Upward-looking acoustic Doppler current profilers (300 kHz) and echo sounders (125 kHz) were deployed on moorings on- and off-shelf to the northwest of South Georgia between 14 October 2002 and 29 December 2005 to measure density of Antarctic krill and environmental parameters continuously. A distinct seasonal pattern in krill density, recurring consistently over all three years, was detected. Krill densities in winter were predominantly low (mean = 18.7 g m^–2, SD = 24.3), but had risen substantially by summer in each year (mean = 89.5 g m^–2, SD = 64.2). A sinusoidal regression model (period = 52 weeks) with time as the independent variable explained 64% of the observed week-to-week variation. Estimates of krill density from moored instruments were not statistically
different (P > 0.05) from estimates derived from standard ship-based krill surveys in adjacent time periods, suggesting that the point estimates from moored instruments were representative of krill density in a wider spatial context (ship surveys cover c. 100 x 100 km). Data from moored instruments were used to explore whether high-frequency temporal variation (i.e. within-year) could have led to the perceived between-year variation in krill density arising from previous summer surveys in the South Georgia western core box region between 1990 and 2005. Comparison of these ‘snap-shot’ ship survey estimates with the observed pattern of within-year variability showed that some of the apparent ‘year-to-year’ variation could simply be attributed to sampling on different dates of the year (e.g. November cf. February). However, there were some survey estimates that were signifi cantly different (P < 0.01) from the regression-predicted within-year variation. Years that stand out for markedly low krill density (i.e. densities below the range expected due to intra-annual variation) were 1993/94, 1998/99 and 1999/2000. Moored instruments provide valuable data that could be important for ecosystem-based management at South Georgia because, for example, they will enable predator–prey functional responses to be explored there for the first time at appropriate temporal scales, and will enable hypotheses relating variation in krill abundance to physical oceanographic variability to be tested.