers in young ice (less than 10 kyr), but the entire region is equally suitable for older ice. The southwest quadrant is best for young ice because of the higher accumulation rates there, but since deeper ice comes from closer to the divide the original layer thicknesses all approach the same value, and so there is less variation in the results the older ice.

- For the most reliable dating, the entire region is more or less equally suitable. Assuming a constant accumulation rate with time, 20 mm thick layers are about 30 kyr old and 5 mm thick layers are about 60 kyr old. These values do not vary significantly over the entire grid. Disturbances due to basal hills and bumps should affect only ice older than about 100 kyr in most places. Migration of the ice divide is not important for flank sites (as long as they remain a flank site).

Except for the northeast comer (which is a poor choice anyway because of very mountainous bed topography), the accumulation rate satisfies the minimum requirement (20 cm/yr) for ice coring everywhere throughout the region. This, combined with the fact that divide and flank sites probably have more or less equal advantages and disadvantages (see next section), means that the choice of site probably boils down to simply choosing an area where the bed topography is as smooth as possible. Fortuitously, a reasonably flat, smooth, wide it plateau," or "bench," extends west from the current summit dome, along the ice flow direction. Thus we recommend drilling the two holes somewhere on this bench, with the exact location within this area being determined by the hole separation criteria. 

Reasons for two holes:

We feel that there are strong justifications for drilling two holes, and that the scientific results obtained from either core will be enormously enhanced by the results from the other core. T"he increase in our understanding of the ice flow and the age-depth relationship that win result from two carefully positioned coreholes, will, apart from any other advantages, warrant drilling of both holes, by the same party if necessary. The following reasons are not in any particular order by priority.

(1) To assess the reproducibility of core measurements and discriminate non-climatic signals.

(2) To determine and correct for, the effects of migration of the ice divide.

(3) To greatly improve the reliability of the calculated time scale.

(4) To compensate for the fact that divide/flank sites have complementary 
advantages/disadvantages.

The last point is important to keep in mind. Neither location is ideal. For example, divide sites give potentially longer records and low shear strain, but have the disadvantages of possibly undecipherable complex flow changes with time due to ice divide migration effects and lower resolution in young ice (back to the Holocene-Wisconsin transition). On the other hand, flank sites give better resolution in young ice (Figure 6) and do not suffer from divide complexities, but the old ice could be subject to disturbances from flow over or around basal bumps. The best ice flow dating of the old ice in the divide core will depend strongly on the time scale derived for the flank core, where presumably direct dating methods can be carried back further into the past. Deformation measurements in both holes will greatly aid the dating of both ice cores.

From an ice dynamics point of view, the two holes should clearly be located along the same flow line. Although our modeling ability has improved considerably, all existing ice flow models which could be applied to the detailed small-scale flow analysis required here are nevertheless still just two-dimensional "flow line" models. This condition, that the two holes be on the same ice flow line, is implicit in the following discussions. 

Separation criteria:

We recommend a separation of about 30 km. Although partly a compromise, it is nevertheless a reasonable compromise, one which is clearly superior, overall, than larger or smaller separations. The points we considered are:

(a) If two holes are to be used to determine the effects of divide migration, then they must be at least 20 ice thicknesses (20h) apart to ensure that they were always in different flow regimes (divide versus flank) at any time in the past. This result comes from finite element modeling of the flow at an ice divide by Raymond (1983), which shows that one must go at least 10h away from the ice divide before all traces of divide flow vanish and the flow becomes simple laminar flow characteristic of flank flow. This distance must then be doubled to allow for the situations where the divide could have been between the two holes. For an average ice thickness of h = 3 km, this translates into a minimum hole separation of 60 km.

(b) If one requires that only the vertical component of velocity must become characteristic of flank flow, instead of both components, then this minimum separation can be reduced to 4h, or about 12 km. If divide migration were the only consideration, this separation might be acceptable; it would be a matter of deciding how much of the disadvantage pointed out in (c) could be tolerated. However, since criteria (d), (e) and (f) would not be met as well, such a separation would not be a satisfactory compromise scientifically.

(c) We estimate maximum divide migration (Holocene to Wisconsin) to be of the order of 100 km (see Appendix). Even though the suggested separation of 30 km is less than this, and thus both holes may still be subject to both types of flow at some time in their history, this distance is probably large enough to ensure that stratigraphy characteristic of ice divide regimes will occur at distinctly different depths in the two cores. However, if very small separations, of the order of 1-3h (3-10 km), are used, then it will be difficult, if not impossible, to distinguish one hole as divide flow and the other as flank flow at any time in the past. On this distance scale, the divide is likely to be migrating back and forth through both hole sites so frequently that, for all practical purposes, they are in the same flow regime. Under such conditions, it would be very difficult to use the two holes to determine the effects of divide migration.

