3.2 Materials and methods
3.4.5 Inter-annual variation
Although the numbers of predators ( ˆN) was higher in the WCB, the absolute values and inter-annual variation in ˆp, was low in the WCB when compared to the ECB (Table 5.7 and Figures 3.3 and 3.4). This suggests that the inter-annual variation in WCB ˆN are not driven by changes in ˆp; at least by those reflected in the snapshot measurements made in these surveys. From the three years of data, only ˆρ during the 1999 survey was higher in the WCB than the ECB. A pattern of higher ˆp in the WCB was observed by Brierley et al. (1999b) for only one year (1994) out of five (1990, 1994, 1996, 1997 and 1998). As with 1999, the low ˆp in 1994 occurred during an overall low krill density year (low krill density classified as ˆp = 33.4 g/m2, as defined by Brierley et al. (1999b) rescaled using the Demer and Conti (2005) target strength model).
Using the sinusoidal model (Figure ) for within year temporal krill density variation created by Saunders et al. (2007) the minimum and maximum line transect survey ˆp
measurements can be estimated. Whilst this model ignores inter-annual variation in drivers of krill density, the Saunders et al. (2007) model is instructive for showing that the line transect surveys used to calculate ˆp in JR28 (late January 1998) in this investigation take place near the predicted time of maximum ˆp, 5 weeks after 1st January (Table 3.5). Surveys JR17 and JR38 were conducted in late December and early January during which the Saunders et al. (2007) model predicts ˆp is at 65% of its maximum (Table). Consequently, the high ˆp observed during the later JR28 cruise, may have partially been caused by the time of year that the cruise was conducted. However, the time of year corrected ˆp (Table 3.5) show that 1998 was still a year of exceptionally high ˆp. Also, it should be remembered that for the purposes of studying krill predator-prey interactions it would be inappropriate to use the time of year scaled ˆpbecause this would introduce a temporal mismatch between the krill and air-breathing predator observations.
Western core box Eastern core box
Year Cruise pˆ scale factor scaled ˆp pˆ scale factor scaled ˆp
1997 JR17 53.16 1.52 80.80 101.91 1.52 154.90
1998 JR28 54.39 1.07 58.20 350.51 1.07 374.50
1999 JR38 43.00 1.52 65.36 32.58 1.52 49.52
Table 3.5: Seasonal peak mean krill density estimates (ˆp, g/m2) using the Saunders et al.
(2007) model and the line-transect ˆppoint estimates and time of survey. The scale factor was estimated using the Saunders et al. (2007) sinusoidal model for ˆptemporal variation. Note: using the Saunders et al. (2007) scaling 1999 in the ECB is no longer considered a year of low krill density.
Predicted krill density week 52 Predicted krill density week 0
Figure 3.7: The sinusoidal model developed by Saunders et al. (2007). This model was used to rescale the one month later JR28 mean areal krill density estimates ( ˆρ). Using this model the difference between ˆρ at week 52 (blue line) and ˆρ at week 0 (red line) of 35% was used to calculate the scale factor in Table 3.5.
local (small-scale) predator sightings (n) and ˆp may result from local physical processes or could be biologically driven as suggested by Meredith et al. (2005). Finer-scale models of the local-scale physical processes around the South Georgia shelf should help determine the role of biological forcing in ˆpdynamics. For example, the higher number of predators in the WCB may have been caused by a lower inter-annual variation in ˆp, while this might appear to make foraging potentially less energetically profitable than the ECB, this might be balanced by making the distribution of prey more predictable. This lower variation in ˆp may be caused by consistent advection of krill into the WCB or, in contrast to the ECB, a higher level of predation which dampens variation in ˆp, so a high density of krill never accumulates in the WCB.
The lower density of krill in the WCB compared to the ECB (Table 3.3; Brierley et al. 1999b) is sufficient to sustain predators that forage in the WCB (e.g. Hunt et al. 1992a; Reid et al. 2000b), suggesting there are other reasons why predators chose to forage in this lower region of ˆp. For example, krill may have a lower detectability in the ECB by predators. This may be caused by krill forming smaller swarms, or multiple krill swarms clustering differently in the ECB (Trathan et al., 2003). Alternatively, krill may be distributed at deeper depths, making swarms harder to locate, preventing surface feeders from foraging and requiring greater energy expenditure by divers (Mori and Boyd, 2004). Finally, the spatial location of krill may be more predictable to predators in the WCB, making a simple foraging strategy effective (e.g. Trathan et al. 2006) enabling
individual animals to optimize foraging, using experience from previous trips (Staniland et al., 2004). The swarm characteristics reported in Chapter 2 go some way to supporting this, particularly swarms in the WCB. Swarm types in the WCB show a consistent split inter-annually and between the proportion of a swarm type that is found on and off shelf (Table 2.8).
