Australian magpies exhibit increased tolerance of aircraft noise on an airport, and are more responsive to take-off than to landing noises

Continual expansion of the global aviation industry means the rate of collisions between aircraft and wildlife (bird strike) is increasing (Allan and Orosz 2001). Bird strikes are a source of human and wildlife fatalities, can damage equipment and create major delays. Globally, methods such as lethal control and habitat manipulation are used to manage strike risk (Dolbeer et al. 1993; Dolbeer 1998; Baxter and Robinson 2007; Thomson 2007; Patrick and Shaw 2012; McKee et al. 2016).

Understanding bird behaviour in response to certain stimuli has major ramifications for how to best manage and reduce the chance of bird strikes. The classic paradigm that applies to bird–aircraft strikes is that encapsulated in a series of theories related to wildlife escape (Cooper and Blumstein 2015). Essentially, birds, like other animals, monitor their environment for risk (‘stimuli’, usually predators) and initiate a response that manages that risk (usually escape). Avian responses to humans and their machines are stereotypically similar to responses to predators although the degree of risk associated with these stimuli (inferred through response distances) varies among stimuli (Weston et al. 2012; Cooper and Blumstein 2015; Lima et al. 2015). Avian behavioural characteristics can mediate the likelihood of a bird strike. Escape behaviour has the potential to avoid or to cause bird–aircraft collisions. Postmortem on birds struck by aircraft has shown that many strikes occur on the posterior and ventral side suggesting that many species exhibit escape responses to aircraft but are apparently unable to escape due to speed of the approach (Bernhardt et al. 2010). Additionally, escaping birds face into the wind, so, initially, they fly in the same direction as the approaching aircraft (Burger 1983). Timely responses of birds presumably enable birds to avoid collisions with aircraft.

To avoid danger in their environment, birds must perceive approaching threats. Bird visual systems and responses are complex and response times are, at least in some species, linked to the speed of the oncoming stimulus or are specific to areas where stimuli travel at different speeds. Thus, birds may struggle to escape from rapidly moving stimuli (Weston et al. 2012). However, aircraft noise is a key aspect of the stimulus available to birds, which might enable them to respond to aircraft in a timely fashion. Noise is known to elicit a variety of responses in birds, which can be dramatically different among species, and which depend on the sound volume (Dunnet 1977; Burger 1981, 1983; Ellis et al. 1991; Arévalo and Newhard 2011). Aircraft are associated with different noise profiles when taking off and landing, those phases of the flight when they are most prone to bird strikes (Burger 1983). A long-term dataset from the USA indicated that 61% of strikes occur on landing and 35% during take-off (Dolbeer et al. 2015). However, available studies that show the effect of take-off and landing noise on bird escape responses do not include Australian birds or conditions, and studies of avian responses to aircraft noise are limited. The available studies that have examined the effect of aircraft on birds have generally involved military aircraft, and low overflights, and have almost always been away from airports (Burger 1981; Brown 1990; Ward and Stehn 1990; Ellis et al. 1991; Conomy et al. 1998b; Delaney et al. 1999; Goudie and Jones 2004). No study known to us has examined whether responses to noise on and around an airport vary in relation to the degree of exposure to aircraft, or whether different phases of aircraft flight (landing versus take-off) elicit different responsiveness.

Birds make complex decisions regarding escape, and apparently modify their responses on the basis of the prevailing risk environment (Weston et al. 2012). Rock doves (Columba livia) exposed to vehicles reduce responsiveness over time compared with naive individuals (DeVault et al. 2017). One process that may affect escape is habituation and habituation-type processes (Blumstein 2016), where responsiveness to a stimulus decreases in areas where the stimulus is common and generally benign. This may be due to individual learning or habituation-like processes such as selection (van Dongen et al. 2015), and has implications for effective avoidance of aircraft, and for effectiveness of bird management (e.g. hazing). Birds living near aircraft may reduce responsiveness, which is adaptive insofar as risky escape flights are less likely, but may mean birds approach or tolerate aircraft at closer proximities.