(d) If two holes are to be used to improve the reliability of the calculated time scale they must have a separation of at least 3 cycles of the characteristic wavelength of the ice flow pattern. This result comes from an analysis of surface strain measurements around the Dye 3 and Byrd coreholes (Whillans and Jezek, 1984; Whillans and Johnsen, 198?), where the characteristic wavelengths are 8 and 10 km, respectively. This wavelength is also typical of the internal layering and bed topography in central Greenland, based on the ice radar data. Hence the hole separation should be 25-30 km.

(e) Two holes can also be used to greatly improve our ability to test and fine-tune ice flow models, by allowing one hole to be used as "control" and the other to provide known data to be "predicted" by the model. The hole separation must encompass several cycles of topographic relief in order for the bed topography to affect realistically and correctly the calculated ice flow. For reasons similar to those referenced in (c), three cycles are the minimum number required for an adequate test, so once again this points to a separation of 25-30 km. Even on a flat bed, the flow at any given point is affected by conditions up to several ice thicknesses upstream or downstream, so the separation should be of the order of 10h to optimize the benefits of information from two holes. Again, 25-30 km is ideal.

(f) In order to distinguish non-climatic signals in the core records, the two cores must be separated by a distance greater than the scale length typical of microclimate processes in central Greenland. This scale length is of the order of 10 km (Reeh, 198?) and these zones may be transient and may move with the divide crest. Depending on the "noise" source, one core or the other can be treated as "control" to make corrections in the second core, but in order for this to be possible, ice in at most one of the cores can have originated in the affected zone. Assuming that one site is the divide, the second site must be sufficiently far away that all ice of interest for these comparison originates more than 10 km from the divide. Preliminary flow models suggest that at a distance of 30 km from the divide, ice which originates at 10 km from the divide will be about 14 kyr old, which should be just enough to reach the Wisconsin period. Smaller separations will open up the possibility that some Holocene ice in both holes will have originated from the same microclimate source area.

(g) Because the regional climatic signal, as distinct from the local, transient microclimate, is the primary target of GISP II, the two holes should be close enough together that they both experience the same regional climate at all times. This regional climate should be coherent over large distances, but the coherence drops as separation increases, and thus we would lose the ability to test signal reproducibility by using the two cores. This is the counter argument to (f) and is one of the reasons for not using too large a separation.

(h) As the separation of the two holes is increased there is an ever-increasing accumulation of flow complexities in the deeper, older ice, due simply to the ice having flowed over more and more bed topography. There is also an increasing chance of an especially complex perturbation occurring at some point, which could destroy any chance of cross-correlation of data between the two cores. If one hole is placed at the current summit of the ice sheet, and the other one down the ice flow line to the west, as suggested earlier, then a separation of significantly more than 30 km starts to run into this problem: the bed topography has a saddle, a small bump, and then, at 50-60 km from the summit, starts a steady decent of over 200 m into a "canyon" whose bottom is below sea level.

(i) In a similar vein, as the separation of the two holes increases, there is an ever-increasing chance that the actual ice flow line will deviate, either now or in the past, from the one which is estimated based on the current topography. Errors in the measurement of this topography, as well as the uncertainties inherent in any interpolation and contouring technique, progressively degrade our ability to define the flow line, and therefore the relative orientation of the two holes with respect to the ice flow, as the holes become further and further apart. Like the two previous points, this is another reason for making the separation too large. Beyond 30 km or so, in fact, the proposed plateau narrows considerably, and changes direction slightly, so this problem could be significantly more important for separations greater than 30 km.

Ideally, the effects of divide migration would be best determined by a large hole separation, 60 km or more. However, given the constraints imposed by the actual topography, the increasing loss of coherence in the regional climate as separation increases, and the fact that the other criteria are either completely, or at least reasonably closely, satisfied by a separation of 30 km., we have selected the latter figure instead. All of the vertical velocity and most of the horizontal velocity would still be characteristic of different flow regimes at this spacing. However, if the separation is less than this value, then the two holes will not be glaciologically distinct, and it will be difficult to use ice flow to help interpret the ice core records. Advantages lost by having two holes clo5e together:

If the holes are situated close together, less than a few ice thicknesses apart, then reasons (2) and (3) are lost for either a flank or divide location, and reason (4) reduces to just the advantages and disadvantages of the flank or divide site chosen. In addition, it will be much more difficult to determine which flow regime the chosen location actually had throughout most of its history, regardless of where it happens to lie at the present time.

Recommended sites:

Figures 3, 4 and 5 show our recommended drill site locations. They are at either end of the bench referred to earlier.

References:

Raymond, C.F., 1983. Deformation in the vicinity of ice divides. Journal of Glaciology, Vol. 29, No. 103, p. 357-373.

Reeh, N., 198?..

Whillans, I.M., and others. 1984. Ice flow leading to the deep core hole at Dye 3, Greenland, by I.M. Whillans, K.C. Jezek, A.R. Drew and N. Gundestrup. Annals of Glaciology, Vol. 5, p. 185-190.

Whillans and Johnsen, 198?.

<- Back | Table of Contents | Next ->