Behavioural processes

The high estimate of mortality may in part be explained by seasonal variation in krill behaviour. Behavioural mechanisms that may influence seasonal variation in abundance include changes in krill swarming patterns and diel vertical migration (DVM) regimes. Foraging behaviour and predator avoidance are often cited as biological factors that influence the formation and maintena

Time (month from 1st January)

ind. m -2 ) Krill abundance ( 1 M = 0.05 mo-1 M = 0.28 mo-1 0 20 80 2_{M = 0.10 mo}-1 3 M = 0.16 mo-1 40 60 0 1 2 3 4 5 6 7

shown that most krill biomass often occurs in small, dense aggregations in the summer, whereas in winter, krill biomass tends to be distributed in larger and more diffuse assemblages (Lascara et al. 1999). Seasonal changes in krill swarming behaviour will directly affect small-scale distributions, which in turn alter the ability to estimate accurately the local abundance of krill according to the capabilities of the sampling equipment. For example, a change in behaviour that favours dispersal rather than swarming would lead to a more uniform spatial distribution, and krill densities may become too low to be detected acoustically using our instruments (minimum detection thresholds for the WCPs and ADCPs are around -85 and -82 dB, respectively).

Vertical migration of krill has been reported from the summer period and occurs generally within the upper 150 m of the water column (Demer and Hewitt 1995; Miller and Hampton 1989), while in more northern latitudes like South Georgia, krill may occur deeper, although still above 200 m (Marr 1962). However, some observations have indicated that during the winter, krill might be benthopelagic or live

close inter

deepe ved

by Ta eep

as 30 ical

migra krill

densities in the winter, as individuals may have been situated below the moorings e hours in which our acoustic observations were made. Thus there f krill associated with the SACCF that arrived close to e moorings in the winter seasons, but our moorings were not deep enough to detect

to the bottom even at depths of around 400 m (Gutt and Siegel 1994). A w ning in the vertical distribution of krill across the Scotia Sea was also obser ki et al. (2005) in fishery data, with the highest concentrations occurring as d 0 m during the day. It is therefore possible that seasonal changes in vert tion might be another possible reason to explain the apparent reduced

during the daytim

could have been a large influx o th

it.

A further possibility is that there could be an active migration process occurring that could have significant effect. Marked seasonal changes in distribution and regional biomass has been recognised around the Antarctic Peninsula, as a result of adult krill migrating on/off shore in summer/winter (Lascara et al. 1999). However, there is currently no evidence to substantiate such a mechanism at South Georgia. These behavioural mechanisms warrant further detailed investigation in order to understand

temporal and spatial variability in krill abundance at South Georgia and throughout the Scotia Sea (Nicol 2003; Siegel 2005).

4.6. Conclusions

In conclusion, a regular annual cycle of variation in krill density was detected on the South Georgian Shelf between 14 October 2002 and 29 December 2005, with krill density being high in the summer and low during the winter. The available data suggest that this pattern also occurred at the off-shelf site, although krill densities were substantially lower than those detected on-shelf. Explanations for the observed pattern of seasonal variation in krill density may be complex, and there is a range of ossible factors that could act synergistically to mediate such change. Cross-

sistent with a pattern of seasonal growth, production and mortality of a resident onsistent with the notion of large influxes p

correlation analysis showed a significant negative correlation between on-shelf temperature at 200 m and on-shelf krill density following a temperature lag of 1-2 months. Limited data from the off-shelf mooring suggested a similar correlation. Data from CTDs deployed on Argo floats and elephant seals provided valuable insight into the timing of the occurrence of the SACCF in proximity to the moorings between 1 January 2004 and 1 November 2005. Peaks in krill density did not coincide with times when the SACCF was detected in proximity to the moorings, and there was no evidence that seasonal variations in krill density were driven by oscillations in the SACCF. There was some evidence to suggest that seasonal variation in krill density at the off-shelf site was linked to changes in current velocity. Krill densities decreased towards the end of the summer season in 2004 and 2005, and this coincided with an increase in current velocity and a reduction in water temperature. However, velocities were slower and more uniform at the on-shelf site, and seasonal variation in krill density appeared not to be mediated by prevailing current velocity. Our data are not con

krill population at South Georgia, but are c

of krill in early summer, and of a predator-driven reduction at between mid- and late- summer. We suggest that seasonal changes in krill behaviour, such as DVM and swarming behaviour might also by important factors, and this warrants further investigation. Our observations did not coincide with seasons of poor krill availability at the island (Brierley et al. 1997).

Moorings provide valuable insight into the coupled biological-physical ecosystem at South Georgia. We envisage that continued use of moorings will provide further data that will be important for understanding, and ultimately predicting, the complex

s underlying intra- and inter-annual variability in krill density. Such causal mechanism

data are prerequisite for the implementation of robust ecosystem-based management strategies at South Georgia.

5. Regional variation in distribution pattern, population