Avoidance of seismic survey activities by penguins

Seismic surveys in search for oil or gas under the seabed, produce the most intense man-made ocean noise with known impacts on invertebrates, fish and marine mammals. No evidence to date exists, however, about potential impacts on seabirds. Penguins may be expected to be particularly affected by loud underwater sounds, due to their largely aquatic existence. This study investigated the behavioural response of breeding endangered African Penguins Spheniscus demersus to seismic surveys within 100 km of their colony in South Africa, using a multi-year GPS tracking dataset. Penguins showed a strong avoidance of their preferred foraging areas during seismic activities, foraging significantly further from the survey vessel when in operation, while increasing their overall foraging effort. The birds reverted to normal behaviour when the operation ceased, although longer-term repercussions on hearing capacities cannot be precluded. The rapid industrialization of the oceans has increased levels of underwater anthropogenic noises globally, a growing concern for a wide range of taxa, now also including seabirds. African penguin numbers have decreased by 70% in the last 10 years, a strong motivation for precautionary management decisions, including the exclusion of seismic exploratory activities within at least 100 km of their breeding colonies.


Results
In total, 333 complete individual foraging tracks were recorded from Bird and St Croix islands between March and May 2009-2013 from breeding African penguins during and outside seismic activities at St Croix Island (n = 31 and n = 74 respectively) and Bird Island (n = 20 and n = 208 respectively).
Foraging effort (duration of trip at sea, foraging path length, maximum distance from the colony) varied among years and was generally greater for birds from St Croix Island (( Table 1, Fig. 1). Compared to other years, foraging effort was slightly lower in 2013 in the absence of seismic activity but increased for penguins from both colonies when the seismic survey was taking place (Table 1). Maximum foraging distance from the colony increased significantly for St Croix birds during seismic activities (p = 0.007, Table 2, Fig. 2a).
Over the entire study period, St Croix Island penguins generally foraged towards the south east of their colony, or due south mostly within the 100 m bathymetric contour of the continental shelf (Fig. 1). Therefore, their preferred foraging areas were closer to where the seismic survey vessel was located in 2013 compared to that of Bird Island birds (ca 65 km on average versus >100 km for St Croix and Bird islands respectively, Table 1, Fig. 1). When seismic activities took place in March 2013, St Croix birds switched to foraging due east or north east of their colony (Fig. 1), constituting a significant change in bearing (Watson 2-sample test = 0.47, p < 0.001). As a result, the birds foraged significantly further away from the centroid of the seismic activities during that period (77 km, compared to ca 65 km on average in the absence of seismic activity, p = 0.008, Fig. 2b, Table 2). By contrast, Bird Island penguins consistently travelled due east to south-southwest of their colonies, also within the 100 m bathymetric contour of the continental shelf ( Fig. 1), regardless of seismic activities. Thus, there was no significant change in bearing for birds from Bird Island (Watson 2-sample test = 0.14, p > 0.1).
Comparing penguin's foraging effort within 2013 only, once the seismic operations ceased the maximum distance travelled by St Croix penguins significantly decreased (Table 1, Mann Whitney U test, w = 258.5, p = 0.03), as well as their foraging distance to the centroid of the positions of seismic vessel (SEISDIST, Table 1, w = 254, p = 0.02). Trip duration and foraging path length remained similar (w = 97, p = 0.2; w = 138, p = 0.17, respectively).

Discussion
Penguins foraging <100 km from active seismic operations showed a clear change of foraging direction during seismic periods, increasing their distance between their feeding area and the location of the seismic vessel. To our knowledge, this is the first record of avoidance behaviour by a seabird to sounds generated from anthropogenic activities at sea. African penguins from St Croix Island seemed to have avoided airgun sounds by foraging east of their colony, diverting from their traditional feeding grounds located in a more southerly direction. Avoidance behaviour as a response to seismic operations has been documented in many cetaceans (see 13 for a review). For example, bowhead whales Balaena mysticetus avoided the area of seismic sources by >20-30 km and showed signs of altered diving and surface behaviour at distances up to 73 km from seismic vessels 39 . Similarly, avoidance behaviour by gray whales Eschrichtius robustus were recorded at ranges up to 24 km from seismic activity and  Table 2. Coefficients (β) and standard errors (SE) of general linear models fitted to assess the influence of seismic activity on four responses, three path metrics (trip duration, path length and maximum distance to the colony) and the distance between the location of penguin fixes at the maximum distance from the colony to the centroid of seismic activity (SEISDIST) for birds breeding on St Croix and Bird islands.
