Summer foraging behaviour of a generalist predator, the New Zealand fur seal (Arctocephalus forsteri)

Abstract

This study examined the dive behaviour of 20 lactating New Zealand fur seals (Arctocephalus forsteri) breeding at Fuchsia Gully (Ohinepuha, 45˚52S, 170˚44E), Otago Peninsula, New Zealand, over five consecutive austral summers (1993/94–1997/98). We examined annual variation in dive behaviour by classifying series of dives into dive bouts using an iterative statistical technique. We found a non-random pattern of dive bouts and bout classification was relatively insensitive to changes in the clustering parameters used. Minimum bouts consisted of at least three dives 10 m occurring within a 20-min period. Bouts were classified into three bout types (clusters) using a multi-variate clustering procedure. These clusters described bouts of: (1) long duration with many dives of medium depth (LONG); (2) short duration with few, shallow dives (SHALLOW); and (3) short duration consisting of long, deep dives and long surface intervals and bottom times (DEEP). Diving was primarily nocturnal, and bout type varied significantly with time of day. The proportion of LONG bouts was greatest at dusk and least near dawn, SHALLOW bouts predominated during the night, and DEEP bouts were of importance near dawn. Few dives occurred during the day. We detected no annual differences in individual parameters of dive behaviour due to low statistical power. We used randomisation tests to assess whether the proportion of each bout type might vary in years of differing prey consumption, but no significant differences were found. Changes in prey composition were detected in two of these years, which suggests that using the dive behaviour of generalist predators to detect changes in resource availability may be a poor option. The high degree of flexibility in foraging behaviour of the New Zealand fur seal means that, inevitably, analyses of dive behaviour will have low statistical power. Changes in foraging behaviour may only be useful to detect very large changes in resource availability. Alternatively, very large sample sizes may be able to detect more subtle changes.