There is considerable spatial structure within the South Georgia krill-base
d ecosystem in terms of predator demand, commercial fishery pressure and kr
ill abundance. Here we investigate the hypothesis that there is also spatia
l structure in the krill population composition, finding differences in len
gth-frequency histograms between the eastern and western ends of the island
. Between 1981 and 1997 the British Antarctic Survey carried out 6 major cr
uises where krill were taken from both ends of South Georgia. Cluster analy
sis revealed that the length-frequency histograms could be grouped into 4 m
ain types with relatively simple biological characteristics: Cluster 1 cont
ained small krill (mean size 23.9 mm) up to 2 yr old (1+ yr class); Cluster
2 (mean size 31.1 mm) contained a mixture of 1+ with some 2+ and 3+ yr cla
sses; Cluster 3 contained medium-sized krill (mean size 41.4 mm) probably 2
+ and 3+ yr classes; Cluster 4 contained large krill (mean size 50.3 mm) li
kely to be 3+ or older. Principal components analysis (PCA) provided good s
eparation of these clusters using the first 2 axes (80 % of the total varia
nce). There were no obvious differences in the length composition of krill
sampled from different water depths, although there were some indications t
hat differences did occur between different water masses. Detailed inspecti
on of the individual cruises revealed that the length-frequency histograms
at the western end of the island contained more large krill than those from
the eastern end of the island and also that when Weddell Sea water was fou
nd within a cruise then this contained the smallest krill. We consider that
such differences may arise not only because krill may experience different
conditions at each end of the island but also may originate from 2 separat
e sources which may have different population structures.