altered behaviour (faster and straighter swimming and shorter blow intervals during seismic noise) at ranges >30 km 40 . Humpback whales Megaptera novaeangliae showed avoidance behaviour at a range of 5-8 km from a full-scale seismic array and maintained a stand-off range of 3-4 km 12, 18 . Avoidance behaviour has also been noted in fish, although behavioural studies on unrestrained fish exposed to airgun sounds are scarce (see 20 for a review). The avoidance behaviour by penguins observed in this study may be explained by either a direct disturbance from the noise generated by the operation or a change in fish distribution during that period (possibly as a result of seismic activities). The present study cannot disentangle the two effects. A possible decrease in prey availability following seismic operations was previously raised as a cause of concern as an indirect impact of surveys on marine mammals 13 . Incidences of reduced commercial fish catches have been recorded in areas where seismic survey were active or directly after the cessation of activities, suggesting avoidance of the area by the targeted fish species (e.g. 22 ). However, several in situ studies showed limited direct response of fish to seismic activities 20 , and when there was a response, the vertical rather than the horizontal distribution of fish was generally influenced (e.g. 26,41 ). Consequently, reduced commercial catch rates associated with seismic activity may possibly have resulted from a vertical displacement of fish. Small-scale acoustic fish surveys assessing distribution and abundance of small pelagic fish in Algoa Bay around both penguin colonies 42 did not show a significant change in distribution and/or abundance of small pelagic fish in the region in March 2013 compared to a few months prior to or after the seismic operations 43 . Therefore, African penguins likely relocated away from their traditional feeding zone to avoid the disturbance generated by the noise of the seismic vessels, rather than to follow their prey.
The exposure to intense sounds, such as the shooting of airguns during seismic operations, can adversely affect the hearing capacity of marine mammals and other species, either temporarily or permanently 1 . This impairment can reduce individual foraging performance, by diminishing prey detection capabilities, but also indirectly by reducing their ability to detect predators or assess their environment, thereby reducing the overall fitness of the individuals affected 44 . Such threshold shifts have been demonstrated experimentally in several species of fish and invertebrates, either in the laboratory or in cages placed in the wild (see review in 20 ) but are generally difficult to assess in wild populations. The hearing capabilities of birds are complex and poorly understood 45 . Although some information is available on underwater hearing capacities of cormorants 46 , virtually no research has been conducted on hearing in penguins in particular 47 . The impact of noise on terrestrial birds is, however, well known and noisy anthropogenic activities can reduce the abundance of passerines, although the mechanisms are unclear 48 . A potential cause could be related to interference with vital life histories involving acoustic communication, such as mate selection or territorial defense, which may ultimately affect breeding success 48 . At sea, however, such mechanisms are unlikely. Loud underwater sounds, such as airgun shooting, may be uncomfortable for birds, especially as sounds travel five times faster in water than in air and cover much greater distances at higher amplitude levels. Pingers emitting sounds of 1 kHz at 120 dB attached to driftnets significantly reduced by-catch of common murres Uria aalge 49 , although the study could not establish if the sounds emitted by the pingers were repulsing birds or their prey. African penguins are known to be sensitive to sounds as low as 100 Hz 34 , therefore it is possible that the sounds emitted by the surveys were a direct disturbance to them.
Noises from seismic operations may also have disrupted communication between African penguin individuals and groups, leading to a change in foraging behaviour, especially considering that the fundamental frequency (i.e. the lowest frequency component) of their vocalisations is around 250 Hz and plays a key role for individual discrimination 50 . A number of marine top predators rely on acoustic signaling for communication, orientation, locating prey and predators 2 . While knowledge of their use of vocalisation for communication at sea remains very limited, it is known that penguins use sound extensively on land for intraspecific communication including mate and chick recognition 51,52 . Contact calls have been primarily recorded for penguins at the surface when at sea [refs 51,53 ,McInnes unpubl.]. Lessening an individual's ability to detect socially relevant signals could affect biologically important processes (e.g. 6,13 ). African penguins often forage in groups 54 , which improves their prey capture efficiency 55 . It is therefore possible that they may use acoustic signals to coordinate their movement at sea and may be disturbed by loud anthropogenic activities. African penguins are also known to respond to underwater vocalisations of predators 56 . Anthropogenic noise pollution may therefore also affect their capacity to detect the presence of a predator, with potential negative consequence on their survival.
African penguins quickly reverted to normal foraging behaviour after cessation of seismic activities during this study, which suggest a relatively short-term influence of seismic activity on these birds' behaviour and/or that of their prey. Most bird and many fish species have the capacity to regenerate lost or damaged sensory cells of the ear 57 , although we cannot rule out potential longer-terms impacts on their hearing ability. Longer or repeated exposure to elevated underwater noise levels can affect reproductive and growth processes in some marine organisms 58 and lead to chronic stress 59 , which in turn can lead to a depressed immune function 60 . The potential for disturbance from cumulative impacts is particularly high for resident species with limited dispersal abilities 7 . This might be particularly true for African penguins breeding on St Croix Island, the largest African penguin colony 61 , as it is located in the vicinity of two large industrial harbours in the bay.
The biological significance of altered behaviours during seismic surveys remains difficult to measure. Some behavioural responses have been associated with reduced rate of foraging or of predator avoidance (e.g. 27 ), others with increased energy expenditure (e.g. 62 ). African penguins increased their foraging effort during seismic periods, particularly when their general foraging area was <100 km from the seismic operations. Increasing energy expenditure at sea to locate food can negatively affect penguins' reproductive output 63 . Breeding success of African penguins is currently at very low levels due to a suite of threats, from predation to extreme weather events 64 , and reduced food availability due to local competition with fisheries 37,65 . As long-lived species, biologically important changes in rates of population trends are difficult to identify, particularly over a short time scale. Following of a recent drastic decrease in their population numbers 61 , a Biodiversity Management Plan has been drafted by the South African Department of Environmental Affairs 38 , to assess and manage the threats to African penguins. Relevant to this management plan, results of the current study demonstrate that seismic survey operations may negatively impact penguins within 100 km of their feeding localities, and should be restricted to areas >100 km from African penguin colonies.
