Human activities have a negative impact on natural habitats in many parts of the world (Marzluff et al. 2001, McKinney 2002). The growth of human populations is linked to urban development that frequently engulfs habitats of wild animals with a variety of effects on their density, distribution and behaviour (Palmer 2003). Urban sites are occupied by wild species of animals (known as “urban tolerant” species), which are able to cope and adapt to changing conditions (Blair 1996, Partecke et al. 2006). Species living in urban conditions have adapted to the proximity of humans and show specific life history characteristics that facilitate colonization of novel habitats: high rates of innovation, a high level of risk-taking and a fast pace of life (Møller 2009). Urban sites are also characterized by higher food availability (Klausnitzer 1989, Stephan 1999). Urban animals could, therefore, live in higher densities than in natural habitats (Batten 1973, Klausnitzer 1989, Stephan 1999), using smaller territories. Urban environments are characterized by higher temperatures and precipitation (Klausnitzer 1989), offering birds longer nesting periods, which should consequently serve as a selective advantage for multicycle species (Batten 1973).

In addition to increased human density, urban environments contain a high number of domestic animals e.g., cats Felis catus and dogs Canis lupus familiaris (Sorace 2002, Møller & Ibáñez-Álamo 2012), as well as many generalist avian predators (Sorace & Gustin 2009, Díaz et al. 2013). In winter, large numbers of avian predators invade cities (e.g. Eurasian Sparrowhawks Accipiter nisus and Common Kestrels Falco tinnunculus and Common Buzzards Buteo buteo). In Prague, Common Kestrels, Common Buzzards and Northern Goshawks Accipiter gentilis were recorded as regular breeding birds (Št'astný et al. 2006), whereas urban Sparrowhawks nested in higher breeding densities than in the surrounding woodlands, with an average density of 600 ha/pair, but locally even of 150–210 ha/pair (Peške 1987, 1999). Urban areas are also occupied by high numbers of cats (Baker et al. 2005) and dogs (more than 86,400 individuals were registered in Prague in 2010, i.e. an average density of 174 individuals/km²; Blažek 2010). In fact, mammalian predators in urban areas reach higher densities than avian predators. Cat predation is more common in urban than in rural habitats (Baker et al. 2008; Sims et al. 2008; Stracey 2011), contrary to avian predation (Stracey 2011).

Birds react to potential predators, including man, by adopting escape behaviour (Scheuerlein et al. 2001, Møller & Ibáñez-Álamo 2012). Flight initiation distance (FID) can serve as a simple way of testing predation risk under the assumption that stress responses are costly, especially if elicited frequently. Urban conditions could select for a reduction in sensitivity to omnipresent stressors such as humans (Kenney & Knight 1992). Analysis of intra- or interspecific differences in wariness as reflected by flight distances will, therefore, provide important information about the ability of individual/species to adapt to local conditions (Rodgers & Smith 1997, Blumstein et al. 2003, Blumstein 2006, Møller 2008).

The objective of this study was to test how FID changed among 20 species of birds in relation to pedestrian activity. I assumed that individuals or species which are exposed to higher levels of human disturbance (expressed as pedestrian density) should have smaller flight distances due to habituation (Webb & Blumstein 2005, Eason et al. 2006) or may distribute themselves among breeding sites depending on their individual susceptibility to the local level of human disturbance (Carrete & Telia 2010).

MATERIAL & METHODS

Statistical analyses

FID and HPI were log₁₀ transformed to achieve distributions that did not differ from normality. In all analyses, I assumed that estimates based on larger sample sizes are closer to the true population estimate and, therefore, analyses were weighted by sample size to adjust for unequal sampling effort among observed species.

All measured variables such as species, HPI, and habitat type were inserted in a General Linear Model (GLM) to detect their effect on the resulting FID. When GLM results showed that FID is a species-specific trait, as per Blumstein et al. (2003) and Møller (2008), I tested if habitat type or HPI played an important role in the resulting FID, using a paired t-test. Although data for 20 species were obtained, the sample size was reduced in some subsequent analyses. Firstly, I compared levels of intraspecific variation between seventeen species distributed to different levels of pedestrian density. Secondly, I compared paired populations of the fourteen species from the different parts of same habitat (park), which distinctly differed in HPI (1.33 and 40.73). Thirdly, I compared paired populations of the thirteen species from different habitats (park and green area between prefabs), which had a very similar HPI (1.33 and 1.73).

Table 1.

Flight initiation distance (FID) in relation to species, pedestrian density (HPI), habitat, and interactions between species, pedestrian density and habitat. Three-way and pedestrian density × habitat interactions were excluded as they were not significant.

Analyses were done using SPSS 20.0 and MS Excel 2007. A Linear model was fitted using the GLM univariate method. I interpret two-sided P-values of less than 0.05 as significant. Residuals from GLM were visually checked and did not deviate substantially from normal. Moreover, I reported partial eta-squared value as a measure of effect size (Cohen 1988). I used suggested norms for partial eta-squared effect sizes: small effect (0.01), intermediate effect (0.06) and large effect (0.14; Stevens 2002).