In Australia, 16 096 reported bird strikes on aircraft occurred in 2006–2015 (Australian Transport Safety Bureau 2017). Of all identified bird taxa, Australian magpies (Cracticus tibicen) are ranked the 6th most struck bird, with 513 strikes 2006–2015 (5.9% of 8717 strikes); 22 magpie strikes involved multiple birds (4.4% of 502 strikes; Australian Transport Safety Bureau 2017). Thus, magpies are a species involved in substantial numbers of collisions with aircraft in Australia. Light aircraft in Australia are common, and smaller-bodied aircraft are particularly vulnerable to bird strike as resultant damage can be catastrophic (Burger 1983; Dolbeer et al. 2015).

We investigated whether Australian magpies exhibited different responses to light aircraft noise in three areas that differed in distance from, and exposure to, aircraft. We aimed to (1) discern whether Australian magpies living in close proximity to aircraft exhibit reduced responsiveness in comparison to individuals that are less frequently exposed to close passes by aircraft, and (2) determine whether responses varied between noise associated with take-off and that with landing.

Point Cook RAAF Base (henceforth, ‘the base’) is a small aerodrome located 29 km south-east of Melbourne (37.930757ʹS, 144.754115ʹE). The predominant air traffic consists of Cessna 172, a single-engine, low-airspeed aircraft. The base consists of the following two areas: (1) ‘airside’, the part of the aerodrome that is made up of a mixture of short, grassed areas, runways; and (2) ‘airport but not airside’, a large area away from actively operating aircraft dominated by roads, short-grass areas and buildings. The third zone, ‘off airport’, the Point Cook Coastal Park, consists of a large area of native vegetation and saltmarsh that abuts the airport. All areas consisted of open grassed areas used for magpie foraging, with a similar availability of perches.

Behavioural responses of Australian magpies to audible playback stimuli were recorded between June 2016 and February 2017. We performed ‘trials’, whereby we played randomly allocated recordings of take-off, landing or neither (‘control’, i.e. we broadcast a blank soundtrack) to a variety of bird species. We avoided repeat sampling of birds within any session, and did not perform a trial within 50 m of any other trial. Only magpies (which were common) were sampled sufficiently and are the focus of the present paper (see Table S1, available as Supplementary Material to this paper, for summaries of data for other species). Trials were all car-based, and consisted of three 30-s phases: pre-, during and post-playback. Focal birds were haphazardly selected if they were within 50 m of the vehicle, were foraging (i.e. non-vigilant), and no real aircraft or other stimuli were active. We used the pre- and post-phases to ensure that any responses that occurred were due to the aircraft playback. The playback stimulus consisted of an audio recording of a Cessna 172 landing and taking off, as well as a control (such aircraft can cause substantive responses in birds; Conomy et al. 1998a). Both recordings were made using a Roland, R-26 portable digital recorder (sensitivity set to high), and ME62/K-6 omni-directional Sennheiser microphone, set 30 m from the place of take-off or touchdown and featured little background noise (for each recording, wind and external noise were reduced using the noise-reduction effect in Audacity 2.1.2 without any loss of quality). The recordings captured the entire noise associated with incoming or departing planes from a stationary point, and were made at Point Cook airfield. Playbacks were of the cleaned but original recordings, at a set amplification (same settings), and were played through a hand-held megaphone (TOA er-1215). Thus, the intrinsic differences in volume between take-off and landing were preserved; however, these were slight (means, at megaphone; take-off, 109.9 dB; landing, 109.7 dB). Volume was set at a realistic level (at the mean distance between broadcast and magpies (30 m), the average peak volume of playbacks was 97 dB, being at the upper end of the range (61–96 dB) of volumes associated with Cessna 172 operations. All trials were in fine weather, with slight or no wind. All trials were recorded by a video, using a Nikon 5100 DSLR with a 70–300-mm lens on a window mount. The distance between the bird and the observer (Dist._Car) was measured with a laser rangefinder.