Rapid industrialization associated with resource extraction in the oceans has increased levels of underwater anthropogenic noises, a growing concern for the survival of a wide range of taxa 1,21 . In addition to over-fishing, habitat destruction and chemical pollution, underwater noise pollution is now recognized as a significant threat to marine wildlife 19 . Many underwater animals from invertebrates to marine mammals, rely on sound-based cues to forage, attract a mate or avoid predation 1,66-68 . Therefore anthropogenic sounds may perturb crucial life history traits 21 . Direct evidences for impacts of noise pollution on marine wildlife remain scarce (e.g. 28 ), even if deemed very likely 14,69 . This is largely due to the difficulty in acquiring the necessary data to demonstrate such effects, despite their potential negative impact on another major marine economic sector, commercial fishing (e.g. 70,71 ). Consequently, the existing evidences are largely anecdotal (e.g. 24,72 ) and there is a crucial need for additional studies of impact of loud noises, such as generated by oil and gas exploration activities, on hearing capabilities, avoidance behaviour and prey dynamics of animals including seabirds. Penguins are currently the most threatened seabird family, and based on the findings of this study, prudent planning of seismic exploration surveys in their habitat is required 7 .

Methods
Foraging behaviour data collection. The foraging behaviour of adult African penguins raising chicks <10 m, and weighing <2.5% of adult body mass. Birds were caught at their nest site, the loggers were attached to their lower back feathers with waterproof tape, and they were released at the nest within <6 min (see details in 35 ). Nest sites were then monitored until the birds returned and the devices were removed. If several foraging tracks were recorded per individual bird, only one (the first one recorded) was included in the analyses to avoid pseudo-replication.
Seismic sound source. Seismic surveys (2D) took place in South Africa in the Algoa Bay/Gamtoos river mouth area from 15th of February to 22nd of March 2013, covering an estimated distance of 1 527 km and a total area of 6 700 km 2 (Fig. 1). Airguns were shot at point intervals of 25 m at an average of 169 airgun shots per hour. Acquisitions were done 24 hours a day, at an average of 11 acquisitions per hour. The source was made up of 4 sub-arrays of airguns (Bolt Long Life 1 900 XT) with a total volume of 4 230 in³ at 2 000 psi ± 10% pressure for the array. Data acquisition was only paused in the event of the presence of marine mammals in close proximity of the ship and when changing lines.
Statistical analyses. From the GPS tracks, we estimated foraging effort (i.e. foraging trip duration, foraging path length and maximum distance from the colony) and the distance of the furthest GPS position, i.e. location of maximum distance for each individual, to the centre of seismic activity (hereafter referred to as SEISDIST), which was determined by calculating the centroid of all georeferenced seismic activities in March 2013 (Fig. 1). Tracks were filtered to exclude erroneous fixes that exceeded the potential distance covered given their mean maximum speed (12.4 km h −1 , ref. 73 ). Trip duration was only calculated for tracks with start and end fixes <5 km from colonies and foraging path length for tracks that had gaps <2 h. When start and/or end fixes were not at the colony but within 5 km, distance travelled and duration were calculated from the average travelling speed of African Penguins in Algoa Bay (2.5 km h −1 , ref. 74 ). The bearing of individual tracks from the island to the farthest point of their trip was calculated in software R 75 (R Core Team, 2015) using package 'Geosphere' 76 . Kernel density analysis was performed in ArcGIS 10.4 on the totality of the individual tracks, using the adaptive kernel method with smoothing parameters selected based on least-squares-cross-validation. Estimates were created for foraging ranges based on 50, 75, and 90% utilisation distribution.
We used generalised mixed effects models (GLMM) with a Gamma error distribution and a log link function ('lme4' package 77 ,) to assess the influence of seismic activity on penguin foraging effort and SEISDIST, with presence/absence of seismic activity and year as fixed effects. Year was included to account for annual variability in oceanographic conditions and fishing intensity, which are known to influence prey availability and penguin foraging performance in this region 33,34,41 . For the models using variables of foraging effort as responses we included colony as a fixed effect; for SEISDIST as the dependant variable models were fitted separately for each colony due to the bi-modal nature of the response. In addition, in order to establish if conditions potentially reverted back to 'normal' once the operations ceased, a within year effect was assessed for the significant foraging parameters, (maximum foraging distance and SEISDIST, see below) against seismic activity in 2013 only for St Croix penguins.
Non-parametric circular statistics using Watson's two-sample test of homogeneity were used to assess the differences in bearing from penguins colonies to their maximum distance location with and without seismic activity. This was done for each colony separately using package 'circular' in R 78 .
Data availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.