RESULTS

I recorded 2117 flight distances, and present data for 20 species. The overall number of estimates per species ranged from 1 to 389 (median = 31.5). Mean FID ranged from 1 to 38 m (mean ± SD: 6.70 ± 6.31). The four species with the shortest relative flight distances were Rock Pigeon Columba livia, Common Chaffinch Fringilla coelebs, Blue Tit Cyanistes caeruleus and Common Wood Pigeon Columba palumbus. The four species with the longest relative flight distances were Common Starling Sturnus vulgaris, European Green Woodpecker Picus viridis, Common Kestrel and Eurasian Sparrowhawk (Appendix 1).

Figure 1.

Flight initiation distance (FID) in relation to pedestrian density (HPI) for different populations and species of birds (all data were log₁₀-transformed). The line shows a linear regression.

Figure 2.

Flight distances across different habitats: city forest (CF), gardens (G), built-up area (BA), green area between prefabs (GAP) and park (P).

Log₁₀-transformed mean FIDs of ten bird populations (combined from all localities) varied among observed species, habitat types and HPI (Table 1). GLM results were highly significant (F _46,2070 = 243.3, r ² = 0.84, P < 0.0001, partial eta-squared = 0.844). Log₁₀-transformed mean FID of bird populations strongly correlated with HPI and decreased with increasing pedestrian density (Figure 1). As expected, FID strongly differed between species. I also found considerable variation in FID among individuals within the same species (Fisher's LSD test, F _1,16 = 39.4, P < 0.0001). Likewise, FID weakly correlated with habitat type. However, post-hoc analyses indicated that FID did not significantly differ between city forest and gardens (Tukey's test, P = 0.98), unlike the other habitat types (Tukey's test, P < 0.01; Figure 2).

In order to test the impact of pedestrian density on FID within various species I compared two parts of the same park with a similar species composition but different HPI. The analysis of paired populations showed that fourteen species responded to pedestrian density in different ways (Table 2). Contrarily, thirteen species in different habitats with similar HPIs did not significantly differ in their FIDs (mean ± SD for gardens: 12.88 m ± 3.53, built-up area 5.53 ± 2.80; paired t-test based on log₁₀-transformed data: mean: t = 1.35, df = 12, P = 0.20; SD: t = 0.31, df = 12, P = 0.76).

DISCUSSION

In many studies, pedestrian density has been found to correlate negatively with FID (Lord et al. 2001, Miller et al. 2001, Webb & Blumstein 2005, Møller et al. 2013). Similar to my observations, Rodgers & Smith (1997), Rodgers & Schwikert (2002), Webb & Blumstein (2005), Møller (2008), Møller et al. (2008, 2013) and Díaz et al. (2013) found differences in FID between and within species correlating to various levels of disturbance or human density. Several hypotheses have been proposed to explain this phenomenon by pointing to adaptive or heritable components of such behaviour. The most frequent explanation for changes in FID is that birds habituate to the local level of disturbance and became more tolerant in heavily disturbed areas, as shown in many previous studies (Lord et al. 2001, Runyan & Blumstein 2004, Stankowich & Blumstein 2005, Eason et al. 2006). An alternative explanation, which does not exclude the first hypothesis, suggests that the distribution of birds is linked, besides to food supply and nesting opportunities, to their tolerance of human disturbance, and thus must have some heritable components (Carrete & Telia 2010). The third hypothesis, suggested by Webb & Blumstein (2005), almost completely lacks empirical evidence and proposes that birds may remember areas with high disturbance levels and adjust their degree of tolerance depending on where they are at the time.

Table 2.

Mean and SD of FID (m) for pairs of populations of fourteen bird species in city parks with low (1) and high (2) density of pedestrians.

Many previous studies found a strong species effect on FID (Rodgers & Smith 1997, Rodgers & Schwikert 2002, Blumstein et al. 2003, Møller 2008), linked perhaps to species-specific life history traits (Johnston et al. 1997, Ghalambor & Martin 2001, Møller & Liang 2013). My finding shows that much variation in FID can be explained by species, which also supports the hypothesis that FID can be considered as a species-specific trait with important implications for management (Blumstein et al. 2003). It has been suggested that FID among species may have coevolved with life history traits, showing an increase in FID with decreasing clutch size (Møller & Liang 2013), fewer offspring (Ghalambor & Martin 2001), higher adult survival (Ghalambor & Martin 2001) and with larger body size (Blumstein 2006). Species-specific FIDs may also have coevolved with the size of sense organs, such as brain and eyes, as suggested by Møller & Erritzøe (2014); however, empirical evidence is lacking in this case.

