The foraging behaviour of adult penguins raising chicks of one to three weeks old was studied at St Croix Island (33°48′ S, 25°46′ E, the ‘treatment colony’) and at Bird Island (33°50′ S, 26°17′ E, the ‘control colony’), before and after closure to fishing, in May–June 2008 and April–May 2009. The positions of purse-seine vessels were monitored constantly via satellite telemetry, ensuring compliance within the experimental closure. African penguins share the care of their brood of one or two chicks between March and August, with typically one adult attending the nest when the partner is at sea. Birds were equipped with GPS-TD loggers (a global positioning system recorder combined with a time-depth recorder; GPS-TD 96 × 39 × 26.5 mm; earth&OCEAN Technologies, Germany), which record latitude and longitude at 1 min intervals to an accuracy of less than 10 m, and depth at 1 s intervals to the nearest 0.1 m. The devices weighed less than 2.5 per cent of adult body mass and were housed in streamlined fibre-composite containers. They were attached to the penguins' lower back feathers with waterproof tape, causing no damage to the plumage. Handling lasted less than 6 min from capture to release, and these methods were approved by University of Cape Town's animal ethics committee. After deployment, nest sites of instrumented birds were monitored until the adult carrying the GPS returned, allowing it to be recaptured and the logger removed. Previous studies showed no significant difference in the foraging behaviour of instrumented versus control African penguins (Petersen et al. 2006
). This was confirmed in 2008 on St Croix Island, where adults at control nests with chicks of similar age were marked without being handled, using bio-compatible dye, and had their foraging trip durations recorded. Handling and instrumentation did not affect the duration of foraging trips (control birds: 22.7 ± 3.1 h, n
= 9, H
= 0.76, p
On retrieval of the devices, trip duration, path length, maximum distance from the colony and diving effort were calculated in order to compare foraging effort at each island between years (before and after closure) and between islands (treatment versus control) using general linear models (GLMs). Each parameter (trip duration, path length, maximum distance from the colony and diving behaviour) was tested as an explanatory variable, while year or colony were taken as dependent factors. Data were only recorded for a single foraging trip per bird to avoid pseudo-replication. There was no difference in brood size between nests across years or colonies (GLM with brood size as the explanatory variable, year and colony as dependent factors, colony: F1,90 = 0.61, p = 0.54; year: F1,90 = 0.08, p = 0.78), so the data from nests of different brood size were pooled.
A GPS position was associated with each feeding dive (>3 m and diurnal, as defined by Wilson & Wilson 1990
). Technical failures of the devices (no GPS positions or no dives recorded) decreased the sample of complete foraging trips where a GPS position could be attributed to each feeding dive. However, all other data were used for the calculation of foraging and diving effort. Adaptive kernel analyses were conducted using Arcview
GIS 3.1 with the smoothing factor chosen according to the least-squares-cross-validation method (Worton 1989
) to estimate contour levels covering 50–75–90% of the foraging locations.