We derived the following response variables from the ‘during’ phase of each trial:

1. Presence of a response (binary, analysed with GzLM-specifying binomial distribution and logit link);

2. Response intensity using an ordinal response scale (1 = no response, 2 = vigilance, 3 = walk or run away, 4 = fly away), which described the response of birds to the stimulus and was analysed using a GzLM (specifying an ordinal distribution and logit link); and,

3. Latency of response (the interval, in seconds, between the start of the playback and the initiation of a response, analysed using a general linear model, GLM).

We also explored the volumes involved in the playback experiment, so as to ensure that these were equivalent among areas. The peak (in dB) of both audio recordings was measured with Roland R-26 portable digital recorder (sensitivity set to high), and ME62/K-6 omni-directional Sennheiser microphone at increments of 10 m, starting at zero and ending at 50 m within the three areas. A repeated-measures GLM (predictor was the volume of take-off and landing (repeated)) showed no effect of area (F_2,14 = 2.590, P = 0.110), but an expected significant negative effect of distance (F_1,14 = 139.099, P < 0.001).

Models included terms for the area in which the trial was conducted (‘region’ i.e. airside, on airport but not airside, and off airport) and the type of noise used (‘soundtrack’ i.e. take-off, landing or control). An interaction between these terms was initially included in all models; however, because none of the interactions was significant, all models were run with main effects only. All models included the covariate of the distance between the observer and the bird (Dist._Car) to adjust for variation in this measure, which is also associated with the sound volume experienced by the bird.

Finally, we wished to examine whether any observed responses to aircraft noise mirrored responses to other stimuli. We performed standard pedestrian approaches to magpies across the three areas, according to the protocols described in Weston et al. (2012). This involved recording a starting distance (SD) and flight-initiation distance (FID), using a laser rangefinder, and analysing FID (response variable) against the covariate SD and the fixed factor region. Means are presented with standard errors throughout.

The model examining the presence of a response showed main effects of region and soundtrack (Table 1). The only significant difference in the probability of a response among regions was that responses on airside (0.20 ± 0.07) were less likely than those on airport but not airside (0.86 ± 0.78; P < 0.001). The probability of a response outside the airport did not differ from that on airport but not airside (0.67 ± 0.12; P = 0.198). All pairwise comparisons within soundtrack were significant (the highest P = 0.007), with the probability of a response to take-off, landing and control being 0.95 ± 0.02, 0.82 ± 0.06 and 0.03 ± 0.02 respectively.

The model exploring the variation in response scale showed no effect of region but a significant effect of soundtrack (all pairwise comparisons had P ≤ 0.002) and Dist._Car. (β, –0.04 ± 0.02; Table 1). Median response scores were 1.0 for airside, 2.0 for on airport but not airside and 2.0 for off airport.

For trials where a response occurred, we analysed the time to a response (s) against region, soundtrack and Dist._Car. Dist._Car (β, 0.08 ± 0.03) and region had significant effects (Table 1). Responses on airside (8.17 ± 0.71 s) were delayed longer than those on airport but not airside (6.06 ± 0.60 s) or off airport (5.68 ± 0.56 s). The only pairwise comparison with a low P value (0.074) was between off airport and the non-airside areas of the airport.

Flight-initiation distances (FID) to a human stimulus did not differ across the three regions (GLM, R² = 0.183; region, F_2,46 = 0.817, P = 0.448; starting distance, F_1,46 = 9.445, P = 0.004) and overall was 27.98 ± 1.88 m.

The present study has shown patterns of responses consistent with the idea that Australian magpies exhibit decreased responsiveness to aircraft noise on airports (reduced occurrence, and increased latency, of responses). It also suggests that they respond more to the louder noise associated with take-offs than to landings. However, even on airside, responses remained common and birds discriminated between the soundtracks (perhaps a generalised response to sound volume). Overall, birds continued to respond to aircraft in a manner that was consistent with discriminatory anti-predator behaviour. We cannot attribute these adaptations to specific processes, such as within-animal learning, local selection or recruitment, or a situation where magpies alter their responsiveness in a spatially dynamic fashion, i.e. by adjusting responsiveness to the prevailing acoustic environment or according to specific localities (Legagneux and Ducatez 2013).