My results indicate that habitat type did not play as important a role as other variables, although this outcome may be biased by the species used in this study (inhabiting a wide variety of habitats offered by urban environment). Large flight distances are to be expected mainly in individuals living in an environment inhabited by stable and dense populations of avian predators, e.g. in city parks or forest. Although habitat type influenced overall FID, a significantly greater proportion of the variation was explained by species or pedestrian density and, importantly, there was no significant interaction between habitat type and other variables. Moreover, additional testing of paired bird populations showed significant differences in FID only between birds which inhabited the same habitat type with different pedestrian densities. However, FID did not differ significantly between different habitat types with similar pedestrian density.

I found a strong correlation between FID and species as well as pedestrian density (both explained approximately a similar proportion of variation), and there was a significant interaction between them which explained an even higher proportion of variation than just habitat type. It has been suggested that some antipredator behaviour may be heritable (Riechert & Hedrick 1990), but behaviour like FID may be much more phenotypically plastic. Although FID seems to represent species-specific antipredator behaviour (Blumstein et al. 2003) and could have heritable components (Murphey et al. 1980, Carrete & Telia 2010), it is more likely to be experience-dependent and to vary within species with predator type or density (Møller & Ibáñez-Álamo 2012, Díaz et al. 2013), season (Møller et al. 2013), food availability (Fernández-Juricic et al. 2002), habitat type (Díaz et al. 2013, Møller & Liang 2013; but see Blumstein et al. 2006) or individual's age (Eason et al. 2006). The strong effect of pedestrian density clearly supports the idea that FID may be primarily affected by the bird’s individuality and its ability to adapt to local levels of disturbance by learning (Eason et al. 2006). Individuals which are not able to adapt to human presence will lose access to resources and incur costs related to flight, increased vigilance (Ward & Low 1997) and reduced foraging (Lord et al. 1997). Consequently, the resistance to human-related stress response will determine what bird species are more likely to live in urban environments. Moreover, the large intraspecific variability suggest that some birds are able to change their habits plastically under local conditions in order to reduce costs of escape behaviour with predation risk and/or distribution and abundance of available food resources (Webb & Blumstein 2005, Møller et al. 2013). Risk assessment thus appears to be a highly plastic process for predator-savvy prey.

Although urban environments were relatively free of avian predators a few decades ago, nowadays several raptors (mostly hawks Accipiter spp., Common Kestrel and Common Buzzard) have colonised cities (reviewed in Chace & Walsh 2006; Peške 1987, 1999, Klausnitzer 1989, Díaz et al. 2013, Mikula et al. 2013). Variation in food supply and predators’ habituation to novel nesting substrates may further influence the spectrum of predators in urban habitats and subsequently change predation risk for urban birds (reviewed in Chace & Walsh 2006). Despite large populations of avian predators inhabiting cities (Peške 1987, 1999), urban birds are threatened more often by mammalian predators such as cats and dogs (Møller & Ibáñez-Álamo 2012) or by disturbance from humans (Fernández-Juricic 2000). Stankowich & Blumstein (2005) found that predator traits that were associated with greater risk such as speed or directness of approach, increased prey escaping distance across animal species. Urban birds are typically exposed to slower and more indirectly moving cats, dogs or humans and, therefore, could have consistently lower FIDs than birds from rural sites (more often threatened by avian predators). Avian communities could be differentially affected when city districts differ in the composition (predominantly mammalian or avian) and density of their predator fauna (Lepczyk et al. 2004, Chace & Walsh 2006).

Monitoring predation risk involves cognitive mechanisms allowing birds to distinguish between the dangerous approach of avian predators, the less dangerous approach of mammalian predators, and the marginally dangerous approach of humans. Moreover, similar cognitive mechanisms may allow birds to respond optimally with respect to distribution and accessibility of food resources (Møller et al. 2013). In this way, the birds' ability to learn what level of risk is acceptable and what is not could be connected with relative brain size. Large-brained birds seem to be more successful at establishing themselves in novel conditions by adjusting breeding behaviour to the prevailing ecological conditions (Brooker et al. 1998), finding novel food resources (Sol et al. 2005) or reducing adult mortality (Sol et al. 2007). Factors which influence cognitive abilities of species, and subsequently FID, could be used as an important predictor of variability between species. Apparently, species with better cognitive abilities rapidly adjust their response to predation risk and exhibit a larger intraspecific variation depending on local conditions.

Finally, my results confirm that pedestrian density can be used as a predictor of individual risk taking. Moreover, results show that FID differed between and within species; this conclusion supports an idea that FID may have some heritable components but can be changed plastically depending on local conditions. This inter- and intraspecific variation suggests: (1) data for interspecific analyses should be combined from different sites in future studies, (2) intraspecific differences in FID could be used as an important predictor of a species's ability to optimize their escape behaviour and (3) the ability to change antipredator behaviour seems to be plastic and, therefore, further study is needed to distinguish which mechanism allows this phenomenal variety of adaptive traits.