1. INTRODUCTION 
The island of South Georgia is located in the South Atlantic 
centred on 54° S, 37° W (Figure 1). It is isolated, and heavily 
glaciated, with 60% of the island being covered by glaciers or 
ice sheets. Its location and isolation make it the primary 
breeding ground for many species of marine birds, such as 
albatrosses, petrels and penguins. 
  
  
  
  
  
    
  
  
        
   
=> 2 
/ ios Falkland 
Je Islands 
South 
NS 
- "w Georgia 
ges ; 
South 
America 
    
0 kilometres 1000 
« Ce) eral 
Figure 1: Location map for South Georgia 
  
  
  
  
Recent studies (Cook et al, 2009) have shown that the 
majority of coastal glaciers on South Georgia are in retreat. A 
small minority are, however, advancing, and a further small 
number are retreating far more rapidly than the average. As 
Cook et al. (2009) show, these changes in glacier extent may 
have significant consequences on the breeding success of 
these iconic birds. Introduced terrestrial predators, in 
particular rats, are currently blocked from major breeding 
grounds by glacier barriers, and continuing retreat of glaciers 
threatens these breeding grounds. Comparison of areas 
occupied by rats and those not shows that most species of 
bird nesting in South Georgia cannot breed successfully in 
areas inhabited by rats. 
In order to understand glacier dynamics, a crucial parameter 
is ice thickness. This can be measured in a variety of ways 
using standard techniques such as seismic sounding, ice- 
penetrating radar or even by drilling. However, all these 
techniques require substantial logistic support, which is not 
available in South Georgia. South Georgia is only accessible 
by ship; there is no landing ground for fixed-wing aircraft. 
While an over-snow expedition could potentially carry out a 
survey of ice thickness, it would be limited in its areal 
coverage compared with airborne survey. Aircraft equipped 
for ice-penetrating radar surveys do not have sufficient range 
to perform a survey over South Georgia after flying from the 
nearest airfield at Port Stanley in the Falkland Islands. 
Fortunately, South Georgia was covered by the Shuttle Radar 
Topography Mission (SRTM) during February 2000, being 
just north of the southern limit of 56? S. The product used 
was DTED 1, 3 arc-second product (-90m post spacing). 
SRTM elevations have an absolute accuracy of 8 metres 
(90% probability) for islands, and a relative accuracy of 6.2 
metres (Farr et al., 2007). Farr et al. also state that accuracies 
are worst over steep slopes, and better for flat areas, such as 
glaciers and ice fields, though the improvement of accuracy 
over flatter areas is not quantified. These relatively accurate 
elevation data permit an estimate of ice thickness to be made 
from the surface slope of the glaciers (Paterson, 1981, page 
86). 
591 
In places, the ice thickness estimates are clearly substantially 
in error, giving unrealistically high estimates. This can be 
linked to changes at the base of the glacier, providing an 
insight into conditions that are relevant to the pattern of 
retreat of coastal glaciers. 
2. METHOD 
2.1 Theory 
The surface slope of a glacier in a steady state is related to 
the ice thickness by the following relationship: 
TzÉzgg.hsina (1) 
Where 1 is the basal shear stress, p is the density of ice, g is 
the acceleration due to gravity, h is the ice thickness and a is 
the surface slope (Paterson, 1981, page 86). 
This equation can be re-arranged to provide a relationship 
between ice thickness and surface slope, assuming a constant 
basal shear stress: 
T 
Bx ——r 
fe 5. fin er (2) 
So, assuming that the retarding forces at the base of the 
glacier (tr) are unvarying, it is possible to estimate ice 
thickness using surface slope values alone, as all other terms 
in the equation are constant. The value of 1 varies within the 
range 50 kPa to 150 kPa depending on a variety of factors 
including the temperature of the ice and the nature of the 
substrate; a reasonable assumption for its value in the 
absence of other information is therefore 100 kPa. 
Glaciers in South Georgia are constrained by valley walls, so 
additional corrections are required to account for this. 
€ 
h pg. gina.F G) 
Where F is a correction factor that depends on W, the ratio of 
the distance to the valley wall and the ice thickness on the 
centre-line of the glacier. GIS techniques detailed below 
allowed the distance to the valley wall to be computed 
accurately, and F was obtained from Table 1. 
Given that much of South Georgia is covered by perennial 
snow or ice, determining the location of glacier margins is 
not trivial. A variety of techniques were tested, but the most 
reliable was clipping the slope data at a value of 17°. 
Methods based on image analysis using a composite Landsat 
ETM+ image failed due to snow cover on glaciers and heavy 
shadowing, but were used to eliminate areas of low slope that 
are not snow-covered (e.g. deglaciated areas in front of 
retreating glacier snouts). The second derivative of the 
surface (i.e. rate of change of slope) in many areas provided a 
good delineation of the edge of a glacier, but failed in areas 
where the glacier merged into snow-fields and at ice-falls. 
Having determined the glacier margins, the next step was to 
compute the glacier centrelines. This was done by computing 
the Euclidean distance from the glacier margins, the centre- 
line is then the trace of the maximum distances from the 
glacier margins. The distance to the glacier wall is then 
available at every point along the glacier centre-line, and the 
mean ice thickness can be computed by averaging over a 
small region along the centre-line. These parameters are used 
to compute the correction factor in Equation 3 (above).