Habituation, adaptation or tolerance to or of certain stimuli, including noise, can alter bird behaviour both in the short and long term (Brumm 2004; Quinn et al. 2006). We have shown, for the first time, decreased responsiveness of a passerine to aircraft noise among birds closest to aircraft. Conomy et al. (1998b) showed that one of two duck species habituated to jet aircraft noise. Although acute and dramatic responses of birds to aircraft noise are short-lived (e.g. the present study), some bird species exposed to aircraft noise exhibit altered behavioural time budgets for hours (Goudie and Jones 2004). Longer-term adaptations to aircraft noise are also known. Analysis of birdsong in high-noise environments has shown that individuals exposed to high levels of background noise sing with higher sound levels (Brumm 2004; Bermúdez-Cuamatzin et al. 2010), and some even advance their dawn chorus apparently to avoid overlap with aircraft noise (Gil et al. 2015).

The ability to habituate to or tolerate common stimuli, thereby altering behavioural characteristics, has led to certain species adapting to human-dominated environments such as urban areas (Blumstein 2006; Lowry et al. 2011), and the same may apply to airports. For example, aircraft noise may discourage some birds from breeding (Awbrey and Hunsaker 1997) and could conceivably exclude more sensitive species or individuals (van Dongen et al. 2015). Quieter aircraft (Burger 1985) may mean previously excluded species may be increasingly able to inhabit airports. Similarly, selection for experienced birds may occur; young birds may be disproportionately struck by aircraft and vehicles (Burger 1985). Also, experience reduces responsiveness to oncoming vehicles for rock pigeons (DeVault et al. 2015). These processes may contribute to the results we describe here.

Although enhanced tolerance of aircraft noise occurred among magpies, no such pattern was evident for pedestrian approaches. Although we tested only two stimuli (aircraft noise and a pedestrian), results suggest not only that Australian magpies are able to discern effectively among different stimuli (McLeod et al. 2013), but also that tolerance itself varies among stimuli. Stimuli to which habituation or tolerance occurs may be frequent, predictable and benign (Mumme et al. 2000; Blumstein 2016). In our study system, aircraft are likely to be more common and predictable than are pedestrians.

As the aviation industry continues to expand, bird strike is an ever increasing issue and poses major risks to humans and wildlife (Allan and Orosz 2001; Australian Transport Safety Bureau 2017). Globally, a wide range of tools are implemented to effectively reduce the risk of a bird strike (Patrick and Shaw 2012; Dolbeer 2013). Although many countries and airports favour actively culling birds in the vicinity of an aerodrome, a method that may reduce bird strike in the immediate term in some circumstances (Dolbeer et al. 1993; Dolbeer 1998), little evidence is available to suggest that these measures are effective over the long term (i.e. these measures require ongoing management intervention to remain effective). Our work and that of others (Harms et al. 1997; Conomy et al. 1998a, 1998b) has shown that some birds can increase their tolerance to aircraft noise, but it remains to be demonstrated whether this is a result of selection or within-individual learning, or other processes. A critical information gap centres on the strike risk of individuals in relation to their behavioural characteristics.

The aircraft we recorded and used in the present study came from a noisy aircraft, and quieter aircraft may result in different avian responses (Burger 1985). We also note that our study species is known for its intelligence and ability to discriminate among stimuli, such as, for example, among different people (Warne and Jones 2003). Thus, other species may not necessarily habituate, tolerate -or discriminate with respect to aircraft noise (Conomy et al. 1998b).

The authors declare no conflicts of interest.

The supplementary material available from the Journal’s website provides summary information on species other than the Australian magpie which were surveyed but were not used in the main analysis.