3.4.6
Summary
The results of this investigation have shown that core box is more important than year for determining the number of air-breathing predator species present and their abundances. Further, despite the higher density of krill in the ECB, more predators were found in the WCB: the reasons for this remain unclear. The absence of Blue petrels and the elevated abundance of Antarctic fur seals suggest that 1998 was an anomalous year, characterised by colder than average water surrounding South Georgia, and a high density of krill in the ECB.
It is apparent that simply examining snap shot krill density on the scale of a core box cannot explain why more predators appear to be foraging in the WCB, a region of lower instantaneous ˆp. Consequently, it may be necessary to examine the distribution of krill at a finer spatial scale the rate of krill flux in the boxes, to consider in more detail areas that may have high ˆp, such as the continental shelf break (Trathan et al., 2003). Using moored instruments Brierley et al. (2006) showed three step changes in krill abundance in a 28 day time series in the WCB. That study demonstrates that it is important to reconcile differences in temporal variation between predators and prey in the South Georgia ecosystem. This also showed the limitations of the acoustic estimate of ˆp, used in this investigation, which assumed that an instantaneous snap shot of krill over a typical 5-day core box survey was representative of a seasonal krill ˆp, and could not examine temporal variation in krill density or predator encounters during a survey.
Given the openess of the South Georgia ecosystem to local and remote environmental variability this investigation has been useful for detecting the influence of ocean-wide events. Within the limitations of the data available, the qualitative influence of these events has been shown, but small scale observations and analysis is required to examine predator-prey interactions and further assess interactions between krill and air-breathing predator species. These issues will be addressed in the following chapter.
Small-scale spatial and temporal
interactions between Antarctic krill
and air-breathing predators at South
Georgia, 1997 to 1999
4.1
Introduction
In the previous two chapters it has been shown that there was significant variation in the mean areal krill density ( ˆρ) between the two South Georgia study sites. The eastern core box (ECB) had consistently higher ˆρ than the WCB. In chapter 2, it was shown that the number of krill swarms (ns) and ˆρ were strongly correlated. Also, krill swarm types were shown to be different between core boxes and on and off continental shelf regions. This chapter examines the small-scale (<10 km) spatial overlap between krill and air-breathing predators.
The research in this chapter has been conducted to determine the characteristic spa- tial scale (Ls) of krill thereby allowing indirect assessment of the prey-field available to predators. The assessment of krill distribution has been split into depth horizons, thereby enabling the availability of krill to predators to be determined.
The characteristic scale of krill predator species can be used to determine if predators forage in similar sized groups, which gives an indication of foraging search efficiency. The cross-correlation between characteristic scales of krill and predators determine the forag- ing overlap, and may also determine the cues a predator use. For example, facultative feeding, where predators do not use the prey distribution to as a foraging cue, rather an- other foraging predator may be used as a cue, which would be shown by high inter-species characteristic scale (Gr¨enbaum and Veit, 2003). Finally, a negative cross-correlation be- tween predators is suggestive of competition avoidance (Veit et al., 1993).
4.1.1
Marine predator-prey interactions
Observing scale dependent predator-prey interactions in the marine environment can be difficult. Often studies are based on acoustic and visual observations made from research ships conducting line transect surveys (e.g. Hunt et al. 1992a; Croll et al. 1998; Fauchald and Erikstad 2002). Marine surveys are time consuming, with the nominal research vessel survey speed being only 10 knots (equivalent to c 440 km of line transect observations per day). Ship based observers only sample a narrow strip transect along the sea surface (Figure 4.3), and vertically-downward looking conventional echosounders, with a narrow beam width, typically being 7o, only sample a small volume of water (Gerlotto et al., 1999). Analyses are often conducted assuming that the spatial coverage of surveys is achieved instantaneously, thereby providing a snapshot of a predator-prey system, which ignores rapid biotic and abiotic changes that can occur in the marine environment. Further, the relationship between predators and prey is likely to be complex, may be density dependent and occur at multiple spatial scales (e.g. Sims et al. 2008). Also, because the marine predator is likely to use a different search strategy to a research ship, there will be a sampling mismatch: there is no requirement for a predator to search for prey until the maximum available prey is encountered, predators simply need to find patches that where quality is above some minimum threshold. Therefore, simply expecting a positive relationship between predator and prey abundances is generally naive.
When considering predator-prey interactions as presence-absence it is possible to ob- serve a predator-prey system in four states (Table 4.1). Incorporating predator behaviour may help determine the state in which a predator-prey system is observed, but given the many potential behaviours of predators interpreting predator-prey interactions can be difficult. Even if predators and prey are co-located this does not mean that foraging is taking place. For example, predators could be resting between foraging bouts, be tran- siting to or from an alternative foraging site, or be satiated. This makes the observation of predator behaviour vital for interpreting predator-prey interactions (Table 4.1).